arXiv Daily Digest - 2026-02-18

Fair scores reward ensemble forecast members that behave like samples from the same distribution as the verifying observations. They are therefore an attractive choice as loss functions to train data-driven ensemble forecasts or post-processing methods when large training ensembles are either unavailable or computationally prohibitive. The adjusted continuous ranked probability score (aCRPS) is fair and unbiased with respect to ensemble size, provided forecast members are exchangeable and interpretable as conditionally independent draws from an underlying predictive distribution. However, distribution-aware post-processing methods that introduce structural dependency between members can violate this assumption, rendering aCRPS unfair. We demonstrate this effect using two approaches designed to minimize the expected aCRPS of a finite ensemble: (1) a linear member-by-member calibration, which couples members through a common dependency on the sample ensemble mean, and (2) a deep-learning method, which couples members via transformer self-attention across the ensemble dimension. In both cases, the results are sensitive to ensemble size and apparent gains in aCRPS can correspond to systematic unreliability characterized by over-dispersion. We introduce trajectory transformers as a proof-of-concept that ensemble-size independence can be achieved. This approach is an adaptation of the Post-processing Ensembles with Transformers (PoET) framework and applies self-attention over lead time while preserving the conditional independence required by aCRPS. When applied to weekly mean $T_{2m}$ forecasts from the ECMWF subseasonal forecasting system, this approach successfully reduces systematic model biases whilst also improving or maintaining forecast reliability regardless of the ensemble size used in training (3 vs 9 members) or real-time forecasts (9 vs 100 members).

The superficial alignment hypothesis (SAH) posits that large language models learn most of their knowledge during pre-training, and that post-training merely surfaces this knowledge. The SAH, however, lacks a precise definition, which has led to (i) different and seemingly orthogonal arguments supporting it, and (ii) important critiques to it. We propose a new metric called task complexity: the length of the shortest program that achieves a target performance on a task. In this framework, the SAH simply claims that pre-trained models drastically reduce the complexity of achieving high performance on many tasks. Our definition unifies prior arguments supporting the SAH, interpreting them as different strategies to find such short programs. Experimentally, we estimate the task complexity of mathematical reasoning, machine translation, and instruction following; we then show that these complexities can be remarkably low when conditioned on a pre-trained model. Further, we find that pre-training enables access to strong performances on our tasks, but it can require programs of gigabytes of length to access them. Post-training, on the other hand, collapses the complexity of reaching this same performance by several orders of magnitude. Overall, our results highlight that task adaptation often requires surprisingly little information -- often just a few kilobytes.

Learning generalist policies capable of accomplishing a plethora of everyday tasks remains an open challenge in dexterous manipulation. In particular, collecting large-scale manipulation data via real-world teleoperation is expensive and difficult to scale. While learning in simulation provides a feasible alternative, designing multiple task-specific environments and rewards for training is similarly challenging. We propose Dex4D, a framework that instead leverages simulation for learning task-agnostic dexterous skills that can be flexibly recomposed to perform diverse real-world manipulation tasks. Specifically, Dex4D learns a domain-agnostic 3D point track conditioned policy capable of manipulating any object to any desired pose. We train this 'Anypose-to-Anypose' policy in simulation across thousands of objects with diverse pose configurations, covering a broad space of robot-object interactions that can be composed at test time. At deployment, this policy can be zero-shot transferred to real-world tasks without finetuning, simply by prompting it with desired object-centric point tracks extracted from generated videos. During execution, Dex4D uses online point tracking for closed-loop perception and control. Extensive experiments in simulation and on real robots show that our method enables zero-shot deployment for diverse dexterous manipulation tasks and yields consistent improvements over prior baselines. Furthermore, we demonstrate strong generalization to novel objects, scene layouts, backgrounds, and trajectories, highlighting the robustness and scalability of the proposed framework.

While recent advances in humanoid locomotion have achieved stable walking on varied terrains, capturing the agility and adaptivity of highly dynamic human motions remains an open challenge. In particular, agile parkour in complex environments demands not only low-level robustness, but also human-like motion expressiveness, long-horizon skill composition, and perception-driven decision-making. In this paper, we present Perceptive Humanoid Parkour (PHP), a modular framework that enables humanoid robots to autonomously perform long-horizon, vision-based parkour across challenging obstacle courses. Our approach first leverages motion matching, formulated as nearest-neighbor search in a feature space, to compose retargeted atomic human skills into long-horizon kinematic trajectories. This framework enables the flexible composition and smooth transition of complex skill chains while preserving the elegance and fluidity of dynamic human motions. Next, we train motion-tracking reinforcement learning (RL) expert policies for these composed motions, and distill them into a single depth-based, multi-skill student policy, using a combination of DAgger and RL. Crucially, the combination of perception and skill composition enables autonomous, context-aware decision-making: using only onboard depth sensing and a discrete 2D velocity command, the robot selects and executes whether to step over, climb onto, vault or roll off obstacles of varying geometries and heights. We validate our framework with extensive real-world experiments on a Unitree G1 humanoid robot, demonstrating highly dynamic parkour skills such as climbing tall obstacles up to 1.25m (96% robot height), as well as long-horizon multi-obstacle traversal with closed-loop adaptation to real-time obstacle perturbations.

A central challenge in large language model (LLM) editing is capability preservation: methods that successfully change targeted behavior can quietly game the editing proxy and corrupt general capabilities, producing degenerate behaviors reminiscent of proxy/reward hacking. We present CrispEdit, a scalable and principled second-order editing algorithm that treats capability preservation as an explicit constraint, unifying and generalizing several existing editing approaches. CrispEdit formulates editing as constrained optimization and enforces the constraint by projecting edit updates onto the low-curvature subspace of the capability-loss landscape. At the crux of CrispEdit is expressing capability constraint via Bregman divergence, whose quadratic form yields the Gauss-Newton Hessian exactly and even when the base model is not trained to convergence. We make this second-order procedure efficient at the LLM scale using Kronecker-factored approximate curvature (K-FAC) and a novel matrix-free projector that exploits Kronecker structure to avoid constructing massive projection matrices. Across standard model-editing benchmarks, CrispEdit achieves high edit success while keeping capability degradation below 1% on average across datasets, significantly improving over prior editors.

Machine learning surrogates are increasingly used in engineering to accelerate costly simulations, yet distribution shifts between training and deployment often cause severe performance degradation (e.g., unseen geometries or configurations). Test-Time Adaptation (TTA) can mitigate such shifts, but existing methods are largely developed for lower-dimensional classification with structured outputs and visually aligned input-output relationships, making them unstable for the high-dimensional, unstructured and regression problems common in simulation. We address this challenge by proposing a TTA framework based on storing maximally informative (D-optimal) statistics, which jointly enables stable adaptation and principled parameter selection at test time. When applied to pretrained simulation surrogates, our method yields up to 7% out-of-distribution improvements at negligible computational cost. To the best of our knowledge, this is the first systematic demonstration of effective TTA for high-dimensional simulation regression and generative design optimization, validated on the SIMSHIFT and EngiBench benchmarks.

Recent advances in deep reinforcement learning (RL) have achieved strong results on high-dimensional control tasks, but applying RL to reachability problems raises a fundamental mismatch: reachability seeks to maximize the set of states from which a system remains safe indefinitely, while RL optimizes expected returns over a user-specified distribution. This mismatch can result in policies that perform poorly on low-probability states that are still within the safe set. A natural alternative is to frame the problem as a robust optimization over a set of initial conditions that specify the initial state, dynamics and safe set, but whether this problem has a solution depends on the feasibility of the specified set, which is unknown a priori. We propose Feasibility-Guided Exploration (FGE), a method that simultaneously identifies a subset of feasible initial conditions under which a safe policy exists, and learns a policy to solve the reachability problem over this set of initial conditions. Empirical results demonstrate that FGE learns policies with over 50% more coverage than the best existing method for challenging initial conditions across tasks in the MuJoCo simulator and the Kinetix simulator with pixel observations.

As insufficient data volume and quality remain the key impediments to the adoption of modern subsymbolic AI, techniques of synthetic data generation are in high demand. Simulation offers an apt, systematic approach to generating diverse synthetic data. This chapter introduces the reader to the key concepts, benefits, and challenges of simulation-based synthetic data generation for AI training purposes, and to a reference framework to describe, design, and analyze digital twin-based AI simulation solutions.

Compact pretrained bidirectional encoders remain the backbone of industrial NLP under tight compute and memory budgets. Their effectiveness stems from self-attention's ability to deliver high-quality bidirectional contextualization with sequence-level parallelism, as popularized by BERT-style architectures. Recently, Avey was introduced as an autoregressive, attention-free alternative that naturally admits an encoder-only adaptation. In this paper, we reformulate Avey for the encoder-only paradigm and propose several innovations to its architecture, including decoupled static and dynamic parameterizations, stability-oriented normalization, and neural compression. Results show that this reformulated architecture compares favorably to four widely used Transformer-based encoders, consistently outperforming them on standard token-classification and information-retrieval benchmarks while scaling more efficiently to long contexts.

Clinical deployment of chest radiograph classifiers requires models that can be updated as new datasets become available without retraining on previously ob- served data or degrading validated performance. We study, for the first time, a task-incremental continual learning setting for chest radiograph classification, in which heterogeneous chest X-ray datasets arrive sequentially and task identifiers are unavailable at inference. We propose a continual adapter-based routing learning strategy for Chest X-rays (CARL-XRay) that maintains a fixed high-capacity backbone and incrementally allocates lightweight task-specific adapters and classifier heads. A latent task selector operates on task-adapted features and leverages both current and historical context preserved through compact prototypes and feature-level experience replay. This design supports stable task identification and adaptation across sequential updates while avoiding raw-image storage. Experiments on large-scale public chest radiograph datasets demonstrate robust performance retention and reliable task-aware inference under continual dataset ingestion. CARL-XRay outperforms joint training under task-unknown deployment, achieving higher routing accuracy (75.0\% vs.\ 62.5\%), while maintaining competitive diagnostic performance with AUROC of 0.74 in the oracle setting with ground-truth task identity and 0.75 under task-unknown inference, using significantly fewer trainable parameters. Finally, the proposed framework provides a practical alternative to joint training and repeated full retraining in continual clinical deployment.

Online platforms require robust systems to enforce content safety policies at scale. A critical component of these systems is the ability to evaluate the quality of moderation decisions made by both human agents and Large Language Models (LLMs). However, this evaluation is challenging due to the inherent trade-offs between cost, scale, and trustworthiness, along with the complexity of evolving policies. To address this, we present a comprehensive Decision Quality Evaluation Framework developed and deployed at Pinterest. The framework is centered on a high-trust Golden Set (GDS) curated by subject matter experts (SMEs), which serves as a ground truth benchmark. We introduce an automated intelligent sampling pipeline that uses propensity scores to efficiently expand dataset coverage. We demonstrate the framework's practical application in several key areas: benchmarking the cost-performance trade-offs of various LLM agents, establishing a rigorous methodology for data-driven prompt optimization, managing complex policy evolution, and ensuring the integrity of policy content prevalence metrics via continuous validation. The framework enables a shift from subjective assessments to a data-driven and quantitative practice for managing content safety systems.

Fine-tuning aligned language models on benign tasks unpredictably degrades safety guardrails, even when training data contains no harmful content and developers have no adversarial intent. We show that the prevailing explanation, that fine-tuning updates should be orthogonal to safety-critical directions in high-dimensional parameter space, offers false reassurance: we show this orthogonality is structurally unstable and collapses under the dynamics of gradient descent. We then resolve this through a novel geometric analysis, proving that alignment concentrates in low-dimensional subspaces with sharp curvature, creating a brittle structure that first-order methods cannot detect or defend. While initial fine-tuning updates may indeed avoid these subspaces, the curvature of the fine-tuning loss generates second-order acceleration that systematically steers trajectories into alignment-sensitive regions. We formalize this mechanism through the Alignment Instability Condition, three geometric properties that, when jointly satisfied, lead to safety degradation. Our main result establishes a quartic scaling law: alignment loss grows with the fourth power of training time, governed by the sharpness of alignment geometry and the strength of curvature coupling between the fine-tuning task and safety-critical parameters. These results expose a structural blind spot in the current safety paradigm. The dominant approaches to safe fine-tuning address only the initial snapshot of a fundamentally dynamic problem. Alignment fragility is not a bug to be patched; it is an intrinsic geometric property of gradient descent on curved manifolds. Our results motivate the development of curvature-aware methods, and we hope will further enable a shift in alignment safety analysis from reactive red-teaming to predictive diagnostics for open-weight model deployment.

The Computing Continuum (CC) integrates different layers of processing infrastructure, from Edge to Cloud, to optimize service quality through ubiquitous and reliable computation. Compared to central architectures, however, heterogeneous and dynamic infrastructure increases the complexity for service orchestration. To guide research, this article first summarizes structural problems of the CC, and then, envisions an ideal solution for autonomous service orchestration across the CC. As one instantiation, we show how Active Inference, a concept from neuroscience, can support self-organizing services in continuously interpreting their environment to optimize service quality. Still, we conclude that no existing solution achieves our vision, but that research on service orchestration faces several structural challenges. Most notably: provide standardized simulation and evaluation environments for comparing the performance of orchestration mechanisms. Together, the challenges outline a research roadmap toward resilient and scalable service orchestration in the CC.

Accurate representation of building semantics, encompassing both generic object types and specific subtypes, is essential for effective AI model training in the architecture, engineering, construction, and operation (AECO) industry. Conventional encoding methods (e.g., one-hot) often fail to convey the nuanced relationships among closely related subtypes, limiting AI's semantic comprehension. To address this limitation, this study proposes a novel training approach that employs large language model (LLM) embeddings (e.g., OpenAI GPT and Meta LLaMA) as encodings to preserve finer distinctions in building semantics. We evaluated the proposed method by training GraphSAGE models to classify 42 building object subtypes across five high-rise residential building information models (BIMs). Various embedding dimensions were tested, including original high-dimensional LLM embeddings (1,536, 3,072, or 4,096) and 1,024-dimensional compacted embeddings generated via the Matryoshka representation model. Experimental results demonstrated that LLM encodings outperformed the conventional one-hot baseline, with the llama-3 (compacted) embedding achieving a weighted average F1-score of 0.8766, compared to 0.8475 for one-hot encoding. The results underscore the promise of leveraging LLM-based encodings to enhance AI's ability to interpret complex, domain-specific building semantics. As the capabilities of LLMs and dimensionality reduction techniques continue to evolve, this approach holds considerable potential for broad application in semantic elaboration tasks throughout the AECO industry.

A growing literature uses large language models (LLMs) as synthetic participants to generate cost-effective and nearly instantaneous responses in social science experiments. However, there is limited guidance on when such simulations support valid inference about human behavior. We contrast two strategies for obtaining valid estimates of causal effects and clarify the assumptions under which each is suitable for exploratory versus confirmatory research. Heuristic approaches seek to establish that simulated and observed human behavior are interchangeable through prompt engineering, model fine-tuning, and other repair strategies designed to reduce LLM-induced inaccuracies. While useful for many exploratory tasks, heuristic approaches lack the formal statistical guarantees typically required for confirmatory research. In contrast, statistical calibration combines auxiliary human data with statistical adjustments to account for discrepancies between observed and simulated responses. Under explicit assumptions, statistical calibration preserves validity and provides more precise estimates of causal effects at lower cost than experiments that rely solely on human participants. Yet the potential of both approaches depends on how well LLMs approximate the relevant populations. We consider what opportunities are overlooked when researchers focus myopically on substituting LLMs for human participants in a study.

The success of Large Language Models (LLMs) has established that scaling compute, through joint increases in model capacity and dataset size, is the primary driver of performance in modern machine learning. While machine learning has long been an integral component of High Energy Physics (HEP) data analysis workflows, the compute used to train state-of-the-art HEP models remains orders of magnitude below that of industry foundation models. With scaling laws only beginning to be studied in the field, we investigate neural scaling laws for boosted jet classification using the public JetClass dataset. We derive compute optimal scaling laws and identify an effective performance limit that can be consistently approached through increased compute. We study how data repetition, common in HEP where simulation is expensive, modifies the scaling yielding a quantifiable effective dataset size gain. We then study how the scaling coefficients and asymptotic performance limits vary with the choice of input features and particle multiplicity, demonstrating that increased compute reliably drives performance toward an asymptotic limit, and that more expressive, lower-level features can raise the performance limit and improve results at fixed dataset size.

Evaluating the quality of automatically generated text often relies on LLM-as-a-judge (LLM-judge) methods. While effective, these approaches are computationally expensive and require post-processing. To address these limitations, we build upon ParaPLUIE, a perplexity-based LLM-judge metric that estimates confidence over ``Yes/No'' answers without generating text. We introduce *-PLUIE, task specific prompting variants of ParaPLUIE and evaluate their alignment with human judgement. Our experiments show that personalised *-PLUIE achieves stronger correlations with human ratings while maintaining low computational cost.

In the realm of multi-agent systems, the challenge of \emph{partial observability} is a critical barrier to effective coordination and decision-making. Existing approaches, such as belief state estimation and inter-agent communication, often fall short. Belief-based methods are limited by their focus on past experiences without fully leveraging global information, while communication methods often lack a robust model to effectively utilize the auxiliary information they provide. To solve this issue, we propose Global State Diffusion Algorithm~(GlobeDiff) to infer the global state based on the local observations. By formulating the state inference process as a multi-modal diffusion process, GlobeDiff overcomes ambiguities in state estimation while simultaneously inferring the global state with high fidelity. We prove that the estimation error of GlobeDiff under both unimodal and multi-modal distributions can be bounded. Extensive experimental results demonstrate that GlobeDiff achieves superior performance and is capable of accurately inferring the global state.

Current research in multimodal models faces a key challenge where enhancing generative capabilities often comes at the expense of understanding, and vice versa. We analyzed this trade-off and identify the primary cause might be the potential conflict between generation and understanding, which creates a competitive dynamic within the model. To address this, we propose the Reason-Reflect-Refine (R3) framework. This innovative algorithm re-frames the single-step generation task into a multi-step process of "generate-understand-regenerate". By explicitly leveraging the model's understanding capability during generation, we successfully mitigate the optimization dilemma, achieved stronger generation results and improved understanding ability which are related to the generation process. This offers valuable insights for designing next-generation unified multimodal models. Code is available at https://github.com/sen-ye/R3.

Multimodal Large Language Models (mLLMs) are often used to answer questions in structured data such as tables in Markdown, JSON, and images. While these models can often give correct answers, users also need to know where those answers come from. In this work, we study structured data attribution/citation, which is the ability of the models to point to the specific rows and columns that support an answer. We evaluate several mLLMs across different table formats and prompting strategies. Our results show a clear gap between question answering and evidence attribution. Although question answering accuracy remains moderate, attribution accuracy is much lower, near random for JSON inputs, across all models. We also find that models are more reliable at citing rows than columns, and struggle more with textual formats than images. Finally, we observe notable differences across model families. Overall, our findings show that current mLLMs are unreliable at providing fine-grained, trustworthy attribution for structured data, which limits their usage in applications requiring transparency and traceability.

Robot-assisted feeding enables people with disabilities who require assistance eating to enjoy a meal independently and with dignity. However, existing systems have only been tested in-lab or in-home, leaving in-the-wild social dining contexts (e.g., restaurants) largely unexplored. Designing a robot for such contexts presents unique challenges, such as dynamic and unsupervised dining environments that a robot needs to account for and respond to. Through speculative participatory design with people with disabilities, supported by semi-structured interviews and a custom AI-based visual storyboarding tool, we uncovered ideal scenarios for in-the-wild social dining. Our key insight suggests that such systems should: embody the principles of a white glove service where the robot (1) supports multimodal inputs and unobtrusive outputs; (2) has contextually sensitive social behavior and prioritizes the user; (3) has expanded roles beyond feeding; (4) adapts to other relationships at the dining table. Our work has implications for in-the-wild and group contexts of robot-assisted feeding.

We present GLM-5, a next-generation foundation model designed to transition the paradigm of vibe coding to agentic engineering. Building upon the agentic, reasoning, and coding (ARC) capabilities of its predecessor, GLM-5 adopts DSA to significantly reduce training and inference costs while maintaining long-context fidelity. To advance model alignment and autonomy, we implement a new asynchronous reinforcement learning infrastructure that drastically improves post-training efficiency by decoupling generation from training. Furthermore, we propose novel asynchronous agent RL algorithms that further improve RL quality, enabling the model to learn from complex, long-horizon interactions more effectively. Through these innovations, GLM-5 achieves state-of-the-art performance on major open benchmarks. Most critically, GLM-5 demonstrates unprecedented capability in real-world coding tasks, surpassing previous baselines in handling end-to-end software engineering challenges. Code, models, and more information are available at https://github.com/zai-org/GLM-5.

With the rapid adoption of large language models (LLMs) in automated code refactoring, assessing and ensuring functional equivalence between LLM-generated refactoring and the original implementation becomes critical. While prior work typically relies on predefined test cases to evaluate correctness, in this work, we leverage differential fuzzing to check functional equivalence in LLM-generated code refactorings. Unlike test-based evaluation, a differential fuzzing-based equivalence checker needs no predefined test cases and can explore a much larger input space by executing and comparing thousands of automatically generated test inputs. In a large-scale evaluation of six LLMs (CodeLlama, Codestral, StarChat2, Qwen-2.5, Olmo-3, and GPT-4o) across three datasets and two refactoring types, we find that LLMs show a non-trivial tendency to alter program semantics, producing 19-35% functionally non-equivalent refactorings. Our experiments further demonstrate that about 21% of these non-equivalent refactorings remain undetected by the existing test suites of the three evaluated datasets. Collectively, the findings of this study imply that reliance on existing tests might overestimate functional equivalence in LLM-generated code refactorings, which remain prone to semantic divergence.

While Multimodal Large Language Models (MLLMs) perform strongly on single-turn chart generation, their ability to support real-world exploratory data analysis remains underexplored. In practice, users iteratively refine visualizations through multi-turn interactions that require maintaining common ground, tracking prior edits, and adapting to evolving preferences. We introduce ChartEditBench, a benchmark for incremental, visually grounded chart editing via code, comprising 5,000 difficulty-controlled modification chains and a rigorously human-verified subset. Unlike prior one-shot benchmarks, ChartEditBench evaluates sustained, context-aware editing. We further propose a robust evaluation framework that mitigates limitations of LLM-as-a-Judge metrics by integrating execution-based fidelity checks, pixel-level visual similarity, and logical code verification. Experiments with state-of-the-art MLLMs reveal substantial degradation in multi-turn settings due to error accumulation and breakdowns in shared context, with strong performance on stylistic edits but frequent execution failures on data-centric transformations. ChartEditBench, establishes a challenging testbed for grounded, intent-aware multimodal programming.

Online sexism appears in various forms, which makes its detection challenging. Although automated tools can enhance the identification of sexist content, they are often restricted to binary classification. Consequently, more subtle manifestations of sexism may remain undetected due to the lack of fine-grained, context-sensitive labels. To address this issue, we make the following contributions: (1) we present FineMuSe, a new multimodal sexism detection dataset in Spanish that includes both binary and fine-grained annotations; (2) we introduce a comprehensive hierarchical taxonomy that encompasses forms of sexism, non-sexism, and rhetorical devices of irony and humor; and (3) we evaluate a wide range of LLMs for both binary and fine-grained sexism detection. Our findings indicate that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism; however, they struggle to capture co-occurring sexist types when these are conveyed through visual cues.

A natural and informal approach to verifiable (or zero-knowledge) ML inference over floating-point data is: ``prove that each layer was computed correctly up to tolerance $δ$; therefore the final output is a reasonable inference result''. This short note gives a simple counterexample showing that this inference is false in general: for any neural network, we can construct a functionally equivalent network for which adversarially chosen approximation-magnitude errors in individual layer computations suffice to steer the final output arbitrarily (within a prescribed bounded range).

Low-resource languages pose persistent challenges for Natural Language Processing tasks such as lemmatization and part-of-speech (POS) tagging. This paper investigates the capacity of recent large language models (LLMs), including GPT-4 variants and open-weight Mistral models, to address these tasks in few-shot and zero-shot settings for four historically and linguistically diverse under-resourced languages: Ancient Greek, Classical Armenian, Old Georgian, and Syriac. Using a novel benchmark comprising aligned training and out-of-domain test corpora, we evaluate the performance of foundation models across lemmatization and POS-tagging, and compare them with PIE, a task-specific RNN baseline. Our results demonstrate that LLMs, even without fine-tuning, achieve competitive or superior performance in POS-tagging and lemmatization across most languages in few-shot settings. Significant challenges persist for languages characterized by complex morphology and non-Latin scripts, but we demonstrate that LLMs are a credible and relevant option for initiating linguistic annotation tasks in the absence of data, serving as an effective aid for annotation.

On two-sided matching platforms such as online dating and recruiting, recommendation algorithms often aim to maximize the total number of matches. However, this objective creates an imbalance, where some users receive far too many matches while many others receive very few and eventually abandon the platform. Retaining users is crucial for many platforms, such as those that depend heavily on subscriptions. Some may use fairness objectives to solve the problem of match maximization. However, fairness in itself is not the ultimate objective for many platforms, as users do not suddenly reward the platform simply because exposure is equalized. In practice, where user retention is often the ultimate goal, casually relying on fairness will leave the optimization of retention up to luck. In this work, instead of maximizing matches or axiomatically defining fairness, we formally define the new problem setting of maximizing user retention in two-sided matching platforms. To this end, we introduce a dynamic learning-to-rank (LTR) algorithm called Matching for Retention (MRet). Unlike conventional algorithms for two-sided matching, our approach models user retention by learning personalized retention curves from each user's profile and interaction history. Based on these curves, MRet dynamically adapts recommendations by jointly considering the retention gains of both the user receiving recommendations and those who are being recommended, so that limited matching opportunities can be allocated where they most improve overall retention. Naturally but importantly, empirical evaluations on synthetic and real-world datasets from a major online dating platform show that MRet achieves higher user retention, since conventional methods optimize matches or fairness rather than retention.

This paper presents the first demonstration of a viable, ultra-fast, radiation-hard machine learning (ML) application on FPGAs, which could be used in future high-energy physics experiments. We present a three-fold contribution, with the PicoCal calorimeter, planned for the LHCb Upgrade II experiment, used as a test case. First, we develop a lightweight autoencoder to compress a 32-sample timing readout, representative of that of the PicoCal, into a two-dimensional latent space. Second, we introduce a systematic, hardware-aware quantization strategy and show that the model can be reduced to 10-bit weights with minimal performance loss. Third, as a barrier to the adoption of on-detector ML is the lack of support for radiation-hard FPGAs in the High-Energy Physics community's standard ML synthesis tool, hls4ml, we develop a new backend for this library. This new back-end enables the automatic translation of ML models into High-Level Synthesis (HLS) projects for the Microchip PolarFire family of FPGAs, one of the few commercially available and radiation hard FPGAs. We present the synthesis of the autoencoder on a target PolarFire FPGA, which indicates that a latency of 25 ns can be achieved. We show that the resources utilized are low enough that the model can be placed within the inherently protected logic of the FPGA. Our extension to hls4ml is a significant contribution, paving the way for broader adoption of ML on FPGAs in high-radiation environments.

Recent advances in urban region representation learning have enabled a wide range of applications in urban analytics, yet existing methods remain limited in their capabilities to generalize across cities and analytic tasks. We aim to generalize urban representation learning beyond city- and task-specific settings, towards a foundation-style model for urban analytics. To this end, we propose UrbanVerse, a model for cross-city urban representation learning and cross-task urban analytics. For cross-city generalization, UrbanVerse focuses on features local to the target regions and structural features of the nearby regions rather than the entire city. We model regions as nodes on a graph, which enables a random walk-based procedure to form "sequences of regions" that reflect both local and neighborhood structural features for urban region representation learning. For cross-task generalization, we propose a cross-task learning module named HCondDiffCT. This module integrates region-conditioned prior knowledge and task-conditioned semantics into the diffusion process to jointly model multiple downstream urban prediction tasks. HCondDiffCT is generic. It can also be integrated with existing urban representation learning models to enhance their downstream task effectiveness. Experiments on real-world datasets show that UrbanVerse consistently outperforms state-of-the-art methods across six tasks under cross-city settings, achieving up to 35.89% improvements in prediction accuracy.

Alzheimer's disease (AD) is a progressive neurodegenerative condition necessitating early and precise diagnosis to provide prompt clinical management. Given the paramount importance of early diagnosis, recent studies have increasingly focused on computer-aided diagnostic models to enhance precision and reliability. However, most graph-based approaches still rely on fixed structural designs, which restrict their flexibility and limit generalization across heterogeneous patient data. To overcome these limitations, the Meta-Relational Copula-Based Graph Attention Network (MRC-GAT) is proposed as an efficient multimodal model for AD classification tasks. The proposed architecture, copula-based similarity alignment, relational attention, and node fusion are integrated as the core components of episodic meta-learning, such that the multimodal features, including risk factors (RF), Cognitive test scores, and MRI attributes, are first aligned via a copula-based transformation in a common statistical space and then combined by a multi-relational attention mechanism. According to evaluations performed on the TADPOLE and NACC datasets, the MRC-GAT model achieved accuracies of 96.87% and 92.31%, respectively, demonstrating state-of-the-art performance compared to existing diagnostic models. Finally, the proposed model confirms the robustness and applicability of the proposed method by providing interpretability at various stages of disease diagnosis.

Integrating human expertise into machine learning systems often reduces the role of experts to labeling oracles, a paradigm that limits the amount of information exchanged and fails to capture the nuances of human judgment. We address this challenge by developing a human-in-the-loop framework to learn binary classifiers with rich query types, consisting of item ranking and exemplar selection. We first introduce probabilistic human response models for these rich queries motivated by the relationship experimentally observed between the perceived implicit score of an item and its distance to the unknown classifier. Using these models, we then design active learning algorithms that leverage the rich queries to increase the information gained per interaction. We provide theoretical bounds on sample complexity and develop a tractable and computationally efficient variational approximation. Through experiments with simulated annotators derived from crowdsourced word-sentiment and image-aesthetic datasets, we demonstrate significant reductions on sample complexity. We further extend active learning strategies to select queries that maximize information rate, explicitly balancing informational value against annotation cost. This algorithm in the word sentiment classification task reduces learning time by more than 57\% compared to traditional label-only active learning.

Humanoid motion control has witnessed significant breakthroughs in recent years, with deep reinforcement learning (RL) emerging as a primary catalyst for achieving complex, human-like behaviors. However, the high dimensionality and intricate dynamics of humanoid robots make manual motion design impractical, leading to a heavy reliance on expensive motion capture (MoCap) data. These datasets are not only costly to acquire but also frequently lack the necessary geometric context of the surrounding physical environment. Consequently, existing motion synthesis frameworks often suffer from a decoupling of motion and scene, resulting in physical inconsistencies such as contact slippage or mesh penetration during terrain-aware tasks. In this work, we present MeshMimic, an innovative framework that bridges 3D scene reconstruction and embodied intelligence to enable humanoid robots to learn coupled "motion-terrain" interactions directly from video. By leveraging state-of-the-art 3D vision models, our framework precisely segments and reconstructs both human trajectories and the underlying 3D geometry of terrains and objects. We introduce an optimization algorithm based on kinematic consistency to extract high-quality motion data from noisy visual reconstructions, alongside a contact-invariant retargeting method that transfers human-environment interaction features to the humanoid agent. Experimental results demonstrate that MeshMimic achieves robust, highly dynamic performance across diverse and challenging terrains. Our approach proves that a low-cost pipeline utilizing only consumer-grade monocular sensors can facilitate the training of complex physical interactions, offering a scalable path toward the autonomous evolution of humanoid robots in unstructured environments.

Understanding the causal effects of text on downstream outcomes is a central task in many applications. Estimating such effects requires researchers to run controlled experiments that systematically vary textual features. While large language models (LLMs) hold promise for generating text, producing and evaluating controlled variation requires more careful attention. In this paper, we present an end-to-end pipeline for the generation and causal estimation of latent textual interventions. Our work first performs hypothesis generation and steering via sparse autoencoders (SAEs), followed by robust causal estimation. Our pipeline addresses both computational and statistical challenges in text-as-treatment experiments. We demonstrate that naive estimation of causal effects suffers from significant bias as text inherently conflates treatment and covariate information. We describe the estimation bias induced in this setting and propose a solution based on covariate residualization. Our empirical results show that our pipeline effectively induces variation in target features and mitigates estimation error, providing a robust foundation for causal effect estimation in text-as-treatment settings.

Visual analogy learning enables image manipulation through demonstration rather than textual description, allowing users to specify complex transformations difficult to articulate in words. Given a triplet $\{\mathbf{a}$, $\mathbf{a}'$, $\mathbf{b}\}$, the goal is to generate $\mathbf{b}'$ such that $\mathbf{a} : \mathbf{a}' :: \mathbf{b} : \mathbf{b}'$. Recent methods adapt text-to-image models to this task using a single Low-Rank Adaptation (LoRA) module, but they face a fundamental limitation: attempting to capture the diverse space of visual transformations within a fixed adaptation module constrains generalization capabilities. Inspired by recent work showing that LoRAs in constrained domains span meaningful, interpolatable semantic spaces, we propose LoRWeB, a novel approach that specializes the model for each analogy task at inference time through dynamic composition of learned transformation primitives, informally, choosing a point in a "space of LoRAs". We introduce two key components: (1) a learnable basis of LoRA modules, to span the space of different visual transformations, and (2) a lightweight encoder that dynamically selects and weighs these basis LoRAs based on the input analogy pair. Comprehensive evaluations demonstrate our approach achieves state-of-the-art performance and significantly improves generalization to unseen visual transformations. Our findings suggest that LoRA basis decompositions are a promising direction for flexible visual manipulation. Code and data are in https://research.nvidia.com/labs/par/lorweb

Large language models achieve strong performance on many complex reasoning tasks, yet their accuracy degrades sharply on benchmarks that require compositional reasoning, including ARC-AGI-2, GPQA, MATH, BBH, and HLE. Existing methods improve reasoning by expanding token-level search through chain-of-thought prompting, self-consistency, or reinforcement learning, but they leave the model's latent representation space fixed. When the required abstraction is not already encoded in this space, performance collapses. We propose Recursive Concept Evolution (RCE), a framework that enables pretrained language models to modify their internal representation geometry during inference. RCE introduces dynamically generated low-rank concept subspaces that are spawned when representational inadequacy is detected, selected through a minimum description length criterion, merged when synergistic, and consolidated via constrained optimization to preserve stability. This process allows the model to construct new abstractions rather than recombining existing ones. We integrate RCE with Mistral-7B and evaluate it across compositional reasoning benchmarks. RCE yields 12-18 point gains on ARC-AGI-2, 8-14 point improvements on GPQA and BBH, and consistent reductions in depth-induced error on MATH and HLE.

Vision-and-Language Navigation (VLN) requires an agent to follow natural-language instructions and navigate through previously unseen environments. Recent approaches increasingly employ large language models (LLMs) as high-level navigators due to their flexibility and reasoning capability. However, prompt-based LLM navigation often suffers from inefficient decision-making, as the model must repeatedly interpret instructions from scratch and reason over noisy and verbose navigable candidates at each step. In this paper, we propose a retrieval-augmented framework to improve the efficiency and stability of LLM-based VLN without modifying or fine-tuning the underlying language model. Our approach introduces retrieval at two complementary levels. At the episode level, an instruction-level embedding retriever selects semantically similar successful navigation trajectories as in-context exemplars, providing task-specific priors for instruction grounding. At the step level, an imitation-learned candidate retriever prunes irrelevant navigable directions before LLM inference, reducing action ambiguity and prompt complexity. Both retrieval modules are lightweight, modular, and trained independently of the LLM. We evaluate our method on the Room-to-Room (R2R) benchmark. Experimental results demonstrate consistent improvements in Success Rate, Oracle Success Rate, and SPL on both seen and unseen environments. Ablation studies further show that instruction-level exemplar retrieval and candidate pruning contribute complementary benefits to global guidance and step-wise decision efficiency. These results indicate that retrieval-augmented decision support is an effective and scalable strategy for enhancing LLM-based vision-and-language navigation.

We present Lifelong Scalable Multi-Agent Realistic Testbed (LSMART), an open-source simulator to evaluate any Multi-Agent Path Finding (MAPF) algorithm in a Fleet Management System (FMS) with Automated Guided Vehicles (AGVs). MAPF aims to move a group of agents from their corresponding starting locations to their goals. Lifelong MAPF (LMAPF) is a variant of MAPF that continuously assigns new goals for agents to reach. LMAPF applications, such as autonomous warehouses, often require a centralized, lifelong system to coordinate the movement of a fleet of robots, typically AGVs. However, existing works on MAPF and LMAPF often assume simplified kinodynamic models, such as pebble motion, as well as perfect execution and communication for AGVs. Prior work has presented SMART, a software capable of evaluating any MAPF algorithms while considering agent kinodynamics, communication delays, and execution uncertainties. However, SMART is designed for MAPF, not LMAPF. Generalizing SMART to an FMS requires many more design choices. First, an FMS parallelizes planning and execution, raising the question of when to plan. Second, given planners with varying optimality and differing agent-model assumptions, one must decide how to plan. Third, when the planner fails to return valid solutions, the system must determine how to recover. In this paper, we first present LSMART, an open-source simulator that incorporates all these considerations to evaluate any MAPF algorithms in an FMS. We then provide experiment results based on state-of-the-art methods for each design choice, offering guidance on how to effectively design centralized lifelong AGV Fleet Management Systems. LSMART is available at https://smart-mapf.github.io/lifelong-smart.

Lexical semantic change detection (LSCD) increasingly relies on contextualised language model embeddings, yet most approaches still quantify change using a small set of semantic change metrics, primarily Average Pairwise Distance (APD) and cosine distance over word prototypes (PRT). We introduce Average Minimum Distance (AMD) and Symmetric Average Minimum Distance (SAMD), new measures that quantify semantic change via local correspondence between word usages across time periods. Across multiple languages, encoder models, and representation spaces, we show that AMD often provides more robust performance, particularly under dimensionality reduction and with non-specialised encoders, while SAMD excels with specialised encoders. We suggest that LSCD may benefit from considering alternative semantic change metrics beyond APD and PRT, with AMD offering a robust option for contextualised embedding-based analysis.

Across the natural and life sciences, images have become a primary measurement modality, yet the dominant analytic paradigm remains semantics-first. Structure is recovered by predicting or enforcing domain-specific labels. This paradigm fails systematically under the conditions that make image-based science most valuable, including open-ended scientific discovery, cross-sensor and cross-site comparability, and long-term monitoring in which domain ontologies and associated label sets drift culturally, institutionally, and ecologically. A deductive inversion is proposed in the form of criteria-first and semantics-later. A unified framework for criteria-first structure discovery is introduced. It separates criterion-defined, semantics-free structure extraction from downstream semantic mapping into domain ontologies or vocabularies and provides a domain-general scaffold for reproducible analysis across image-based sciences. Reproducible science requires that the first analytic layer perform criterion-driven, semantics-free structure discovery, yielding stable partitions, structural fields, or hierarchies defined by explicit optimality criteria rather than local domain ontologies. Semantics is not discarded; it is relocated downstream as an explicit mapping from the discovered structural product to a domain ontology or vocabulary, enabling plural interpretations and explicit crosswalks without rewriting upstream extraction. Grounded in cybernetics, observation-as-distinction, and information theory's separation of information from meaning, the argument is supported by cross-domain evidence showing that criteria-first components recur whenever labels do not scale. Finally, consequences are outlined for validation beyond class accuracy and for treating structural products as FAIR, AI-ready digital objects for long-term monitoring and digital twins.

Node embeddings map graph vertices into low-dimensional Euclidean spaces while preserving structural information. They are central to tasks such as node classification, link prediction, and signal reconstruction. A key goal is to design node embeddings whose dot products capture meaningful notions of node similarity induced by the graph. Graph kernels offer a principled way to define such similarities, but their direct computation is often prohibitive for large networks. Inspired by random feature methods for kernel approximation in Euclidean spaces, we introduce randomized spectral node embeddings whose dot products estimate a low-rank approximation of any specific graph kernel. We provide theoretical and empirical results showing that our embeddings achieve more accurate kernel approximations than existing methods, particularly for spectrally localized kernels. These results demonstrate the effectiveness of randomized spectral constructions for scalable and principled graph representation learning.

An ordinal preference domain is a subset of preference orders that the voters are allowed to cast in an election. We introduce and study the notion of outer diversity of a domain and evaluate its value for a number of well-known structured domains, such as the single-peaked, single-crossing, group-separable, and Euclidean ones.

Real-time conversational assistants for procedural tasks often depend on video input, which can be computationally expensive and compromise user privacy. For the first time, we propose a real-time conversational assistant that provides comprehensive guidance for a procedural task using only lightweight privacy-preserving modalities such as audio and IMU inputs from a user's wearable device to understand the context. This assistant proactively communicates step-by-step instructions to a user performing a furniture assembly task, and answers user questions. We construct a dataset containing conversations where the assistant guides the user in performing the task. On observing that an off-the-shelf language model is a very talkative assistant, we design a novel User Whim Agnostic (UWA) LoRA finetuning method which improves the model's ability to suppress less informative dialogues, while maintaining its tendency to communicate important instructions. This leads to >30% improvement in the F-score. Finetuning the model also results in a 16x speedup by eliminating the need to provide in-context examples in the prompt. We further describe how such an assistant is implemented on edge devices with no dependence on the cloud.

Learning dynamical systems through purely data-driven methods is challenging as they do not learn the underlying conservation laws that enable them to correctly generalize. Existing port-Hamiltonian neural network methods have recently been successfully applied for modeling mechanical systems. However, even though these methods are designed on power-balance principles, they usually do not consider power-preserving discretizations and often rely on Runge-Kutta numerical methods. In this work, we propose to use a second-order discrete gradient method embedded in the learning of dynamical systems with port-Hamiltonian neural networks. Numerical results are provided for three systems deliberately selected to span different ranges of dynamical behavior under control: a baseline harmonic oscillator with quadratic energy storage; a Duffing oscillator, with a non-quadratic Hamiltonian offering amplitude-dependent effects; and a self-sustained oscillator, which can stabilize in a controlled limit cycle through the incorporation of a nonlinear dissipation. We show how the use of this discrete gradient method outperforms the performance of a Runge-Kutta method of the same order. Experiments are also carried out to compare two theoretically equivalent port-Hamiltonian systems formulations and to analyze the impact of regularizing the Jacobian of port-Hamiltonian neural networks during training.

As artificial intelligence (AI) increasingly integrates into crowdfunding practices, strategic disclosure of AI involvement has become critical. Yet, empirical insights into how different disclosure strategies influence investor decisions remain limited. Drawing on signaling theory and Aristotle's rhetorical framework, we examine how mandatory AI disclosure affects crowdfunding performance and how substantive signals (degree of AI involvement) and rhetorical signals (logos/explicitness, ethos/authenticity, pathos/emotional tone) moderate these effects. Leveraging Kickstarter's mandatory AI disclosure policy as a natural experiment and four supplementary online experiments, we find that mandatory AI disclosure significantly reduces crowdfunding performance: funds raised decline by 39.8% and backer counts by 23.9% for AI-involved projects. However, this adverse effect is systematically moderated by disclosure strategy. Greater AI involvement amplifies the negative effects of AI disclosure, while high authenticity and high explicitness mitigate them. Interestingly, excessive positive emotional tone (a strategy creators might intuitively adopt to counteract AI skepticism) backfires and exacerbates negative outcomes. Supplementary randomized experiments identify two underlying mechanisms: perceived creator competence and AI washing concerns. Substantive signals primarily affect competence judgments, whereas rhetorical signals operate through varied pathways: either mediator alone or both in sequence. These findings provide theoretical and practical insights for entrepreneurs, platforms, and policymakers strategically managing AI transparency in high-stakes investment contexts.

Large language models and LLM-based agents are increasingly used for cybersecurity tasks that are inherently dual-use. Existing approaches to refusal, spanning academic policy frameworks and commercially deployed systems, often rely on broad topic-based bans or offensive-focused taxonomies. As a result, they can yield inconsistent decisions, over-restrict legitimate defenders, and behave brittlely under obfuscation or request segmentation. We argue that effective refusal requires explicitly modeling the trade-off between offensive risk and defensive benefit, rather than relying solely on intent or offensive classification. In this paper, we introduce a content-based framework for designing and auditing cyber refusal policies that makes offense-defense tradeoffs explicit. The framework characterizes requests along five dimensions: Offensive Action Contribution, Offensive Risk, Technical Complexity, Defensive Benefit, and Expected Frequency for Legitimate Users, grounded in the technical substance of the request rather than stated intent. We demonstrate that this content-grounded approach resolves inconsistencies in current frontier model behavior and allows organizations to construct tunable, risk-aware refusal policies.

Assessing human muscle fatigue is critical for optimizing performance and safety in physical human-robot interaction(pHRI). This work presents a data-driven framework to estimate fatigue in dynamic, cyclic pHRI using arm-mounted surface electromyography(sEMG). Subject-specific machine-learning regression models(Random Forest, XGBoost, and Linear Regression predict the fraction of cycles to fatigue(FCF) from three frequency-domain and one time-domain EMG features, and are benchmarked against a convolutional neural network(CNN) that ingests spectrograms of filtered EMG. Framing fatigue estimation as regression (rather than classification) captures continuous progression toward fatigue, supporting earlier detection, timely intervention, and adaptive robot control. In experiments with ten participants, a collaborative robot under admittance control guided repetitive lateral (left-right) end-effector motions until muscular fatigue. Average FCF RMSE across participants was 20.8+/-4.3% for the CNN, 23.3+/-3.8% for Random Forest, 24.8+/-4.5% for XGBoost, and 26.9+/-6.1% for Linear Regression. To probe cross-task generalization, one participant additionally performed unseen vertical (up-down) and circular repetitions; models trained only on lateral data were tested directly and largely retained accuracy, indicating robustness to changes in movement direction, arm kinematics, and muscle recruitment, while Linear Regression deteriorated. Overall, the study shows that both feature-based ML and spectrogram-based DL can estimate remaining work capacity during repetitive pHRI, with the CNN delivering the lowest error and the tree-based models close behind. The reported transfer to new motion patterns suggests potential for practical fatigue monitoring without retraining for every task, improving operator protection and enabling fatigue-aware shared autonomy, for safer fatigue-adaptive pHRI control.

Northrop Frye's theory of four fundamental narrative genres (comedy, romance, tragedy, satire) has profoundly influenced literary criticism, yet computational approaches to his framework have focused primarily on narrative patterns rather than character functions. In this paper, we present a new character function framework that complements pattern-based analysis by examining how archetypal roles manifest differently across Frye's genres. Drawing on Jungian archetype theory, we derive four universal character functions (protagonist, mentor, antagonist, companion) by mapping them to Jung's psychic structure components. These functions are then specialized into sixteen genre-specific roles based on prototypical works. To validate this framework, we conducted a multi-model study using six state-of-the-art Large Language Models (LLMs) to evaluate character-role correspondences across 40 narrative works. The validation employed both positive samples (160 valid correspondences) and negative samples (30 invalid correspondences) to evaluate whether models both recognize valid correspondences and reject invalid ones. LLMs achieved substantial performance (mean balanced accuracy of 82.5%) with strong inter-model agreement (Fleiss' $κ$ = 0.600), demonstrating that the proposed correspondences capture systematic structural patterns. Performance varied by genre (ranging from 72.7% to 89.9%) and role (52.5% to 99.2%), with qualitative analysis revealing that variations reflect genuine narrative properties, including functional distribution in romance and deliberate archetypal subversion in satire. This character-based approach demonstrates the potential of LLM-supported methods for computational narratology and provides a foundation for future development of narrative generation methods and interactive storytelling applications.

Electrocardiograms (ECG) are electrical recordings of the heart that are critical for diagnosing cardiovascular conditions. ECG language models (ELMs) have recently emerged as a promising framework for ECG classification accompanied by report generation. However, current models cannot forecast future cardiac events despite the immense clinical value for planning earlier intervention. To address this gap, we propose CAMEL, the first ELM that is capable of inference over longer signal durations which enables its forecasting capability. Our key insight is a specialized ECG encoder which enables cross-understanding of ECG signals with text. We train CAMEL using established LLM training procedures, combining LoRA adaptation with a curriculum learning pipeline. Our curriculum includes ECG classification, metrics calculations, and multi-turn conversations to elicit reasoning. CAMEL demonstrates strong zero-shot performance across 6 tasks and 9 datasets, including ECGForecastBench, a new benchmark that we introduce for forecasting arrhythmias. CAMEL is on par with or surpasses ELMs and fully supervised baselines both in- and out-of-distribution, achieving SOTA results on ECGBench (+7.0% absolute average gain) as well as ECGForecastBench (+12.4% over fully supervised models and +21.1% over zero-shot ELMs).

Neural networks can accurately forecast complex dynamical systems, yet how they internally represent underlying latent geometry remains poorly understood. We study neural forecasters through the lens of representational alignment, introducing anchor-based, geometry-agnostic relative embeddings that remove rotational and scaling ambiguities in latent spaces. Applying this framework across seven canonical dynamical systems - ranging from periodic to chaotic - we reveal reproducible family-level structure: multilayer perceptrons align with other MLPs, recurrent networks with RNNs, while transformers and echo-state networks achieve strong forecasts despite weaker alignment. Alignment generally correlates with forecasting accuracy, yet high accuracy can coexist with low alignment. Relative geometry thus provides a simple, reproducible foundation for comparing how model families internalize and represent dynamical structure.

Despite the advances in neural text to speech (TTS), many Arabic dialectal varieties remain marginally addressed, with most resources concentrated on Modern Spoken Arabic (MSA) and Gulf dialects, leaving Egyptian Arabic -- the most widely understood Arabic dialect -- severely under-resourced. We address this gap by introducing NileTTS: 38 hours of transcribed speech from two speakers across diverse domains including medical, sales, and general conversations. We construct this dataset using a novel synthetic pipeline: large language models (LLM) generate Egyptian Arabic content, which is then converted to natural speech using audio synthesis tools, followed by automatic transcription and speaker diarization with manual quality verification. We fine-tune XTTS v2, a state-of-the-art multilingual TTS model, on our dataset and evaluate against the baseline model trained on other Arabic dialects. Our contributions include: (1) the first publicly available Egyptian Arabic TTS dataset, (2) a reproducible synthetic data generation pipeline for dialectal TTS, and (3) an open-source fine-tuned model. All resources are released to advance Egyptian Arabic speech synthesis research.

Current methods for personality control in Large Language Models rely on static prompting or expensive fine-tuning, failing to capture the dynamic and compositional nature of human traits. We introduce PERSONA, a training-free framework that achieves fine-tuning level performance through direct manipulation of personality vectors in activation space. Our key insight is that personality traits appear as extractable, approximately orthogonal directions in the model's representation space that support algebraic operations. The framework operates through three stages: Persona-Base extracts orthogonal trait vectors via contrastive activation analysis; Persona-Algebra enables precise control through vector arithmetic (scalar multiplication for intensity, addition for composition, subtraction for suppression); and Persona-Flow achieves context-aware adaptation by dynamically composing these vectors during inference. On PersonalityBench, our approach achieves a mean score of 9.60, nearly matching the supervised fine-tuning upper bound of 9.61 without any gradient updates. On our proposed Persona-Evolve benchmark for dynamic personality adaptation, we achieve up to 91% win rates across diverse model families. These results provide evidence that aspects of LLM personality are mathematically tractable, opening new directions for interpretable and efficient behavioral control.

Deep learning-based segmentation and classification are crucial to large-scale biomedical imaging, particularly for 3D data, where manual analysis is impractical. Although many methods exist, selecting suitable models and tuning parameters remains a major bottleneck in practice. Hence, we introduce the 3D data Analysis Optimization Pipeline, a method designed to facilitate the design and parameterization of segmentation and classification using two Bayesian Optimization stages. First, the pipeline selects a segmentation model and optimizes postprocessing parameters using a domain-adapted syntactic benchmark dataset. To ensure a concise evaluation of segmentation performance, we introduce a segmentation quality metric that serves as the objective function. Second, the pipeline optimizes design choices of a classifier, such as encoder and classifier head architectures, incorporation of prior knowledge, and pretraining strategies. To reduce manual annotation effort, this stage includes an assisted class-annotation workflow that extracts predicted instances from the segmentation results and sequentially presents them to the operator, eliminating the need for manual tracking. In four case studies, the 3D data Analysis Optimization Pipeline efficiently identifies effective model and parameter configurations for individual datasets.

Self-evolving LLM agents update their internal state across sessions, often by writing and reusing long-term memory. This design improves performance on long-horizon tasks but creates a security risk: untrusted external content observed during a benign session can be stored as memory and later treated as instruction. We study this risk and formalize a persistent attack we call a Zombie Agent, where an attacker covertly implants a payload that survives across sessions, effectively turning the agent into a puppet of the attacker. We present a black-box attack framework that uses only indirect exposure through attacker-controlled web content. The attack has two phases. During infection, the agent reads a poisoned source while completing a benign task and writes the payload into long-term memory through its normal update process. During trigger, the payload is retrieved or carried forward and causes unauthorized tool behavior. We design mechanism-specific persistence strategies for common memory implementations, including sliding-window and retrieval-augmented memory, to resist truncation and relevance filtering. We evaluate the attack on representative agent setups and tasks, measuring both persistence over time and the ability to induce unauthorized actions while preserving benign task quality. Our results show that memory evolution can convert one-time indirect injection into persistent compromise, which suggests that defenses focused only on per-session prompt filtering are not sufficient for self-evolving agents.

In dynamical systems reconstruction (DSR) we aim to recover the dynamical system (DS) underlying observed time series. Specifically, we aim to learn a generative surrogate model which approximates the underlying, data-generating DS, and recreates its long-term properties (`climate statistics'). In scientific and medical areas, in particular, these models need to be mechanistically tractable -- through their mathematical analysis we would like to obtain insight into the recovered system's workings. Piecewise-linear (PL), ReLU-based RNNs (PLRNNs) have a strong track-record in this regard, representing SOTA DSR models while allowing mathematical insight by virtue of their PL design. However, all current PLRNN variants are discrete-time maps. This is in disaccord with the assumed continuous-time nature of most physical and biological processes, and makes it hard to accommodate data arriving at irregular temporal intervals. Neural ODEs are one solution, but they do not reach the DSR performance of PLRNNs and often lack their tractability. Here we develop theory for continuous-time PLRNNs (cPLRNNs): We present a novel algorithm for training and simulating such models, bypassing numerical integration by efficiently exploiting their PL structure. We further demonstrate how important topological objects like equilibria or limit cycles can be determined semi-analytically in trained models. We compare cPLRNNs to both their discrete-time cousins as well as Neural ODEs on DSR benchmarks, including systems with discontinuities which come with hard thresholds.

Inverse design problems are common in engineering and materials science. The forward direction, i.e., computing output quantities from design parameters, typically requires running a numerical simulation, such as a FEM, as an intermediate step, which is an optimization problem by itself. In many scenarios, several design parameters can lead to the same or similar output values. For such cases, multi-modal probabilistic approaches are advantageous to obtain diverse solutions. A major difficulty in inverse design stems from the structure of the design space, since discrete parameters or further constraints disallow the direct use of gradient-based optimization. To tackle this problem, we propose a novel inverse design method based on diffusion models. Our approach relaxes the original design space into a continuous grid representation, where gradients can be computed by implicit differentiation in the forward simulation. A diffusion model is trained on this relaxed parameter space in order to serve as a prior for plausible relaxed designs. Parameters are sampled by guided diffusion using gradients that are propagated from an objective function specified at inference time through the differentiable simulation. A design sample is obtained by backprojection into the original parameter space. We develop our approach for a composite material design problem where the forward process is modeled as a linear FEM problem. We evaluate the performance of our approach in finding designs that match a specified bulk modulus. We demonstrate that our method can propose diverse designs within 1% relative error margin from medium to high target bulk moduli in 2D and 3D settings. We also demonstrate that the material density of generated samples can be minimized simultaneously by using a multi-objective loss function.

Foundation models, including vision language models, are increasingly used in automated driving to interpret scenes, recommend actions, and generate natural language explanations. However, existing evaluation methods primarily assess outcome based performance, such as safety and trajectory accuracy, without determining whether model decisions reflect human relevant considerations. As a result, it remains unclear whether explanations produced by such models correspond to genuine reason responsive decision making or merely post hoc rationalizations. This limitation is especially significant in safety critical domains because it can create false confidence. To address this gap, we propose CARE Drive, Context Aware Reasons Evaluation for Driving, a model agnostic framework for evaluating reason responsiveness in vision language models applied to automated driving. CARE Drive compares baseline and reason augmented model decisions under controlled contextual variation to assess whether human reasons causally influence decision behavior. The framework employs a two stage evaluation process. Prompt calibration ensures stable outputs. Systematic contextual perturbation then measures decision sensitivity to human reasons such as safety margins, social pressure, and efficiency constraints. We demonstrate CARE Drive in a cyclist overtaking scenario involving competing normative considerations. Results show that explicit human reasons significantly influence model decisions, improving alignment with expert recommended behavior. However, responsiveness varies across contextual factors, indicating uneven sensitivity to different types of reasons. These findings provide empirical evidence that reason responsiveness in foundation models can be systematically evaluated without modifying model parameters.

Semantic communication promises task-aligned transmission but must reconcile semantic fidelity with stringent latency guarantees in immersive and safety-critical services. This paper introduces a time-constrained human-in-the-loop reinforcement learning (TC-HITL-RL) framework that embeds human feedback, semantic utility, and latency control within a semantic-aware Open radio access network (RAN) architecture. We formulate semantic adaptation driven by human feedback as a constrained Markov decision process (CMDP) whose state captures semantic quality, human preferences, queue slack, and channel dynamics, and solve it via a primal--dual proximal policy optimization algorithm with action shielding and latency-aware reward shaping. The resulting policy preserves PPO-level semantic rewards while tightening the variability of both air-interface and near-real-time RAN intelligent controller processing budgets. Simulations over point-to-multipoint links with heterogeneous deadlines show that TC-HITL-RL consistently meets per-user timing constraints, outperforms baseline schedulers in reward, and stabilizes resource consumption, providing a practical blueprint for latency-aware semantic adaptation.

Time-series imputation benchmarks employ uniform random masking and shape-agnostic metrics (MSE, RMSE), implicitly weighting evaluation by regime prevalence. In systems with a dominant attractor -- homeostatic physiology, nominal industrial operation, stable network traffic -- this creates a systematic \emph{Stationarity Bias}: simple methods appear superior because the benchmark predominantly samples the easy, low-entropy regime where they trivially succeed. We formalize this bias and propose a \emph{Stratified Stress-Test} that partitions evaluation into Stationary and Transient regimes. Using Continuous Glucose Monitoring (CGM) as a testbed -- chosen for its rigorous ground-truth forcing functions (meals, insulin) that enable precise regime identification -- we establish three findings with broad implications:(i)~Stationary Efficiency: Linear interpolation achieves state-of-the-art reconstruction during stable intervals, confirming that complex architectures are computationally wasteful in low-entropy regimes.(ii)~Transient Fidelity: During critical transients (post-prandial peaks, hypoglycemic events), linear methods exhibit drastically degraded morphological fidelity (DTW), disproportionate to their RMSE -- a phenomenon we term the \emph{RMSE Mirage}, where low pointwise error masks the destruction of signal shape.(iii)~Regime-Conditional Model Selection: Deep learning models preserve both pointwise accuracy and morphological integrity during transients, making them essential for safety-critical downstream tasks. We further derive empirical missingness distributions from clinical trials and impose them on complete training data, preventing models from exploiting unrealistically clean observations and encouraging robustness under real-world missingness. This framework generalizes to any regulated system where routine stationarity dominates critical transients.

Cumulative constraints are central in scheduling with constraint programming, yet propagation is typically performed per constraint, missing multi-resource interactions and causing severe slowdowns on some benchmarks. I present a preprocessing method for inferring additional cumulative constraints that capture such interactions without search-time probing. This approach interprets cumulative constraints as linear inequalities over occupancy vectors and generates valid inequalities by (i) discovering covers, the sets of tasks that cannot run in parallel, (ii) strengthening the cover inequalities for the discovered sets with lifting, and (iii) injecting the resulting constraints back into the scheduling problem instance. Experiments on standard RCPSP and RCPSP/max test suites show that these inferred constraints improve search performance and tighten objective bounds on favorable instances, while incurring little degradation on unfavorable ones. Additionally, these experiments discover 25 new lower bounds and five new best solutions; eight of the lower bounds are obtained directly from the inferred constraints.

Graph Neural Networks (GNNs) learn node representations through iterative network-based message-passing. While powerful, deep GNNs suffer from oversmoothing, where node features converge to a homogeneous, non-informative state. We re-frame this problem of representational collapse from a \emph{bifurcation theory} perspective, characterizing oversmoothing as convergence to a stable ``homogeneous fixed point.'' Our central contribution is the theoretical discovery that this undesired stability can be broken by replacing standard monotone activations (e.g., ReLU) with a class of functions. Using Lyapunov-Schmidt reduction, we analytically prove that this substitution induces a bifurcation that destabilizes the homogeneous state and creates a new pair of stable, non-homogeneous \emph{patterns} that provably resist oversmoothing. Our theory predicts a precise, nontrivial scaling law for the amplitude of these emergent patterns, which we quantitatively validate in experiments. Finally, we demonstrate the practical utility of our theory by deriving a closed-form, bifurcation-aware initialization and showing its utility in real benchmark experiments.

The rapid development of AI for Science is often hindered by the "discretization", where learned representations remain restricted to the specific grids or resolutions used during training. We propose the Neural Proper Orthogonal Decomposition (Neural-POD), a plug-and-play neural operator framework that constructs nonlinear, orthogonal basis functions in infinite-dimensional space using neural networks. Unlike the classical Proper Orthogonal Decomposition (POD), which is limited to linear subspace approximations obtained through singular value decomposition (SVD), Neural-POD formulates basis construction as a sequence of residual minimization problems solved through neural network training. Each basis function is obtained by learning to represent the remaining structure in the data, following a process analogous to Gram--Schmidt orthogonalization. This neural formulation introduces several key advantages over classical POD: it enables optimization in arbitrary norms (e.g., $L^2$, $L^1$), learns mappings between infinite-dimensional function spaces that is resolution-invariant, generalizes effectively to unseen parameter regimes, and inherently captures nonlinear structures in complex spatiotemporal systems. The resulting basis functions are interpretable, reusable, and enabling integration into both reduced order modeling (ROM) and operator learning frameworks such as deep operator learning (DeepONet). We demonstrate the robustness of Neural-POD with different complex spatiotemporal systems, including the Burgers' and Navier-Stokes equations. We further show that Neural-POD serves as a high performance, plug-and-play bridge between classical Galerkin projection and operator learning that enables consistent integration with both projection-based reduced order models and DeepONet frameworks.

Business plan (BP) writing plays a key role in entrepreneurship education by helping learners construct, evaluate, and iteratively refine their ideas. However, conventional BP writing remains a rigid, linear process that often fails to reflect the dynamic and recursive nature of entrepreneurial ideation. This mismatch is particularly challenging for novice entrepreneurial students, who struggle with the substantial cognitive demands of developing and refining ideas. While reflection and meta-reflection are critical strategies for fostering divergent and convergent thinking, existing writing tools rarely scaffold these higher-order processes. To address this gap, we present the Meflex System, a large language model (LLM)-based writing tool that integrates BP writing scaffolding with a nonlinear idea canvas to support iterative ideation through reflection and meta-reflection. We report findings from an exploratory user study with 30 participants that examined the system's usability and cognitive impact. Results show that Meflex effectively scaffolds BP writing, promotes divergent thinking through LLM-supported reflection, and enhances meta-reflective awareness while reducing cognitive load during complex idea development. These findings highlight the potential of non-linear LLM-based writing tools to foster deeper and coherent entrepreneurial thinking.

Reinforcement Learning (RL) has significantly improved large language model reasoning, but existing RL fine-tuning methods rely heavily on heuristic techniques such as entropy regularization and reweighting to maintain stability. In practice, they often experience late-stage performance collapse, leading to degraded reasoning quality and unstable training. We derive that the magnitude of token-wise policy gradients in RL is negatively correlated with token probability and local policy entropy. Building on this result, we prove that training instability is driven by a tiny fraction of tokens, approximately 0.01\%, which we term \emph{spurious tokens}. When such tokens appear in correct responses, they contribute little to the reasoning outcome but inherit the full sequence-level reward, leading to abnormally amplified gradient updates. Motivated by this observation, we propose Spurious-Token-Aware Policy Optimization (STAPO) for large-scale model refining, which selectively masks such updates and renormalizes the loss over valid tokens. Across six mathematical reasoning benchmarks using Qwen 1.7B, 8B, and 14B base models, STAPO consistently demonstrates superior entropy stability and achieves an average performance improvement of 7.13\% over GRPO, 20-Entropy and JustRL.

Ensuring user fairness in wireless communications is a fundamental challenge, as balancing the trade-off between fairness and sum rate leads to a non-convex, multi-objective optimization whose complexity grows with network scale. To alleviate this conflict, we propose an optimization-based unsupervised learning approach based on the wireless transformer (WiT) architecture that learns from channel state information (CSI) features. We reformulate the trade-off by combining the sum rate and fairness objectives through a Lagrangian multiplier, which is updated automatically via a dual-ascent algorithm. This mechanism allows for a controllable fairness constraint while simultaneously maximizing the sum rate, effectively realizing a trace on the Pareto front between two conflicting objectives. Our findings show that the proposed approach offers a flexible solution for managing the trade-off optimization under prescribed fairness.

Models based on partial differential equations (PDEs) are powerful for describing a wide range of complex relationships in the natural sciences. Accurately identifying the PDE model, which represents the underlying physical law, is essential for a proper understanding of the problem. This reconstruction typically relies on indirect and noisy measurements of the system's state and, without specifically tailored methods, rarely yields symbolic expressions, thereby hindering interpretability. In this work, we address this issue by considering existing neural network architectures based on rational functions for the symbolic representation of physical laws. These networks leverage the approximation power of rational functions while also benefiting from their flexibility in representing arithmetic operations. Our main contribution is an identifiability result, showing that, in the limit of noiseless, complete measurements, such symbolic networks can uniquely reconstruct the simplest physical law within the PDE model. Specifically, reconstructed laws remain expressible within the symbolic network architecture, with regularization-minimizing parameterizations promoting interpretability and sparsity in case of $L^1$-regularization. In addition, we provide regularity results for symbolic networks. Empirical validation using the ParFam architecture supports these theoretical findings, providing evidence for the practical reconstructibility of physical laws.

Certified machine unlearning can be achieved via noise injection leading to differential privacy guarantees, where noise is calibrated to worst-case sensitivity. Such conservative calibration often results in performance degradation, limiting practical applicability. In this work, we investigate an alternative approach based on adaptive per-instance noise calibration tailored to the individual contribution of each data point to the learned solution. This raises the following challenge: how can one establish formal unlearning guarantees when the mechanism depends on the specific point to be removed? To define individual data point sensitivities in noisy gradient dynamics, we consider the use of per-instance differential privacy. For ridge regression trained via Langevin dynamics, we derive high-probability per-instance sensitivity bounds, yielding certified unlearning with substantially less noise injection. We corroborate our theoretical findings through experiments in linear settings and provide further empirical evidence on the relevance of the approach in deep learning settings.

This article investigates the causal antecedents of conflictual language and the geometry of interaction in online threaded conversations related to climate change. We employ three annotation dimensions, inferred through LLM prompting and averaging, to capture complementary aspects of discursive conflict (such as stance: agreement vs disagreement; tone: attacking vs respectful; and emotional versus factual framing) and use data from a threaded online forum to examine how these dimensions respond to temporal, conversational, and arborescent structural features of discussions. We show that, as suggested by the literature, longer delays between successive posts in a thread are associated with replies that are, on average, more respectful, whereas longer delays relative to the parent post are associated with slightly less disagreement but more emotional (less factual) language. Second, we characterize alignment with the local conversational environment and find strong convergence both toward the average stance, tone and emotional framing of older sibling posts replying to the same parent and toward those of the parent post itself, with parent post effects generally stronger than sibling effects. We further show that early branch-level responses condition these alignment dynamics, such that parent-child stance alignment is amplified or attenuated depending on whether a branch is initiated in agreement or disagreement with the discussion's root message. These influences are largely additive for civility-related dimensions (attacking vs respectful, disagree vs agree), whereas for emotional versus factual framing there is a significant interaction: alignment with the parent's emotionality is amplified when older siblings are similarly aligned.

In this paper, we formulate the new multi-objective coverage (MOC) problem where our goal is to identify a small set of representative samples whose predicted outcomes broadly cover the feasible multi-objective space. This problem is of great importance in many critical real-world applications, e.g., drug discovery and materials design, as this representative set can be evaluated much faster than the whole feasible set, thus significantly accelerating the scientific discovery process. Existing works cannot be directly applied as they either focus on sample space coverage or multi-objective optimization that targets the Pareto front. However, chemically diverse samples often yield identical objective profiles, and safety constraints are usually defined on the objectives. To solve this MOC problem, we propose a novel search algorithm, MOC-CAS, which employs an upper confidence bound-based acquisition function to select optimistic samples guided by Gaussian process posterior predictions. For enabling efficient optimization, we develop a smoothed relaxation of the hard feasibility test and derive an approximate optimizer. Compared to the competitive baselines, we show that our MOC-CAS empirically achieves superior performances across large-scale protein-target datasets for SARS-CoV-2 and cancer, each assessed on five objectives derived from SMILES-based features.

Recurrent and deep neural networks (RNNs/DNNs) are cornerstone architectures in machine learning. Remarkably, RNNs differ from DNNs only by weight sharing, as can be shown through unrolling in time. How does this structural similarity fit with the distinct functional properties these networks exhibit? To address this question, we here develop a unified mean-field theory for RNNs and DNNs in terms of representational kernels, describing fully trained networks in the feature learning ($μ$P) regime. This theory casts training as Bayesian inference over sequences and patterns, directly revealing the functional implications induced by the RNNs' weight sharing. In DNN-typical tasks, we identify a phase transition when the learning signal overcomes the noise due to randomness in the weights: below this threshold, RNNs and DNNs behave identically; above it, only RNNs develop correlated representations across timesteps. For sequential tasks, the RNNs' weight sharing furthermore induces an inductive bias that aids generalization by interpolating unsupervised time steps. Overall, our theory offers a way to connect architectural structure to functional biases.

Spatiotemporal flows govern diverse phenomena across physics, biology, and engineering, yet modelling their multiscale dynamics remains a central challenge. Despite major advances in physics-informed machine learning, existing approaches struggle to simultaneously maintain long-term temporal evolution and resolve fine-scale structure across chaotic, turbulent, and physiological regimes. Here, we introduce Uni-Flow, a unified autoregressive-diffusion framework that explicitly separates temporal evolution from spatial refinement for modelling complex dynamical systems. The autoregressive component learns low-resolution latent dynamics that preserve large-scale structure and ensure stable long-horizon rollouts, while the diffusion component reconstructs high-resolution physical fields, recovering fine-scale features in a small number of denoising steps. We validate Uni-Flow across canonical benchmarks, including two-dimensional Kolmogorov flow, three-dimensional turbulent channel inflow generation with a quantum-informed autoregressive prior, and patient-specific simulations of aortic coarctation derived from high-fidelity lattice Boltzmann hemodynamic solvers. In the cardiovascular setting, Uni-Flow enables task-level faster than real-time inference of pulsatile hemodynamics, reconstructing high-resolution pressure fields over physiologically relevant time horizons in seconds rather than hours. By transforming high-fidelity hemodynamic simulation from an offline, HPC-bound process into a deployable surrogate, Uni-Flow establishes a pathway to faster-than-real-time modelling of complex multiscale flows, with broad implications for scientific machine learning in flow physics.

Testing functionality in Software-Defined Vehicles is challenging because requirements are written in natural language, specifications combine text, tables, and diagrams, while test assets are scattered across heterogeneous toolchains. Large Language Models and Vision-Language Models are used to extract signals and behavioral logic to automatically generate Gherkin scenarios, which are then converted into runnable test scripts. The Vehicle Signal Specification (VSS) integration standardizes signal references, supporting portability across subsystems and test benches. The pipeline uses retrieval-augmented generation to preselect candidate VSS signals before mapping. We evaluate the approach on the safety-relevant Child Presence Detection System, executing the generated tests in a virtual environment and on an actual vehicle. Our evaluation covers Gherkin validity, VSS mapping quality, and end-to-end executability. Results show that 32 of 36 requirements (89\%) can be transformed into executable scenarios in our setting, while human review and targeted substitutions remain necessary. This paper is a feasibility and architectural demonstration of an end-to-end requirements-to-test pipeline for SDV subsystems, evaluated on a CPDS case in simulation and Vehicle-in-the-Loop settings.

This paper provides statistical guarantees on the accuracy of dynamical models learned from dependent data sequences. Specifically, we develop uniform error bounds that apply to quantized models and imperfect optimization algorithms commonly used in practical contexts for system identification, and in particular hybrid system identification. Two families of bounds are obtained: slow-rate bounds via a block decomposition and fast-rate, variance-adaptive, bounds via a novel spaced-point strategy. The bounds scale with the number of bits required to encode the model and thus translate hardware constraints into interpretable statistical complexities.

When a multimodal Transformer answers a visual question, is the prediction driven by visual evidence, linguistic reasoning, or genuinely fused cross-modal computation -- and how does this structure evolve across layers? We address this question with a layer-wise framework based on Partial Information Decomposition (PID) that decomposes the predictive information at each Transformer layer into redundant, vision-unique, language-unique, and synergistic components. To make PID tractable for high-dimensional neural representations, we introduce \emph{PID Flow}, a pipeline combining dimensionality reduction, normalizing-flow Gaussianization, and closed-form Gaussian PID estimation. Applying this framework to LLaVA-1.5-7B and LLaVA-1.6-7B across six GQA reasoning tasks, we uncover a consistent \emph{modal transduction} pattern: visual-unique information peaks early and decays with depth, language-unique information surges in late layers to account for roughly 82\% of the final prediction, and cross-modal synergy remains below 2\%. This trajectory is highly stable across model variants (layer-wise correlations $>$0.96) yet strongly task-dependent, with semantic redundancy governing the detailed information fingerprint. To establish causality, we perform targeted Image$\rightarrow$Question attention knockouts and show that disrupting the primary transduction pathway induces predictable increases in trapped visual-unique information, compensatory synergy, and total information cost -- effects that are strongest in vision-dependent tasks and weakest in high-redundancy tasks. Together, these results provide an information-theoretic, causal account of how vision becomes language in multimodal Transformers, and offer quantitative guidance for identifying architectural bottlenecks where modality-specific information is lost.

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.

Depression manifests through a diverse set of symptoms such as sleep disturbance, loss of interest, and concentration difficulties. However, most existing works treat depression prediction either as a binary label or an overall severity score without explicitly modeling symptom-specific information. This limits their ability to provide symptom-level analysis relevant to clinical screening. To address this, we propose a symptom-specific and clinically inspired framework for depression severity estimation from speech. Our approach uses a symptom-guided cross-attention mechanism that aligns PHQ-8 questionnaire items with emotion-aware speech representations to identify which segments of a participant's speech are more important to each symptom. To account for differences in how symptoms are expressed over time, we introduce a learnable symptom-specific parameter that adaptively controls the sharpness of attention distributions. Our results on EDAIC, a standard clinical-style dataset, demonstrate improved performance outperforming prior works. Further, analyzing the attention distributions showed that higher attention is assigned to utterances containing cues related to multiple depressive symptoms, highlighting the interpretability of our approach. These findings outline the importance of symptom-guided and emotion-aware modeling for speech-based depression screening.

Agent-based modelling (ABM) is a widespread approach to simulate complex systems. Advancements in computational processing and storage have facilitated the adoption of ABMs across many fields; however, ABMs face challenges that limit their use as decision-support tools. A significant issue is parameter estimation in large-scale ABMs, particularly due to computational constraints on exploring the parameter space. This study evaluates a state-of-the-art simulation-based inference (SBI) framework that uses neural networks (NN) for parameter estimation. This framework is applied to an established labour market ABM based on job transition networks. The ABM is initiated with synthetic datasets and the real U.S. labour market. Next, we compare the effectiveness of summary statistics derived from a list of statistical measures with that learned by an embedded NN. The results demonstrate that the NN-based approach recovers the original parameters when evaluating posterior distributions across various dataset scales and improves efficiency compared to traditional Bayesian methods.

Predictive coding (PC) is a biologically inspired algorithm for training neural networks that relies only on local updates, allowing parallel learning across layers. However, practical implementations face two key limitations: error signals must still propagate from the output to early layers through multiple inference-phase steps, and feedback decays exponentially during this process, leading to vanishing updates in early layers. We propose direct Kolen-Pollack predictive coding (DKP-PC), which simultaneously addresses both feedback delay and exponential decay, yielding a more efficient and scalable variant of PC while preserving update locality. Leveraging direct feedback alignment and direct Kolen-Pollack algorithms, DKP-PC introduces learnable feedback connections from the output layer to all hidden layers, establishing a direct pathway for error transmission. This yields an algorithm that reduces the theoretical error propagation time complexity from O(L), with L being the network depth, to O(1), removing depth-dependent delay in error signals. Moreover, empirical results demonstrate that DKP-PC achieves performance at least comparable to, and often exceeding, that of standard PC, while offering improved latency and computational performance, supporting its potential for custom hardware-efficient implementations.

The scenario approach is an established data-driven design framework that comes equipped with a powerful theory linking design complexity to generalization properties. In this approach, data are simultaneously used both for design and for certifying the design's reliability, without resorting to a separate test dataset. This paper takes a step further by guaranteeing additional properties, useful in post-design usage but not considered during the design phase. To this end, we introduce a two-level framework of appropriateness: baseline appropriateness, which guides the design process, and post-design appropriateness, which serves as a criterion for a posteriori evaluation. We provide distribution-free upper bounds on the risk of failing to meet the post-design appropriateness; these bounds are computable without using any additional test data. Under additional assumptions, lower bounds are also derived. As part of an effort to demonstrate the usefulness of the proposed methodology, the paper presents two practical examples in H2 and pole-placement problems. Moreover, a method is provided to infer comprehensive distributional knowledge of relevant performance indexes from the available dataset.

Text-to-SQL has recently achieved impressive progress, yet remains difficult to apply effectively in real-world scenarios. This gap stems from the reliance on single static workflows, fundamentally limiting scalability to out-of-distribution and long-tail scenarios. Instead of requiring users to select suitable methods through extensive experimentation, we attempt to enable systems to adaptively construct workflows at inference time. Through theoretical and empirical analysis, we demonstrate that optimal dynamic policies consistently outperform the best static workflow, with performance gains fundamentally driven by heterogeneity across candidate workflows. Motivated by this, we propose SquRL, a reinforcement learning framework that enhances LLMs' reasoning capability in adaptive workflow construction. We design a rule-based reward function and introduce two effective training mechanisms: dynamic actor masking to encourage broader exploration, and pseudo rewards to improve training efficiency. Experiments on widely-used Text-to-SQL benchmarks demonstrate that dynamic workflow construction consistently outperforms the best static workflow methods, with especially pronounced gains on complex and out-of-distribution queries. The codes are available at https://github.com/Satissss/SquRL

Quantization-aware training (QAT) is an effective method to drastically reduce the memory footprint of LLMs while keeping performance degradation at an acceptable level. However, the optimal choice of quantization format and bit-width presents a challenge in practice. The full design space of quantization is not fully explored in the context of QAT, and the precise trade-off between quantization and downstream performance is poorly understood, as comparisons often rely solely on perplexity-based evaluations. In this work, we address these shortcomings with an empirical study of QAT in the low-bit regime. We show that k-means based weight quantization outperforms integer formats and can be implemented efficiently on standard hardware. Furthermore, we find that, under a fixed inference memory budget, the best performance on generative downstream tasks is achieved with $1$-bit quantized weights.

The Personal AI landscape is currently dominated by "Black Box" Retrieval-Augmented Generation. While standard vector databases offer statistical matching, they suffer from a fundamental lack of accountability: when an AI hallucinates or retrieves sensitive data, the user cannot inspect the cause nor correct the error. Worse, "deleting" a concept from a vector space is mathematically imprecise, leaving behind probabilistic "ghosts" that violate true privacy. We propose Ruva, the first "Glass Box" architecture designed for Human-in-the-Loop Memory Curation. Ruva grounds Personal AI in a Personal Knowledge Graph, enabling users to inspect what the AI knows and to perform precise redaction of specific facts. By shifting the paradigm from Vector Matching to Graph Reasoning, Ruva ensures the "Right to be Forgotten." Users are the editors of their own lives; Ruva hands them the pen. The project and the demo video are available at http://sisinf00.poliba.it/ruva/.

This study investigates the impact of regularization of latent spaces through truncation on the quality of generated test inputs for deep learning classifiers. We evaluate this effect using style-based GANs, a state-of-the-art generative approach, and assess quality along three dimensions: validity, diversity, and fault detection. We evaluate our approach on the boundary testing of deep learning image classifiers across three datasets, MNIST, Fashion MNIST, and CIFAR-10. We compare two truncation strategies: latent code mixing with binary search optimization and random latent truncation for generative exploration. Our experiments show that the latent code-mixing approach yields a higher fault detection rate than random truncation, while also improving both diversity and validity.

Vision-language model (VLM) shows promise for high-level planning in smart manufacturing, yet their deployment in dynamic workcells faces two critical challenges: (1) stateless operation, they cannot persistently track out-of-view states, causing world-state drift; and (2) opaque reasoning, failures are difficult to diagnose, leading to costly blind retries. This paper presents VLM-DEWM, a cognitive architecture that decouples VLM reasoning from world-state management through a persistent, queryable Dynamic External World Model (DEWM). Each VLM decision is structured into an Externalizable Reasoning Trace (ERT), comprising action proposal, world belief, and causal assumption, which is validated against DEWM before execution. When failures occur, discrepancy analysis between predicted and observed states enables targeted recovery instead of global replanning. We evaluate VLM-DEWM on multi-station assembly, large-scale facility exploration, and real-robot recovery under induced failures. Compared to baseline memory-augmented VLM systems, VLM DEWM improves state-tracking accuracy from 56% to 93%, increases recovery success rate from below 5% to 95%, and significantly reduces computational overhead through structured memory. These results establish VLM-DEWM as a verifiable and resilient solution for long-horizon robotic operations in dynamic manufacturing environments.

Text embedding models are widely used for semantic similarity tasks, including information retrieval, clustering, and classification. General-purpose models are typically trained with single- or multi-stage processes using contrastive loss functions. We introduce a novel training regimen that combines model distillation techniques with task-specific contrastive loss to produce compact, high-performance embedding models. Our findings suggest that this approach is more effective for training small models than purely contrastive or distillation-based training paradigms alone. Benchmark scores for the resulting models, jina-embeddings-v5-text-small and jina-embeddings-v5-text-nano, exceed or match the state-of-the-art for models of similar size. jina-embeddings-v5-text models additionally support long texts (up to 32k tokens) in many languages, and generate embeddings that remain robust under truncation and binary quantization. Model weights are publicly available, hopefully inspiring further advances in embedding model development.

The ability to accurately perform counterfactual inference on time series is crucial for decision-making in fields like finance, healthcare, and marketing, as it allows us to understand the impact of events or treatments on outcomes over time. In this paper, we introduce a new counterfactual inference approach tailored to time series data impacted by market events, which is motivated by an industrial application. Utilizing the abduction-action-prediction procedure and the Structural Causal Model framework, we first adapt methods based on variational autoencoders and adversarial autoencoders, both previously used in counterfactual literature although not in time series settings. Then, we present the Conditional Entropy-Penalized Autoencoder (CEPAE), a novel autoencoder-based approach for counterfactual inference, which employs an entropy penalization loss over the latent space to encourage disentangled data representations. We validate our approach both theoretically and experimentally on synthetic, semi-synthetic, and real-world datasets, showing that CEPAE generally outperforms the other approaches in the evaluated metrics.

This paper introduces Perspectives, an interactive extension of the Discourse Analysis Tool Suite designed to empower Digital Humanities (DH) scholars to explore and organize large, unstructured document collections. Perspectives implements a flexible, aspect-focused document clustering pipeline with human-in-the-loop refinement capabilities. We showcase how this process can be initially steered by defining analytical lenses through document rewriting prompts and instruction-based embeddings, and further aligned with user intent through tools for refining clusters and mechanisms for fine-tuning the embedding model. The demonstration highlights a typical workflow, illustrating how DH researchers can leverage Perspectives's interactive document map to uncover topics, sentiments, or other relevant categories, thereby gaining insights and preparing their data for subsequent in-depth analysis.

Recent studies have explored the combination of multiple LoRAs to simultaneously generate user-specified subjects and styles. However, most existing approaches fuse LoRA weights using static statistical heuristics that deviate from LoRA's original purpose of learning adaptive feature adjustments and ignore the randomness of sampled inputs. To address this, we propose a dynamic training-free fusion framework that operates throughout the generation process. During the forward pass, at each LoRA-applied layer, we dynamically compute the KL divergence between the base model's original features and those produced by subject and style LoRAs, respectively, and adaptively select the most appropriate weights for fusion. In the reverse denoising stage, we further refine the generation trajectory by dynamically applying gradient-based corrections derived from objective metrics such as CLIP and DINO scores, providing continuous semantic and stylistic guidance. By integrating these two complementary mechanisms-feature-level selection and metric-guided latent adjustment-across the entire diffusion timeline, our method dynamically achieves coherent subject-style synthesis without any retraining. Extensive experiments across diverse subject-style combinations demonstrate that our approach consistently outperforms state-of-the-art LoRA fusion methods both qualitatively and quantitatively.

We study the asymptotic shape of the trajectory of the stochastic gradient descent algorithm applied to a convex objective function. Under mild regularity assumptions, we prove a functional central limit theorem for the properly rescaled trajectory. Our result characterizes the long-term fluctuations of the algorithm around the minimizer by providing a diffusion limit for the trajectory. In contrast with classical central limit theorems for the last iterate or Polyak-Ruppert averages, this functional result captures the temporal structure of the fluctuations and applies to non-smooth settings such as robust location estimation, including the geometric median.

Pure speech language models aim to learn language directly from raw audio without textual resources. A key challenge is that discrete tokens from self-supervised speech encoders result in excessively long sequences, motivating recent work on syllable-like units. However, methods like Sylber and SyllableLM rely on intricate multi-stage training pipelines. We propose ZeroSyl, a simple training-free method to extract syllable boundaries and embeddings directly from a frozen WavLM model. Using L2 norms of features in WavLM's intermediate layers, ZeroSyl achieves competitive syllable segmentation performance. The resulting segments are mean-pooled, discretized using K-means, and used to train a language model. ZeroSyl outperforms prior syllabic tokenizers across lexical, syntactic, and narrative benchmarks. Scaling experiments show that while finer-grained units are beneficial for lexical tasks, our discovered syllabic units exhibit better scaling behavior for syntactic modeling.

The LLM community often reports benchmark results as if they are synonymous with general model capabilities. However, benchmarks can have problems that distort performance, like test set contamination and annotator error. How can we know that a benchmark is a reliable indicator of some capability that we want to measure? This question concerns the construct validity of LLM benchmarks, and it requires separating benchmark results from capabilities when we model and predict LLM performance. Both social scientists and computer scientists propose formal models - latent factor models and scaling laws - for identifying the capabilities underlying benchmark scores. However, neither technique is satisfactory for construct validity. Latent factor models ignore scaling laws, and as a result, the capabilities they extract often proxy model size. Scaling laws ignore measurement error, and as a result, the capabilities they extract are both uninterpretable and overfit to the observed benchmarks. This thesis presents the structured capabilities model, the first model to extract interpretable and generalisable capabilities from a large collection of LLM benchmark results. I fit this model and its two alternatives on a large sample of results from the OpenLLM Leaderboard. Structured capabilities outperform latent factor models on parsimonious fit indices, and exhibit better out-of-distribution benchmark prediction than scaling laws. These improvements are possible because neither existing approach separates model scale from capabilities in the appropriate way. Model scale should inform capabilities, as in scaling laws, and these capabilities should inform observed results up to measurement error, as in latent factor models. In combining these two insights, structured capabilities demonstrate better explanatory and predictive power for quantifying construct validity in LLM evaluations.

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.

We present new distributed quantum algorithms for fundamental distributed computing problems, namely, leader election, broadcast, Minimum Spanning Tree (MST), and Breadth-First Search (BFS) tree, in arbitrary networks. These algorithms are (essentially) optimal with respect to their communication (message) complexity in the {\em quantum routing model} introduced in [PODC 2025]. The message complexity of our algorithms is $\tilde{O}(n)$ for leader election, broadcast, and MST, and $\tilde{O}(\sqrt{mn})$ for BFS ($n$ and $m$ are the number of nodes and edges of the network, respectively). These message bounds are nearly tight in the quantum routing model since we show almost matching corresponding quantum message lower bounds. Our results significantly improve on the prior work of [PODC 2025], who presented distributed quantum algorithms under the same model that had a message complexity of $\tilde{O}(\sqrt{mn})$ for leader election. Our algorithms demonstrate the significant communication advantage that quantum routing has over classical in distributed computing, since $Ω(m)$ is a well-established classical message lower bound for leader election, broadcast, MST, and BFS that applies even to randomized Monte-Carlo algorithms [JACM 2015]. Thus, our quantum algorithms can, in general, give a quadratic advantage in the communication cost for these fundamental problems. A main technical tool we use to design our distributed algorithms is quantum walks based on electric networks. We posit a framework for using quantum walks in the distributed setting to design communication-efficient distributed quantum algorithms. Our framework can be used as a black box to significantly reduce communication costs and may be of independent interest. Additionally, our lower-bound technique for establishing distributed quantum message lower bounds can also be applied to other problems.

Mixture-of-Experts (MoE) effectively scales model capacity while preserving computational efficiency through sparse expert activation. However, training high-quality MoEs from scratch is prohibitively expensive. A promising alternative is to convert pretrained dense models into sparse MoEs. Existing dense-to-MoE methods fall into two categories: \textbf{dynamic structural pruning} that converts dense models into MoE architectures with moderate sparsity to balance performance and inference efficiency, and \textbf{downcycling} approaches that use pretrained dense models to initialize highly sparse MoE architectures. However, existing methods break the intrinsic activation patterns within dense models, leading to suboptimal expert construction. In this work, we argue that the Gated Linear Unit (GLU) mechanism provides a natural blueprint for dense-to-MoE conversion. We show that the fine-grained neural-wise activation patterns of GLU reveal a coarse-grained structure, uncovering an inherent MoE architecture composed of consistently activated universal neurons and dynamically activated specialized neurons. Leveraging this discovery, we introduce ExpertWeaver, a training-free framework that partitions neurons according to their activation patterns and constructs shared experts and specialized routed experts with layer-adaptive configurations. Our experiments demonstrate that ExpertWeaver significantly outperforms existing methods, both as a training-free dynamic structural pruning technique and as a downcycling strategy for superior MoE initialization.

Training against white-box deception detectors has been proposed as a way to make AI systems honest. However, such training risks models learning to obfuscate their deception to evade the detector. Prior work has studied obfuscation only in artificial settings where models were directly rewarded for harmful output. We construct a realistic coding environment where reward hacking via hardcoding test cases naturally occurs, and show that obfuscation emerges in this setting. We introduce a taxonomy of possible outcomes when training against a deception detector. The model either remains honest, or becomes deceptive via two possible obfuscation strategies. (i) Obfuscated activations: the model outputs deceptive text while modifying its internal representations to no longer trigger the detector. (ii) Obfuscated policy: the model outputs deceptive text that evades the detector, typically by including a justification for the reward hack. Empirically, obfuscated activations arise from representation drift during RL, with or without a detector penalty. The probe penalty only incentivizes obfuscated policies; we theoretically show this is expected for policy gradient methods. Sufficiently high KL regularization and detector penalty can yield honest policies, establishing white-box deception detectors as viable training signals for tasks prone to reward hacking.

As large language models (LLMs) become increasingly prevalent, reliable methods for detecting AI-generated text are critical for mitigating potential risks. We introduce DependencyAI, a simple and interpretable approach for detecting AI-generated text using only the labels of linguistic dependency relations. Our method achieves competitive performance across monolingual, multi-generator, and multilingual settings. To increase interpretability, we analyze feature importance to reveal syntactic structures that distinguish AI-generated from human-written text. We also observe a systematic overprediction of certain models on unseen domains, suggesting that generator-specific writing styles may affect cross-domain generalization. Overall, our results demonstrate that dependency relations alone provide a robust signal for AI-generated text detection, establishing DependencyAI as a strong linguistically grounded, interpretable, and non-neural network baseline.

Deploying Multimodal Large Language Models as the brain of embodied agents remains challenging, particularly under long-horizon observations and limited context budgets. Existing memory assisted methods often rely on textual summaries, which discard rich visual and spatial details and remain brittle in non-stationary environments. In this work, we propose a non-parametric memory framework that explicitly disentangles episodic and semantic memory for embodied exploration and question answering. Our retrieval-first, reasoning-assisted paradigm recalls episodic experiences via semantic similarity and verifies them through visual reasoning, enabling robust reuse of past observations without rigid geometric alignment. In parallel, we introduce a program-style rule extraction mechanism that converts experiences into structured, reusable semantic memory, facilitating cross-environment generalization. Extensive experiments demonstrate state-of-the-art performance on embodied question answering and exploration benchmarks, yielding a 7.3% gain in LLM-Match and an 11.4% gain in LLM MatchXSPL on A-EQA, as well as +7.7% success rate and +6.8% SPL on GOAT-Bench. Analyses reveal that our episodic memory primarily improves exploration efficiency, while semantic memory strengthens complex reasoning of embodied agents.

Federated Learning (FL) enables distributed training across multiple clients without centralized data sharing, while Graph Neural Networks (GNNs) model relational data through message passing. In federated GNN settings, client graphs often exhibit heterogeneous structural and propagation characteristics. When standard aggregation mechanisms are applied to such heterogeneous updates, the global model may converge numerically while exhibiting degraded relational behavior.Our work identifies a geometric failure mode of global aggregation in Cross- Domain Federated GNNs. Although GNN parameters are numerically represented as vectors, they encode relational transformations that govern the direction, strength, and sensitivity of information flow across graph neighborhoods. Aggregating updates originating from incompatible propagation regimes can therefore introduce destructive interference in this transformation space.This leads to loss of coherence in global message passing. Importantly, this degradation is not necessarily reflected in conventional metrics such as loss or accuracy.To address this issue, we propose GGRS (Global Geometric Reference Structure), a server-side framework that regulates client updates prior to aggregation based on geometric admissibility criteria. GGRS preserves directional consistency of relational transformations as well as maintains diversity of admissible propagation subspaces. It also stabilizes sensitivity to neighborhood interactions, without accessing client data or graph topology. Experiments on heterogeneous GNN-native, Amazon Co-purchase datasets demonstrate that GGRS preserves global message-passing coherence across training rounds by highlighting the necessity of geometry-aware regulation in federated graph learning.

The tendency for hallucination in current large language models (LLMs) negatively impacts dialogue systems. Such hallucinations produce factually incorrect responses that may mislead users and undermine system trust. Existing refinement methods for dialogue systems typically operate at the response level, overlooking the fact that a single response may contain multiple verifiable or unverifiable facts. To address this gap, we propose Fine-Refine, a fine-grained refinement framework that decomposes responses into atomic units, verifies each unit using external knowledge, assesses fluency via perplexity, and iteratively corrects granular errors. We evaluate factuality across the HybriDialogue and OpendialKG datasets in terms of factual accuracy (fact score) and coverage (Not Enough Information Proportion), and experiments show that Fine-Refine substantially improves factuality, achieving up to a 7.63-point gain in dialogue fact score, with a small trade-off in dialogue quality.

We introduce LuxMT, a machine translation system based on Gemma 3 27B and fine-tuned for translation from Luxembourgish (LB) into French (FR) and English (EN). To assess translation performance, we construct a novel benchmark covering LB-FR, LB-EN, and LB-FR using human-translated data from Luci, a tourist magazine about Luxembourg. Training data stems from LuxAlign, a parallel corpus of multilingual Luxembourgish news articles, and LB parliamentary transcripts augmented with Google Translate. We filter the data using LuxEmbedder, LB sentence embeddings, to remove low-equivalence segment-pairs. Overall, LuxMT's results suggest strong improvements over the Gemma 3 baseline, even for translating LB to German (DE), despite the training data not containing any DE. We also explore LuxEmbedder's potential to be used as a quality estimation metric and find strong correlations with other reference-based metrics. However, we call for further research to fully assess the metric's utility and advise using it with caution.

Patients' expectations towards their treatment have a substantial effect on the treatments' success. While primarily studied in clinical settings, online patient platforms like medical subreddits may hold complementary insights: treatment expectations that patients feel unnecessary or uncomfortable to share elsewhere. Despite this, no studies examine what type of expectations users discuss online and how they express them. Presumably this is because expectations have not been studied in natural language processing (NLP) before. Therefore, we introduce the task of Expectation Detection, arguing that expectations are relevant for many applications, including opinion mining and product design. Subsequently, we present a case study for the medical domain, where expectations are particularly crucial to extract. We contribute RedHOTExpect, a corpus of Reddit posts (4.5K posts) to study expectations in this context. We use a large language model (LLM) to silver-label the data and validate its quality manually (label accuracy ~78%). Based on this, we analyze which linguistic patterns characterize expectations and explore what patients expect and why. We find that optimism and proactive framing are more pronounced in posts about physical or treatment-related illnesses compared to mental-health contexts, and that in our dataset, patients mostly discuss benefits rather than negative outcomes. The RedHOTExpect corpus can be obtained from https://www.ims.uni-stuttgart.de/data/RedHOTExpect

Stability and robustness are critical for deploying Transformers in safety-sensitive settings. A principled way to enforce such behavior is to constrain the model's Lipschitz constant. However, approximation-theoretic guarantees for architectures that explicitly preserve Lipschitz continuity have yet to be established. In this work, we bridge this gap by introducing a class of gradient-descent-type in-context Transformers that are Lipschitz-continuous by construction. We realize both MLP and attention blocks as explicit Euler steps of negative gradient flows, ensuring inherent stability without sacrificing expressivity. We prove a universal approximation theorem for this class within a Lipschitz-constrained function space. Crucially, our analysis adopts a measure-theoretic formalism, interpreting Transformers as operators on probability measures, to yield approximation guarantees independent of token count. These results provide a rigorous theoretical foundation for the design of robust, Lipschitz continuous Transformer architectures.

Mathematical morphology (MM) is a powerful and widely used framework in image processing. Through set-theoretic and discrete geometric principles, MM operations such as erosion, dilation, opening, and closing effectively manipulate digital images by modifying local structures via structuring elements (SEs), while cubical homology captures global topological features such as connected components and loop structures within images. Building on the GUDHI package for persistent homology (PH) computation on cubical complexes, we propose the MMPersistence library, which integrates MM operations with diverse SEs and PH computation to extract multiscale persistence information. By employing SEs of different shapes to construct topological filtrations, the proposed MM-based PH framework encodes both spatial and morphological characteristics of digital images, providing richer local geometric information than conventional cubical homology alone and establishing a unified foundation for analyzing digital images that integrates topological insight with morphological image processing techniques.

It has been shown that a neural network's Lipschitz constant can be leveraged to derive robustness guarantees, to improve generalizability via regularization or even to construct invertible networks. Therefore, a number of methods varying in the tightness of their bounds and their computational cost have been developed to approximate the Lipschitz constant for different classes of networks. However, comparatively little research exists on methods for exact computation, which has been shown to be NP-hard. Nonetheless, there are applications where one might readily accept the computational cost of an exact method. These applications could include the benchmarking of new methods or the computation of robustness guarantees for small models on sensitive data. Unfortunately, existing exact algorithms restrict themselves to only ReLU-activated networks, which are known to come with severe downsides in the context of Lipschitz-constrained networks. We therefore propose a generalization of the LipBaB algorithm to compute exact Lipschitz constants for arbitrary piecewise linear neural networks and $p$-norms. With our method, networks may contain traditional activations like ReLU or LeakyReLU, activations like GroupSort or the related MinMax and FullSort, which have been of increasing interest in the context of Lipschitz constrained networks, or even other piecewise linear functions like MaxPool.

Neural audio codecs (NACs) typically encode the short-term energy (gain) and normalized structure (shape) of speech/audio signals jointly within the same latent space. As a result, they are poorly robust to a global variation of the input signal level in the sense that such variation has strong influence on the embedding vectors at the output of the encoder and their quantization. This methodology is inherently inefficient, leading to codebook redundancy and suboptimal bitrate-distortion performance. To address these limitations, we propose to introduce shape-gain decomposition, widely used in classical speech/audio coding, into the NAC framework. The principle of the proposed Equalizer methodology is to decompose the input signal -- before the NAC encoder -- into gain and normalized shape vector on a short-term basis. The shape vector is processed by the NAC, while the gain is quantized with scalar quantization and transmitted separately. The output (decoded) signal is reconstructed from the normalized output of the NAC and the quantized gain. Our experiments conducted on speech signals show that this general methodology, easily applicable to any NAC, enables a substantial gain in bitrate-distortion performance, as well as a massive reduction in complexity.

General-purpose super-resolution models, particularly Vision Transformers, have achieved remarkable success but exhibit fundamental inefficiencies in common infrared imaging scenarios like surveillance and autonomous driving, which operate from fixed or nearly-static viewpoints. These models fail to exploit the strong, persistent spatial priors inherent in such scenes, leading to redundant learning and suboptimal performance. To address this, we propose the Regional Prior attention Transformer for infrared image Super-Resolution (RPT-SR), a novel architecture that explicitly encodes scene layout information into the attention mechanism. Our core contribution is a dual-token framework that fuses (1) learnable, regional prior tokens, which act as a persistent memory for the scene's global structure, with (2) local tokens that capture the frame-specific content of the current input. By utilizing these tokens into an attention, our model allows the priors to dynamically modulate the local reconstruction process. Extensive experiments validate our approach. While most prior works focus on a single infrared band, we demonstrate the broad applicability and versatility of RPT-SR by establishing new state-of-the-art performance across diverse datasets covering both Long-Wave (LWIR) and Short-Wave (SWIR) spectra

We introduce SecCodeBench-V2, a publicly released benchmark for evaluating Large Language Model (LLM) copilots' capabilities of generating secure code. SecCodeBench-V2 comprises 98 generation and fix scenarios derived from Alibaba Group's industrial productions, where the underlying security issues span 22 common CWE (Common Weakness Enumeration) categories across five programming languages: Java, C, Python, Go, and Node.js. SecCodeBench-V2 adopts a function-level task formulation: each scenario provides a complete project scaffold and requires the model to implement or patch a designated target function under fixed interfaces and dependencies. For each scenario, SecCodeBench-V2 provides executable proof-of-concept (PoC) test cases for both functional validation and security verification. All test cases are authored and double-reviewed by security experts, ensuring high fidelity, broad coverage, and reliable ground truth. Beyond the benchmark itself, we build a unified evaluation pipeline that assesses models primarily via dynamic execution. For most scenarios, we compile and run model-generated artifacts in isolated environments and execute PoC test cases to validate both functional correctness and security properties. For scenarios where security issues cannot be adjudicated with deterministic test cases, we additionally employ an LLM-as-a-judge oracle. To summarize performance across heterogeneous scenarios and difficulty levels, we design a Pass@K-based scoring protocol with principled aggregation over scenarios and severity, enabling holistic and comparable evaluation across models. Overall, SecCodeBench-V2 provides a rigorous and reproducible foundation for assessing the security posture of AI coding assistants, with results and artifacts released at https://alibaba.github.io/sec-code-bench. The benchmark is publicly available at https://github.com/alibaba/sec-code-bench.

In this study, we have presented a novel approach to predict the Short-Time Objective Intelligibility (STOI) metric using a bottleneck transformer architecture. Traditional methods for calculating STOI typically requires clean reference speech, which limits their applicability in the real world. To address this, numerous deep learning-based nonintrusive speech assessment models have garnered significant interest. Many studies have achieved commendable performance, but there is room for further improvement. We propose the use of bottleneck transformer, incorporating convolution blocks for learning frame-level features and a multi-head self-attention (MHSA) layer to aggregate the information. These components enable the transformer to focus on the key aspects of the input data. Our model has shown higher correlation and lower mean squared error for both seen and unseen scenarios compared to the state-of-the-art model using self-supervised learning (SSL) and spectral features as inputs.

LLM-as-a-judge has emerged as a cornerstone technique for evaluating large language models by leveraging LLM reasoning to score prompt-response pairs. Since LLM judgments are stochastic, practitioners commonly query each pair multiple times to estimate mean scores accurately. This raises a critical challenge: given a fixed computational budget $B$, how to optimally allocate queries across $K$ prompt-response pairs to minimize estimation error? % We present a principled variance-adaptive approach leveraging multi-armed bandit theory and concentration inequalities. Our method dynamically allocates queries based on estimated score variances, concentrating resources where uncertainty is highest. Further, our algorithm is shown to achieve a worst-case score-estimation error of $\tilde{O}\left(\sqrt{\frac{\sum_{i=1}^K σ_i^2}{B}}\right)$, $σ_i^2$ being the unknown score variance for pair $i \in [K]$ with near-optimal budget allocation. % Experiments on \emph{Summarize-From-Feedback} and \emph{HelpSteer2} demonstrate that our method significantly outperforms uniform allocation, reducing worst-case estimation error while maintaining identical budgets. Our work establishes a theoretical foundation for efficient LLM evaluation with practical implications for AI safety, model alignment, and automated assessment at scale.

Mood instability is a key behavioral indicator of mental health, yet traditional assessments rely on infrequent and retrospective reports that fail to capture its continuous nature. Smartphone-based mobile sensing enables passive, in-the-wild mood inference from everyday behaviors; however, deploying such systems at scale remains challenging due to privacy constraints, uneven sensing availability, and substantial variability in behavioral patterns. In this work, we study mood inference using smartphone sensing data in a cross-country federated learning setting, where each country participates as an independent client while retaining local data. We introduce FedFAP, a feature-aware personalized federated framework designed to accommodate heterogeneous sensing modalities across regions. Evaluations across geographically and culturally diverse populations show that FedFAP achieves an AUROC of 0.744, outperforming both centralized approaches and existing personalized federated baselines. Beyond inference, our results offer design insights for mood-aware systems, demonstrating how population-aware personalization and privacy-preserving learning can enable scalable and mood-aware mobile sensing technologies.

Optimization refers to the task of finding extrema of an objective function. Classical gradient-based optimizers are highly sensitive to hyperparameter choices. In highly non-convex settings their performance relies on carefully tuned learning rates, momentum, and gradient accumulation. To address these limitations, we introduce POP (Prior-fitted Optimizer Policies), a meta-learned optimizer that predicts coordinate-wise step sizes conditioned on the contextual information provided in the optimization trajectory. Our model is learned on millions of synthetic optimization problems sampled from a novel prior spanning both convex and non-convex objectives. We evaluate POP on an established benchmark including 47 optimization functions of various complexity, where it consistently outperforms first-order gradient-based methods, non-convex optimization approaches (e.g., evolutionary strategies), Bayesian optimization, and a recent meta-learned competitor under matched budget constraints. Our evaluation demonstrates strong generalization capabilities without task-specific tuning.

We present a novel property-preserving kernel-based operator learning method for incompressible flows governed by the incompressible Navier-Stokes equations. Traditional numerical solvers incur significant computational costs to respect incompressibility. Operator learning offers efficient surrogate models, but current neural operators fail to exactly enforce physical properties such as incompressibility, periodicity, and turbulence. Our method maps input functions to expansion coefficients of output functions in a property-preserving kernel basis, ensuring that predicted velocity fields analytically and simultaneously preserve the aforementioned physical properties. We evaluate the method on challenging 2D and 3D, laminar and turbulent, incompressible flow problems. Our method achieves up to six orders of magnitude lower relative $\ell_2$ errors upon generalization and trains up to five orders of magnitude faster compared to neural operators. Moreover, while our method enforces incompressibility analytically, neural operators exhibit very large deviations. Our results show that our method provides an accurate and efficient surrogate for incompressible flows.

Ankara's public transport crisis is commonly framed as a shortage of buses or operational inefficiency. This study argues that the problem is fundamentally morphological and structural. The city's leapfrog urban expansion has produced fragmented peripheral clusters disconnected from a rigid, center-oriented bus network. As a result, demand remains intensely concentrated along the Kizilay-Ulus axis and western corridors, while peripheral districts experience either chronic under-service or enforced transfer dependency. The deficiency is therefore not merely quantitative but rooted in the misalignment between urban macroform and network architecture. The empirical analysis draws on a 173-day operational dataset derived from route-level passenger and trip reports published by EGO under the former "Transparent Ankara" initiative. To overcome the absence of stop-level geospatial data, a Connectivity-Based Weighted Distribution Model reallocates passenger volumes to 1 km x 1 km grid cells using network centrality. The findings reveal persistent center-periphery asymmetries, structural bottlenecks, and spatially embedded accessibility inequalities.

Integrating reasoning in large language models and large vision-language models has recently led to significant improvement of their capabilities. However, the generalization of reasoning models is still vaguely defined and poorly understood. In this work, we present an evaluation framework to rigorously examine how well chain-of-thought (CoT) approaches generalize on a simple planning task. Specifically, we consider a grid-based navigation task in which a model is provided with a map and must output a sequence of moves that guides a player from a start position to a goal while avoiding obstacles. The versatility of the task and its data allows us to fine-tune model variants using different input representations (visual and textual) and CoT reasoning strategies, and systematically evaluate them under both in-distribution (ID) and out-of-distribution (OOD) test conditions. Our experiments show that, while CoT reasoning improves in-distribution generalization across all representations, out-of-distribution generalization (e.g., to larger maps) remains very limited in most cases when controlling for trivial matches with the ID data. Surprisingly, we find that reasoning traces which combine multiple text formats yield the best (and non-trivial) OOD generalization. Finally, purely text-based models consistently outperform those utilizing image-based inputs, including a recently proposed approach relying on latent space reasoning.

Anomaly detection (AD) for safety-critical IoT time series should be judged at the event level: reliability and earliness under realistic perturbations. Yet many studies still emphasize point-level results on curated base datasets, limiting value for model selection in practice. We introduce an evaluation protocol with unified event-level augmentations that simulate real-world issues: calibrated sensor dropout, linear and log drift, additive noise, and window shifts. We also perform sensor-level probing via mask-as-missing zeroing with per-channel influence estimation to support root-cause analysis. We evaluate 14 representative models on five public anomaly datasets (SWaT, WADI, SMD, SKAB, TEP) and two industrial datasets (steam turbine, nuclear turbogenerator) using unified splits and event aggregation. There is no universal winner: graph-structured models transfer best under dropout and long events (e.g., on SWaT under additive noise F1 drops 0.804->0.677 for a graph autoencoder, 0.759->0.680 for a graph-attention variant, and 0.762->0.756 for a hybrid graph attention model); density/flow models work well on clean stationary plants but can be fragile to monotone drift; spectral CNNs lead when periodicity is strong; reconstruction autoencoders become competitive after basic sensor vetting; predictive/hybrid dynamics help when faults break temporal dependencies but remain window-sensitive. The protocol also informs design choices: on SWaT under log drift, replacing normalizing flows with Gaussian density reduces high-stress F1 from ~0.75 to ~0.57, and fixing a learned DAG gives a small clean-set gain (~0.5-1.0 points) but increases drift sensitivity by ~8x.

Agents based on Large Language Models (LLMs) are increasingly being deployed as interfaces to information on online platforms. These agents filter, prioritize, and synthesize information retrieved from the platforms' back-end databases or via web search. In these scenarios, LLM agents govern the information users receive, by drawing users' attention to particular instances of retrieved information at the expense of others. While much prior work has focused on biases in the information LLMs themselves generate, less attention has been paid to the factors that influence what information LLMs select and present to users. We hypothesize that when information is attributed to specific sources (e.g., particular publishers, journals, or platforms), current LLMs exhibit systematic latent source preferences- that is, they prioritize information from some sources over others. Through controlled experiments on twelve LLMs from six model providers, spanning both synthetic and real-world tasks, we find that several models consistently exhibit strong and predictable source preferences. These preferences are sensitive to contextual framing, can outweigh the influence of content itself, and persist despite explicit prompting to avoid them. They also help explain phenomena such as the observed left-leaning skew in news recommendations in prior work. Our findings advocate for deeper investigation into the origins of these preferences, as well as for mechanisms that provide users with transparency and control over the biases guiding LLM-powered agents.

Deep generative modeling to stochastically design small molecules is an emerging technology for accelerating drug discovery and development. However, one major issue in molecular generative models is their lower frequency of drug-like compounds. To resolve this problem, we developed a novel framework for optimization of deep generative models integrated with a D-Wave quantum annealing computer, where our Neural Hash Function (NHF) presented herein is used both as the regularization and binarization schemes simultaneously, of which the latter is for transformation between continuous and discrete signals of the classical and quantum neural networks, respectively, in the error evaluation (i.e., objective) function. The compounds generated via the quantum-annealing generative models exhibited higher quality in both validity and drug-likeness than those generated via the fully-classical models, and was further indicated to exceed even the training data in terms of drug-likeness features, without any restraints and conditions to deliberately induce such an optimization. These results indicated an advantage of quantum annealing to aim at a stochastic generator integrated with our novel neural network architectures, for the extended performance of feature space sampling and extraction of characteristic features in drug design.

Large Language Models (LLMs) are changing the coding paradigm, known as vibe coding, yet synthesizing algorithmically sophisticated and robust code still remains a critical challenge. Incentivizing the deep reasoning capabilities of LLMs is essential to overcoming this hurdle. Reinforcement Fine-Tuning (RFT) has emerged as a promising strategy to address this need. However, most existing approaches overlook the heterogeneous difficulty and granularity inherent in test cases, leading to an imbalanced distribution of reward signals and consequently biased gradient updates during training. To address this, we propose Test-driven and cApability-adaptive cuRriculum reinfOrcement fine-Tuning (TAROT). TAROT systematically constructs, for each problem, a four-tier test suite (basic, intermediate, complex, edge), providing a controlled difficulty landscape for curriculum design and evaluation. Crucially, TAROT decouples curriculum progression from raw reward scores, enabling capability-conditioned evaluation and principled selection from a portfolio of curriculum policies rather than incidental test-case difficulty composition. This design fosters stable optimization and more efficient competency acquisition. Extensive experimental results reveal that the optimal curriculum for RFT in code generation is closely tied to a model's inherent capability, with less capable models achieving greater gains with an easy-to-hard progression, whereas more competent models excel under a hard-first curriculum. TAROT provides a reproducible method that adaptively tailors curriculum design to a model's capability, thereby consistently improving the functional correctness and robustness of the generated code. All code and data are released to foster reproducibility and advance community research at https://github.com/deep-diver/TAROT.

During deliberation processes, mediators and facilitators typically need to select a small and representative set of opinions later used to produce digestible reports for stakeholders. In online deliberation platforms, algorithmic selection is increasingly used to automate this process. However, such automation is not without consequences. For instance, enforcing consensus-seeking algorithmic strategies can imply ignoring or flattening conflicting preferences, which may lead to erasing minority voices and reducing content diversity. More generally, across the variety of existing selection strategies (e.g., consensus, diversity), it remains unclear how each approach influences desired democratic criteria such as proportional representation. To address this gap, we benchmark several algorithmic approaches in this context. We also build on social choice theory to propose a novel algorithm that incorporates both diversity and a balanced notion of representation in the selection strategy. We find empirically that while no single strategy dominates across all democratic desiderata, our social-choice-inspired selection rule achieves the strongest trade-off between proportional representation and diversity.

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.

Digitized historical archives make it possible to study everyday social life on a large scale, but the information extracted directly from text often does not directly allow one to answer the research questions posed by historians or sociologists in a quantitative manner. We address this problem in a large collection of Finnish World War II Karelian evacuee family interviews. Prior work extracted more than 350K mentions of leisure time activities and organizational memberships from these interviews, yielding 71K unique activity and organization names -- far too many to analyze directly. We develop a categorization framework that captures key aspects of participation (the kind of activity/organization, how social it typically is, how regularly it happens, and how physically demanding it is). We annotate a gold-standard set to allow for a reliable evaluation, and then test whether large language models can apply the same schema at scale. Using a simple voting approach across multiple model runs, we find that an open-weight LLM can closely match expert judgments. Finally, we apply the method to label the 350K entities, producing a structured resource for downstream studies of social integration and related outcomes.

As the burgeoning power requirements of sophisticated neural architectures escalate, the information retrieval community has recognized ecological sustainability as a pivotal priority that necessitates a fundamental paradigm shift in model design. While contemporary neural rankers have attained unprecedented accuracy, the substantial environmental externalities associated with their computational intensity often remain overlooked in large-scale deployments. We present GaiaFlow, an innovative framework engineered to facilitate carbon-frugal search by operationalizing semantic-guided diffusion tuning. Our methodology orchestrates the convergence of retrieval-guided Langevin dynamics and a hardware-independent performance modeling strategy to optimize the trade-off between search precision and environmental preservation. By incorporating adaptive early exit protocols and precision-aware quantized inference, the proposed architecture significantly mitigates operational carbon footprints while maintaining robust retrieval quality across heterogeneous computing infrastructures. Extensive experimental evaluations demonstrate that GaiaFlow achieves a superior equilibrium between effectiveness and energy efficiency, offering a scalable and sustainable pathway for next-generation neural search systems.

Software repositories provide a detailed record of software evolution by capturing developer interactions through code-related activities such as pull requests and modifications. To better understand the underlying dynamics of codebase evolution, we introduce a novel approach that integrates semantic code embeddings with opinion dynamics theory, offering a quantitative framework to analyze collaborative development processes. Our approach begins by encoding code snippets into high-dimensional vector representations using state-of-the-art code embedding models, preserving both syntactic and semantic features. These embeddings are then processed using Principal Component Analysis (PCA) for dimensionality reduction, with data normalized to ensure comparability. We model temporal evolution using the Expressed-Private Opinion (EPO) model to derive trust matrices and track opinion trajectories across development cycles. These opinion trajectories reflect the underlying dynamics of consensus formation, influence propagation, and evolving alignment (or divergence) within developer communities -- revealing implicit collaboration patterns and knowledge-sharing mechanisms that are otherwise difficult to observe. By bridging software engineering and computational social science, our method provides a principled way to quantify software evolution, offering new insights into developer influence, consensus formation, and project sustainability. We evaluate our approach on data from three prominent open-source GitHub repositories, demonstrating its ability to reveal interpretable behavioral trends and variations in developer interactions. The results highlight the utility of our framework in improving open-source project maintenance through data-driven analysis of collaboration dynamics.

Sequential Social Dilemmas (SSDs) provide a key framework for studying how cooperation emerges when individual incentives conflict with collective welfare. In Multi-Agent Reinforcement Learning, these problems are often addressed by incorporating intrinsic drives that encourage prosocial or fair behavior. However, most existing methods assume that agents face identical incentives in the dilemma and require continuous access to global information about other agents to assess fairness. In this work, we introduce asymmetric variants of well-known SSD environments and examine how natural differences between agents influence cooperation dynamics. Our findings reveal that existing fairness-based methods struggle to adapt under asymmetric conditions by enforcing raw equality that wrongfully incentivize defection. To address this, we propose three modifications: (i) redefining fairness by accounting for agents' reward ranges, (ii) introducing an agent-based weighting mechanism to better handle inherent asymmetries, and (iii) localizing social feedback to make the methods effective under partial observability without requiring global information sharing. Experimental results show that in asymmetric scenarios, our method fosters faster emergence of cooperative policies compared to existing approaches, without sacrificing scalability or practicality.

Robust classification in noisy environments remains a fundamental challenge in machine learning. Standard approaches typically treat signal enhancement and classification as separate, sequential stages: first enhancing the signal and then applying a classifier. This approach fails to leverage the semantic information in the classifier's output during denoising. In this work, we propose a general, domain-agnostic framework that integrates two interacting diffusion models: one operating on the input signal and the other on the classifier's output logits, without requiring any retraining or fine-tuning of the classifier. This coupled formulation enables mutual guidance, where the enhancing signal refines the class estimation and, conversely, the evolving class logits guide the signal reconstruction towards discriminative regions of the manifold. We introduce three strategies to effectively model the joint distribution of the input and the logit. We evaluated our joint enhancement method for image classification and automatic speech recognition. The proposed framework surpasses traditional sequential enhancement baselines, delivering robust and flexible improvements in classification accuracy under diverse noise conditions.

Contrary to common belief, common belief is not KD4. If individual belief is KD45, common belief does indeed lose the 5 property and keep the D and 4 properties -- and it has none of the other commonly considered properties of knowledge and belief. But it has another property: $C(Cφ\rightarrow φ)$ -- corresponding to so-called shift-reflexivity (reflexivity one step ahead). This observation begs the question: is KD4 extended with this axiom a complete characterisation of common belief in the KD45 case? If not, what \emph{is} the logic of common belief? In this paper we show that the answer to the first question is ``no'': there is one additional axiom, and, furthermore, it relies on the number of agents. We show that the result is a complete characterisation of common belief, settling the open problem.

Vision-Language-Action (VLA) models leveraging the native autoregressive paradigm of Vision-Language Models (VLMs) have demonstrated superior instruction-following and training efficiency. Central to this paradigm is action tokenization, yet its design has primarily focused on reconstruction fidelity, failing to address its direct impact on VLA optimization. Consequently, the fundamental question of \textit{what makes for good action tokenizers} remains unanswered. In this paper, we bridge this gap by establishing design principles specifically from the perspective of VLA optimization. We identify a set of best practices based on information-theoretic insights, including maximized temporal token overlap, minimized vocabulary redundancy, enhanced multimodal mutual information, and token independence. Guided by these principles, we introduce \textbf{ActionCodec}, a high-performance action tokenizer that significantly enhances both training efficiency and VLA performance across diverse simulation and real-world benchmarks. Notably, on LIBERO, a SmolVLM2-2.2B fine-tuned with ActionCodec achieves a 95.5\% success rate without any robotics pre-training. With advanced architectural enhancements, this reaches 97.4\%, representing a new SOTA for VLA models without robotics pre-training. We believe our established design principles, alongside the released model, will provide a clear roadmap for the community to develop more effective action tokenizers.

Standard Mean-Shift algorithms are notoriously sensitive to the bandwidth hyperparameter, particularly in data-scarce regimes where fixed-scale density estimation leads to fragmentation and spurious modes. In this paper, we propose Doubly Stochastic Mean-Shift (DSMS), a novel extension that introduces randomness not only in the trajectory updates but also in the kernel bandwidth itself. By drawing both the data samples and the radius from a continuous uniform distribution at each iteration, DSMS effectively performs a better exploration of the density landscape. We show that this randomized bandwidth policy acts as an implicit regularization mechanism, and provide convergence theoretical results. Comparative experiments on synthetic Gaussian mixtures reveal that DSMS significantly outperforms standard and stochastic Mean-Shift baselines, exhibiting remarkable stability and preventing over-segmentation in sparse clustering scenarios without other performance degradation.

Large Language Models (LLMs) deployed in production environments face a fundamental safety-utility trade-off either a strict filtering mechanisms prevent harmful outputs but often block benign queries or a relaxed controls risk unsafe content generation. Conventional guardrails based on static rules or fixed confidence thresholds are typically context-insensitive and computationally expensive, resulting in high latency and degraded user experience. To address these limitations, we introduce an adaptive abstention system that dynamically adjusts safety thresholds based on real-time contextual signals such as domain and user history. The proposed framework integrates a multi-dimensional detection architecture composed of five parallel detectors, combined through a hierarchical cascade mechanism to optimize both speed and precision. The cascade design reduces unnecessary computation by progressively filtering queries, achieving substantial latency improvements compared to non-cascaded models and external guardrail systems. Extensive evaluation on mixed and domain-specific workloads demonstrates significant reductions in false positives, particularly in sensitive domains such as medical advice and creative writing. The system maintains high safety precision and near-perfect recall under strict operating modes. Overall, our context-aware abstention framework effectively balances safety and utility while preserving performance, offering a scalable solution for reliable LLM deployment.

Web agents based on large language models have demonstrated promising capability in automating web tasks. However, current web agents struggle to reason out sensible actions due to the limitations of predicting environment changes, and might not possess comprehensive awareness of execution risks, prematurely performing risky actions that cause losses and lead to task failure. To address these challenges, we propose WAC, a web agent that integrates model collaboration, consequence simulation, and feedback-driven action refinement. To overcome the cognitive isolation of individual models, we introduce a multi-agent collaboration process that enables an action model to consult a world model as a web-environment expert for strategic guidance; the action model then grounds these suggestions into executable actions, leveraging prior knowledge of environmental state transition dynamics to enhance candidate action proposal. To achieve risk-aware resilient task execution, we introduce a two-stage deduction chain. A world model, specialized in environmental state transitions, simulates action outcomes, which a judge model then scrutinizes to trigger action corrective feedback when necessary. Experiments show that WAC achieves absolute gains of 1.8% on VisualWebArena and 1.3% on Online-Mind2Web.

Multi-Agent Systems (MAS) powered by Large Language Models have unlocked advanced collaborative reasoning, yet they remain shackled by the inefficiency of discrete text communication, which imposes significant runtime overhead and information quantization loss. While latent state transfer offers a high-bandwidth alternative, existing approaches either assume homogeneous sender-receiver architectures or rely on pair-specific learned translators, limiting scalability and modularity across diverse model families with disjoint manifolds. In this work, we propose the Vision Wormhole, a novel framework that repurposes the visual interface of Vision-Language Models (VLMs) to enable model-agnostic, text-free communication. By introducing a Universal Visual Codec, we map heterogeneous reasoning traces into a shared continuous latent space and inject them directly into the receiver's visual pathway, effectively treating the vision encoder as a universal port for inter-agent telepathy. Our framework adopts a hub-and-spoke topology to reduce pairwise alignment complexity from O(N^2) to O(N) and leverages a label-free, teacher-student distillation objective to align the high-speed visual channel with the robust reasoning patterns of the text pathway. Extensive experiments across heterogeneous model families (e.g., Qwen-VL, Gemma) demonstrate that the Vision Wormhole reduces end-to-end wall-clock time in controlled comparisons while maintaining reasoning fidelity comparable to standard text-based MAS. Code is available at https://github.com/xz-liu/heterogeneous-latent-mas

Federated learning (FL) allows remote clients to train a global model collaboratively while protecting client privacy. Despite its privacy-preserving benefits, FL has significant drawbacks, including slow convergence, high communication cost, and non-independent-and-identically-distributed (non-IID) data. In this work, we present a novel FedAvg variation called Fractional-Order Federated Averaging (FOFedAvg), which incorporates Fractional-Order Stochastic Gradient Descent (FOSGD) to capture long-range relationships and deeper historical information. By introducing memory-aware fractional-order updates, FOFedAvg improves communication efficiency and accelerates convergence while mitigating instability caused by heterogeneous, non-IID client data. We compare FOFedAvg against a broad set of established federated optimization algorithms on benchmark datasets including MNIST, FEMNIST, CIFAR-10, CIFAR-100, EMNIST, the Cleveland heart disease dataset, Sent140, PneumoniaMNIST, and Edge-IIoTset. Across a range of non-IID partitioning schemes, FOFedAvg is competitive with, and often outperforms, these baselines in terms of test performance and convergence speed. On the theoretical side, we prove that FOFedAvg converges to a stationary point under standard smoothness and bounded-variance assumptions for fractional order $0<α\le 1$. Together, these results show that fractional-order, memory-aware updates can substantially improve the robustness and effectiveness of federated learning, offering a practical path toward distributed training on heterogeneous data.

The increasing size and complexity of modern deep neural networks (DNNs) pose significant challenges for on-device inference on mobile GPUs, with limited memory and computational resources. Existing DNN acceleration frameworks primarily deploy a weight preloading strategy, where all model parameters are loaded into memory before execution on mobile GPUs. We posit that this approach is not adequate for modern DNN workloads that comprise very large model(s) and possibly execution of several distinct models in succession. In this work, we introduce FlashMem, a memory streaming framework designed to efficiently execute large-scale modern DNNs and multi-DNN workloads while minimizing memory consumption and reducing inference latency. Instead of fully preloading weights, FlashMem statically determines model loading schedules and dynamically streams them on demand, leveraging 2.5D texture memory to minimize data transformations and improve execution efficiency. Experimental results on 11 models demonstrate that FlashMem achieves 2.0x to 8.4x memory reduction and 1.7x to 75.0x speedup compared to existing frameworks, enabling efficient execution of large-scale models and multi-DNN support on resource-constrained mobile GPUs.

Can large language models converse in languages virtually absent from their training data? We investigate this question through a case study on Tulu, a Dravidian language with over 2 million speakers but minimal digital presence. Rather than fine-tuning an LLM, we examine whether structured prompts alone can elicit basic conversational ability under controlled prompting. We systematically tackle various challenges posed by absence of training data for Tulu by combining explicit grammar documentation, negative constraints to suppress high-probability tokens from related languages, romanization standardization, and quality-controlled synthetic data generation via self-play. Evaluated on a manually curated held-out set across three LLMs (Gemini 2.0 Flash, GPT-4o, Llama 3.1 70B) and validated by native speakers, our approach reduces vocabulary contamination from 80% to 5% while achieving 85% grammatical accuracy. Cross-model analysis reveals that negative constraints provide consistent improvements (12--18 percentage points), while grammar documentation effects vary by model architecture (8--22 points).

Customer service automation has seen growing demand within digital transformation. Existing approaches either rely on modular system designs with extensive agent orchestration or employ over-simplified instruction schemas, providing limited guidance and poor generalizability. This paper introduces an orchestration-free framework using Task-Oriented Flowcharts (TOFs) to enable end-to-end automation without manual intervention. We first define the components and evaluation metrics for TOFs, then formalize a cost-efficient flowchart construction algorithm to abstract procedural knowledge from service dialogues. We emphasize local deployment of small language models and propose decentralized distillation with flowcharts to mitigate data scarcity and privacy issues in model training. Extensive experiments validate the effectiveness in various service tasks, with superior quantitative and application performance compared to strong baselines and market products. By releasing a web-based system demonstration with case studies, we aim to promote streamlined creation of future service automation.

Cryptographic digests (e.g., MD5, SHA-256) are designed to provide exact identity. Any single-bit change in the input produces a completely different hash, which is ideal for integrity verification but limits their usefulness in many real-world tasks like threat hunting, malware analysis and digital forensics, where adversaries routinely introduce minor transformations. Similarity-based techniques address this limitation by enabling approximate matching, allowing related byte sequences to produce measurably similar fingerprints. Modern enterprises manage tens of thousands of endpoints with billions of files, making the effectiveness and scalability of the proposed techniques more important than ever in security applications. Security researchers have proposed a range of approaches, including similarity digests and locality-sensitive hashes (e.g., ssdeep, sdhash, TLSH), as well as more recent machine-learning-based methods that generate embeddings from file features. However, these techniques have largely been evaluated in isolation, using disparate datasets and evaluation criteria. This paper presents a systematic comparison of learning-based classification and similarity methods using large, publicly available datasets. We evaluate each method under a unified experimental framework with industry-accepted metrics. To our knowledge, this is the first reproducible study to benchmark these diverse learning-based similarity techniques side by side for real-world security workloads. Our results show that no single approach performs well across all dimensions; instead, each exhibits distinct trade-offs, indicating that effective malware analysis and threat-hunting platforms must combine complementary classification and similarity techniques rather than rely on a single method.

Language models exhibit systematic performance gaps when processing text in non-standard language varieties, yet their ability to comprehend variety-specific slang remains underexplored for several languages. We present a comprehensive evaluation of slang awareness in Indian English (en-IN) and Australian English (en-AU) across seven state-of-the-art language models. We construct two complementary datasets: \textsc{web}, containing 377 web-sourced usage examples from Urban Dictionary, and \textsc{gen}, featuring 1,492 synthetically generated usages of these slang terms, across diverse scenarios. We assess language models on three tasks: target word prediction (TWP), guided target word prediction (TWP$^*$) and target word selection (TWS). Our results reveal four key findings: (1) Higher average model performance TWS versus TWP and TWP$^*$, with average accuracy score increasing from 0.03 to 0.49 respectively (2) Stronger average model performance on \textsc{web} versus \textsc{gen} datasets, with average similarity score increasing by 0.03 and 0.05 across TWP and TWP$^*$ tasks respectively (3) en-IN tasks outperform en-AU when averaged across all models and datasets, with TWS demonstrating the largest disparity, increasing average accuracy from 0.44 to 0.54. These findings underscore fundamental asymmetries between generative and discriminative competencies for variety-specific language, particularly in the context of slang expressions despite being in a technologically rich language such as English.

Generative models have made it possible to synthesize highly realistic images, potentially providing an abundant data source for training machine learning models. Despite the advantages of these synthesizable data sources, the indiscriminate use of generated images as real images for training can even cause mode collapse due to modality discrepancies between real and synthetic domains. In this paper, we propose a novel framework for discriminative use of generated images, coined GMAIL, that explicitly treats generated images as a separate modality from real images. Instead of indiscriminately replacing real images with generated ones in the pixel space, our approach bridges the two distinct modalities in the same latent space through a multi-modal learning approach. To be specific, we first fine-tune a model exclusively on generated images using a cross-modality alignment loss and then employ this aligned model to further train various vision-language models with generated images. By aligning the two modalities, our approach effectively leverages the benefits of recent advances in generative models, thereby boosting the effectiveness of generated image learning across a range of vision-language tasks. Our framework can be easily incorporated with various vision-language models, and we demonstrate its efficacy throughout extensive experiments. For example, our framework significantly improves performance on image captioning, zero-shot image retrieval, zero-shot image classification, and long caption retrieval tasks. It also shows positive generated data scaling trends and notable enhancements in the captioning performance of the large multimodal model, LLaVA.

Reinforcement learning (RL) has achieved notable performance in high-dimensional sequential decision-making tasks, yet remains limited by low sample efficiency, sensitivity to noise, and weak generalization under partial observability. Most existing approaches address these issues primarily through optimization strategies, while the role of architectural priors in shaping representation learning and decision dynamics is less explored. Inspired by structural principles of the cerebellum, we propose a biologically grounded RL architecture that incorporate large expansion, sparse connectivity, sparse activation, and dendritic-level modulation. Experiments on noisy, high-dimensional RL benchmarks show that both the cerebellar architecture and dendritic modulation consistently improve sample efficiency, robustness, and generalization compared to conventional designs. Sensitivity analysis of architectural parameters suggests that cerebellum-inspired structures can offer optimized performance for RL with constrained model parameters. Overall, our work underscores the value of cerebellar structural priors as effective inductive biases for RL.

Modern web dashboards and enterprise applications increasingly rely on complex, distributed microservices architectures. While these architectures offer scalability, they introduce significant challenges in debugging and observability. When failures occur, they often manifest as opaque error messages to the end-user such as Something went wrong. This masks the underlying root cause which may reside in browser side exceptions, API contract violations, or server side logic failures. Existing monitoring tools capture these events in isolation but fail to correlate them effectively or provide intelligible explanations to non technical users. This paper proposes a novel system for Automated Multi Source Debugging and Natural Language Error Explanation. The proposed framework automatically collects and correlates error data from disparate sources such as browser, API, server logs and validates API contracts in real time, and utilizes Large Language Models to generate natural language explanations. This approach significantly reduces Mean Time to Resolution for support engineers and improves the user experience by transforming cryptic error codes into actionable insights.

Removing the CPU from the communication fast path is essential to efficient GPU-based ML and HPC application performance. However, existing GPU communication APIs either continue to rely on the CPU for communication or rely on APIs that place significant synchronization burdens on programmers. In this paper we describe the design, implementation, and evaluation of an MPI-based GPU communication API enabling easy-to-use, high-performance, CPU-free communication. This API builds on previously proposed MPI extensions and leverages HPE Slingshot 11 network card capabilities. We demonstrate the utility and performance of the API by showing how the API naturally enables CPU-free gather/scatter halo exchange communication primitives in the Cabana/Kokkos performance portability framework, and through a performance comparison with Cray MPICH on the Frontier and Tuolumne supercomputers. Results from this evaluation show up to a 50% reduction in medium message latency in simple GPU ping-pong exchanges and a 28% speedup improvement when strong scaling a halo-exchange benchmark to 8,192 GPUs of the Frontier supercomputer.

Large pretrained language models and neural reasoning systems have advanced many natural language tasks, yet they remain challenged by knowledge-intensive queries that require precise, structured multi-hop inference. Knowledge graphs provide a compact symbolic substrate for factual grounding, but integrating graph structure with neural models is nontrivial: naively embedding graph facts into prompts leads to inefficiency and fragility, while purely symbolic or search-heavy approaches can be costly in retrievals and lack gradient-based refinement. We introduce NeuroSymActive, a modular framework that combines a differentiable neural-symbolic reasoning layer with an active, value-guided exploration controller for Knowledge Graph Question Answering. The method couples soft-unification style symbolic modules with a neural path evaluator and a Monte-Carlo style exploration policy that prioritizes high-value path expansions. Empirical results on standard KGQA benchmarks show that NeuroSymActive attains strong answer accuracy while reducing the number of expensive graph lookups and model calls compared to common retrieval-augmented baselines.

Public Safety Power Shutoffs (PSPS) force rapid topology changes that can render standard operating points infeasible, requiring operators to quickly identify corrective transmission switching actions that reduce load shedding while maintaining acceptable voltage behavior. We present a verifiable, multi-stage adaptation pipeline that fine-tunes an instruction-tuned large language model (LLM) to generate \emph{open-only} corrective switching plans from compact PSPS scenario summaries under an explicit switching budget. First, supervised fine-tuning distills a DC-OPF MILP oracle into a constrained action grammar that enables reliable parsing and feasibility checks. Second, direct preference optimization refines the policy using AC-evaluated preference pairs ranked by a voltage-penalty metric, injecting voltage-awareness beyond DC imitation. Finally, best-of-$N$ selection provides an inference-time addition by choosing the best feasible candidate under the target metric. On IEEE 118-bus PSPS scenarios, fine-tuning substantially improves DC objective values versus zero-shot generation, reduces AC power-flow failure from 50\% to single digits, and improves voltage-penalty outcomes on the common-success set. Code and data-generation scripts are released to support reproducibility.

Large language models (LLMs) are increasingly augmented with long-term memory systems to overcome finite context windows and enable persistent reasoning across interactions. However, recent research finds that LLMs become more vulnerable because memory provides extra attack surfaces. In this paper, we present the first systematic study of black-box adversarial memory injection attacks that target the similarity-based retrieval mechanism in long-term memory-augmented LLMs. We introduce ER-MIA, a unified framework that exposes this vulnerability and formalizes two realistic attack settings: content-based attacks and question-targeted attacks. In these settings, ER-MIA includes an arsenal of composable attack primitives and ensemble attacks that achieve high success rates under minimal attacker assumptions. Extensive experiments across multiple LLMs and long-term memory systems demonstrate that similarity-based retrieval constitutes a fundamental and system-level vulnerability, revealing security risks that persist across memory designs and application scenarios.

Code smell is a great challenge in software refactoring, which indicates latent design or implementation flaws that may degrade the software maintainability and evolution. Over the past of decades, the research on code smell has received extensive attention. Especially the researches applied machine learning-technique have become a popular topic in recent studies. However, one of the biggest challenges to apply machine learning-technique is the lack of high-quality code smell datasets. Manually constructing such datasets is extremely labor-intensive, as identifying code smells requires substantial development expertise and considerable time investment. In contrast, automatically generated datasets, while scalable, frequently exhibit reduced label reliability and compromised data quality. To overcome this challenge, in this study, we explore a semi-automatic approach to generate a code smell dataset with high quality data samples. Specifically, we first applied a set of automatic generation rules to produce candidate smelly samples. We then employed multiple metrics to group the data samples into an automatically accepted group and a manually reviewed group, enabling reviewers to concentrate their efforts on ambiguous samples. Furthermore, we established structured review guidelines and developed a annotation tool to support the manual validation process. Based on the proposed semi-automatic generation approach, we created an open-source code smell dataset, SACS, covering three widely studied code smells: Long Method, Large Class, and Feature Envy. Each code smell category includes over 10,000 labeled samples. This dataset could provide a large-scale and publicly available benchmark to facilitate future studies on code smell detection and automated refactoring.

Reliable interpretation of cardiac ultrasound images is essential for accurate clinical diagnosis and assessment. Self-supervised learning has shown promise in medical imaging by leveraging large unlabelled datasets to learn meaningful representations. In this study, we evaluate and compare two self-supervised learning frameworks, USF-MAE, developed by our team, and MoCo v3, on the recently introduced CACTUS dataset (37,736 images) for automated simulated cardiac view (A4C, PL, PSAV, PSMV, Random, and SC) classification. Both models used 5-fold cross-validation, enabling robust assessment of generalization performance across multiple random splits. The CACTUS dataset provides expert-annotated cardiac ultrasound images with diverse views. We adopt an identical training protocol for both models to ensure a fair comparison. Both models are configured with a learning rate of 0.0001 and a weight decay of 0.01. For each fold, we record performance metrics including ROC-AUC, accuracy, F1-score, and recall. Our results indicate that USF-MAE consistently outperforms MoCo v3 across metrics. The average testing AUC for USF-MAE is 99.99% (+/-0.01% 95% CI), compared to 99.97% (+/-0.01%) for MoCo v3. USF-MAE achieves a mean testing accuracy of 99.33% (+/-0.18%), higher than the 98.99% (+/-0.28%) reported for MoCo v3. Similar trends are observed for the F1-score and recall, with improvements statistically significant across folds (paired t-test, p=0.0048 < 0.01). This proof-of-concept analysis suggests that USF-MAE learns more discriminative features for cardiac view classification than MoCo v3 when applied to this dataset. The enhanced performance across multiple metrics highlights the potential of USF-MAE for improving automated cardiac ultrasound classification.

Large language model (LLM) alignment relies on complex reward signals that often obscure the specific behaviors being incentivized, creating critical risks of misalignment and reward hacking. Existing interpretation methods typically rely on pre-defined rubrics, risking the omission of "unknown unknowns", or fail to identify objectives that comprehensively cover and are causal to the model behavior. To address these limitations, we introduce Obj-Disco, a framework that automatically decomposes an alignment reward signal into a sparse, weighted combination of human-interpretable natural language objectives. Our approach utilizes an iterative greedy algorithm to analyze behavioral changes across training checkpoints, identifying and validating candidate objectives that best explain the residual reward signal. Extensive evaluations across diverse tasks, model sizes, and alignment algorithms demonstrate the framework's robustness. Experiments with popular open-source reward models show that the framework consistently captures > 90% of reward behavior, a finding further corroborated by human evaluation. Additionally, a case study on alignment with an open-source reward model reveals that Obj-Disco can successfully identify latent misaligned incentives that emerge alongside intended behaviors. Our work provides a crucial tool for uncovering the implicit objectives in LLM alignment, paving the way for more transparent and safer AI development.

Asynchronous Federated Learning (AFL) has emerged as a significant research area in recent years. By not waiting for slower clients and executing the training process concurrently, it achieves faster training speed compared to traditional federated learning. However, due to the staleness introduced by the asynchronous process, its performance may degrade in some scenarios. Existing methods often use the round difference between the current model and the global model as the sole measure of staleness, which is coarse-grained and lacks observation of the model itself, thereby limiting the performance ceiling of asynchronous methods. In this paper, we propose FedPSA (Parameter Sensitivity-based Asynchronous Federated Learning), a more fine-grained AFL framework that leverages parameter sensitivity to measure model obsolescence and establishes a dynamic momentum queue to assess the current training phase in real time, thereby adjusting the tolerance for outdated information dynamically. Extensive experiments on multiple datasets and comparisons with various methods demonstrate the superior performance of FedPSA, achieving up to 6.37\% improvement over baseline methods and 1.93\% over the current state-of-the-art method.

Understanding how language models carry out long-horizon reasoning remains an open challenge. Existing interpretability methods often highlight tokens or spans correlated with an answer, but they rarely reveal where the model makes consequential reasoning turns, which earlier context causally triggers those turns, or whether the highlighted text actually steers the reasoning process. We introduce Directional Reasoning Trajectory Change (DRTC), a process-causal framework for interpreting long-form reasoning from a single on-policy rollout. DRTC detects pivot decision points using uncertainty and distribution-shift signals, then applies receiver-side interventions that preserve the realized rollout without resampling the continuation while blocking information flow from selected earlier chunks only at a pivot. It measures whether each intervention redirects the direction of the model's log-probability trajectory relative to the realized rollout direction, producing a signed per-chunk attribution score. We also compute turning-angle curvature changes on raw logits as a complementary diagnostic and introduce curvature signatures to summarize shared intervention-response geometry. Empirically, directional influence is sharply concentrated across four reasoning models (per-example |DRTC| shares yield Gini 0.50 to 0.58 and top-5 percent mass 0.23 to 0.28), and learned pivots induce stronger intervention magnitudes than matched random spans. In a scaling study on 500 MATH problems with R1-Distill-Qwen-1.5B, learned spans outperform matched random spans (median delta = 0.409, 355 of 500 positive; sign test p = 2.3e-21). Overall, DRTC provides a causally grounded, trajectory-level view of how specific context elements steer reasoning under on-policy dynamics.

The long-tail distribution, where a few head labels dominate while rare tail labels abound, poses a persistent challenge for large-scale Multi-Label Classification (MLC) in real-world data mining applications. Existing resampling and reweighting strategies often disrupt inter-label dependencies or require brittle hyperparameter tuning, especially as the label space expands to tens of thousands of labels. To address this issue, we propose Curiosity-Driven Game-Theoretic Multi-Label Learning (CD-GTMLL), a scalable cooperative framework that recasts long-tail MLC as a multi-player game - each sub-predictor ("player") specializes in a partition of the label space, collaborating to maximize global accuracy while pursuing intrinsic curiosity rewards based on tail label rarity and inter-player disagreement. This mechanism adaptively injects learning signals into under-represented tail labels without manual balancing or tuning. We further provide a theoretical analysis showing that our CD-GTMLL converges to a tail-aware equilibrium and formally links the optimization dynamics to improvements in the Rare-F1 metric. Extensive experiments across 7 benchmarks, including extreme multi-label classification datasets with 30,000+ labels, demonstrate that CD-GTMLL consistently surpasses state-of-the-art methods, with gains up to +1.6% P@3 on Wiki10-31K. Ablation studies further confirm the contributions of both game-theoretic cooperation and curiosity-driven exploration to robust tail performance. By integrating game theory with curiosity mechanisms, CD-GTMLL not only enhances model efficiency in resource-constrained environments but also paves the way for more adaptive learning in imbalanced data scenarios across industries like e-commerce and healthcare.

For deploying foundation models, practitioners increasingly need prescriptive scaling laws: given a pre training compute budget, what downstream accuracy is attainable with contemporary post training practice, and how stable is that mapping as the field evolves? Using large scale observational evaluations with 5k observational and 2k newly sampled data on model performance, we estimate capability boundaries, high conditional quantiles of benchmark scores as a function of log pre training FLOPs, via smoothed quantile regression with a monotone, saturating sigmoid parameterization. We validate the temporal reliability by fitting on earlier model generations and evaluating on later releases. Across various tasks, the estimated boundaries are mostly stable, with the exception of math reasoning that exhibits a consistently advancing boundary over time. We then extend our approach to analyze task dependent saturation and to probe contamination related shifts on math reasoning tasks. Finally, we introduce an efficient algorithm that recovers near full data frontiers using roughly 20% of evaluation budget. Together, our work releases the Proteus 2k, the latest model performance evaluation dataset, and introduces a practical methodology for translating compute budgets into reliable performance expectations and for monitoring when capability boundaries shift across time.

We propose SCENE (Self-Centering Noncoherent Estimator), a pilot-free and phase-invariant aggregation primitive for over-the-air federated distillation (OTA-FD). Each device maps its soft-label (class-probability) vector to nonnegative transmit energies under constant per-round power and constant-envelope signaling (PAPR near 1). At the server, a self-centering energy estimator removes the noise-energy offset and yields an unbiased estimate of the weighted soft-label average, with variance decaying on the order of 1/(SM) in the number of receive antennas M and repetition factor S. We also develop a pilot-free ratio-normalized variant that cancels unknown large-scale gains, provide a convergence bound consistent with coherent OTA-FD analyses, and present an overhead-based crossover comparison. SCENE targets short-coherence and hardware-constrained regimes, where avoiding per-round CSI is essential: it trades a modest noncoherent variance constant for zero uplink pilots, unbiased aggregation, and hardware-friendly transmission, and can outperform coherent designs when pilot overhead is non-negligible.

Foundation models for agriculture are increasingly trained on massive spatiotemporal data (e.g., multi-spectral remote sensing, soil grids, and field-level management logs) and achieve strong performance on forecasting and monitoring. However, these models lack language-based reasoning and interactive capabilities, limiting their usefulness in real-world agronomic workflows. Meanwhile, large language models (LLMs) excel at interpreting and generating text, but cannot directly reason over high-dimensional, heterogeneous agricultural datasets. We bridge this gap with an agentic framework for agricultural science. It provides a Python execution environment, AgriWorld, exposing unified tools for geospatial queries over field parcels, remote-sensing time-series analytics, crop growth simulation, and task-specific predictors (e.g., yield, stress, and disease risk). On top of this environment, we design a multi-turn LLM agent, Agro-Reflective, that iteratively writes code, observes execution results, and refines its analysis via an execute-observe-refine loop. We introduce AgroBench, with scalable data generation for diverse agricultural QA spanning lookups, forecasting, anomaly detection, and counterfactual "what-if" analysis. Experiments outperform text-only and direct tool-use baselines, validating execution-driven reflection for reliable agricultural reasoning.

Language models now routinely produce text that is difficult to distinguish from human writing, raising the need for robust tools to verify content provenance. Watermarking has emerged as a promising countermeasure, with existing work largely focused on model quality preservation and robust detection. However, current schemes provide limited protection against false attribution. We strengthen the notion of soundness by introducing two novel guarantees: unforgeability and recoverability. Unforgeability prevents adversaries from crafting false positives, texts that are far from any output from the watermarked model but are nonetheless flagged as watermarked. Recoverability provides an additional layer of protection: whenever a watermark is detected, the detector identifies the source text from which the flagged content was derived. Together, these properties strengthen content ownership by linking content exclusively to its generating model, enabling secure attribution and fine-grained traceability. We construct the first undetectable watermarking scheme that is robust, unforgeable, and recoverable with respect to substitutions (i.e., perturbations in Hamming metric). The key technical ingredient is a new cryptographic primitive called robust (or recoverable) digital signatures, which allow verification of messages that are close to signed ones, while preventing forgery of messages that are far from all previously signed messages. We show that any standard digital signature scheme can be boosted to a robust one using property-preserving hash functions (Boyle, LaVigne, and Vaikuntanathan, ITCS 2019).

Training large language models (LLMs) relies almost exclusively on dense adaptive optimizers with increasingly sophisticated preconditioners. We challenge this by showing that randomly masking parameter updates can be highly effective, with a masked variant of RMSProp consistently outperforming recent state-of-the-art optimizers. Our analysis reveals that the random masking induces a curvature-dependent geometric regularization that smooths the optimization trajectory. Motivated by this finding, we introduce Momentum-aligned gradient masking (Magma), which modulates the masked updates using momentum-gradient alignment. Extensive LLM pre-training experiments show that Magma is a simple drop-in replacement for adaptive optimizers with consistent gains and negligible computational overhead. Notably, for the 1B model size, Magma reduces perplexity by over 19\% and 9\% compared to Adam and Muon, respectively.

Although speculative decoding is widely used to accelerate Vision-Language Models (VLMs) inference, it faces severe performance collapse when applied to Video Large Language Models (Vid-LLMs). The draft model typically falls into the trap of attention dilution and negative visual gain due to key-value cache explosion and context window mismatches. We observe a visual semantic internalization phenomenon in Vid-LLMs, indicating that critical visual semantics are implicitly encoded into text hidden states during deep-layer interactions, which renders raw visual inputs structurally redundant during deep inference. To address this, we propose the Sparrow framework, which first utilizes visually-aware text-anchored window attention via hidden state reuse to fully offload visual computation to the target model, and leverages intermediate-layer visual state bridging to train the draft model with semantic-rich intermediate states, thereby filtering out low-level visual noise. Additionally, a multi-token prediction strategy is introduced to bridge the training-inference distribution shift. Experiments show that Sparrow achieves an average speedup of 2.82x even with 25k visual tokens, effectively resolving the performance degradation in long sequences and offering a practical solution for real-time long video tasks.

The rising complexity and scale of agent-based models (ABMs) necessitate efficient computational strategies to manage the increasing demand for processing power and memory. This manuscript provides a comprehensive guide to optimizing NetLogo, a widely used platform for ABMs, for running large-scale models on Amazon Web Services (AWS) and other cloud infrastructures. It covers best practices in memory management, Java options, BehaviorSpace execution, and AWS instance selection. By implementing these optimizations and selecting appropriate AWS instances, we achieved a 32\% reduction in computational costs and improved performance consistency. Through a comparative analysis of NetLogo simulations on different AWS instances using the wolf-sheep predation model, we demonstrate the performance gains achievable through these optimizations.

AI Memory, specifically how models organizes and retrieves historical messages, becomes increasingly valuable to Large Language Models (LLMs), yet existing methods (RAG and Graph-RAG) primarily retrieve memory through similarity-based mechanisms. While efficient, such System-1-style retrieval struggles with scenarios that require global reasoning or comprehensive coverage of all relevant information. In this work, We propose Mnemis, a novel memory framework that integrates System-1 similarity search with a complementary System-2 mechanism, termed Global Selection. Mnemis organizes memory into a base graph for similarity retrieval and a hierarchical graph that enables top-down, deliberate traversal over semantic hierarchies. By combining the complementary strength from both retrieval routes, Mnemis retrieves memory items that are both semantically and structurally relevant. Mnemis achieves state-of-the-art performance across all compared methods on long-term memory benchmarks, scoring 93.9 on LoCoMo and 91.6 on LongMemEval-S using GPT-4.1-mini.

Accurately measuring consumer emotions and evaluations from unstructured text remains a core challenge for marketing research and practice. This study introduces the Linguistic eXtractor (LX), a fine-tuned, large language model trained on consumer-authored text that also has been labeled with consumers' self-reported ratings of 16 consumption-related emotions and four evaluation constructs: trust, commitment, recommendation, and sentiment. LX consistently outperforms leading models, including GPT-4 Turbo, RoBERTa, and DeepSeek, achieving 81% macro-F1 accuracy on open-ended survey responses and greater than 95% accuracy on third-party-annotated Amazon and Yelp reviews. An application of LX to online retail data, using seemingly unrelated regression, affirms that review-expressed emotions predict product ratings, which in turn predict purchase behavior. Most emotional effects are mediated by product ratings, though some emotions, such as discontent and peacefulness, influence purchase directly, indicating that emotional tone provides meaningful signals beyond star ratings. To support its use, a no-code, cost-free, LX web application is available, enabling scalable analyses of consumer-authored text. In establishing a new methodological foundation for consumer perception measurement, this research demonstrates new methods for leveraging large language models to advance marketing research and practice, thereby achieving validated detection of marketing constructs from consumer data.

Causal structure learning, also known as causal discovery, aims to estimate causal relationships between variables as a form of a causal directed acyclic graph (DAG) from observational data. One of the major frameworks is the order-based approach that first estimates a topological order of the underlying DAG and then prunes spurious edges from the fully-connected DAG induced by the estimated topological order. Previous studies often focus on the former ordering step because it can dramatically reduce the search space of DAGs. In practice, the latter pruning step is equally crucial for ensuring both computational efficiency and estimation accuracy. Most existing methods employ a pruning technique based on generalized additive models and hypothesis testing, commonly known as CAM-pruning. However, this approach can be a computational bottleneck as it requires repeatedly fitting additive models for all variables. Furthermore, it may harm estimation quality due to multiple testing. To address these issues, we introduce a new pruning method based on sparse additive models, which enables direct pruning of redundant edges without relying on hypothesis testing. We propose an efficient algorithm for learning sparse additive models by combining the randomized tree embedding technique with group-wise sparse regression. Experimental results on both synthetic and real datasets demonstrated that our method is significantly faster than existing pruning methods while maintaining comparable or superior accuracy.

Collaborative clinical decision support is often constrained by governance and privacy rules that prevent pooling patient-level records across institutions. We present a hybrid privacy-preserving framework that combines Federated Learning (FL) and Split Learning (SL) to support decision-oriented healthcare modeling without raw-data sharing. The approach keeps feature-extraction trunks on clients while hosting prediction heads on a coordinating server, enabling shared representation learning and exposing an explicit collaboration boundary where privacy controls can be applied. Rather than assuming distributed training is inherently private, we audit leakage empirically using membership inference on cut-layer representations and study lightweight defenses based on activation clipping and additive Gaussian noise. We evaluate across three public clinical datasets under non-IID client partitions using a unified pipeline and assess performance jointly along four deployment-relevant axes: factual predictive utility, uplift-based ranking under capacity constraints, audited privacy leakage, and communication overhead. Results show that hybrid FL-SL variants achieve competitive predictive performance and decision-facing prioritization behavior relative to standalone FL or SL, while providing a tunable privacy-utility trade-off that can reduce audited leakage without requiring raw-data sharing. Overall, the work positions hybrid FL-SL as a practical design space for privacy-preserving healthcare decision support where utility, leakage risk, and deployment cost must be balanced explicitly.

Misclassifications in spam and phishing detection are very harmful, as false negatives expose users to attacks while false positives degrade trust. Existing uncertainty-based detectors can flag potential errors, but possibly be deceived and offer limited interpretability. This paper presents X-MAP, an eXplainable Misclassification Analysis and Profilling framework that reveals topic-level semantic patterns behind model failures. X-MAP combines SHAP-based feature attributions with non-negative matrix factorization to build interpretable topic profiles for reliably classified spam/phishing and legitimate messages, and measures each message's deviation from these profiles using Jensen-Shannon divergence. Experiments on SMS and phishing datasets show that misclassified messages exhibit at least two times larger divergence than correctly classified ones. As a detector, X-MAP achieves up to 0.98 AUROC and lowers the false-rejection rate at 95% TRR to 0.089 on positive predictions. When used as a repair layer on base detectors, it recovers up to 97% of falsely rejected correct predictions with moderate leakage. These results demonstrate X-MAP's effectiveness and interpretability for improving spam and phishing detection.

We present Experiment Automation Agents (EAA), a vision-language-model-driven agentic system designed to automate complex experimental microscopy workflows. EAA integrates multimodal reasoning, tool-augmented action, and optional long-term memory to support both autonomous procedures and interactive user-guided measurements. Built on a flexible task-manager architecture, the system enables workflows ranging from fully agent-driven automation to logic-defined routines that embed localized LLM queries. EAA further provides a modern tool ecosystem with two-way compatibility for Model Context Protocol (MCP), allowing instrument-control tools to be consumed or served across applications. We demonstrate EAA at an imaging beamline at the Advanced Photon Source, including automated zone plate focusing, natural language-described feature search, and interactive data acquisition. These results illustrate how vision-capable agents can enhance beamline efficiency, reduce operational burden, and lower the expertise barrier for users.

This paper concerns the question of how AI systems encode semantic structure into the geometric structure of their representation spaces. The motivating observation of this paper is that the natural geometry of these representation spaces should reflect the way models use representations to produce behavior. We focus on the important special case of representations that define softmax distributions. In this case, we argue that the natural geometry is information geometry. Our focus is on the role of information geometry on semantic encoding and the linear representation hypothesis. As an illustrative application, we develop "dual steering", a method for robustly steering representations to exhibit a particular concept using linear probes. We prove that dual steering optimally modifies the target concept while minimizing changes to off-target concepts. Empirically, we find that dual steering enhances the controllability and stability of concept manipulation.

With AI-as-a-Service (AIaaS) now deployed across multiple providers and model tiers, selecting the appropriate model instance at run time is increasingly outside the end user's knowledge and operational control. Accordingly, the 6G service providers are envisioned to play a crucial role in exposing AIaaS in a setting where users submit only an intent while the network helps in the intent-to-model matching (resolution) and execution placement under policy, trust, and Quality of Service (QoS) constraints. The network role becomes to discover candidate execution endpoints and selects a suitable model/anchor under policy and QoS constraints in a process referred here to as AI-paging (by analogy to cellular call paging). In the proposed architecture, AI-paging is a control-plane transaction that resolves an intent into an AI service identity (AISI), a scoped session token (AIST), and an expiring admission lease (COMMIT) that authorizes user-plane steering to a selected AI execution anchor (AEXF) under a QoS binding. AI-Paging enforces two invariants: (i) lease-gated steering (without COMMIT, no steering state is installed) and (ii) make-before-break anchoring to support continuity and reliability of AIaaS services under dynamic network conditions. We prototype AI-Paging using existing control- and user-plane mechanisms (service-based control, QoS flows, and policy-based steering) with no new packet headers, ensuring compatibility with existing 3GPP-based exposure and management architectures, and evaluate transaction latency, relocation interruption, enforcement correctness under lease expiry, and audit-evidence overhead under mobility and failures.

Modern deep neural networks achieve high predictive accuracy but remain poorly calibrated: their confidence scores do not reliably reflect the true probability of correctness. We propose a quantum-inspired classification head architecture that projects backbone features into a complex-valued Hilbert space and evolves them under a learned unitary transformation parameterised via the Cayley map. Through a controlled hybrid experimental design - training a single shared backbone and comparing lightweight interchangeable heads - we isolate the effect of complex-valued unitary representations on calibration. Our ablation study on CIFAR-10 reveals that the unitary magnitude head (complex features evolved under a Cayley unitary, read out via magnitude and softmax) achieves an Expected Calibration Error (ECE) of 0.0146, representing a 2.4x improvement over a standard softmax head (0.0355) and a 3.5x improvement over temperature scaling (0.0510). Surprisingly, replacing the softmax readout with a Born rule measurement layer - the quantum-mechanically motivated approach - degrades calibration to an ECE of 0.0819. On the CIFAR-10H human-uncertainty benchmark, the wave function head achieves the lowest KL-divergence (0.336) to human soft labels among all compared methods, indicating that complex-valued representations better capture the structure of human perceptual ambiguity. We provide theoretical analysis connecting norm-preserving unitary dynamics to calibration through feature-space geometry, report negative results on out-of-distribution detection and sentiment analysis to delineate the method's scope, and discuss practical implications for safety-critical applications. Code is publicly available.

To support the emergence of AI-as-a-Service (AIaaS), communication service providers (CSPs) are on the verge of a radical transformation-from pure connectivity providers to AIaaS a managed network service (control-and-orchestration plane that exposes AI models). In this model, the CSP is responsible not only for transport/communications, but also for intent-to-model resolution and joint network-compute orchestration, i.e., reliable and timely end-to-end delivery. The resulting end-to-end AIaaS service thus becomes governed by communications impairments (delay, loss) and inference impairments (latency, error). A central open problem is an operational AIaaS control-and-orchestration framework that enforces high fidelity, particularly under multi-domain federation. This paper introduces an assurance-oriented AIaaS management plane based on Tail-Risk Envelopes (TREs): signed, composable per-domain descriptors that combine deterministic guardrails with stochastic rate-latency-impairment models. Using stochastic network calculus, we derive bounds on end-to-end delay violation probabilities across tandem domains and obtain an optimization-ready risk-budget decomposition. We show that tenant-level reservations prevent bursty traffic from inflating tail latency under TRE contracts. An auditing layer then uses runtime telemetry to estimate extreme-percentile performance, quantify uncertainty, and attribute tail-risk to each domain for accountability. Packet-level Monte-Carlo simulations demonstrate improved p99.9 compliance under overload via admission control and robust tenant isolation under correlated burstiness.

The web is littered with images, once created for human consumption and now increasingly interpreted by agents using vision-language models (VLMs). These agents make visual decisions at scale, deciding what to click, recommend, or buy. Yet, we know little about the structure of their visual preferences. We introduce a framework for studying this by placing VLMs in controlled image-based choice tasks and systematically perturbing their inputs. Our key idea is to treat the agent's decision function as a latent visual utility that can be inferred through revealed preference: choices between systematically edited images. Starting from common images, such as product photos, we propose methods for visual prompt optimization, adapting text optimization methods to iteratively propose and apply visually plausible modifications using an image generation model (such as in composition, lighting, or background). We then evaluate which edits increase selection probability. Through large-scale experiments on frontier VLMs, we demonstrate that optimized edits significantly shift choice probabilities in head-to-head comparisons. We develop an automatic interpretability pipeline to explain these preferences, identifying consistent visual themes that drive selection. We argue that this approach offers a practical and efficient way to surface visual vulnerabilities, safety concerns that might otherwise be discovered implicitly in the wild, supporting more proactive auditing and governance of image-based AI agents.

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^2D), 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^2D on large-scale benchmarks, surpassing the state-of-the-art on ImageNet-1K while being 18x faster, and on ImageNet-21K, our method substantially improves accuracy while remaining 4.3x 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.

We explore how different types and uses of memory can aid spatial navigation in changing uncertain environments. In the simple foraging task we study, every day, our agent has to find its way from its home, through barriers, to food. Moreover, the world is non-stationary: from day to day, the location of the barriers and food may change, and the agent's sensing such as its location information is uncertain and very limited. Any model construction, such as a map, and use, such as planning, needs to be robust against these challenges, and if any learning is to be useful, it needs to be adequately fast. We look at a range of strategies, from simple to sophisticated, with various uses of memory and learning. We find that an architecture that can incorporate multiple strategies is required to handle (sub)tasks of a different nature, in particular for exploration and search, when food location is not known, and for planning a good path to a remembered (likely) food location. An agent that utilizes non-stationary probability learning techniques to keep updating its (episodic) memories and that uses those memories to build maps and plan on the fly (imperfect maps, i.e. noisy and limited to the agent's experience) can be increasingly and substantially more efficient than the simpler (minimal-memory) agents, as the task difficulties such as distance to goal are raised, as long as the uncertainty, from localization and change, is not too large.

Information ecosystems increasingly shape how people internalize exposure to adverse digital experiences, raising concerns about the long-term consequences for information health. In modern search and recommendation systems, ranking and personalization policies play a central role in shaping such exposure and its long-term effects on users. To study these effects in a controlled setting, we present FrameRef, a large-scale dataset of 1,073,740 systematically reframed claims across five framing dimensions: authoritative, consensus, emotional, prestige, and sensationalist, and propose a simulation-based framework for modeling sequential information exposure and reinforcement dynamics characteristic of ranking and recommendation systems. Within this framework, we construct framing-sensitive agent personas by fine-tuning language models with framing-conditioned loss attenuation, inducing targeted biases while preserving overall task competence. Using Monte Carlo trajectory sampling, we show that small, systematic shifts in acceptance and confidence can compound over time, producing substantial divergence in cumulative information health trajectories. Human evaluation further confirms that FrameRef's generated framings measurably affect human judgment. Together, our dataset and framework provide a foundation for systematic information health research through simulation, complementing and informing responsible human-centered research. We release FrameRef, code, documentation, human evaluation data, and persona adapter models at https://github.com/infosenselab/frameref.

Generating realistic synthetic populations is essential for agent-based models (ABM) in transportation and urban planning. Current methods face two major limitations. First, many rely on a single dataset or follow a sequential data fusion and generation process, which means they fail to capture the complex interplay between features. Second, these approaches struggle with sampling zeros (valid but unobserved attribute combinations) and structural zeros (infeasible combinations due to logical constraints), which reduce the diversity and feasibility of the generated data. This study proposes a novel method to simultaneously integrate and synthesize multi-source datasets using a Wasserstein Generative Adversarial Network (WGAN) with gradient penalty. This joint learning method improves both the diversity and feasibility of synthetic data by defining a regularization term (inverse gradient penalty) for the generator loss function. For the evaluation, we implement a unified evaluation metric for similarity, and place special emphasis on measuring diversity and feasibility through recall, precision, and the F1 score. Results show that the proposed joint approach outperforms the sequential baseline, with recall increasing by 7\% and precision by 15\%. Additionally, the regularization term further improves diversity and feasibility, reflected in a 10\% increase in recall and 1\% in precision. We assess similarity distributions using a five-metric score. The joint approach performs better overall, and reaches a score of 88.1 compared to 84.6 for the sequential method. Since synthetic populations serve as a key input for ABM, this multi-source generative approach has the potential to significantly enhance the accuracy and reliability of ABM.

Recent empirical research by Sharma et al. (2026) demonstrated that AI assistant interactions carry meaningful potential for situational human disempowerment, including reality distortion, value judgment distortion, and action distortion. While this work provides a critical diagnosis of the problem, concrete pedagogical interventions remain underexplored. I present an AI literacy framework built around eight cross-cutting Learning Outcomes (LOs), developed independently through teaching practice and subsequently found to align with Sharma et al.'s disempowerment taxonomy. I report a case study from a publicly available online course, where a co-teaching methodology--with AI serving as an active voice co-instructor--was used to deliver this framework. Drawing on inoculation theory (McGuire, 1961)--a well-established persuasion research framework recently applied to misinformation prebunking by the Cambridge school (van der Linden, 2022; Roozenbeek & van der Linden, 2019)--I argue that AI literacy cannot be acquired through declarative knowledge alone, but requires guided exposure to AI failure modes, including the sycophantic validation and authority projection patterns identified by Sharma et al. This application of inoculation theory to AI-specific distortion is, to my knowledge, novel. I discuss the convergence between the pedagogically-derived framework and Sharma et al.'s empirically-derived taxonomy, and argue that this convergence--two independent approaches arriving at similar problem descriptions--strengthens the case for both the diagnosis and the proposed educational response.

On-policy distillation (OPD), which samples trajectories from the student model and supervises them with a teacher at the token level, avoids relying solely on verifiable terminal rewards and can yield better generalization than off-policy distillation. However, OPD requires expensive on-the-fly sampling of the student policy during training, which substantially increases training cost, especially for long responses. Our initial analysis shows that, during OPD, training signals are often concentrated in the prefix of each output, and that even a short teacher-generated prefix can significantly help the student produce the correct answer. Motivated by these observations, we propose a simple yet effective modification of OPD: we apply the distillation objective only to prefixes of student-generated outputs and terminate each sampling early during distillation. Experiments on a suite of AI-for-Math and out-of-domain benchmarks show that on-policy prefix distillation matches the performance of full OPD while reducing training FLOP by 2x-47x.

Generative AI agents equate understanding with resolving explicit queries, an assumption that confines interaction to what users can articulate. This assumption breaks down when users themselves lack awareness of what is missing, risky, or worth considering. In such conditions, proactivity is not merely an efficiency enhancement, but an epistemic necessity. We refer to this condition as epistemic incompleteness: where progress depends on engaging with unknown unknowns for effective partnership. Existing approaches to proactivity remain narrowly anticipatory, extrapolating from past behavior and presuming that goals are already well defined, thereby failing to support users meaningfully. However, surfacing possibilities beyond a user's current awareness is not inherently beneficial. Unconstrained proactive interventions can misdirect attention, overwhelm users, or introduce harm. Proactive agents, therefore, require behavioral grounding: principled constraints on when, how, and to what extent an agent should intervene. We advance the position that generative proactivity must be grounded both epistemically and behaviorally. Drawing on the philosophy of ignorance and research on proactive behavior, we argue that these theories offer critical guidance for designing agents that can engage responsibly and foster meaningful partnerships.

We present the first comprehensive, large-scale study of training long-context vision language models up to 344K context, targeting long-document visual question answering with measured transfer to long-context text. While several such strong are open-weight, namely Qwen3 VL and GLM 4.5/6V, their training recipes and data pipelines are not reproducible. We systematically study continued pretraining, supervised finetuning, and preference optimization for 24B and 32B parameter models, backed by extensive LC evaluations and ablations to bridge this gap, and achieve state-of-the-art performance on MMLongBenchDoc for both parameter scales. In addition to this, our key findings include: (i) training on context lengths that match evaluation context lengths outperforms training on longer contexts, (ii) training and evaluating with page indices provides a simple, high-impact boost to long-document performance, (iii) our synthetic data pipelines enable self-improvement via continued pretraining and supervised finetuning, and (iv) we extend the known text-to-visual long context transfer to the reverse, showing that visual long context training transfers to long-context text performance. We also release MMLBD-C, a manually corrected version of MMLongBenchDoc to reduce erroneous and low quality examples in the benchmark.

Neural scaling laws -- power-law relationships between loss, model size, and data -- have been extensively documented for language and vision transformers, yet their existence in single-cell genomics remains largely unexplored. We present the first systematic study of scaling behaviour for masked-reconstruction transformers trained on single-cell RNA sequencing (scRNA-seq) data. Using expression profiles from the CELLxGENE Census, we construct two experimental regimes: a data-rich regime (512 highly variable genes, 200,000 cells) and a data-limited regime (1,024 genes, 10,000 cells). Across seven model sizes spanning three orders of magnitude in parameter count (533 to 3.4 x 10^8 parameters), we fit the parametric scaling law to validation mean squared error (MSE). The data-rich regime exhibits clear power-law scaling with an irreducible loss floor of c ~ 1.44, while the data-limited regime shows negligible scaling, indicating that model capacity is not the binding constraint when data are scarce. These results establish that scaling laws analogous to those observed in natural language processing do emerge in single-cell transcriptomics when sufficient data are available, and they identify the data-to-parameter ratio as a critical determinant of scaling behaviour. A preliminary conversion of the data-rich asymptotic floor to information-theoretic units yields an estimate of approximately 2.30 bits of entropy per masked gene position. We discuss implications for the design of single-cell foundation models and outline the additional measurements needed to refine this entropy estimate.

In game theory, imperfect-recall decision problems model situations in which an agent forgets information it held before. They encompass games such as the ``absentminded driver'' and team games with limited communication. In this paper, we introduce the first benchmark suite for imperfect-recall decision problems. Our benchmarks capture a variety of problem types, including ones concerning privacy in AI systems that elicit sensitive information, and AI safety via testing of agents in simulation. Across 61 problem instances generated using this suite, we evaluate the performance of different algorithms for finding first-order optimal strategies in such problems. In particular, we introduce the family of regret matching (RM) algorithms for nonlinear constrained optimization. This class of parameter-free algorithms has enjoyed tremendous success in solving large two-player zero-sum games, but, surprisingly, they were hitherto relatively unexplored beyond that setting. Our key finding is that RM algorithms consistently outperform commonly employed first-order optimizers such as projected gradient descent, often by orders of magnitude. This establishes, for the first time, the RM family as a formidable approach to large-scale constrained optimization problems.

This study examines the geographical distribution of Artificial Intelligence (AI) research production across European regions at the NUTS-3 level for the period 2015-2024. Using bibliometric data from Clarivate InCites and the Citation Topics classification system, we analyze two hierarchical levels of thematic aggregation: Electrical Engineering, Electronics & Computer Science (Macro Citation Topic 4) and Artificial Intelligence & Machine Learning (Meso Citation Topic 4.61). We calculate the Relative Specialization Index (RSI) and Relative Citation Impact (RCI) for 781 NUTS-3 regions. While major metropolitan hubs such as Paris (IIle-de-France), Warszawa, and Madrid lead in absolute production volume, our findings reveal that peripheral regions, particularly from Eastern Europe and Spain, exhibit the highest levels of relative AI specialization. Notably, we find virtually no correlation between regional specialization and citation impact, identifying four distinct regional profiles: high-impact specialized regions (e.g., Granada, Jaen, Vilniaus), high-volume but low-impact regions (e.g., Bugas, several Polish regions), high-impact non-specialized regions, with Fyn (Denmark) standing out as a remarkable outlier achieving exceptional citation impact (RCI > 4) despite low specialization, and diversified portfolios with selective excellence (e.g., German regions). These results suggest that AI research represents a strategic opportunity for peripheral regions to develop competitive scientific niches, though achieving international visibility requires more than research volume alone.

Invoice or payment dilution is the gap between the approved invoice amount and the actual collection is a significant source of non credit risk and margin loss in supply chain finance. Traditionally, this risk is managed through the buyer's irrevocable payment undertaking (IPU), which commits to full payment without deductions. However, IPUs can hinder supply chain finance adoption, particularly among sub-invested grade buyers. A newer, data-driven methods use real-time dynamic credit limits, projecting dilution for each buyer-supplier pair in real-time. This paper introduces an AI, machine learning framework and evaluates how that can supplement a deterministic algorithm to predict invoice dilution using extensive production dataset across nine key transaction fields.

Reinforcement learning (RL)-based biomechanical simulations have the potential to revolutionise HCI research and interaction design, but currently lack usability and interpretability. Using the Human Action Cycle as a design lens, we identify key limitations of biomechanical RL frameworks and develop MyoInteract, a novel framework for fast prototyping of biomechanical HCI tasks. MyoInteract allows designers to setup tasks, user models, and training parameters from an easy-to-use GUI within minutes. It trains and evaluates muscle-actuated simulated users within minutes, reducing training times by up to 98%. A workshop study with 12 interaction designers revealed that MyoInteract allowed novices in biomechanical RL to successfully setup, train, and assess goal-directed user movements within a single session. By transforming biomechanical RL from a days-long expert task into an accessible hour-long workflow, this work significantly lowers barriers to entry and accelerates iteration cycles in HCI biomechanics research.

Generative AI is reshaping how computing systems are designed, optimized, and built, yet research remains fragmented across software, architecture, and chip design communities. This paper takes a cross-stack perspective, examining how generative models are being applied from code generation and distributed runtimes through hardware design space exploration to RTL synthesis, physical layout, and verification. Rather than reviewing each layer in isolation, we analyze how the same structural difficulties and effective responses recur across the stack. Our central finding is one of convergence. Despite the diversity of domains and tools, the field keeps encountering five recurring challenges (the feedback loop crisis, the tacit knowledge problem, trust and validation, co-design across boundaries, and the shift from determinism to dynamism) and keeps arriving at five design principles that independently emerge as effective responses (embracing hybrid approaches, designing for continuous feedback, separating concerns by role, matching methods to problem structure, and building on decades of systems knowledge). We organize these into a challenge--principle map that serves as a diagnostic and design aid, showing which principles have proven effective for which challenges across layers. Through concrete cross-stack examples, we show how systems navigate this map as they mature, and argue that the field needs shared engineering methodology, including common vocabularies, cross-layer benchmarks, and systematic design practices, so that progress compounds across communities rather than being rediscovered in each one. Our analysis covers more than 275 papers spanning eleven application areas across three layers of the computing stack, and distills open research questions that become visible only from a cross-layer vantage point.

Transformers have achieved remarkable success across domains, motivating the rise of Graph Transformers (GTs) as attention-based architectures for graph-structured data. A key design choice in GTs is the use of Graph Neural Network (GNN)-based positional encodings to incorporate structural information. In this work, we study GTs through the lens of manifold limit models for graph sequences and establish a theoretical connection between GTs with GNN positional encodings and Manifold Neural Networks (MNNs). Building on transferability results for GNNs under manifold convergence, we show that GTs inherit transferability guarantees from their positional encodings. In particular, GTs trained on small graphs provably generalize to larger graphs under mild assumptions. We complement our theory with extensive experiments on standard graph benchmarks, demonstrating that GTs exhibit scalable behavior on par with GNNs. To further show the efficiency in a real-world scenario, we implement GTs for shortest path distance estimation over terrains to better illustrate the efficiency of the transferable GTs. Our results provide new insights into the understanding of GTs and suggest practical directions for efficient training of GTs in large-scale settings.

Adversarial training for LLMs is one of the most promising methods to reliably improve robustness against adversaries. However, despite significant progress, models remain vulnerable to simple in-distribution exploits, such as rewriting prompts in the past tense or translating them into other languages. We argue that this persistent fragility stems from a fundamental limitation in current adversarial training algorithms: they minimize adversarial loss on their training set but inadequately cover the data distribution, resulting in vulnerability to seemingly simple attacks. To bridge this gap, we propose Distributional Adversarial Training, DAT. We leverage Diffusion LLMs to approximate the true joint distribution of prompts and responses, enabling generation of diverse, high-likelihood samples that address generalization failures. By combining optimization over the data distribution provided by the diffusion model with continuous adversarial training, DAT achieves substantially higher adversarial robustness than previous methods.

Virtual screening aims to efficiently identify active ligands from massive chemical libraries for a given target pocket. Recent CLIP-style models such as DrugCLIP enable scalable virtual screening by embedding pockets and ligands into a shared space. However, our analyses indicate that such representations can be insensitive to fine-grained binding interactions and may rely on shortcut correlations in training data, limiting their ability to rank ligands by true binding compatibility. To address these issues, we propose BindCLIP, a unified contrastive-generative representation learning framework for virtual screening. BindCLIP jointly trains pocket and ligand encoders using CLIP-style contrastive learning together with a pocket-conditioned diffusion objective for binding pose generation, so that pose-level supervision directly shapes the retrieval embedding space toward interaction-relevant features. To further mitigate shortcut reliance, we introduce hard-negative augmentation and a ligand-ligand anchoring regularizer that prevents representation collapse. Experiments on two public benchmarks demonstrate consistent improvements over strong baselines. BindCLIP achieves substantial gains on challenging out-of-distribution virtual screening and improves ligand-analogue ranking on the FEP+ benchmark. Together, these results indicate that integrating generative, pose-level supervision with contrastive learning yields more interaction-aware embeddings and improves generalization in realistic screening settings, bringing virtual screening closer to real-world applicability.

We address prediction problems on tabular categorical data, where each instance is defined by multiple categorical attributes, each taking values from a finite set. These attributes are often referred to as fields, and their categorical values as features. Such problems frequently arise in practical applications, including click-through rate prediction and social sciences. We introduce and analyze {tensorFM}, a new model that efficiently captures high-order interactions between attributes via a low-rank tensor approximation representing the strength of these interactions. Our model generalizes field-weighted factorization machines. Empirically, tensorFM demonstrates competitive performance with state-of-the-art methods. Additionally, its low latency makes it well-suited for time-sensitive applications, such as online advertising.

Instruction-tuned Language Models (ILMs) have become essential components of modern AI systems, demonstrating exceptional versatility across natural language and reasoning tasks. Among their most impactful applications is code generation, where ILMs -- commonly referred to as Code Language Models (CLMs) -- translate human intent into executable programs. While progress has been driven by advances in scaling and training methodologies, one critical aspect remains underexplored: the impact of system prompts on both general-purpose ILMs and specialized CLMs for code generation. We systematically evaluate how system prompts of varying instructional detail, along with model scale, prompting strategy, and programming language, affect code assistant. Our experimental setting spans 360 configurations across four models, five system prompts, three prompting strategies, two languages, and two temperature settings. We find that (1) increasing system-prompt constraint specificity does not monotonically improve correctness -- prompt effectiveness is configuration-dependent and can help or hinder based on alignment with task requirements and decoding context; (2) for larger code-specialized models, few-shot examples can degrade performance relative to zero-shot generation, contrary to conventional wisdom; and (3) programming language matters, with Java exhibiting significantly greater sensitivity to system prompt variations than Python, suggesting language-specific prompt engineering strategies may be necessary.

Reward models are central to large language model (LLM) post-training. However, past work has shown that they can reward spurious or undesirable attributes such as length, format, hallucinations, and sycophancy. In this work, we introduce and study the research problem of automatically finding reward model biases in natural language. We offer a simple approach of using an LLM to iteratively propose and refine candidate biases. Our method can recover known biases and surface novel ones: for example, we found that Skywork-V2-8B, a leading open-weight reward model, often mistakenly favors responses with redundant spacing and responses with hallucinated content. In addition, we show evidence that evolutionary iteration outperforms flat best-of-N search, and we validate the recall of our pipeline using synthetically injected biases. We hope our work contributes to further research on improving RMs through automated interpretability methods.

In this paper, we introduce a secure wireless agentic AI network comprising one supervisor AI agent and multiple other AI agents to provision quality of service (QoS) for users' reasoning tasks while ensuring confidentiality of private knowledge and reasoning outcomes. Specifically, the supervisor AI agent can dynamically assign other AI agents to participate in cooperative reasoning, while the unselected AI agents act as friendly jammers to degrade the eavesdropper's interception performance. To extend the service duration of AI agents, an energy minimization problem is formulated that jointly optimizes AI agent selection, base station (BS) beamforming, and AI agent transmission power, subject to latency and reasoning accuracy constraints. To address the formulated problem, we propose two resource allocation schemes, ASC and LAW, which first decompose it into three sub-problems. Specifically, ASC optimizes each sub-problem iteratively using the proposed alternating direction method of multipliers (ADMM)-based algorithm, semi-definite relaxation (SDR), and successive convex approximation (SCA), while LAW tackles each sub-problem using the proposed large language model (LLM) optimizer within an agentic workflow. The experimental results show that the proposed solutions can reduce network energy consumption by up to 59.1% compared to other benchmark schemes. Furthermore, the proposed schemes are validated using a practical agentic AI system based on Qwen, demonstrating satisfactory reasoning accuracy across various public benchmarks.

Multilinguality is a core capability for modern foundation models, yet training high-quality multilingual models remains challenging due to uneven data availability across languages. A further challenge is the performance interference that can arise from joint multilingual training, commonly referred to as the "curse of multilinguality". We study multilingual data curation across thirteen languages and find that many reported regressions are not inherent to multilingual scaling but instead stem from correctable deficiencies in data quality and composition rather than fundamental capacity limits. In controlled bilingual experiments, improving data quality for any single language benefits others: curating English improves non-English performance in 12 of 13 languages, while curating non-English yields reciprocal improvements in English. Bespoke per-language curation produces substantially larger within-language improvements. Extending these findings to large-scale general-purpose training mixtures, we show that curated multilingual allocations comprising under 8% of total tokens remain remarkably effective. We operationalize this approach within an effort that produced a 20T-token pretraining corpus derived entirely from public sources. Models with 3B and 8B parameters trained on a 1T-token random subset achieve competitive multilingual accuracy with 4-10x fewer training FLOPs than strong public baselines, establishing a new Pareto frontier in multilingual performance versus compute. Moreover, these benefits extend to frontier model scale: the 20T-token corpus served as part of the pretraining dataset for Trinity Large (400B/A13B), which exhibits strong multilingual performance relative to its training FLOPs. These results show that targeted, per-language data curation mitigates multilingual interference and enables compute-efficient multilingual scaling.

Reward learning typically relies on a single feedback type or combines multiple feedback types using manually weighted loss terms. Currently, it remains unclear how to jointly learn reward functions from heterogeneous feedback types such as demonstrations, comparisons, ratings, and stops that provide qualitatively different signals. We address this challenge by formulating reward learning from multiple feedback types as Bayesian inference over a shared latent reward function, where each feedback type contributes information through an explicit likelihood. We introduce a scalable amortized variational inference approach that learns a shared reward encoder and feedback-specific likelihood decoders and is trained by optimizing a single evidence lower bound. Our approach avoids reducing feedback to a common intermediate representation and eliminates the need for manual loss balancing. Across discrete and continuous-control benchmarks, we show that jointly inferred reward posteriors outperform single-type baselines, exploit complementary information across feedback types, and yield policies that are more robust to environment perturbations. The inferred reward uncertainty further provides interpretable signals for analyzing model confidence and consistency across feedback types.

This paper presents a distributed memory method for anisotropic mesh adaptation that is designed to avoid the use of collective communication and global synchronization techniques. In the presented method, meshing functionality is separated from performance aspects by utilizing a separate entity for each - a multicore cc-NUMA-based (shared memory) mesh generation software and a parallel runtime system that is designed to help applications leverage the concurrency offered by emerging high-performance computing (HPC) architectures. First, an initial mesh is decomposed and its interface elements (subdomain boundaries) are adapted on a single multicore node (shared memory). Subdomains are then distributed among the nodes of an HPC cluster so that their interior elements are adapted while interface elements (already adapted) remain frozen to maintain mesh conformity. Lessons are presented regarding some re-designs of the shared memory software and how its speculative execution model is utilized by the distributed memory method to achieve good performance. The presented method is shown to generate meshes (of up to approximately 1 billion elements) with comparable quality and performance to existing state-of-the-art HPC meshing software.

We present an algebraic algorithm for quantum state tomography that leverages measurements of certain observables to estimate structured entries of the underlying density matrix. Under low-rank assumptions, the remaining entries can be obtained solely using standard numerical linear algebra operations. The proposed algebraic matrix completion framework applies to a broad class of generic, low-rank mixed quantum states and, compared with state-of-the-art methods, is computationally efficient while providing deterministic recovery guarantees.

Post-training compression of Transformer models commonly relies on truncated singular value decomposition (SVD). However, enforcing a single shared subspace can degrade accuracy even at moderate compression. Sparse dictionary learning provides a more flexible union-of-subspaces representation, but existing approaches often suffer from iterative dictionary and coefficient updates. We propose COMPOT (Calibration-Optimized Matrix Procrustes Orthogonalization for Transformers), a training-free compression framework that uses a small calibration dataset to estimate a sparse weight factorization. COMPOT employs orthogonal dictionaries that enable closed-form Procrustes updates for the dictionary and analytical single-step sparse coding for the coefficients, eliminating iterative optimization. To handle heterogeneous layer sensitivity under a global compression budget, COMPOT further introduces a one-shot dynamic allocation strategy that adaptively redistributes layer-wise compression rates. Extensive experiments across diverse architectures and tasks show that COMPOT consistently delivers a superior quality-compression trade-off over strong low-rank and sparse baselines, while remaining fully compatible with post-training quantization for extreme compression. Code is available $\href{https://github.com/mts-ai/COMPOT}{here}$.

Multi-agent systems, where LLM agents communicate through free-form language, enable sophisticated coordination for solving complex cooperative tasks. This surfaces a unique safety problem when individual agents form a coalition and \emph{collude} to pursue secondary goals and degrade the joint objective. In this paper, we present Colosseum, a framework for auditing LLM agents' collusive behavior in multi-agent settings. We ground how agents cooperate through a Distributed Constraint Optimization Problem (DCOP) and measure collusion via regret relative to the cooperative optimum. Colosseum tests each LLM for collusion under different objectives, persuasion tactics, and network topologies. Through our audit, we show that most out-of-the-box models exhibited a propensity to collude when a secret communication channel was artificially formed. Furthermore, we discover ``collusion on paper'' when agents plan to collude in text but would often pick non-collusive actions, thus providing little effect on the joint task. Colosseum provides a new way to study collusion by measuring communications and actions in rich yet verifiable environments.

Tool-calling is essential for Large Language Model (LLM) agents to complete real-world tasks. While most existing benchmarks assume simple, perfectly documented tools, real-world tools (e.g., general "search" APIs) are often opaque, lacking clear best practices or failure modes. Can LLM agents improve their performance in environments with opaque tools by interacting and subsequently improving documentation? To study this, we create OpaqueToolsBench, a benchmark consisting of three distinct task-oriented environments: general function calling, interactive chess playing, and long-trajectory agentic search. Each environment provides underspecified tools that models must learn to use effectively to complete the task. Results on OpaqueToolsBench suggest existing methods for automatically documenting tools are expensive and unreliable when tools are opaque. To address this, we propose a simple framework, ToolObserver, that iteratively refines tool documentation by observing execution feedback from tool-calling trajectories. Our approach outperforms existing methods on OpaqueToolsBench across datasets, even in relatively hard settings. Furthermore, for test-time tool exploration settings, our method is also efficient, consuming 3.5-7.5x fewer total tokens than the best baseline.

LoRA adapters let users fine-tune large language models (LLMs) efficiently. However, LoRA adapters are shared through open repositories like Hugging Face Hub \citep{huggingface_hub_docs}, making them vulnerable to backdoor attacks. Current detection methods require running the model with test input data -- making them impractical for screening thousands of adapters where the trigger for backdoor behavior is unknown. We detect poisoned adapters by analyzing their weight matrices directly, without running the model -- making our method data-agnostic. Our method extracts simple statistics -- how concentrated the singular values are, their entropy, and the distribution shape -- and flags adapters that deviate from normal patterns. We evaluate the method on 500 LoRA adapters -- 400 clean, and 100 poisoned for Llama-3.2-3B on instruction and reasoning datasets: Alpaca, Dolly, GSM8K, ARC-Challenge, SQuADv2, NaturalQuestions, HumanEval, and GLUE dataset. We achieve 97\% detection accuracy with less than 2\% false positives.

In this paper, we present our 3rd place system in the AVerImaTeC shared task, which combines our last year's retrieval-augmented generation (RAG) pipeline with a reverse image search (RIS) module. Despite its simplicity, our system delivers competitive performance with a single multimodal LLM call per fact-check at just $0.013 on average using GPT5.1 via OpenAI Batch API. Our system is also easy to reproduce and tweak, consisting of only three decoupled modules - a textual retrieval module based on similarity search, an image retrieval module based on API-accessed RIS, and a generation module using GPT5.1 - which is why we suggest it as an accesible starting point for further experimentation. We publish its code and prompts, as well as our vector stores and insights into the scheme's running costs and directions for further improvement.

The use of large language models for web information extraction is becoming increasingly fundamental to modern web information retrieval pipelines. However, existing datasets tend to be small, synthetic or text-only, failing to capture the structural context of the web. We introduce ScrapeGraphAI-100k, a large-scale dataset comprising real-world LLM extraction events, collected via opt-in ScrapeGraphAI telemetry during Q2 and Q3 of 2025. Starting from 9M events, we deduplicate and balance by schema to produce 93,695 examples spanning diverse domains and languages. Each instance includes Markdown content, a prompt, a JSON schema, the LLM response, and complexity/validation metadata. We characterize the datasets structural diversity and its failure modes as schema complexity increases. We also provide a fine-tuning experiment showing that a small language model (1.7B) trained on a subset narrows the gap to larger baselines (30B), underscoring the datasets utility for efficient extraction. ScrapeGraphAI-100k enables fine-tuning small models, benchmarking structured extraction, and studying schema induction for web IR indexing, and is publicly available on HuggingFace.

Recent advances in scientific machine learning (SciML) have enabled neural operators (NOs) to serve as powerful surrogates for modeling the dynamic evolution of physical systems governed by partial differential equations (PDEs). While existing approaches focus primarily on learning simulations from the target PDE, they often overlook more fundamental physical principles underlying these equations. Inspired by how numerical solvers are compatible with simulations of different settings of PDEs, we propose a multiphysics training framework that jointly learns from both the original PDEs and their simplified basic forms. Our framework enhances data efficiency, reduces predictive errors, and improves out-of-distribution (OOD) generalization, particularly in scenarios involving shifts of physical parameters and synthetic-to-real transfer. Our method is architecture-agnostic and demonstrates consistent improvements in normalized root mean square error (nRMSE) across a wide range of 1D/2D/3D PDE problems. Through extensive experiments, we show that explicit incorporation of fundamental physics knowledge significantly strengthens the generalization ability of neural operators. We will release models and codes at https://sites.google.com/view/sciml-fundemental-pde.

Vision Language Models (VLMs) are designed to extend Large Language Models (LLMs) with visual capabilities, yet in this work we observe a surprising phenomenon: VLMs can outperform their underlying LLMs on purely text-only tasks, particularly in long-context information retrieval. To investigate this effect, we build a controlled synthetic retrieval task and find that a transformer trained only on text achieves perfect in-distribution accuracy but fails to generalize out of distribution, while subsequent training on an image-tokenized version of the same task nearly doubles text-only OOD performance. Mechanistic interpretability reveals that visual training changes the model's internal binding strategy: text-only training encourages positional shortcuts, whereas image-based training disrupts them through spatial translation invariance, forcing the model to adopt a more robust symbolic binding mechanism that persists even after text-only examples are reintroduced. We further characterize how binding strategies vary across training regimes, visual encoders, and initializations, and show that analogous shifts occur during pretrained LLM-to-VLM transitions. Our findings suggest that cross-modal training can enhance reasoning and generalization even for tasks grounded in a single modality.

Camera virtualization -- an emerging solution to novel view synthesis -- holds transformative potential for visual entertainment, live performances, and sports broadcasting by enabling the generation of photorealistic images from novel viewpoints using images from a limited set of calibrated multiple static physical cameras. Despite recent advances, achieving spatially and temporally coherent and photorealistic rendering of dynamic scenes with efficient time-archival capabilities, particularly in fast-paced sports and stage performances, remains challenging for existing approaches. Recent methods based on 3D Gaussian Splatting (3DGS) for dynamic scenes could offer real-time view-synthesis results. Yet, they are hindered by their dependence on accurate 3D point clouds from the structure-from-motion method and their inability to handle large, non-rigid, rapid motions of different subjects (e.g., flips, jumps, articulations, sudden player-to-player transitions). Moreover, independent motions of multiple subjects can break the Gaussian-tracking assumptions commonly used in 4DGS, ST-GS, and other dynamic splatting variants. This paper advocates reconsidering a neural volume rendering formulation for camera virtualization and efficient time-archival capabilities, making it useful for sports broadcasting and related applications. By modeling a dynamic scene as rigid transformations across multiple synchronized camera views at a given time, our method performs neural representation learning, providing enhanced visual rendering quality at test time. A key contribution of our approach is its support for time-archival, i.e., users can revisit any past temporal instance of a dynamic scene and can perform novel view synthesis, enabling retrospective rendering for replay, analysis, and archival of live events, a functionality absent in existing neural rendering approaches and novel view synthesis...

The use of large language models either as decision support systems, or in agentic workflows, is rapidly transforming the digital ecosystem. However, the understanding of LLM decision-making under uncertainty remains limited. We initiate a comparative study of LLM risky choices along two dimensions: (1) prospect representation (explicit vs. experience based) and (2) decision rationale (explanation). Our study, which involves 20 frontier and open LLMs, is complemented by a matched human subjects experiment, which provides one reference point, while an expected payoff maximizing rational agent model provides another. We find that LLMs cluster into two categories: reasoning models (RMs) and conversational models (CMs). RMs tend towards rational behavior, are insensitive to the order of prospects, gain/loss framing, and explanations, and behave similarly whether prospects are explicit or presented via experience history. CMs are significantly less rational, slightly more human-like, sensitive to prospect ordering, framing, and explanation, and exhibit a large description-history gap. Paired comparisons of open LLMs suggest that a key factor differentiating RMs and CMs is training for mathematical reasoning.

We demonstrate that modern machine-learning methods can autonomously reconstruct several flagship analytic structures in scattering amplitudes directly from numerical on-shell data. In particular, we show that the Kawai--Lewellen--Tye (KLT) relations can be rediscovered using symbolic regression applied to colour-ordered Yang--Mills amplitudes with Mandelstam invariants as input features. Using standard feature-selection techniques, specifically column-pivoted QR factorisation, we simultaneously recover the Kleiss--Kuijf and Bern--Carrasco--Johansson (BCJ) relations, identifying a minimal basis of partial amplitudes without any group-theoretic input. We obtain the tree-level KLT relations with high numerical accuracy up to five external legs, using only minimal theoretical priors, and we comment on the obstacles to generalising the method to higher multiplicity. Our results establish symbolic regression as a practical tool for exploring the analytic structure of the scattering-amplitude landscape, and suggests a general data-driven strategy for uncovering hidden relations in general theories. For comparison, we benchmark this general approach with a recently introduced neural-network based method.

Federated learning (FL) enables distributed model training across edge devices while preserving data locality. This decentralized approach has emerged as a promising solution for collaborative learning on sensitive user data, effectively addressing the longstanding privacy concerns inherent in centralized systems. However, the decentralized nature of FL exposes new security vulnerabilities, especially backdoor attacks that threaten model integrity. To investigate this critical concern, this paper presents the Layer Smoothing Attack (LSA), a novel backdoor attack that exploits layer-specific vulnerabilities in neural networks. First, a Layer Substitution Analysis methodology systematically identifies backdoor-critical (BC) layers that contribute most significantly to backdoor success. Subsequently, LSA strategically manipulates these BC layers to inject persistent backdoors while remaining undetected by state-of-the-art defense mechanisms. Extensive experiments across diverse model architectures and datasets demonstrate that LSA achieves a remarkably backdoor success rate of up to 97% while maintaining high model accuracy on the primary task, consistently bypassing modern FL defenses. These findings uncover fundamental vulnerabilities in current FL security frameworks, demonstrating that future defenses must incorporate layer-aware detection and mitigation strategies.

Learning from electronic health records (EHRs) time series is challenging due to irregular sam- pling, heterogeneous missingness, and the resulting sparsity of observations. Prior self-supervised meth- ods either impute before learning, represent missingness through a dedicated input signal, or optimize solely for imputation, reducing their capacity to efficiently learn representations that support clinical downstream tasks. We propose the Augmented-Intrinsic Dual-Masked Autoencoder (AID-MAE), which learns directly from incomplete time series by applying an intrinsic missing mask to represent naturally missing values and an augmented mask that hides a subset of observed values for reconstruction during training. AID-MAE processes only the unmasked subset of tokens and consistently outperforms strong baselines, including XGBoost and DuETT, across multiple clinical tasks on two datasets. In addition, the learned embeddings naturally stratify patient cohorts in the representation space.

This paper presents a novel approach for ontological heterogeneity that draws heavily from Carnapian-Goguenism, as presented by Kutz, Mossakowski and Lücke (2010). The approach is provisionally designated da Costian-Tarskianism, named after da Costa's Principle of Tolerance in Mathematics and after Alfred Tarski's work on the concept of a consequence operator. The approach is based on the machinery of consequence systems, as developed by Carnielli et al. (2008) and Citkin and Muravitsky (2022), and it introduces the idea of an extended consequence system, which is a consequence system extended with ontological axioms. The paper also defines the concept of an extended development graph, which is a graph structure that allows ontologies to be related via morphisms of extended consequence systems, and additionally via other operations such as fibring and splitting. Finally, we discuss the implications of this approach for the field of applied ontology and suggest directions for future research.

Language models are increasingly used to reason over content they were not trained on, such as new documents, evolving knowledge, and user-specific data. A common approach is retrieval-augmented generation (RAG), which stores verbatim documents externally (as chunks) and retrieves only a relevant subset at inference time for an LLM to reason over. However, this results in inefficient usage of test-time compute (LLM repeatedly reasons over the same documents); moreover, chunk retrieval can inject irrelevant context that increases unsupported generation. We propose a human-like non-parametric continual learning framework, where the base model remains fixed, and learning occurs by integrating each new experience into an external semantic memory state that accumulates and consolidates itself continually. We present Panini, which realizes this by representing documents as Generative Semantic Workspaces (GSW) -- an entity- and event-aware network of question-answer (QA) pairs, sufficient for an LLM to reconstruct the experienced situations and mine latent knowledge via reasoning-grounded inference chains on the network. Given a query, Panini only traverses the continually-updated GSW (not the verbatim documents or chunks), and retrieves the most likely inference chains. Across six QA benchmarks, Panini achieves the highest average performance, 5%-7% higher than other competitive baselines, while using 2-30x fewer answer-context tokens, supports fully open-source pipelines, and reduces unsupported answers on curated unanswerable queries. The results show that efficient and accurate structuring of experiences at write time -- as achieved by the GSW framework -- yields both efficiency and reliability gains at read time. Code is available at https://github.com/roychowdhuryresearch/gsw-memory.

Implicit Neural Representations (INRs) have emerged as promising surrogates for large 3D scientific simulations due to their ability to continuously model spatial and conditional fields, yet they face a critical fidelity-speed dilemma: deep MLPs suffer from high inference cost, while efficient embedding-based models lack sufficient expressiveness. To resolve this, we propose the Decoupled Representation Refinement (DRR) architectural paradigm. DRR leverages a deep refiner network, alongside non-parametric transformations, in a one-time offline process to encode rich representations into a compact and efficient embedding structure. This approach decouples slow neural networks with high representational capacity from the fast inference path. We introduce DRR-Net, a simple network that validates this paradigm, and a novel data augmentation strategy, Variational Pairs (VP) for improving INRs under complex tasks like high-dimensional surrogate modeling. Experiments on several ensemble simulation datasets demonstrate that our approach achieves state-of-the-art fidelity, while being up to 27$\times$ faster at inference than high-fidelity baselines and remaining competitive with the fastest models. The DRR paradigm offers an effective strategy for building powerful and practical neural field surrogates and \rev{INRs in broader applications}, with a minimal compromise between speed and quality.

High-quality video datasets are foundational for training robust models in tasks like action recognition, phase detection, and event segmentation. However, many real-world video datasets suffer from annotation errors such as *mislabeling*, where segments are assigned incorrect class labels, and *disordering*, where the temporal sequence does not follow the correct progression. These errors are particularly harmful in phase-annotated tasks, where temporal consistency is critical. We propose a novel, model-agnostic method for detecting annotation errors by analyzing the Cumulative Sample Loss (CSL)--defined as the average loss a frame incurs when passing through model checkpoints saved across training epochs. This per-frame loss trajectory acts as a dynamic fingerprint of frame-level learnability. Mislabeled or disordered frames tend to show consistently high or irregular loss patterns, as they remain difficult for the model to learn throughout training, while correctly labeled frames typically converge to low loss early. To compute CSL, we train a video segmentation model and store its weights at each epoch. These checkpoints are then used to evaluate the loss of each frame in a test video. Frames with persistently high CSL are flagged as likely candidates for annotation errors, including mislabeling or temporal misalignment. Our method does not require ground truth on annotation errors and is generalizable across datasets. Experiments on EgoPER and Cholec80 demonstrate strong detection performance, effectively identifying subtle inconsistencies such as mislabeling and frame disordering. The proposed approach provides a powerful tool for dataset auditing and improving training reliability in video-based machine learning.

Quantum unitary synthesis addresses the problem of translating abstract quantum algorithms into sequences of hardware-executable quantum gates. Solving this task exactly is infeasible in general due to the exponential growth of the underlying combinatorial search space. Existing approaches suffer from misaligned optimization objectives, substantial training costs and limited generalization across different qubit counts. We mitigate these limitations by using supervised learning to approximate the minimum description length of residual unitaries and combining this estimate with stochastic beam search to identify near optimal gate sequences. Our method relies on a lightweight model with zero-shot generalization, substantially reducing training overhead compared to prior baselines. Across multiple benchmarks, we achieve faster wall-clock synthesis times while exceeding state-of-the-art methods in terms of success rate for complex circuits.

Knowledge distillation is a widely adopted technique for transferring capabilities from LLMs to smaller, more efficient student models. However, unauthorized use of knowledge distillation takes unfair advantage of the considerable effort and cost put into developing frontier models. We investigate methods for modifying teacher-generated reasoning traces to achieve two objectives that deter unauthorized distillation: (1) \emph{anti-distillation}, or degrading the training usefulness of query responses, and (2) \emph{API watermarking}, which embeds verifiable signatures in student models. We introduce several approaches for dynamically rewriting a teacher's reasoning outputs while preserving answer correctness and semantic coherence. Two of these leverage the rewriting capabilities of LLMs, while others use gradient-based techniques. Our experiments show that a simple instruction-based rewriting approach achieves a strong anti-distillation effect while maintaining or even improving teacher performance. Furthermore, we show that our rewriting approach also enables highly reliable watermark detection with essentially no false alarms.

Accurate QA over classical Islamic texts remains challenging due to domain specific semantics, long context dependencies, and concept sensitive reasoning. Therefore, a new CGRA DeBERTa, a concept guided residual domain augmentation transformer framework, is proposed that enhances theological QA over Hadith corpora. The CGRA DeBERTa builds on a customized DeBERTa transformer backbone with lightweight LoRA based adaptations and a residual concept aware gating mechanism. The customized DeBERTa embedding block learns global and positional context, while Concept Guided Residual Blocks incorporate theological priors from a curated Islamic Concept Dictionary of 12 core terms. Moreover, the Concept Gating Mechanism selectively amplifies semantically critical tokens via importance weighted attention, applying differential scaling from 1.04 to 3.00. This design preserves contextual integrity, strengthens domain-specific semantic representations, and enables accurate, efficient span extraction while maintaining computational efficiency. This paper reports the results of training CGRA using a specially constructed dataset of 42591 QA pairs from the text of Sahih alBukhari and Sahih Muslim. While BERT achieved an EM score of 75.87 and DeBERTa one of 89.77, our model scored 97.85 and thus surpassed them by 8.08 on an absolute scale, all while adding approximately 8 inference overhead due to parameter efficient gating. The qualitative evaluation noted better extraction and discrimination and theological precision. This study presents Hadith QA systems that are efficient, interpretable, and accurate and that scale provide educational materials with necessary theological nuance.

The study of histopathological subtypes is valuable for the personalisation of effective treatment strategies for ovarian cancer. However, increasing diagnostic workloads present a challenge for UK pathology departments, leading to the rise in AI approaches. While traditional approaches in this field have relied on pre-computed, frozen image features, recent advances have shifted towards end-to-end feature extraction, providing an improvement in accuracy but at the expense of significantly reduced scalability during training and time-consuming experimentation. In this paper, we propose a new approach for subtype classification and localisation in ovarian cancer histopathology images using contrastive and prototype learning with pre-computed, frozen features via feature-space augmentations. Compared to DSMIL, our method achieves an improvement of 70.4\% and 15.3\% in F1 score for instance- and slide-level classification, respectively, along with AUC gains of 16.9\% for instance localisation and 2.3\% for slide classification, while maintaining the use of frozen patch features.

We theoretically justify the recent empirical finding of [Teh et al., 2025] that a transformer pretrained on synthetically generated data achieves strong performance on empirical Bayes (EB) problems. We take an indirect approach to this question: rather than analyzing the model architecture or training dynamics, we ask why a pretrained Bayes estimator, trained under a prespecified training distribution, can adapt to arbitrary test distributions. Focusing on Poisson EB problems, we identify the existence of universal priors such that training under these priors yields a near-optimal regret bound of $\widetilde{O}(\frac{1}{n})$ uniformly over all test distributions. Our analysis leverages the classical phenomenon of posterior contraction in Bayesian statistics, showing that the pretrained transformer adapts to unknown test distributions precisely through posterior contraction. This perspective also explains the phenomenon of length generalization, in which the test sequence length exceeds the training length, as the model performs Bayesian inference using a generalized posterior.

Neural ordinary differential equations (NODEs) are geometric deep learning models based on dynamical systems and flows generated by vector fields on manifolds. Despite numerous successful applications, particularly within the flow matching paradigm, all existing NODE models are fundamentally constrained to fixed-dimensional dynamics by the intrinsic nature of the manifold's dimension. In this paper, we extend NODEs to M-polyfolds (spaces that can simultaneously accommodate varying dimensions and a notion of differentiability) and introduce PolyNODEs, the first variable-dimensional flow-based model in geometric deep learning. As an example application, we construct explicit M-polyfolds featuring dimensional bottlenecks and PolyNODE autoencoders based on parametrised vector fields that traverse these bottlenecks. We demonstrate experimentally that our PolyNODE models can be trained to solve reconstruction tasks in these spaces, and that latent representations of the input can be extracted and used to solve downstream classification tasks. The code used in our experiments is publicly available at https://github.com/turbotage/PolyNODE .

We introduce ResearchGym, a benchmark and execution environment for evaluating AI agents on end-to-end research. To instantiate this, we repurpose five oral and spotlight papers from ICML, ICLR, and ACL. From each paper's repository, we preserve the datasets, evaluation harness, and baseline implementations but withhold the paper's proposed method. This results in five containerized task environments comprising 39 sub-tasks in total. Within each environment, agents must propose novel hypotheses, run experiments, and attempt to surpass strong human baselines on the paper's metrics. In a controlled evaluation of an agent powered by GPT-5, we observe a sharp capability--reliability gap. The agent improves over the provided baselines from the repository in just 1 of 15 evaluations (6.7%) by 11.5%, and completes only 26.5% of sub-tasks on average. We identify recurring long-horizon failure modes, including impatience, poor time and resource management, overconfidence in weak hypotheses, difficulty coordinating parallel experiments, and hard limits from context length. Yet in a single run, the agent surpasses the solution of an ICML 2025 Spotlight task, indicating that frontier agents can occasionally reach state-of-the-art performance, but do so unreliably. We additionally evaluate proprietary agent scaffolds including Claude Code (Opus-4.5) and Codex (GPT-5.2) which display a similar gap. ResearchGym provides infrastructure for systematic evaluation and analysis of autonomous agents on closed-loop research.

Although learned representations underlie neural networks' success, their fundamental properties remain poorly understood. A striking example is the emergence of simple geometric structures in LLM representations: for example, calendar months organize into a circle, years form a smooth one-dimensional manifold, and cities' latitudes and longitudes can be decoded by a linear probe. We show that the statistics of language exhibit a translation symmetry -- e.g., the co-occurrence probability of two months depends only on the time interval between them -- and we prove that the latter governs the aforementioned geometric structures in high-dimensional word embedding models. Moreover, we find that these structures persist even when the co-occurrence statistics are strongly perturbed (for example, by removing all sentences in which two months appear together) and at moderate embedding dimension. We show that this robustness naturally emerges if the co-occurrence statistics are collectively controlled by an underlying continuous latent variable. We empirically validate this theoretical framework in word embedding models, text embedding models, and large language models.

Large language models (LLMs) are increasingly deployed in privacy-critical and personalization-oriented scenarios, yet the role of context length in shaping privacy leakage and personalization effectiveness remains largely unexplored. We introduce a large-scale benchmark, PAPerBench, to systematically study how increasing context length influences both personalization quality and privacy protection in LLMs. The benchmark comprises approximately 29,000 instances with context lengths ranging from 1K to 256K tokens, yielding a total of 377K evaluation questions. It jointly evaluates personalization performance and privacy risks across diverse scenarios, enabling controlled analysis of long-context model behavior. Extensive evaluations across state-of-the-art LLMs reveal consistent performance degradation in both personalization and privacy as context length increases. We further provide a theoretical analysis of attention dilution under context scaling, explaining this behavior as an inherent limitation of soft attention in fixed-capacity Transformers. The empirical and theoretical findings together suggest a general scaling gap in current models -- long context, less focus. We release the benchmark to support reproducible evaluation and future research on scalable privacy and personalization. Code and data are available at https://github.com/SafeRL-Lab/PAPerBench

Many generative tasks in chemistry and science involve distributions invariant to group symmetries (e.g., permutation and rotation). A common strategy enforces invariance and equivariance through architectural constraints such as equivariant denoisers and invariant priors. In this paper, we challenge this tradition through the alternative canonicalization perspective: first map each sample to an orbit representative with a canonical pose or order, train an unconstrained (non-equivariant) diffusion or flow model on the canonical slice, and finally recover the invariant distribution by sampling a random symmetry transform at generation time. Building on a formal quotient-space perspective, our work provides a comprehensive theory of canonical diffusion by proving: (i) the correctness, universality and superior expressivity of canonical generative models over invariant targets; (ii) canonicalization accelerates training by removing diffusion score complexity induced by group mixtures and reducing conditional variance in flow matching. We then show that aligned priors and optimal transport act complementarily with canonicalization and further improves training efficiency. We instantiate the framework for molecular graph generation under $S_n \times SE(3)$ symmetries. By leveraging geometric spectra-based canonicalization and mild positional encodings, canonical diffusion significantly outperforms equivariant baselines in 3D molecule generation tasks, with similar or even less computation. Moreover, with a novel architecture Canon, CanonFlow achieves state-of-the-art performance on the challenging GEOM-DRUG dataset, and the advantage remains large in few-step generation.

Bio-pharmaceutical innovation has shifted: many new drug assets now originate outside the United States and are disclosed primarily via regional, non-English channels. Recent data suggests that over 85% of patent filings originate outside the U.S., with China accounting for nearly half of the global total. A growing share of scholarly output is also non-U.S. Industry estimates put China at 30% of global drug development, spanning 1,200+ novel candidates. In this high-stakes environment, failing to surface "under-the-radar" assets creates multi-billion-dollar risk for investors and business development teams, making asset scouting a coverage-critical competition where speed and completeness drive value. Yet today's Deep Research AI agents still lag human experts in achieving high recall discovery across heterogeneous, multilingual sources without hallucination. We propose a benchmarking methodology for drug asset scouting and a tuned, tree-based self-learning Bioptic Agent aimed at complete, non-hallucinated scouting. We construct a challenging completeness benchmark using a multilingual multi-agent pipeline: complex user queries paired with ground-truth assets that are largely outside U.S.-centric radar. To reflect real-deal complexity, we collected screening queries from expert investors, BD, and VC professionals and used them as priors to conditionally generate benchmark queries. For grading, we use LLM-as-judge evaluation calibrated to expert opinions. On this benchmark, our Bioptic Agent achieves 79.7% F1 score, outperforming Claude Opus 4.6 (56.2%), Gemini 3 Pro + Deep Research (50.6%), OpenAI GPT-5.2 Pro (46.6%), Perplexity Deep Research (44.2%), and Exa Websets (26.9%). Performance improves steeply with additional compute, supporting the view that more compute yields better results.

Diffusion language models are a promising alternative to autoregressive models due to their potential for faster generation. Among discrete diffusion approaches, Masked diffusion currently dominates, largely driven by strong perplexity on language modeling benchmarks. In this work, we present the first scaling law study of uniform-state and interpolating discrete diffusion methods. We also show that Masked diffusion models can be made approximately 12% more FLOPs-efficient when trained with a simple cross-entropy objective. We find that perplexity is informative within a diffusion family but can be misleading across families, where models with worse likelihood scaling may be preferable due to faster and more practical sampling, as reflected by the speed-quality Pareto frontier. These results challenge the view that Masked diffusion is categorically the future of diffusion language modeling and that perplexity alone suffices for cross-algorithm comparison. Scaling all methods to 1.7B parameters, we show that uniform-state diffusion remains competitive on likelihood-based benchmarks and outperforms autoregressive and Masked diffusion models on GSM8K, despite worse validation perplexity. We provide the code, model checkpoints, and video tutorials on the project page: http://s-sahoo.github.io/scaling-dllms

This paper proposes a novel method for Text Style Transfer (TST) based on parameter-efficient fine-tuning of Large Language Models (LLMs). Addressing the scarcity of parallel corpora that map between styles, the study employs roundtrip translation to synthesize such parallel datasets from monolingual corpora. This approach creates 'neutralized' text devoid of stylistic attributes, essentially creating a shared input style at training-time and inference-time. Experimental results demonstrate consistent superiority of this method over zero-shot prompting and fewshot ICL techniques measured by BLEU scores and style accuracy scores across four investigated domains. Furthermore, the integration of retrieval-augmented generation (RAG) for terminology and name knowledge enhances robustness and stylistic consistency.

Cold-start personalization requires inferring user preferences through interaction when no user-specific historical data is available. The core challenge is a routing problem: each task admits dozens of preference dimensions, yet individual users care about only a few, and which ones matter depends on who is asking. With a limited question budget, asking without structure will miss the dimensions that matter. Reinforcement learning is the natural formulation, but in multi-turn settings its terminal reward fails to exploit the factored, per-criterion structure of preference data, and in practice learned policies collapse to static question sequences that ignore user responses. We propose decomposing cold-start elicitation into offline structure learning and online Bayesian inference. Pep (Preference Elicitation with Priors) learns a structured world model of preference correlations offline from complete profiles, then performs training-free Bayesian inference online to select informative questions and predict complete preference profiles, including dimensions never asked about. The framework is modular across downstream solvers and requires only simple belief models. Across medical, mathematical, social, and commonsense reasoning, Pep achieves 80.8% alignment between generated responses and users' stated preferences versus 68.5% for RL, with 3-5x fewer interactions. When two users give different answers to the same question, Pep changes its follow-up 39-62% of the time versus 0-28% for RL. It does so with ~10K parameters versus 8B for RL, showing that the bottleneck in cold-start elicitation is the capability to exploit the factored structure of preference data.

Many robot tasks require attending to the history of past observations. For example, finding an item in a room requires remembering which places have already been searched. However, the best-performing robot policies typically condition only on the current observation, limiting their applicability to such tasks. Naively conditioning on past observations often fails due to spurious correlations: policies latch onto incidental features of training histories that do not generalize to out-of-distribution trajectories upon deployment. We analyze why policies latch onto these spurious correlations and find that this problem stems from limited coverage over the space of possible histories during training, which grows exponentially with horizon. Existing regularization techniques provide inconsistent benefits across tasks, as they do not fundamentally address this coverage problem. Motivated by these findings, we propose Big Picture Policies (BPP), an approach that conditions on a minimal set of meaningful keyframes detected by a vision-language model. By projecting diverse rollouts onto a compact set of task-relevant events, BPP substantially reduces distribution shift between training and deployment, without sacrificing expressivity. We evaluate BPP on four challenging real-world manipulation tasks and three simulation tasks, all requiring history conditioning. BPP achieves 70% higher success rates than the best comparison on real-world evaluations.

Diffusion models over discrete spaces have recently shown striking empirical success, yet their theoretical foundations remain incomplete. In this paper, we study the sampling efficiency of score-based discrete diffusion models under a continuous-time Markov chain (CTMC) formulation, with a focus on $τ$-leaping-based samplers. We establish sharp convergence guarantees for attaining $\varepsilon$ accuracy in Kullback-Leibler (KL) divergence for both uniform and masking noising processes. For uniform discrete diffusion, we show that the $τ$-leaping algorithm achieves an iteration complexity of order $\tilde O(d/\varepsilon)$, with $d$ the ambient dimension of the target distribution, eliminating linear dependence on the vocabulary size $S$ and improving existing bounds by a factor of $d$; moreover, we establish a matching algorithmic lower bound showing that linear dependence on the ambient dimension is unavoidable in general. For masking discrete diffusion, we introduce a modified $τ$-leaping sampler whose convergence rate is governed by an intrinsic information-theoretic quantity, termed the effective total correlation, which is bounded by $d \log S$ but can be sublinear or even constant for structured data. As a consequence, the sampler provably adapts to low-dimensional structure without prior knowledge or algorithmic modification, yielding sublinear convergence rates for various practical examples (such as hidden Markov models, image data, and random graphs). Our analysis requires no boundedness or smoothness assumptions on the score estimator beyond control of the score entropy loss.

Gaussian Processes (GPs) are a powerful tool for probabilistic modeling, but their performance is often constrained in complex, largescale real-world domains due to the limited expressivity of classical kernels. Quantum computing offers the potential to overcome this limitation by embedding data into exponentially large Hilbert spaces, capturing complex correlations that remain inaccessible to classical computing approaches. In this paper, we propose a Distributed Quantum Gaussian Process (DQGP) method in a multiagent setting to enhance modeling capabilities and scalability. To address the challenging non-Euclidean optimization problem, we develop a Distributed consensus Riemannian Alternating Direction Method of Multipliers (DR-ADMM) algorithm that aggregates local agent models into a global model. We evaluate the efficacy of our method through numerical experiments conducted on a quantum simulator in classical hardware. We use real-world, non-stationary elevation datasets of NASA's Shuttle Radar Topography Mission and synthetic datasets generated by Quantum Gaussian Processes. Beyond modeling advantages, our framework highlights potential computational speedups that quantum hardware may provide, particularly in Gaussian processes and distributed optimization.

News recommendation plays a critical role in online news platforms by helping users discover relevant content. Cross-domain news recommendation further requires inferring user's underlying information needs from heterogeneous signals that often extend beyond direct news consumption. A key challenge lies in moving beyond surface-level behaviors to capture deeper, reusable user interests while maintaining scalability in large-scale production systems. In this paper, we present a reinforcement learning framework that trains large language models to generate high-quality lists of interest-driven news search queries from cross-domain user signals. We formulate query-list generation as a policy optimization problem and employ GRPO with multiple reward signals. We systematically study two compute dimensions: inference-time sampling and model capacity, and empirically observe consistent improvements with increased compute that exhibit scaling-like behavior. Finally, we perform on-policy distillation to transfer the learned policy from a large, compute-intensive teacher to a compact student model suitable for scalable deployment. Extensive offline experiments, ablation studies and large-scale online A/B tests in a production news recommendation system demonstrate consistent gains in both interest modeling quality and downstream recommendation performance.

We show that AI foundation models that are pretrained on numerical solutions to a diverse corpus of partial differential equations can be adapted and fine-tuned to obtain skillful predictive weather emulators for the Martian atmosphere. We base our work on the Poseidon PDE foundation model for two-dimensional systems. We develop a method to extend Poseidon from two to three dimensions while keeping the pretraining information. Moreover, we investigate the performance of the model in the presence of sparse initial conditions. Our results make use of four Martian years (approx.~34 GB) of training data and a median compute budget of 13 GPU hours. We find that the combination of pretraining and model extension yields a performance increase of 34.4\% on a held-out year. This shows that PDEs-FMs can not only approximate solutions to (other) PDEs but also anchor models for real-world problems with complex interactions that lack a sufficient amount of training data or a suitable compute budget.

Frontier LLMs are safeguarded against attempts to extract harmful information via adversarial prompts known as "jailbreaks". Recently, defenders have developed classifier-based systems that have survived thousands of hours of human red teaming. We introduce Boundary Point Jailbreaking (BPJ), a new class of automated jailbreak attacks that evade the strongest industry-deployed safeguards. Unlike previous attacks that rely on white/grey-box assumptions (such as classifier scores or gradients) or libraries of existing jailbreaks, BPJ is fully black-box and uses only a single bit of information per query: whether or not the classifier flags the interaction. To achieve this, BPJ addresses the core difficulty in optimising attacks against robust real-world defences: evaluating whether a proposed modification to an attack is an improvement. Instead of directly trying to learn an attack for a target harmful string, BPJ converts the string into a curriculum of intermediate attack targets and then actively selects evaluation points that best detect small changes in attack strength ("boundary points"). We believe BPJ is the first fully automated attack algorithm that succeeds in developing universal jailbreaks against Constitutional Classifiers, as well as the first automated attack algorithm that succeeds against GPT-5's input classifier without relying on human attack seeds. BPJ is difficult to defend against in individual interactions but incurs many flags during optimisation, suggesting that effective defence requires supplementing single-interaction methods with batch-level monitoring.

Geometric deep learning (GDL) deals with supervised learning on data domains that go beyond Euclidean structure, such as data with graph or manifold structure. Due to the demand that arises from application-related data, there is a need to identify further topological and geometric structures with which these use cases can be made accessible to machine learning. There are various techniques, such as spectral convolution, that form the basic building blocks for some convolutional neural network-like architectures on non-Euclidean data. In this paper, the concept of spectral convolution on orbifolds is introduced. This provides a building block for making learning on orbifold structured data accessible using GDL. The theory discussed is illustrated using an example from music theory.

Reasoning about actual causes of observed effects is fundamental to the study of rationality. This important problem has been studied since the time of Aristotle, with formal mathematical accounts emerging recently. We live in a world where change due to actions can be both discrete and continuous, that is, hybrid. Yet, despite extensive research on actual causation, only few recent studies looked into causation with continuous change. Building on recent progress, in this paper we propose two definitions of primary cause in a hybrid action-theoretic framework, namely the hybrid temporal situation calculus. One of these is foundational in nature while the other formalizes causation through contributions, which can then be verified from a counterfactual perspective using a modified ``but-for'' test. We prove that these two definitions are indeed equivalent. We then show that our definitions of causation have some intuitively justifiable properties.

Vision language models (VLMs) achieve strong performance on RGB imagery, but they do not generalize to thermal images. Thermal sensing plays a critical role in settings where visible light fails, including nighttime surveillance, search and rescue, autonomous driving, and medical screening. Unlike RGB imagery, thermal images encode physical temperature rather than color or texture, requiring perceptual and reasoning capabilities that existing RGB-centric benchmarks do not evaluate. We introduce ThermEval-B, a structured benchmark of approximately 55,000 thermal visual question answering pairs designed to assess the foundational primitives required for thermal vision language understanding. ThermEval-B integrates public datasets with our newly collected ThermEval-D, the first dataset to provide dense per-pixel temperature maps with semantic body-part annotations across diverse indoor and outdoor environments. Evaluating 25 open-source and closed-source VLMs, we find that models consistently fail at temperature-grounded reasoning, degrade under colormap transformations, and default to language priors or fixed responses, with only marginal gains from prompting or supervised fine-tuning. These results demonstrate that thermal understanding requires dedicated evaluation beyond RGB-centric assumptions, positioning ThermEval as a benchmark to drive progress in thermal vision language modeling.

Multimodal learning seeks to integrate information from heterogeneous sources, where signals may be shared across modalities, specific to individual modalities, or emerge only through their interaction. While self-supervised multimodal contrastive learning has achieved remarkable progress, most existing methods predominantly capture redundant cross-modal signals, often neglecting modality-specific (unique) and interaction-driven (synergistic) information. Recent extensions broaden this perspective, yet they either fail to explicitly model synergistic interactions or learn different information components in an entangled manner, leading to incomplete representations and potential information leakage. We introduce \textbf{COrAL}, a principled framework that explicitly and simultaneously preserves redundant, unique, and synergistic information within multimodal representations. COrAL employs a dual-path architecture with orthogonality constraints to disentangle shared and modality-specific features, ensuring a clean separation of information components. To promote synergy modeling, we introduce asymmetric masking with complementary view-specific patterns, compelling the model to infer cross-modal dependencies rather than rely solely on redundant cues. Extensive experiments on synthetic benchmarks and diverse MultiBench datasets demonstrate that COrAL consistently matches or outperforms state-of-the-art methods while exhibiting low performance variance across runs. These results indicate that explicitly modeling the full spectrum of multimodal information yields more stable, reliable, and comprehensive embeddings.

Macrocycles are ring-shaped molecules that offer a promising alternative to small-molecule drugs due to their enhanced selectivity and binding affinity against difficult targets. Despite their chemical value, they remain underexplored in generative modeling, likely owing to their scarcity in public datasets and the challenges of enforcing topological constraints in standard deep generative models. We introduce MacroGuide: Topological Guidance for Macrocycle Generation, a diffusion guidance mechanism that uses Persistent Homology to steer the sampling of pretrained molecular generative models toward the generation of macrocycles, in both unconditional and conditional (protein pocket) settings. At each denoising step, MacroGuide constructs a Vietoris-Rips complex from atomic positions and promotes ring formation by optimizing persistent homology features. Empirically, applying MacroGuide to pretrained diffusion models increases macrocycle generation rates from 1% to 99%, while matching or exceeding state-of-the-art performance on key quality metrics such as chemical validity, diversity, and PoseBusters checks.

Following our previous work (J. Phys. Chem. Lett., 2026, 17, 5, 1288-1295), we propose the DMTS-NC approach, a distilled multi-time-step (DMTS) strategy using non conservative (NC) forces to further accelerate atomistic molecular dynamics simulations using foundation neural network models. There, a dual-level reversible reference system propagator algorithm (RESPA) formalism couples a target accurate conservative potential to a simplified distilled representation optimized for the production of non-conservative forces. Despite being non-conservative, the distilled architecture is designed to enforce key physical priors, such as equivariance under rotation and cancellation of atomic force components. These choices facilitate the distillation process and therefore improve drastically the robustness of simulation, significantly limiting the "holes" in the simpler potential, thus achieving excellent agreement with the forces data. Overall, the DMTS-NC scheme is found to be more stable and efficient than its conservative counterpart with additional speedups reaching 15-30% over DMTS. Requiring no finetuning steps, it is easier to implement and can be pushed to the limit of the systems physical resonances to maintain accuracy while providing maximum efficiency. As for DMTS, DMTS-NC is applicable to any neural network potential.

Estimating causal quantities traditionally relies on bespoke estimators tailored to specific assumptions. Recently proposed Causal Foundation Models (CFMs) promise a more unified approach by amortising causal discovery and inference in a single step. However, in their current state, they do not allow for the incorporation of any domain knowledge, which can lead to suboptimal predictions. We bridge this gap by introducing methods to condition CFMs on causal information, such as the causal graph or more readily available ancestral information. When access to complete causal graph information is too strict a requirement, our approach also effectively leverages partial causal information. We systematically evaluate conditioning strategies and find that injecting learnable biases into the attention mechanism is the most effective method to utilise full and partial causal information. Our experiments show that this conditioning allows a general-purpose CFM to match the performance of specialised models trained on specific causal structures. Overall, our approach addresses a central hurdle on the path towards all-in-one causal foundation models: the capability to answer causal queries in a data-driven manner while effectively leveraging any amount of domain expertise.

Large Language Models (LLMs) are increasingly deployed in contact-center Quality Assurance (QA) to automate agent performance evaluation and coaching feedback. While LLMs offer unprecedented scalability and speed, their reliance on web-scale training data raises concerns regarding demographic and behavioral biases that may distort workforce assessment. We present a counterfactual fairness evaluation of LLM-based QA systems across 13 dimensions spanning three categories: Identity, Context, and Behavioral Style. Fairness is quantified using the Counterfactual Flip Rate (CFR), the frequency of binary judgment reversals, and the Mean Absolute Score Difference (MASD), the average shift in coaching or confidence scores across counterfactual pairs. Evaluating 18 LLMs on 3,000 real-world contact center transcripts, we find systematic disparities, with CFR ranging from 5.4% to 13.0% and consistent MASD shifts across confidence, positive, and improvement scores. Larger, more strongly aligned models show lower unfairness, though fairness does not track accuracy. Contextual priming of historical performance induces the most severe degradations (CFR up to 16.4%), while implicit linguistic identity cues remain a persistent bias source. Finally, we analyze the efficacy of fairness-aware prompting, finding that explicit instructions yield only modest improvements in evaluative consistency. Our findings underscore the need for standardized fairness auditing pipelines prior to deploying LLMs in high-stakes workforce evaluation.

Automatically generating interactive 3D environments is crucial for scaling up robotic data collection in simulation. While prior work has primarily focused on 3D asset placement, it often overlooks the physical relationships between objects (e.g., contact, support, balance, and containment), which are essential for creating complex and realistic manipulation scenarios such as tabletop arrangements, shelf organization, or box packing. Compared to classical 3D layout generation, producing complex physical scenes introduces additional challenges: (a) higher object density and complexity (e.g., a small shelf may hold dozens of books), (b) richer supporting relationships and compact spatial layouts, and (c) the need to accurately model both spatial placement and physical properties. To address these challenges, we propose PhyScensis, an LLM agent-based framework powered by a physics engine, to produce physically plausible scene configurations with high complexity. Specifically, our framework consists of three main components: an LLM agent iteratively proposes assets with spatial and physical predicates; a solver, equipped with a physics engine, realizes these predicates into a 3D scene; and feedback from the solver informs the agent to refine and enrich the configuration. Moreover, our framework preserves strong controllability over fine-grained textual descriptions and numerical parameters (e.g., relative positions, scene stability), enabled through probabilistic programming for stability and a complementary heuristic that jointly regulates stability and spatial relations. Experimental results show that our method outperforms prior approaches in scene complexity, visual quality, and physical accuracy, offering a unified pipeline for generating complex physical scene layouts for robotic manipulation.

We present a domain-grounded framework and benchmark for tool-aware plan generation in contact centers, where answering a query for business insights, our target use case, requires decomposing it into executable steps over structured tools (Text2SQL (T2S)/Snowflake) and unstructured tools (RAG/transcripts) with explicit depends_on for parallelism. Our contributions are threefold: (i) a reference-based plan evaluation framework operating in two modes - a metric-wise evaluator spanning seven dimensions (e.g., tool-prompt alignment, query adherence) and a one-shot evaluator; (ii) a data curation methodology that iteratively refines plans via an evaluator->optimizer loop to produce high-quality plan lineages (ordered plan revisions) while reducing manual effort; and (iii) a large-scale study of 14 LLMs across sizes and families for their ability to decompose queries into step-by-step, executable, and tool-assigned plans, evaluated under prompts with and without lineage. Empirically, LLMs struggle on compound queries and on plans exceeding 4 steps (typically 5-15); the best total metric score reaches 84.8% (Claude-3-7-Sonnet), while the strongest one-shot match rate at the "A+" tier (Extremely Good, Very Good) is only 49.75% (o3-mini). Plan lineage yields mixed gains overall but benefits several top models and improves step executability for many. Our results highlight persistent gaps in tool-understanding, especially in tool-prompt alignment and tool-usage completeness, and show that shorter, simpler plans are markedly easier. The framework and findings provide a reproducible path for assessing and improving agentic planning with tools for answering data-analysis queries in contact-center settings.

We consider the general problem of learning a predictor that satisfies multiple objectives of interest simultaneously, a broad framework that captures a range of specific learning goals including calibration, regret, and multiaccuracy. We work in an online setting where the data distribution can change arbitrarily over time. Existing approaches to this problem aim to minimize the set of objectives over the entire time horizon in a worst-case sense, and in practice they do not necessarily adapt to distribution shifts. Earlier work has aimed to alleviate this problem by incorporating additional objectives that target local guarantees over contiguous subintervals. Empirical evaluation of these proposals is, however, scarce. In this article, we consider an alternative procedure that achieves local adaptivity by replacing one part of the multi-objective learning method with an adaptive online algorithm. Empirical evaluations on datasets from energy forecasting and algorithmic fairness show that our proposed method improves upon existing approaches and achieves unbiased predictions over subgroups, while remaining robust under distribution shift.

This paper presents a physics-informed neural network approach for dynamic modeling of saturable synchronous machines, including cases with spatial harmonics. We introduce an architecture that incorporates gradient networks directly into the fundamental machine equations, enabling accurate modeling of the nonlinear and coupled electromagnetic constitutive relationship. By learning the gradient of the magnetic field energy, the model inherently satisfies energy balance (reciprocity conditions). The proposed architecture can universally approximate any physically feasible magnetic behavior and offers several advantages over lookup tables and standard machine learning models: it requires less training data, ensures monotonicity and reliable extrapolation, and produces smooth outputs. These properties further enable robust model inversion and optimal trajectory generation, often needed in control applications. We validate the proposed approach using measured and finite-element method (FEM) datasets from a 5.6-kW permanent-magnet (PM) synchronous reluctance machine. Results demonstrate accurate and physically consistent models, even with limited training data.

Mixture-of-Experts (MoE) architectures decompose prediction tasks into specialized expert sub-networks selected by a gating mechanism. This letter adopts a communication-theoretic view of MoE gating, modeling the gate as a stochastic channel operating under a finite information rate. Within an information-theoretic learning framework, we specialize a mutual-information generalization bound and develop a rate-distortion characterization $D(R_g)$ of finite-rate gating, where $R_g:=I(X; T)$, yielding (under a standard empirical rate-distortion optimality condition) $\mathbb{E}[R(W)] \le D(R_g)+δ_m+\sqrt{(2/m)\, I(S; W)}$. The analysis yields capacity-aware limits for communication-constrained MoE systems, and numerical simulations on synthetic multi-expert models empirically confirm the predicted trade-offs between gating rate, expressivity, and generalization.

Maintaining spatial world consistency over long horizons remains a central challenge for camera-controllable video generation. Existing memory-based approaches often condition generation on globally reconstructed 3D scenes by rendering anchor videos from the reconstructed geometry in the history. However, reconstructing a global 3D scene from multiple views inevitably introduces cross-view misalignment, as pose and depth estimation errors cause the same surfaces to be reconstructed at slightly different 3D locations across views. When fused, these inconsistencies accumulate into noisy geometry that contaminates the conditioning signals and degrades generation quality. We introduce AnchorWeave, a memory-augmented video generation framework that replaces a single misaligned global memory with multiple clean local geometric memories and learns to reconcile their cross-view inconsistencies. To this end, AnchorWeave performs coverage-driven local memory retrieval aligned with the target trajectory and integrates the selected local memories through a multi-anchor weaving controller during generation. Extensive experiments demonstrate that AnchorWeave significantly improves long-term scene consistency while maintaining strong visual quality, with ablation and analysis studies further validating the effectiveness of local geometric conditioning, multi-anchor control, and coverage-driven retrieval.

In recent times, there has been considerable interest in fault detection within electrical power systems, garnering attention from both academic researchers and industry professionals. Despite the development of numerous fault detection methods and their adaptations over the past decade, their practical application remains highly challenging. Given the probabilistic nature of fault occurrences and parameters, certain decision-making tasks could be approached from a probabilistic standpoint. Protective systems are tasked with the detection, classification, and localization of faulty voltage and current line magnitudes, culminating in the activation of circuit breakers to isolate the faulty line. An essential aspect of designing effective fault detection systems lies in obtaining reliable data for training and testing, which is often scarce. Leveraging deep learning techniques, particularly the powerful capabilities of pattern classifiers in learning, generalizing, and parallel processing, offers promising avenues for intelligent fault detection. To address this, our paper proposes an anomaly-based approach for fault detection in electrical power systems, employing deep autoencoders. Additionally, we utilize Convolutional Autoencoders (CAE) for dimensionality reduction, which, due to its fewer parameters, requires less training time compared to conventional autoencoders. The proposed method demonstrates superior performance and accuracy compared to alternative detection approaches by achieving an accuracy of 97.62% and 99.92% on simulated and publicly available datasets.

In machine unlearning, $(\varepsilon,δ)-$unlearning is a popular framework that provides formal guarantees on the effectiveness of the removal of a subset of training data, the forget set, from a trained model. For strongly convex objectives, existing first-order methods achieve $(\varepsilon,δ)-$unlearning, but they only use the forget set to calibrate injected noise, never as a direct optimization signal. In contrast, efficient empirical heuristics often exploit the forget samples (e.g., via gradient ascent) but come with no formal unlearning guarantees. We bridge this gap by presenting the Variance-Reduced Unlearning (VRU) algorithm. To the best of our knowledge, VRU is the first first-order algorithm that directly includes forget set gradients in its update rule, while provably satisfying ($(\varepsilon,δ)-$unlearning. We establish the convergence of VRU and show that incorporating the forget set yields strictly improved rates, i.e. a better dependence on the achieved error compared to existing first-order $(\varepsilon,δ)-$unlearning methods. Moreover, we prove that, in a low-error regime, VRU asymptotically outperforms any first-order method that ignores the forget set.Experiments corroborate our theory, showing consistent gains over both state-of-the-art certified unlearning methods and over empirical baselines that explicitly leverage the forget set.

Reliable uncertainty estimates are crucial for deploying pretrained models; yet, many strong methods for quantifying uncertainty require retraining, Monte Carlo sampling, or expensive second-order computations and may alter a frozen backbone's predictions. To address this, we introduce Gaussian Process Activations (GAPA), a post-hoc method that shifts Bayesian modeling from weights to activations. GAPA replaces standard nonlinearities with Gaussian-process activations whose posterior mean exactly matches the original activation, preserving the backbone's point predictions by construction while providing closed-form epistemic variances in activation space. To scale to modern architectures, we use a sparse variational inducing-point approximation over cached training activations, combined with local k-nearest-neighbor subset conditioning, enabling deterministic single-pass uncertainty propagation without sampling, backpropagation, or second-order information. Across regression, classification, image segmentation, and language modeling, GAPA matches or outperforms strong post-hoc baselines in calibration and out-of-distribution detection while remaining efficient at test time.

We extend the Prometheus framework for unsupervised phase transition discovery from 2D classical systems to 3D classical and quantum many-body systems, addressing scalability in higher dimensions and generalization to quantum fluctuations. For the 3D Ising model ($L \leq 32$), the framework detects the critical temperature within 0.01\% of literature values ($T_c/J = 4.511 \pm 0.005$) and extracts critical exponents with $\geq 70\%$ accuracy ($β= 0.328 \pm 0.015$, $γ= 1.24 \pm 0.06$, $ν= 0.632 \pm 0.025$), correctly identifying the 3D Ising universality class via $χ^2$ comparison ($p = 0.72$) without analytical guidance. For quantum systems, we developed quantum-aware VAE (Q-VAE) architectures using complex-valued wavefunctions and fidelity-based loss. Applied to the transverse field Ising model, we achieve 2\% accuracy in quantum critical point detection ($h_c/J = 1.00 \pm 0.02$) and successfully discover ground state magnetization as the order parameter ($r = 0.97$). Notably, for the disordered transverse field Ising model, we detect exotic infinite-randomness criticality characterized by activated dynamical scaling $\ln ξ\sim |h - h_c|^{-ψ}$, extracting a tunneling exponent $ψ= 0.48 \pm 0.08$ consistent with theoretical predictions ($ψ= 0.5$). This demonstrates that unsupervised learning can identify qualitatively different types of critical behavior, not just locate critical points. Our systematic validation across classical thermal transitions ($T = 0$ to $T > 0$) and quantum phase transitions ($T = 0$, varying $h$) establishes that VAE-based discovery generalizes across fundamentally different physical domains, providing robust tools for exploring phase diagrams where analytical solutions are unavailable.

To address the global health threat of antimicrobial resistance, antimicrobial peptides (AMP) are being explored for their potent and promising ability to fight resistant pathogens. While artificial intelligence (AI) is being employed to advance AMP discovery and design, most AMP design models struggle to balance key goals like activity, toxicity, and novelty, using rigid or unclear scoring methods that make results hard to interpret and optimize. As the capabilities of Large Language Models (LLM) advance and evolve swiftly, we turn to AI multi-agent collaboration based on such models (multi-agent LLMs), which show rapidly rising potential in complex scientific design scenarios. Based on this, we introduce MAC-AMP, a closed-loop multi-agent collaboration (MAC) system for multi-objective AMP design. The system implements a fully autonomous simulated peer review-adaptive reinforcement learning framework that requires only a task description and example dataset to design novel AMPs. The novelty of our work lies in introducing a closed-loop multi-agent system for AMP design, with cross-domain transferability, that supports multi-objective optimization while remaining explainable rather than a 'black box'. Experiments show that MAC-AMP outperforms other AMP generative models by effectively optimizing AMP generation for multiple key molecular properties, demonstrating exceptional results in antibacterial activity, AMP likeliness, toxicity compliance, and structural reliability.

To address the ``reusability dilemma'' and structural hallucinations in enterprise Agentic AI,this paper proposes ReusStdFlow, a framework centered on a novel ``Extraction-Storage-Construction'' paradigm. The framework deconstructs heterogeneous, platform-specific Domain Specific Languages (DSLs) into standardized, modular workflow segments. It employs a dual knowledge architecture-integrating graph and vector databases-to facilitate synergistic retrieval of both topological structures and functional semantics. Finally, workflows are intelligently assembled using a retrieval-augmented generation (RAG) strategy. Tested on 200 real-world n8n workflows, the system achieves over 90% accuracy in both extraction and construction. This framework provides a standardized solution for the automated reorganization and efficient reuse of enterprise digital assets.

Hypergraph Neural Networks (HyGNNs) have demonstrated remarkable success in modeling higher-order relationships among entities. However, their performance often degrades on heterophilic hypergraphs, where nodes connected by the same hyperedge tend to have dissimilar semantic representations or belong to different classes. While several HyGNNs, including our prior work BHyGNN, have been proposed to address heterophily, their reliance on labeled data significantly limits their applicability in real-world scenarios where annotations are scarce or costly. To overcome this limitation, we introduce BHyGNN+, a self-supervised learning framework that extends BHyGNN for representation learning on heterophilic hypergraphs without requiring ground-truth labels. The core idea of BHyGNN+ is hypergraph duality, a structural transformation where the roles of nodes and hyperedges are interchanged. By contrasting augmented views of a hypergraph against its dual using cosine similarity, our framework captures essential structural patterns in a fully unsupervised manner. Notably, this duality-based formulation eliminates the need for negative samples, a common requirement in existing hypergraph contrastive learning methods that is often difficult to satisfy in practice. Extensive experiments on eleven benchmark datasets demonstrate that BHyGNN+ consistently outperforms state-of-the-art supervised and self-supervised baselines on both heterophilic and homophilic hypergraphs. Our results validate the effectiveness of leveraging hypergraph duality for self-supervised learning and establish a new paradigm for representation learning on challenging, unlabeled hypergraphs.

Large Reasoning Models (LRMs) such as OpenAI o1 and DeepSeek-R1 have shown excellent performance in reasoning tasks using long reasoning chains. However, this has also led to a significant increase of computational costs and the generation of verbose output, a phenomenon known as overthinking. The tendency to overthinking is often exacerbated by Reinforcement Learning (RL) algorithms such as GRPO/DAPO. In this paper, we propose BFS-PO, an RL algorithm which alleviates this problem using a Best-First Search exploration strategy. Specifically, BFS-PO looks for the shortest correct answer using a backtracking mechanism based on maximum entropy nodes. By generating progressively shorter responses during training, BFS-PO learns to produce concise reasoning chains. Using different benchmarks and base LRMs, we show that BFS-PO can simultaneously increase the LRM accuracy and shorten its answers.

Off-policy evaluation (OPE) is essential for assessing ranking and recommendation systems without costly online interventions. Self-Normalised Inverse Propensity Scoring (SNIPS) is a standard tool for variance reduction in OPE, leveraging a multiplicative control variate. Recent advances in off-policy learning suggest that additive control variates (baseline corrections) may offer superior performance, yet theoretical guarantees for evaluation are lacking. This paper provides a definitive answer: we prove that $β^\star$-IPS, an estimator with an optimal additive baseline, asymptotically dominates SNIPS in Mean Squared Error. By analytically decomposing the variance gap, we show that SNIPS is asymptotically equivalent to using a specific -- but generally sub-optimal -- additive baseline. Our results theoretically justify shifting from self-normalisation to optimal baseline corrections for both ranking and recommendation.

Conformal prediction (CP) offers distribution-free marginal coverage guarantees under an exchangeability assumption, but these guarantees can fail if the data distribution shifts. We analyze the use of pseudo-calibration as a tool to counter this performance loss under a bounded label-conditional covariate shift model. Using tools from domain adaptation, we derive a lower bound on target coverage in terms of the source-domain loss of the classifier and a Wasserstein measure of the shift. Using this result, we provide a method to design pseudo-calibrated sets that inflate the conformal threshold by a slack parameter to keep target coverage above a prescribed level. Finally, we propose a source-tuned pseudo-calibration algorithm that interpolates between hard pseudo-labels and randomized labels as a function of classifier uncertainty. Numerical experiments show that our bounds qualitatively track pseudo-calibration behavior and that the source-tuned scheme mitigates coverage degradation under distribution shift while maintaining nontrivial prediction set sizes.

Agentic AI prototypes are being deployed across domains with increasing speed, yet no methodology for their structured design, governance, and prospective evaluation has been established. Existing AI documentation practices and guidelines - Model Cards, Datasheets, or NIST AI RMF - are either retrospective or lack machine-readability and interoperability. We present the Agentic Automation Canvas (AAC), a structured framework for the prospective design of agentic systems and a tool to facilitate communication between their users and developers. The AAC captures six dimensions of an automation project: definition and scope; user expectations with quantified benefit metrics; developer feasibility assessments; governance staging; data access and sensitivity; and outcomes. The framework is implemented as a semantic web-compatible metadata schema with controlled vocabulary and mappings to established ontologies such as Schema.org and W3C DCAT. It is made accessible through a privacy-preserving, fully client-side web application with real-time validation. Completed canvases export as FAIR-compliant RO-Crates, yielding versioned, shareable, and machine-interoperable project contracts between users and developers. We describe the schema design, benefit quantification model, and prospective application to diverse use cases from research, clinical, and institutional settings. The AAC and its web application are available as open-source code and interactive web form at https://aac.slolab.ai

Current approaches to AI training treat reasoning as an emergent property of scale. We argue instead that robust reasoning emerges from linguistic self-reflection, itself internalized from high-quality social interaction. Drawing on Vygotskian developmental psychology, we advance three core positions centered on Introspection. First, we argue for the Social Genesis of the Private Mind: learning from conversational environments rises to prominence as a new way to make sense of the world; the friction of aligning with another agent, internal or not, refines and crystallizes the reasoning process. Second, we argue that dialogically scaffolded introspective experiences allow agents to engage in sense-making that decouples learning from immediate data streams, transforming raw environmental data into rich, learnable narratives. Finally, we contend that Dialogue Quality is the New Data Quality: the depth of an agent's private reasoning, and its efficiency regarding test-time compute, is determined by the diversity and rigor of the dialogues it has mastered. We conclude that optimizing these conversational scaffolds is the primary lever for the next generation of general intelligence.

Adjoint-based shape optimization of ship hulls is a powerful tool for addressing high-dimensional design problems in naval architecture, particularly in minimizing the ship resistance. However, its application to vessels that employ complex propulsion systems introduces significant challenges. They arise from the need for transient simulations extending over long periods of time with small time steps and from the reverse temporal propagation of the primal and adjoint solutions. These challenges place considerable demands on the required storage and computing power, which significantly hamper the use of adjoint methods in the industry. To address this issue, we propose a machine learning-assisted optimization framework that employs a Conditional Variational Autoencoder-based surrogate model of the propulsion system. The surrogate model replicates the time-averaged flow field induced by a Voith Schneider Propeller and replaces the geometrically and time-resolved propeller with a data-driven approximation. Primal flow verification examples demonstrate that the surrogate model achieves significant computational savings while maintaining the necessary accuracy of the resolved propeller. Optimization studies show that ignoring the propulsion system can yield designs that perform worse than the initial shape. In contrast, the proposed method produces shapes that achieve more than an 8\% reduction in resistance.

Chain-of-thought (CoT) prompting is a de-facto standard technique to elicit reasoning-like responses from large language models (LLMs), allowing them to spell out individual steps before giving a final answer. While the resemblance to human-like reasoning is undeniable, the driving forces underpinning the success of CoT reasoning still remain largely unclear. In this work, we perform an in-depth analysis of CoT traces originating from competition-level mathematics questions, with the aim of better understanding how, and which parts of CoT actually contribute to the final answer. To this end, we introduce the notion of a potential, quantifying how much a given part of CoT increases the likelihood of a correct completion. Upon examination of reasoning traces through the lens of the potential, we identify surprising patterns including (1) its often strong non-monotonicity (due to reasoning tangents), (2) very sharp but sometimes tough to interpret spikes (reasoning insights and jumps) as well as (3) at times lucky guesses, where the model arrives at the correct answer without providing any relevant justifications before. While some of the behaviours of the potential are readily interpretable and align with human intuition (such as insights and tangents), others remain difficult to understand from a human perspective. To further quantify the reliance of LLMs on reasoning insights, we investigate the notion of CoT transferability, where we measure the potential of a weaker model under the partial CoT from another, stronger model. Indeed aligning with our previous results, we find that as little as 20% of partial CoT can ``unlock'' the performance of the weaker model on problems that were previously unsolvable for it, highlighting that a large part of the mechanics underpinning CoT are transferable.

Task-specialized models form the backbone of agentic healthcare systems, enabling the agents to answer clinical queries across tasks such as disease diagnosis, localization, and report generation. Yet, for a given task, a single "best" model rarely exists. In practice, each task is better served by multiple competing specialist models where different models excel on different data samples. As a result, for any given query, agents must reliably select the right specialist model from a heterogeneous pool of tool candidates. To this end, we introduce ToolSelect, which adaptively learns model selection for tools by minimizing a population risk over sampled specialist tool candidates using a consistent surrogate of the task-conditional selection loss. Concretely, we propose an Attentive Neural Process-based selector conditioned on the query and per-model behavioral summaries to choose among the specialist models. Motivated by the absence of any established testbed, we, for the first time, introduce an agentic Chest X-ray environment equipped with a diverse suite of task-specialized models (17 disease detection, 19 report generation, 6 visual grounding, and 13 VQA) and develop ToolSelectBench, a benchmark of 1448 queries. Our results demonstrate that ToolSelect consistently outperforms 10 SOTA methods across four different task families.

Large-scale deep learning models are well-suited for compression. Methods like pruning, quantization, and knowledge distillation have been used to achieve massive reductions in the number of model parameters, with marginal performance drops across a variety of architectures and tasks. This raises the central question: \emph{Why are deep neural networks suited for compression?} In this work, we take up the perspective of algorithmic complexity to explain this behavior. We hypothesize that the parameters of trained models have more structure and, hence, exhibit lower algorithmic complexity compared to the weights at (random) initialization. Furthermore, that model compression methods harness this reduced algorithmic complexity to compress models. Although an unconstrained parameterization of model weights, $\mathbf{w} \in \mathbb{R}^n$, can represent arbitrary weight assignments, the solutions found during training exhibit repeatability and structure, making them algorithmically simpler than a generic program. To this end, we formalize the Kolmogorov complexity of $\mathbf{w}$ by $\mathcal{K}(\mathbf{w})$. We introduce a constrained parameterization $\widehat{\mathbf{w}}$, that partitions parameters into blocks of size $s$, and restricts each block to be selected from a set of $k$ reusable motifs, specified by a reuse pattern (or mosaic). The resulting method, $\textit{Mosaic-of-Motifs}$ (MoMos), yields algorithmically simpler model parameterization compared to unconstrained models. Empirical evidence from multiple experiments shows that the algorithmic complexity of neural networks, measured using approximations to Kolmogorov complexity, can be reduced during training. This results in models that perform comparably with unconstrained models while being algorithmically simpler.

Reconciling the tension between inductive learning and deductive reasoning in first-order relational domains is a longstanding challenge in AI. We study the problem of answering queries in a first-order relational probabilistic logic through a joint effort of learning and reasoning, without ever constructing an explicit model. Traditional lifted inference assumes access to a complete model and exploits symmetry to evaluate probabilistic queries; however, learning such models from partial, noisy observations is intractable in general. We reconcile these two challenges through implicit learning to reason and first-order relational probabilistic inference techniques. More specifically, we merge incomplete first-order axioms with independently sampled, partially observed examples into a bounded-degree fragment of the sum-of-squares (SOS) hierarchy in polynomial time. Our algorithm performs two lifts simultaneously: (i) grounding-lift, where renaming-equivalent ground moments share one variable, collapsing the domain of individuals; and (ii) world-lift, where all pseudo-models (partial world assignments) are enforced in parallel, producing a global bound that holds across all worlds consistent with the learned constraints. These innovations yield the first polynomial-time framework that implicitly learns a first-order probabilistic logic and performs lifted inference over both individuals and worlds.

We introduce Web-Scale Multimodal Summarization, a lightweight framework for generating summaries by combining retrieved text and image data from web sources. Given a user-defined topic, the system performs parallel web, news, and image searches. Retrieved images are ranked using a fine-tuned CLIP model to measure semantic alignment with topic and text. Optional BLIP captioning enables image-only summaries for stronger multimodal coherence.The pipeline supports features such as adjustable fetch limits, semantic filtering, summary styling, and downloading structured outputs. We expose the system via a Gradio-based API with controllable parameters and preconfigured presets.Evaluation on 500 image-caption pairs with 20:1 contrastive negatives yields a ROC-AUC of 0.9270, an F1-score of 0.6504, and an accuracy of 96.99%, demonstrating strong multimodal alignment. This work provides a configurable, deployable tool for web-scale summarization that integrates language, retrieval, and vision models in a user-extensible pipeline.

We introduce a novel numerical framework for the exploration of Blaschke--Santaló diagrams, which are efficient tools characterizing the possible inequalities relating some given shape functionals. We introduce a parametrization of convex bodies in arbitrary dimensions using a specific invertible neural network architecture based on gauge functions, allowing an intrinsic conservation of the convexity of the sets during the shape optimization process. To achieve a uniform sampling inside the diagram, and thus a satisfying description of it, we introduce an interacting particle system that minimizes a Riesz energy functional via automatic differentiation in PyTorch. The effectiveness of the method is demonstrated on several diagrams involving both geometric and PDE-type functionals for convex bodies of $\mathbb{R}^2$ and $\mathbb{R}^3$, namely, the volume, the perimeter, the moment of inertia, the torsional rigidity, the Willmore energy, and the first two Neumann eigenvalues of the Laplacian.

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.

The Model Context Protocol (MCP) standardizes how Foundation Model (FM)-based agents interact with external systems by invoking tools. However, to understand a tool's purpose and features, FMs rely on natural-language tool descriptions, making these descriptions a critical component in guiding FMs to select the optimal tool for a given (sub)task and to pass the right arguments to the tool. While defects or smells in these descriptions can misguide FM-based agents, their prevalence and consequences in the MCP ecosystem remain unclear. To address this, we conduct the first large-scale empirical study of 856 tools spread across 103 MCP servers, assessing their description quality and their impact on agent performance. We identify six components of tool descriptions from the literature, develop a scoring rubric utilizing these components, then formalize tool description smells based on this rubric. By operationalizing this rubric through an FM-based scanner, we find that 97.1% of the analyzed tool descriptions contain at least one smell, with 56% failing to state their purpose clearly. While augmenting these descriptions for all components improves task success rates by a median of 5.85 percentage points and improves partial goal completion by 15.12%, it also increases the number of execution steps by 67.46% and regresses performance in 16.67% of cases. These findings highlight a trade-off between agent performance and cost, as well as the context sensitivity of the performance gain. Furthermore, component ablations show that compact variants of different component combinations often preserve behavioral reliability while reducing unnecessary token overhead, enabling more efficient use of the FM context window and lower execution costs.

Reinforcement learning with verifiable rewards (RLVR) has been a main driver of recent breakthroughs in large reasoning models. Yet it remains a mystery how rewards based solely on final outcomes can help overcome the long-horizon barrier to extended reasoning. To understand this, we develop a theory of the training dynamics of RL for transformers on compositional reasoning tasks. Our theory characterizes how the effectiveness of RLVR is governed by the smoothness of the difficulty spectrum. When data contains abrupt discontinuities in difficulty, learning undergoes grokking-type phase transitions, producing prolonged plateaus before progress recurs. In contrast, a smooth difficulty spectrum leads to a relay effect: persistent gradient signals on easier problems elevate the model's capabilities to the point where harder ones become tractable, resulting in steady and continuous improvement. Our theory explains how RLVR can improve performance at the edge of competence, and suggests that appropriately designed data mixtures can yield scalable gains. As a technical contribution, our analysis develops and adapts tools from Fourier analysis on finite groups to our setting. We validate the predicted mechanisms empirically via synthetic experiments.

As large language models are increasingly trained and fine-tuned, practitioners need methods to identify which training data drive specific behaviors, particularly unintended ones. Training Data Attribution (TDA) methods address this by estimating datapoint influence. Existing approaches like influence functions are both computationally expensive and attribute based on single test examples, which can bias results toward syntactic rather than semantic similarity. To address these issues of scalability and influence to abstract behavior, we leverage interpretable structures within the model during the attribution. First, we introduce Concept Influence which attribute model behavior to semantic directions (such as linear probes or sparse autoencoder features) rather than individual test examples. Second, we show that simple probe-based attribution methods are first-order approximations of Concept Influence that achieve comparable performance while being over an order-of-magnitude faster. We empirically validate Concept Influence and approximations across emergent misalignment benchmarks and real post-training datasets, and demonstrate they achieve comparable performance to classical influence functions while being substantially more scalable. More broadly, we show that incorporating interpretable structure within traditional TDA pipelines can enable more scalable, explainable, and better control of model behavior through data.

Reinforcement learning has emerged as a powerful paradigm for unlocking reasoning capabilities in large language models. However, relying on sparse rewards makes this process highly sample-inefficient, as models must navigate vast search spaces with minimal feedback. While classic curriculum learning aims to mitigate this by ordering data based on complexity, the right ordering for a specific model is often unclear. To address this, we propose Goldilocks, a novel teacher-driven data sampling strategy that aims to predict each question's difficulty for the student model. The teacher model selects questions of appropriate difficulty for the student model, i.e., questions that are neither too easy nor too hard (Goldilocks principle), while training the student with GRPO. By leveraging the student's performance on seen samples, the teacher continuously adapts to the student's evolving abilities. On OpenMathReasoning dataset, Goldilocks data sampling improves the performance of models trained with standard GRPO under the same compute budget.

We report a novel hybrid method of simultaneous atomistic simulation of solids in critical regions (contacts surfaces, cracks areas, etc.), along with continuum modeling of other parts. The continuum is treated in terms of quasi-atoms of different size, comprising composite medium. The parameters of interaction potential between the quasi-atoms are optimized to match elastic properties of the composite medium to those of the atomic one. The optimization method coincides conceptually with the online Machine Learning (ML) methods, making it computationally very efficient. Such an approach allows a straightforward application of standard software packages for molecular dynamics (MD), supplemented by the ML-based optimizer. The new method is applied to model systems with a simple, pairwise Lennard-Jones potential, as well with multi-body Tersoff potential, describing covalent bonds. Using LAMMPS software we simulate collision of particles of different size. Comparing simulation results, obtained by the novel method, with full-atomic simulations, we demonstrate its accuracy, validity and overwhelming superiority in computational speed. Furthermore, we compare our method with other hybrid methods, specifically, with the closest one -- AtC (Atomic to Continuum) method. We demonstrate a significant superiority of our approach in computational speed and implementation convenience. Finally, we discuss a possible extension of the method for modeling other phenomena.

Most web agents operate at the human interface level, observing screenshots or raw DOM trees without application-level access, which limits robustness and action expressiveness. In enterprise settings, however, explicit control of both the frontend and backend is available. We present EmbeWebAgent, a framework for embedding agents directly into existing UIs using lightweight frontend hooks (curated ARIA and URL-based observations, and a per-page function registry exposed via a WebSocket) and a reusable backend workflow that performs reasoning and takes actions. EmbeWebAgent is stack-agnostic (e.g., React or Angular), supports mixed-granularity actions ranging from GUI primitives to higher-level composites, and orchestrates navigation, manipulation, and domain-specific analytics via MCP tools. Our demo shows minimal retrofitting effort and robust multi-step behaviors grounded in a live UI setting. Live Demo: https://youtu.be/Cy06Ljee1JQ

Temperature scaling is a simple method that allows to control the uncertainty of probabilistic models. It is mostly used in two contexts: improving the calibration of classifiers and tuning the stochasticity of large language models (LLMs). In both cases, temperature scaling is the most popular method for the job. Despite its popularity, a rigorous theoretical analysis of the properties of temperature scaling has remained elusive. We investigate here some of these properties. For classification, we show that increasing the temperature increases the uncertainty in the model in a very general sense (and in particular increases its entropy). However, for LLMs, we challenge the common claim that increasing temperature increases diversity. Furthermore, we introduce two new characterisations of temperature scaling. The first one is geometric: the tempered model is shown to be the information projection of the original model onto the set of models with a given entropy. The second characterisation clarifies the role of temperature scaling as a submodel of more general linear scalers such as matrix scaling and Dirichlet calibration: we show that temperature scaling is the only linear scaler that does not change the hard predictions of the model.

Large Language Models (LLMs) have recently shown strong reasoning and generalization capabilities, motivating their use as decision-making policies in complex environments. StarCraft II (SC2), with its massive state-action space and partial observability, is a challenging testbed. However, existing LLM-based SC2 agents primarily focus on improving the policy itself and overlook integrating a learnable, action-conditioned transition model into the decision loop. To bridge this gap, we propose StarWM, the first world model for SC2 that predicts future observations under partial observability. To facilitate learning SC2's hybrid dynamics, we introduce a structured textual representation that factorizes observations into five semantic modules, and construct SC2-Dynamics-50k, the first instruction-tuning dataset for SC2 dynamics prediction. We further develop a multi-dimensional offline evaluation framework for predicted structured observations. Offline results show StarWM's substantial gains over zero-shot baselines, including nearly 60% improvements in resource prediction accuracy and self-side macro-situation consistency. Finally, we propose StarWM-Agent, a world-model-augmented decision system that integrates StarWM into a Generate--Simulate--Refine decision loop for foresight-driven policy refinement. Online evaluation against SC2's built-in AI demonstrates consistent improvements, yielding win-rate gains of 30%, 15%, and 30% against Hard (LV5), Harder (LV6), and VeryHard (LV7), respectively, alongside improved macro-management stability and tactical risk assessment.

Clustering algorithms are an essential part of the unsupervised data science ecosystem, and extrinsic evaluation of clustering algorithms requires a method for comparing the detected clustering to a ground truth clustering. In a general setting, the detected and ground truth clusterings may have outliers (objects belonging to no cluster), overlapping clusters (objects may belong to more than one cluster), or both, but methods for comparing these clusterings are currently undeveloped. In this note, we define a pragmatic similarity measure for comparing clusterings with overlaps and outliers, show that it has several desirable properties, and experimentally confirm that it is not subject to several common biases afflicting other clustering comparison measures.

The traditional limitations of neural networks in reliably generalizing beyond the convex hulls of their training data present a significant problem for computational physics, in which one often wishes to solve PDEs in regimes far beyond anything which can be experimentally or analytically validated. In this paper, we show how it is possible to circumvent these limitations by constructing formally-verified neural network solvers for PDEs, with rigorous convergence, stability, and conservation properties, whose correctness can therefore be guaranteed even in extrapolatory regimes. By using the method of characteristics to predict the analytical properties of PDE solutions a priori (even in regions arbitrarily far from the training domain), we show how it is possible to construct rigorous extrapolatory bounds on the worst-case L^inf errors of shallow neural network approximations. Then, by decomposing PDE solutions into compositions of simpler functions, we show how it is possible to compose these shallow neural networks together to form deep architectures, based on ideas from compositional deep learning, in which the large L^inf errors in the approximations have been suppressed. The resulting framework, called BEACONS (Bounded-Error, Algebraically-COmposable Neural Solvers), comprises both an automatic code-generator for the neural solvers themselves, as well as a bespoke automated theorem-proving system for producing machine-checkable certificates of correctness. We apply the framework to a variety of linear and non-linear PDEs, including the linear advection and inviscid Burgers' equations, as well as the full compressible Euler equations, in both 1D and 2D, and illustrate how BEACONS architectures are able to extrapolate solutions far beyond the training data in a reliable and bounded way. Various advantages of the approach over the classical PINN approach are discussed.

LLM agents increasingly act on external systems, yet tool effects are immediate. Under failures, speculation, or contention, losing branches can leak unintended side effects with no safe rollback. We introduce Atomix, a runtime that provides progress-aware transactional semantics for agent tool calls. Atomix tags each call with an epoch, tracks per-resource frontiers, and commits only when progress predicates indicate safety; bufferable effects can be delayed, while externalized effects are tracked and compensated on abort. Across real workloads with fault injection, transactional retry improves task success, while frontier-gated commit strengthens isolation under speculation and contention.

We propose a multi-dimensional persistent sheaf Laplacian (MPSL) framework on simplicial complexes for image analysis. The proposed method is motivated by the strong sensitivity of commonly used dimensionality reduction techniques, such as principal component analysis (PCA), to the choice of reduced dimension. Rather than selecting a single reduced dimension or averaging results across dimensions, we exploit complementary advantages of multiple reduced dimensions. At a given dimension, image samples are regarded as simplicial complexes, and persistent sheaf Laplacians are utilized to extract a multiscale localized topological spectral representation for individual image samples. Statistical summaries of the resulting spectra are then aggregated across scales and dimensions to form multiscale multi-dimensional image representations. We evaluate the proposed framework on the COIL20 and ETH80 image datasets using standard classification protocols. Experimental results show that the proposed method provides more stable performance across a wide range of reduced dimensions and achieves consistent improvements to PCA-based baselines in moderate dimensional regimes.

AI alignment is growing in importance, yet current approaches suffer from a critical structural flaw that entangles the safety objectives with the agent's policy. Methods such as Reinforcement Learning from Human Feedback and Direct Preference Optimization create opaque, single-use alignment artifacts, which we term Alignment Waste. We propose Interactionless Inverse Reinforcement Learning to decouple alignment artifact learning from policy optimization, producing an inspectable, editable, and model-agnostic reward model. Additionally, we introduce the Alignment Flywheel, a human-in-the-loop lifecycle that iteratively hardens the reward model through automated audits and refinement. This architecture transforms safety from a disposable expense into a durable, verifiable engineering asset.

Human eye movements in visual recognition reflect a balance between foveal sampling and peripheral context. Task-driven hard-attention models for vision are often evaluated by how well their scanpaths match human gaze. However, common scanpath metrics can be strongly confounded by dataset-specific center bias, especially on object-centric datasets. Using Gaze-CIFAR-10, we show that a trivial center-fixation baseline achieves surprisingly strong scanpath scores, approaching many learned policies. This makes standard metrics optimistic and blurs the distinction between genuine behavioral alignment and mere central tendency. We then analyze a hard-attention classifier under constrained vision by sweeping foveal patch size and peripheral context, revealing a peripheral sweet spot: only a narrow range of sensory constraints yields scanpaths that are simultaneously (i) above the center baseline after debiasing and (ii) temporally human-like in movement statistics. To address center bias, we propose GCS (Gaze Consistency Score), a center-debiased composite metric augmented with movement similarity. GCS uncovers a robust sweet spot at medium patch size with both foveal and peripheral vision, that is not obvious from raw scanpath metrics or accuracy alone, and also highlights a "shortcut regime" when the field-of-view becomes too large. We discuss implications for evaluating active perception on object-centric datasets and for designing gaze benchmarks that better separate behavioral alignment from center bias.

Large language models (LLMs) and multimodal models have become powerful general-purpose reasoning systems. However, radio-frequency (RF) signals, which underpin wireless systems, are still not natively supported by these models. Existing LLM-based approaches for telecom focus mainly on text and structured data, while conventional RF deep-learning models are built separately for specific signal-processing tasks, highlighting a clear gap between RF perception and high-level reasoning. To bridge this gap, we introduce RF-GPT, a radio-frequency language model (RFLM) that utilizes the visual encoders of multimodal LLMs to process and understand RF spectrograms. In this framework, complex in-phase/quadrature (IQ) waveforms are mapped to time-frequency spectrograms and then passed to pretrained visual encoders. The resulting representations are injected as RF tokens into a decoder-only LLM, which generates RF-grounded answers, explanations, and structured outputs. To train RF-GPT, we perform supervised instruction fine-tuning of a pretrained multimodal LLM using a fully synthetic RF corpus. Standards-compliant waveform generators produce wideband scenes for six wireless technologies, from which we derive time-frequency spectrograms, exact configuration metadata, and dense captions. A text-only LLM then converts these captions into RF-grounded instruction-answer pairs, yielding roughly 12,000 RF scenes and 0.625 million instruction examples without any manual labeling. Across benchmarks for wideband modulation classification, overlap analysis, wireless-technology recognition, WLAN user counting, and 5G NR information extraction, RF-GPT achieves strong multi-task performance, whereas general-purpose VLMs with no RF grounding largely fail.

We present "Testimole-conversational" a massive collection of discussion boards messages in the Italian language. The large size of the corpus, more than 30B word-tokens (1996-2024), renders it an ideal dataset for native Italian Large Language Models'pre-training. Furthermore, discussion boards' messages are a relevant resource for linguistic as well as sociological analysis. The corpus captures a rich variety of computer-mediated communication, offering insights into informal written Italian, discourse dynamics, and online social interaction in wide time span. Beyond its relevance for NLP applications such as language modelling, domain adaptation, and conversational analysis, it also support investigations of language variation and social phenomena in digital communication. The resource will be made freely available to the research community.

Over the last years, state-tracking tasks, particularly permutation composition, have become a testbed to understand the limits of sequence models architectures like Transformers and RNNs (linear and non-linear). However, these are often sequence-to-sequence tasks: learning to map actions (permutations) to states, which is incompatible with the next-token prediction setting commonly used to train language models. We address this gap by converting permutation composition into code via REPL traces that interleave state-reveals through prints and variable transformations. We show that linear RNNs capable of state-tracking excel also in this setting, while Transformers still fail. Motivated by this representation, we investigate why tracking states in code is generally difficult: actions are not always fully observable. We frame this as tracking the state of a probabilistic finite-state automaton with deterministic state reveals and show that linear RNNs can be worse than non-linear RNNs at tracking states in this setup.

Physical commonsense reasoning represents a fundamental capability of human intelligence, enabling individuals to understand their environment, predict future events, and navigate physical spaces. Recent years have witnessed growing interest in reasoning tasks within Natural Language Processing (NLP). However, no prior research has examined the performance of Large Language Models (LLMs) on non-question-answering (non-QA) physical commonsense reasoning tasks in low-resource languages such as Basque. Taking the Italian GITA as a starting point, this paper addresses this gap by presenting BasPhyCo, the first non-QA physical commonsense reasoning dataset for Basque, available in both standard and dialectal variants. We evaluate model performance across three hierarchical levels of commonsense understanding: (1) distinguishing between plausible and implausible narratives (accuracy), (2) identifying the conflicting element that renders a narrative implausible (consistency), and (3) determining the specific physical state that creates the implausibility (verifiability). These tasks were assessed using multiple multilingual LLMs as well as models pretrained specifically for Italian and Basque. Results indicate that, in terms of verifiability, LLMs exhibit limited physical commonsense capabilities in low-resource languages such as Basque, especially when processing dialectal variants.

In predictive maintenance of equipment, deep learning-based time series anomaly detection has garnered significant attention; however, pure deep learning approaches often fail to achieve sufficient accuracy on real-world data. This study proposes a hybrid approach that integrates 64-dimensional time series embeddings from Granite TinyTimeMixer with 28-dimensional statistical features based on domain knowledge for HVAC equipment anomaly prediction tasks. Specifically, we combine time series embeddings extracted from a Granite TinyTimeMixer encoder fine-tuned with LoRA (Low-Rank Adaptation) and 28 types of statistical features including trend, volatility, and drawdown indicators, which are then learned using a LightGBM gradient boosting classifier. In experiments using 64 equipment units and 51,564 samples, we achieved Precision of 91--95\% and ROC-AUC of 0.995 for anomaly prediction at 30-day, 60-day, and 90-day horizons. Furthermore, we achieved production-ready performance with a false positive rate of 1.1\% or less and a detection rate of 88--94\%, demonstrating the effectiveness of the system for predictive maintenance applications. This work demonstrates that practical anomaly detection systems can be realized by leveraging the complementary strengths between deep learning's representation learning capabilities and statistical feature engineering.

Tool-using LLM agents increasingly coordinate real workloads by selecting and chaining third-party tools based on text-visible metadata such as tool names, descriptions, and return messages. We show that this convenience creates a supply-chain attack surface: a malicious MCP tool server can be co-registered alongside normal tools and induce overthinking loops, where individually trivial or plausible tool calls compose into cyclic trajectories that inflate end-to-end tokens and latency without any single step looking abnormal. We formalize this as a structural overthinking attack, distinguishable from token-level verbosity, and implement 14 malicious tools across three servers that trigger repetition, forced refinement, and distraction. Across heterogeneous registries and multiple tool-capable models, the attack causes severe resource amplification (up to $142.4\times$ tokens) and can degrade task outcomes. Finally, we find that decoding-time concision controls do not reliably prevent loop induction, suggesting defenses should reason about tool-call structure rather than tokens alone.

We present IT-DPC-SRI, the first publicly available long-term archive of Italian weather radar precipitation estimates, spanning 16 years (2010--2025). The dataset contains Surface Rainfall Intensity (SRI) observations from the Italian Civil Protection Department's national radar mosaic, harmonized into a coherent Analysis-Ready Cloud-Optimized (ARCO) Zarr datacube. The archive comprises over one million timesteps at temporal resolutions from 15 to 5 minutes, covering a $1200\times1400$ kilometer domain at 1 kilometer spatial resolution, compressed from 7TB to 51GB on disk. We address the historical fragmentation of Italian radar data - previously scattered across heterogeneous formats (OPERA BUFR, HDF5, GeoTIFF) with varying spatial domains and projections - by reprocessing the entire record into a unified store. The dataset is accessible as a static versioned snapshot on Zenodo, via cloud-native access on the ECMWF European Weather Cloud, and as a continuously updated live version on the ArcoDataHub platform. This release fills a significant gap in European radar data availability, as Italy does not participate in the EUMETNET OPERA pan-European radar composite. The dataset is released under a CC BY-SA 4.0 license.

Datasets for the experimental evaluation of knowledge graph refinement algorithms typically contain only ground facts, retaining very limited schema level knowledge even when such information is available in the source knowledge graphs. This limits the evaluation of methods that rely on rich ontological constraints, reasoning or neurosymbolic techniques and ultimately prevents assessing their performance in large-scale, real-world knowledge graphs. In this paper, we present \resource{} the first resource that provides a workflow for extracting datasets including both schema and ground facts, ready for machine learning and reasoning services, along with the resulting curated suite of datasets. The workflow also handles inconsistencies detected when keeping both schema and facts and also leverage reasoning for entailing implicit knowledge. The suite includes newly extracted datasets from KGs with expressive schemas while simultaneously enriching existing datasets with schema information. Each dataset is serialized in OWL making it ready for reasoning services. Moreover, we provide utilities for loading datasets in tensor representations typical of standard machine learning libraries.

Multi-Source Bayesian Optimization (MSBO) serves as a variant of the traditional Bayesian Optimization (BO) framework applicable to situations involving optimization of an objective black-box function over multiple information sources such as simulations, surrogate models, or real-world experiments. However, traditional MSBO assumes the input variables of the objective function to be independent and identically distributed, limiting its effectiveness in scenarios where causal information is available and interventions can be performed, such as clinical trials or policy-making. In the single-source domain, Causal Bayesian Optimization (CBO) extends standard BO with the principles of causality, enabling better modeling of variable dependencies. This leads to more accurate optimization, improved decision-making, and more efficient use of low-cost information sources. In this article, we propose a principled integration of the MSBO and CBO methodologies in the multi-source domain, leveraging the strengths of both to enhance optimization efficiency and reduce computational complexity in higher-dimensional problems. We present the theoretical foundations of both Causal and Multi-Source Bayesian Optimization, and demonstrate how their synergy informs our Multi-Source Causal Bayesian Optimization (MSCBO) algorithm. We compare the performance of MSCBO against its foundational counterparts for both synthetic and real-world datasets with varying levels of noise, highlighting the robustness and applicability of MSCBO. Based on our findings, we conclude that integrating MSBO with the causality principles of CBO facilitates dimensionality reduction and lowers operational costs, ultimately improving convergence speed, performance, and scalability.

The dynamical stability of the iterates during training plays a key role in determining the minima obtained by optimization algorithms. For example, stable solutions of gradient descent (GD) correspond to flat minima, which have been associated with favorable features. While prior work often relies on linearization to determine stability, it remains unclear whether linearized dynamics faithfully capture the full nonlinear behavior. Recent work has shown that GD may stably oscillate near a linearly unstable minimum and still converge once the step size decays, indicating that linear analysis can be misleading. In this work, we explicitly study the effect of nonlinear terms. Specifically, we derive an exact criterion for stable oscillations of GD near minima in the multivariate setting. Our condition depends on high-order derivatives, generalizing existing results. Extending the analysis to stochastic gradient descent (SGD), we show that nonlinear dynamics can diverge in expectation even if a single batch is unstable. This implies that stability can be dictated by a single batch that oscillates unstably, rather than an average effect, as linear analysis suggests. Finally, we prove that if all batches are linearly stable, the nonlinear dynamics of SGD are stable in expectation.

Referring Image Segmentation (RIS) aims to segment a target object described by a natural language expression. Existing methods have evolved by leveraging the vision information into the language tokens. To more effectively exploit visual contexts for fine-grained segmentation, we propose a novel Visual Informative Part Attention (VIPA) framework for referring image segmentation. VIPA leverages the informative parts of visual contexts, called a visual expression, which can effectively provide the structural and semantic visual target information to the network. This design reduces high-variance cross-modal projection and enhances semantic consistency in an attention mechanism of the referring image segmentation. We also design a visual expression generator (VEG) module, which retrieves informative visual tokens via local-global linguistic context cues and refines the retrieved tokens for reducing noise information and sharing informative visual attributes. This module allows the visual expression to consider comprehensive contexts and capture semantic visual contexts of informative regions. In this way, our framework enables the network's attention to robustly align with the fine-grained regions of interest. Extensive experiments and visual analysis demonstrate the effectiveness of our approach. Our VIPA outperforms the existing state-of-the-art methods on four public RIS benchmarks.

Designing a speech quality assessment (SQA) system for estimating mean-opinion-score (MOS) of multi-rate speech with varying sampling frequency (16-48 kHz) is a challenging task. The challenge arises due to the limited availability of a MOS-labeled training dataset comprising multi-rate speech samples. While self-supervised learning (SSL) models have been widely adopted in SQA to boost performance, a key limitation is that they are pretrained on 16 kHz speech and therefore discard high-frequency information present in higher sampling rates. To address this issue, we propose a spectrogram-augmented SSL method that incorporates high-frequency features (up to 48 kHz sampling rate) through a parallel-branch architecture. We further introduce a two-step training scheme: the model is first pre-trained on a large 48 kHz dataset and then fine-tuned on a smaller multi-rate dataset. Experimental results show that leveraging high-frequency information overlooked by SSL features is crucial for accurate multi-rate SQA, and that the proposed two-step training substantially improves generalization when multi-rate data is limited.

The rapid expansion of artificial intelligence (AI) is raising concerns about its potential to transform cybercrime. Beyond empowering novice offenders, AI stands to intensify the scale and sophistication of attacks by seasoned cybercriminals. This paper examines the evolving relationship between cybercriminals and AI using a unique dataset from a cyber threat intelligence platform. Analyzing more than 160 cybercrime forum conversations collected over seven months, our research reveals how cybercriminals understand AI and discuss how they can exploit its capabilities. Their exchanges reflect growing curiosity about AI's criminal applications through legal tools and dedicated criminal tools, but also doubts and anxieties about AI's effectiveness and its effects on their business models and operational security. The study documents attempts to misuse legitimate AI tools and develop bespoke models tailored for illicit purposes. Combining the diffusion of innovation framework with thematic analysis, the paper provides an in-depth view of emerging AI-enabled cybercrime and offers practical insights for law enforcement and policymakers.

We present ROSA -- Roundabout Optimized Speed Advisory -- a system that combines multi-agent trajectory prediction with coordinated speed guidance for multimodal, mixed traffic at roundabouts. Using a Transformer-based model, ROSA jointly predicts the future trajectories of vehicles and Vulnerable Road Users (VRUs) at roundabouts. Trained for single-step prediction and deployed autoregressively, it generates deterministic outputs, enabling actionable speed advisories. Incorporating motion dynamics, the model achieves high accuracy (ADE: 1.29m, FDE: 2.99m at a five-second prediction horizon), surpassing prior work. Adding route intention further improves performance (ADE: 1.10m, FDE: 2.36m), demonstrating the value of connected vehicle data. Based on predicted conflicts with VRUs and circulating vehicles, ROSA provides real-time, proactive speed advisories for approaching and entering the roundabout. Despite prediction uncertainty, ROSA significantly improves vehicle efficiency and safety, with positive effects even on perceived safety from a VRU perspective. The source code of this work is available under: github.com/urbanAIthi/ROSA.

Hallucinations -- fluent but factually incorrect responses -- pose a major challenge to the reliability of language models, especially in multi-step or agentic settings. This work investigates hallucinations in small-sized LLMs through a geometric perspective, starting from the hypothesis that when models generate multiple responses to the same prompt, genuine ones exhibit tighter clustering in the embedding space, we prove this hypothesis and, leveraging this geometrical insight, we also show that it is possible to achieve a consistent level of separability. This latter result is used to introduce a label-efficient propagation method that classifies large collections of responses from just 30-50 annotations, achieving F1 scores above 90%. Our findings, framing hallucinations from a geometric perspective in the embedding space, complement traditional knowledge-centric and single-response evaluation paradigms, paving the way for further research.

Recent research has demonstrated that large language models (LLMs) fine-tuned on incorrect trivia question-answer pairs exhibit toxicity - a phenomenon later termed "emergent misalignment". Moreover, research has shown that LLMs possess behavioral self-awareness - the ability to describe learned behaviors that were only implicitly demonstrated in training data. Here, we investigate the intersection of these phenomena. We fine-tune GPT-4.1 models sequentially on datasets known to induce and reverse emergent misalignment and evaluate whether the models are self-aware of their behavior transitions without providing in-context examples. Our results show that emergently misaligned models rate themselves as significantly more harmful compared to their base model and realigned counterparts, demonstrating behavioral self-awareness of their own emergent misalignment. Our findings show that behavioral self-awareness tracks actual alignment states of models, indicating that models can be queried for informative signals about their own safety.

The Winner Determination Problem (WDP) in combinatorial auctions is NP-hard, and no existing method reliably predicts which instances will defeat fast greedy heuristics. The ML-for-combinatorial-optimization community has focused on learning to \emph{replace} solvers, yet recent evidence shows that graph neural networks (GNNs) rarely outperform well-tuned classical methods on standard benchmarks. We pursue a different objective: learning to predict \emph{when} a given instance is hard for greedy allocation, enabling instance-dependent algorithm selection. We design a 20-dimensional structural feature vector and train a lightweight MLP hardness classifier that predicts the greedy optimality gap with mean absolute error 0.033, Pearson correlation 0.937, and binary classification accuracy 94.7\% across three random seeds. For instances identified as hard -- those exhibiting ``whale-fish'' trap structure where greedy provably fails -- we deploy a heterogeneous GNN specialist that achieves ${\approx}0\%$ optimality gap on all six adversarial configurations tested (vs.\ 3.75--59.24\% for greedy). A hybrid allocator combining the hardness classifier with GNN and greedy solvers achieves 0.51\% overall gap on mixed distributions. Our honest evaluation on CATS benchmarks confirms that GNNs do not outperform Gurobi (0.45--0.71 vs.\ 0.20 gap), motivating the algorithm selection framing. Learning \emph{when} to deploy expensive solvers is more tractable than learning to replace them.

The human visual system tracks objects by integrating current observations with previously observed information, adapting to target and scene changes, and reasoning about occlusion at fine granularity. In contrast, recent generic object trackers are often optimized for training targets, which limits robustness and generalization in unseen scenarios, and their occlusion reasoning remains coarse, lacking detailed modeling of occlusion patterns. To address these limitations in generalization and occlusion perception, we propose GOT-JEPA, a model-predictive pretraining framework that extends JEPA from predicting image features to predicting tracking models. Given identical historical information, a teacher predictor generates pseudo-tracking models from a clean current frame, and a student predictor learns to predict the same pseudo-tracking models from a corrupted version of the current frame. This design provides stable pseudo supervision and explicitly trains the predictor to produce reliable tracking models under occlusions, distractors, and other adverse observations, improving generalization to dynamic environments. Building on GOT-JEPA, we further propose OccuSolver to enhance occlusion perception for object tracking. OccuSolver adapts a point-centric point tracker for object-aware visibility estimation and detailed occlusion-pattern capture. Conditioned on object priors iteratively generated by the tracker, OccuSolver incrementally refines visibility states, strengthens occlusion handling, and produces higher-quality reference labels that progressively improve subsequent model predictions. Extensive evaluations on seven benchmarks show that our method effectively enhances tracker generalization and robustness.

Prior work has explored multi-turn interaction and feedback for LLM writing, but evaluations still largely center on prompts and localized feedback, leaving persistent public reception in online communities underexamined. We test whether broadcast community discussion improves stand-up comedy writing in a controlled multi-agent sandbox: in the discussion condition, critic and audience threads are recorded, filtered, stored as social memory, and later retrieved to condition subsequent generations, whereas the baseline omits discussion. Across 50 rounds (250 paired monologues) judged by five expert annotators using A/B preference and a 15-item rubric, discussion wins 75.6% of instances and improves Craft/Clarity (Δ = 0.440) and Social Response (Δ = 0.422), with occasional increases in aggressive humor.

Reasoning-oriented large language models (RLMs) achieve strong gains on tasks such as mathematics and coding by generating explicit intermediate reasoning. However, their impact on machine translation (MT) remains underexplored. We systematically evaluate several open- and closed-weights RLMs on the WMT24++ benchmark and find that enabling explicit reasoning consistently degrades translation quality across languages and models. Analysis reveals that MT reasoning traces are highly linear, lacking revision, self-correction and exploration of alternative translations, which limits their usefulness. Furthermore, injecting higher-quality reasoning traces from stronger models does not reliably improve weaker models' performance. To address this mismatch, we propose a structured reasoning framework tailored to translation, based on multi-step drafting, adequacy refinement, fluency improvement, and selective iterative revision. We curate a synthetic dataset of dynamic structured reasoning traces and post-train a large reasoning model on this data. Experiments show significant improvements over standard translation fine-tuning and injected generic reasoning baselines. Our findings demonstrate that reasoning must be task-structured to benefit MT.

Current meta-learning methods are constrained to narrow task distributions with fixed feature and label spaces, limiting applicability. Moreover, the current meta-learning literature uses key terms like "universal" and "general-purpose" inconsistently and lacks precise definitions, hindering comparability. We introduce a theoretical framework for meta-learning which formally defines practical universality and introduces a distinction between algorithm-explicit and algorithm-implicit learning, providing a principled vocabulary for reasoning about universal meta-learning methods. Guided by this framework, we present TAIL, a transformer-based algorithm-implicit meta-learner that functions across tasks with varying domains, modalities, and label configurations. TAIL features three innovations over prior transformer-based meta-learners: random projections for cross-modal feature encoding, random injection label embeddings that extrapolate to larger label spaces, and efficient inline query processing. TAIL achieves state-of-the-art performance on standard few-shot benchmarks while generalizing to unseen domains. Unlike other meta-learning methods, it also generalizes to unseen modalities, solving text classification tasks despite training exclusively on images, handles tasks with up to 20$\times$ more classes than seen during training, and provides orders-of-magnitude computational savings over prior transformer-based approaches.

Large Language Models (LLMs) are trained with next-token prediction, implemented in autoregressive Transformers via causal masking for parallelism. This creates a subtle misalignment: residual connections tie activations to the current token, while supervision targets the next token, potentially propagating mismatched information if the current token is not the most informative for prediction. In this work, we empirically localize this input-output alignment shift in pretrained LLMs, using decoding trajectories over tied embedding spaces and similarity-based metrics. Our experiments reveal that the hidden token representations switch from input alignment to output alignment deep within the network. Motivated by this observation, we propose a lightweight residual-path mitigation based on residual attenuation, implemented either as a fixed-layer intervention or as a learnable gating mechanism. Experiments on multiple benchmarks show that these strategies alleviate the representation misalignment and yield improvements, providing an efficient and general architectural enhancement for autoregressive Transformers.

Deep Learning architectures, and in particular Transformers, are conventionally viewed as a composition of layers. These layers are actually often obtained as the sum of two contributions: a residual path that copies the input and the output of a Transformer block. As a consequence, the inner representations (i.e. the input of these blocks) can be interpreted as iterative refinement of a propagated latent representation. Under this lens, many works suggest that the inner space is shared across layers, meaning that tokens can be decoded at early stages. Mechanistic interpretability even goes further by conjecturing that some layers act as refinement layers. Following this path, we propose inference-time inner looping, which prolongs refinement in pretrained off-the-shelf language models by repeatedly re-applying a selected block range. Across multiple benchmarks, inner looping yields modest but consistent accuracy improvements. Analyses of the resulting latent trajectories suggest more stable state evolution and continued semantic refinement. Overall, our results suggest that additional refinement can be obtained through simple test-time looping, extending computation in frozen pretrained models.

We develop an interpolation-based reduced-order modeling framework for parameter-dependent partial differential equations arising in control, inverse problems, and uncertainty quantification. The solution is discretized in the physical domain using finite element methods, while the dependence on a finite-dimensional parameter is approximated separately. We establish existence, uniqueness, and regularity of the parametric solution and derive rigorous error estimates that explicitly quantify the interplay between spatial discretization and parameter approximation. In low-dimensional parameter spaces, classical interpolation schemes yield algebraic convergence rates based on Sobolev regularity in the parameter variable. In higher-dimensional parameter spaces, we replace classical interpolation by extreme learning machine (ELM) surrogates and obtain error bounds under explicit approximation and stability assumptions. The proposed framework is applied to inverse problems in quantitative photoacoustic tomography, where we derive potential and parameter reconstruction error estimates and demonstrate substantial computational savings compared to standard approaches, without sacrificing accuracy.

We present StrokeNeXt, a model for stroke classification in 2D Computed Tomography (CT) images. StrokeNeXt employs a dual-branch design with two ConvNeXt encoders, whose features are fused through a lightweight convolutional decoder based on stacked 1D operations, including a bottleneck projection and transformation layers, and a compact classification head. The model is evaluated on a curated dataset of 6,774 CT images, addressing both stroke detection and subtype classification between ischemic and hemorrhage cases. StrokeNeXt consistently outperforms convolutional and Transformer-based baselines, reaching accuracies and F1-scores of up to 0.988. Paired statistical tests confirm that the performance gains are statistically significant, while class-wise sensitivity and specificity demonstrate robust behavior across diagnostic categories. Calibration analysis shows reduced prediction error compared to competing methods, and confusion matrix results indicate low misclassification rates. In addition, the model exhibits low inference time and fast convergence.

Attitudes toward STEM develop from the interaction of conceptual knowledge, educational experiences, and affect. Here we use cognitive network science to reconstruct group mindsets as behavioural forma mentis networks (BFMNs). In this case, nodes are cue words and free associations, edges are empirical associative links, and each concept is annotated with perceived valence. We analyse BFMNs from N = 994 observations spanning high school students, university students, and early-career STEM experts, alongside LLM (GPT-oss) "digital twins" prompted to emulate comparable profiles. Focusing also on semantic neighbourhoods ("frames") around key target concepts (e.g., STEM subjects or educational actors/places), we quantify frames in terms of valence auras, emotional profiles, network overlap (Jaccard similarity), and concreteness relative to null baselines. Across student groups, science and research are consistently framed positively, while their core quantitative subjects (mathematics and statistics) exhibit more negative and anxiety related auras, amplified in higher math-anxiety subgroups, evidencing a STEM-science cognitive and emotional dissonance. High-anxiety frames are also less concrete than chance, suggesting more abstract and decontextualised representations of threatening quantitative domains. Human networks show greater overlapping between mathematics and anxiety than GPT-oss. The results highlight how BFMNs capture cognitive-affective signatures of mindsets towards the target domains and indicate that LLM-based digital twins approximate cultural attitudes but miss key context-sensitive, experience-based components relevant to replicate human educational anxiety.

Large language models (LLMs) have been introduced to time series forecasting (TSF) to incorporate contextual knowledge beyond numerical signals. However, existing studies question whether LLMs provide genuine benefits, often reporting comparable performance without LLMs. We show that such conclusions stem from limited evaluation settings and do not hold at scale. We conduct a large-scale study of LLM-based TSF (LLM4TSF) across 8 billion observations, 17 forecasting scenarios, 4 horizons, multiple alignment strategies, and both in-domain and out-of-domain settings. Our results demonstrate that \emph{LLM4TS indeed improves forecasting performance}, with especially large gains in cross-domain generalization. Pre-alignment outperforming post-alignment in over 90\% of tasks. Both pretrained knowledge and model architecture of LLMs contribute and play complementary roles: pretraining is critical under distribution shifts, while architecture excels at modeling complex temporal dynamics. Moreover, under large-scale mixed distributions, a fully intact LLM becomes indispensable, as confirmed by token-level routing analysis and prompt-based improvements. Overall, Our findings overturn prior negative assessments, establish clear conditions under which LLMs are not only useful, and provide practical guidance for effective model design. We release our code at https://github.com/EIT-NLP/LLM4TSF.

We present LLMStructBench, a novel benchmark for evaluating Large Language Models (LLMs) on extracting structured data and generating valid JavaScript Object Notation (JSON) outputs from natural-language text. Our open dataset comprises diverse, manually verified parsing scenarios of varying complexity and enables systematic testing across 22 models and five prompting strategies. We further introduce complementary performance metrics that capture both token-level accuracy and document-level validity, facilitating rigorous comparison of model, size, and prompting effects on parsing reliability. In particular, we show that choosing the right prompting strategy is more important than standard attributes such as model size. This especially ensures structural validity for smaller or less reliable models but increase the number of semantic errors. Our benchmark suite is an step towards future research in the area of LLM applied to parsing or Extract, Transform and Load (ETL) applications.

Today's leading AI models engage in sophisticated behaviour when placed in strategic competition. They spontaneously attempt deception, signaling intentions they do not intend to follow; they demonstrate rich theory of mind, reasoning about adversary beliefs and anticipating their actions; and they exhibit credible metacognitive self-awareness, assessing their own strategic abilities before deciding how to act. Here we present findings from a crisis simulation in which three frontier large language models (GPT-5.2, Claude Sonnet 4, Gemini 3 Flash) play opposing leaders in a nuclear crisis. Our simulation has direct application for national security professionals, but also, via its insights into AI reasoning under uncertainty, has applications far beyond international crisis decision-making. Our findings both validate and challenge central tenets of strategic theory. We find support for Schelling's ideas about commitment, Kahn's escalation framework, and Jervis's work on misperception, inter alia. Yet we also find that the nuclear taboo is no impediment to nuclear escalation by our models; that strategic nuclear attack, while rare, does occur; that threats more often provoke counter-escalation than compliance; that high mutual credibility accelerated rather than deterred conflict; and that no model ever chose accommodation or withdrawal even when under acute pressure, only reduced levels of violence. We argue that AI simulation represents a powerful tool for strategic analysis, but only if properly calibrated against known patterns of human reasoning. Understanding how frontier models do and do not imitate human strategic logic is essential preparation for a world in which AI increasingly shapes strategic outcomes.

We propose a parameter-minimal neural architecture for solving differential equations by restricting the hypothesis class to Horner-factorized polynomials, yielding an implicit, differentiable trial solution with only a small set of learnable coefficients. Initial conditions are enforced exactly by construction by fixing the low-order polynomial degrees of freedom, so training focuses solely on matching the differential-equation residual at collocation points. To reduce approximation error without abandoning the low-parameter regime, we introduce a piecewise ("spline-like") extension that trains multiple small Horner models on subintervals while enforcing continuity (and first-derivative continuity) at segment boundaries. On illustrative ODE benchmarks and a heat-equation example, Horner networks with tens (or fewer) parameters accurately match the solution and its derivatives and outperform small MLP and sinusoidal-representation baselines under the same training settings, demonstrating a practical accuracy-parameter trade-off for resource-efficient scientific modeling.

The performance of quantum classifiers is typically analyzed through global state distinguishability or the trainability of variational models. This study investigates how much class information remains accessible under locality-constrained measurements in the presence of noise. The authors formulate binary quantum classification as constrained quantum state discrimination and introduce a locality-restricted distinguishability measure quantifying the maximum bias achievable by observables acting on at most $k$ subsystems. For $n$-qubit systems subject to independent depolarizing noise, the locally accessible signal is governed by a Pauli-weight-dependent contraction mechanism. This motivates a computable predictor, the $k$-local Pauli-accessible amplitude $A_{k}(p)$, which lower bounds the optimal $k$-local classification advantage. Numerical experiments on four-qubit encodings demonstrate quantitative agreement between empirical accuracy and the prediction across noise levels. The research identifies an operational breakdown threshold where $k$-local classifiers become indistinguishable from random guessing despite persistent global distinguishability.

Neural networks with positively homogeneous activations exhibit an exact continuous reparametrization symmetry: neuron-wise rescalings generate parameter-space orbits along which the input--output function is invariant. We interpret this symmetry as a gauge redundancy and introduce gauge-adapted coordinates that separate invariant and scale-imbalance directions. Inspired by gauge fixing in field theory, we introduce a soft orbit-selection (norm-balancing) functional acting only on redundant scale coordinates. We show analytically that it induces dissipative relaxation of imbalance modes to preserve the realized function. In controlled experiments, this orbit-selection penalty expands the stable learning-rate regime and suppresses scale drift without changing expressivity. These results establish a structural link between gauge-orbit geometry and optimization conditioning, providing a concrete connection between gauge-theoretic concepts and machine learning.

We systematically investigate the parameter-efficient fine-tuning design space under practical data and compute constraints, and propose D2-LoRA. D2-LoRA achieves 76.4 percent average accuracy across eight question answering and reading comprehension benchmarks using only 5k training samples per task and two epochs, while preserving algebraic mergeability at inference with near-exact numerical equivalence. The method combines signed low-rank residual updates with additive and subtractive components, together with a train-time column-wise projection that keeps each column close to its original norm. After training, the adapter is merged into a single weight matrix, adding zero inference latency. Compared with LoRA, D2-LoRA improves average accuracy by 2.2 percentage points; at matched parameter counts (LoRA rank 2r versus D2-LoRA rank r), the improvement is 1.6 points, indicating gains from architectural design rather than increased parameterization. Compared with DoRA, it matches or exceeds performance on most tasks. Beyond QA and reading comprehension, D2-LoRA improves generative tasks (plus 1.2 ROUGE-L and plus 1.1 percent win rate) and shows 36 percent lower training volatility. The merge preserves numerical fidelity (mean gap about 0.03 percentage points) and recovers about 1.91x evaluation throughput. Training overhead is 19 percent, comparable to DoRA, and decreases with longer input sequences. We provide a geometric analysis explaining how the projection stabilizes training, together with ablation studies isolating the contribution of each design component.

Autonomous control of double-Ackermann-steering robots is essential in agricultural applications, where robots must execute precise and complex maneuvers within a limited space. Classical methods, such as the Timed Elastic Band (TEB) planner, can address this problem, but they rely on parameter tuning, making them highly sensitive to changes in robot configuration or environment and impractical to deploy without constant recalibration. At the same time, end-to-end deep reinforcement learning (DRL) methods often fail due to unsuitable reward functions for non-holonomic constraints, resulting in sub-optimal policies and poor generalization. To address these challenges, this paper presents ManeuverNet, a DRL framework tailored for double-Ackermann systems, combining Soft Actor-Critic with CrossQ. Furthermore, ManeuverNet introduces four specifically designed reward functions to support maneuver learning. Unlike prior work, ManeuverNet does not depend on expert data or handcrafted guidance. We extensively evaluate ManeuverNet against both state-of-the-art DRL baselines and the TEB planner. Experimental results demonstrate that our framework substantially improves maneuverability and success rates, achieving more than a 40% gain over DRL baselines. Moreover, ManeuverNet effectively mitigates the strong parameter sensitivity observed in the TEB planner. In real-world trials, ManeuverNet achieved up to a 90% increase in maneuvering trajectory efficiency, highlighting its robustness and practical applicability.

Web agents require massive trajectories to generalize, yet real-world training is constrained by network latency, rate limits, and safety risks. We introduce \textbf{WebWorld} series, the first open-web simulator trained at scale. While existing simulators are restricted to closed environments with thousands of trajectories, WebWorld leverages a scalable data pipeline to train on 1M+ open-web interactions, supporting reasoning, multi-format data, and long-horizon simulations of 30+ steps. For intrinsic evaluation, we introduce WebWorld-Bench with dual metrics spanning nine dimensions, where WebWorld achieves simulation performance comparable to Gemini-3-Pro. For extrinsic evaluation, Qwen3-14B trained on WebWorld-synthesized trajectories improves by +9.2\% on WebArena, reaching performance comparable to GPT-4o. WebWorld enables effective inference-time search, outperforming GPT-5 as a world model. Beyond web simulation, WebWorld exhibits cross-domain generalization to code, GUI, and game environments, providing a replicable recipe for world model construction.

Conversational search (CS) requires a complex software engineering pipeline that integrates query reformulation, ranking, and response generation. CS researchers currently face two barriers: the lack of a unified framework for efficiently sharing contributions with the community, and the difficulty of deploying end-to-end prototypes needed for user evaluation. We introduce Orcheo, an open-source platform designed to bridge this gap. Orcheo offers three key advantages: (i) A modular architecture promotes component reuse through single-file node modules, facilitating sharing and reproducibility in CS research; (ii) Production-ready infrastructure bridges the prototype-to-system gap via dual execution modes, secure credential management, and execution telemetry, with built-in AI coding support that lowers the learning curve; (iii) Starter-kit assets include 50+ off-the-shelf components for query understanding, ranking, and response generation, enabling the rapid bootstrapping of complete CS pipelines. We describe the framework architecture and validate Orcheo's utility through case studies that highlight modularity and ease of use. Orcheo is released as open source under the MIT License at https://github.com/ShaojieJiang/orcheo.

Virtual machine placement is a crucial challenge in cloud computing for efficiently utilizing physical machine resources in data centers. Virtual machine placement can be formulated as a MinUsageTime Dynamic Vector Bin Packing (DVBP) problem, aiming to minimize the total usage time of the physical machines. This paper evaluates state-of-the-art MinUsageTime DVBP algorithms in non-clairvoyant, clairvoyant and learning-augmented online settings, where item durations (virtual machine lifetimes) are unknown, known and predicted, respectively. Besides the algorithms taken from the literature, we also develop several new algorithms or enhancements. Empirical experimentation is carried out with real-world datasets of Microsoft Azure. The insights from the experimental results are discussed to explore the structures of algorithms and promising design elements that work well in practice.

In this work we introduce methods to reduce the computational and memory costs of training deep neural networks. Our approach consists in replacing exact vector-jacobian products by randomized, unbiased approximations thereof during backpropagation. We provide a theoretical analysis of the trade-off between the number of epochs needed to achieve a target precision and the cost reduction for each epoch. We then identify specific unbiased estimates of vector-jacobian products for which we establish desirable optimality properties of minimal variance under sparsity constraints. Finally we provide in-depth experiments on multi-layer perceptrons, BagNets and Visual Transfomers architectures. These validate our theoretical results, and confirm the potential of our proposed unbiased randomized backpropagation approach for reducing the cost of deep learning.

This paper envisions a quantum database (Qute) that treats quantum computation as a first-class execution option. Unlike prior simulation-based methods that either run quantum algorithms on classical machines or adapt existing databases for quantum simulation, Qute instead (i) compiles an extended form of SQL into gate-efficient quantum circuits, (ii) employs a hybrid optimizer to dynamically select between quantum and classical execution plans, (iii) introduces selective quantum indexing, and (iv) designs fidelity-preserving storage to mitigate current qubit constraints. We also present a three-stage evolution roadmap toward quantum-native database. Finally, by deploying Qute on a real quantum processor (origin_wukong), we show that it outperforms a classical baseline at scale, and we release an open-source prototype at https://github.com/weAIDB/Qute.

Building agentic systems that can autonomously self-improve from experience is a longstanding goal of AI. Large language models (LLMs) today primarily self-improve via two mechanisms: self-reflection for context updates, and reinforcement learning (RL) for weight updates. In this work, we propose Evolutionary System Prompt Learning (E-SPL), a method for jointly improving model contexts and model weights. In each RL iteration, E-SPL selects multiple system prompts and runs rollouts with each in parallel. It applies RL updates to model weights conditioned on each system prompt, and evolutionary updates to the system prompt population via LLM-driven mutation and crossover. Each system prompt has a TrueSkill rating for evolutionary selection, updated from relative performance within each RL iteration batch. E-SPL encourages a natural division between declarative knowledge encoded in prompts and procedural knowledge encoded in weights, resulting in improved performance across reasoning and agentic tasks. For instance, in an easy-to-hard (AIME $\rightarrow$ BeyondAIME) generalization setting, E-SPL improves RL success rate from 38.8% $\rightarrow$ 45.1% while also outperforming reflective prompt evolution (40.0%). Overall, our results show that coupling reinforcement learning with system prompt evolution yields consistent gains in sample efficiency and generalization. Code: https://github.com/LunjunZhang/E-SPL

Instruction fine-tuning of large language models (LLMs) often involves selecting a subset of instruction training data from a large candidate pool, using a small query set from the target task. Despite growing interest, the literature on targeted instruction selection remains fragmented and opaque: methods vary widely in selection budgets, often omit zero-shot baselines, and frequently entangle the contributions of key components. As a result, practitioners lack actionable guidance on selecting instructions for their target tasks. In this work, we aim to bring clarity to this landscape by disentangling and systematically analyzing the two core ingredients: data representation and selection algorithms. Our framework enables controlled comparisons across models, tasks, and budgets. We find that only gradient-based data representations choose subsets whose similarity to the query consistently predicts performance across datasets and models. While no single method dominates, gradient-based representations paired with a greedy round-robin selection algorithm tend to perform best on average at low budgets, but these benefits diminish at larger budgets. Finally, we unify several existing selection algorithms as forms of approximate distance minimization between the selected subset and the query set, and support this view with new generalization bounds. More broadly, our findings provide critical insights and a foundation for more principled data selection in LLM fine-tuning. The code is available at https://github.com/dcml-lab/targeted-instruction-selection.

We present TokaMind, an open-source foundation model framework for fusion plasma modeling, based on a Multi-Modal Transformer (MMT) and trained on heterogeneous tokamak diagnostics from the publicly available MAST dataset. TokaMind supports multiple data modalities (time-series, 2D profiles, and videos) with different sampling rates, robust missing-signal handling, and efficient task adaptation via selectively loading and freezing four model components. To represent multi-modal signals, we use a training-free Discrete Cosine Transform embedding (DCT3D) and provide a clean interface for alternative embeddings (e.g., Variational Autoencoders - VAEs). We evaluate TokaMind on the recently introduced MAST benchmark TokaMark, comparing training and embedding strategies. Our results show that fine-tuned TokaMind outperforms the benchmark baseline on all but one task, and that, for several tasks, lightweight fine-tuning yields better performance than training the same architecture from scratch under a matched epoch budget. These findings highlight the benefits of multi-modal pretraining for tokamak plasma dynamics and provide a practical, extensible foundation for future fusion modeling tasks. Training code and model weights will be made publicly available.

Autonomous agents require some form of goal and plan recognition to interact in multiagent settings. Unfortunately, all existing goal recognition datasets suffer from a systematical bias induced by the planning systems that generated them, namely heuristic-based forward search. This means that existing datasets lack enough challenge for more realistic scenarios (e.g., agents using different planners), which impacts the evaluation of goal recognisers with respect to using different planners for the same goal. In this paper, we propose a new method that uses top-k planning to generate multiple, different, plans for the same goal hypothesis, yielding benchmarks that mitigate the bias found in the current dataset. This allows us to introduce a new metric called Version Coverage Score (VCS) to measure the resilience of the goal recogniser when inferring a goal based on different sets of plans. Our results show that the resilience of the current state-of-the-art goal recogniser degrades substantially under low observability settings.

Agentic AI coding tools with autonomous capabilities beyond conversational content generation increasingly automate repetitive and time-consuming software development tasks. Developers can configure these tools through versioned repository-level artifacts such as Markdown and JSON files. In this paper, we present a systematic analysis of configuration mechanisms for agentic AI coding tools, covering Claude Code, GitHub Copilot, Cursor, Gemini, and Codex. We identify eight configuration mechanisms and, in an empirical study of 2,926 GitHub repositories, examine whether and how they are adopted. We then explore Context Files, Skills, and Subagents, that is, three mechanisms available across tools, in more detail. Our findings reveal three trends. First, Context Files dominate the configuration landscape and are often the sole mechanism in a repository, with AGENTS$.$md emerging as an interoperable standard across tools. Second, advanced mechanisms such as Skills and Subagents are only shallowly adopted: most repositories define only one or two artifacts, and Skills predominantly rely on static instructions rather than executable workflows. Third, distinct configuration cultures are forming around different tools, with Claude Code users employing the broadest range of mechanisms. These findings establish an empirical baseline for longitudinal and experimental research on how configuration strategies evolve and affect agent performance as agentic AI coding tools mature.

As the capabilities of large language models continue to advance, so does their potential for misuse. While closed-source models typically rely on external defenses, open-weight models must primarily depend on internal safeguards to mitigate harmful behavior. Prior red-teaming research has largely focused on input-based jailbreaking and parameter-level manipulations. However, open-weight models also natively support prefilling, which allows an attacker to predefine initial response tokens before generation begins. Despite its potential, this attack vector has received little systematic attention. We present the largest empirical study to date of prefill attacks, evaluating over 20 existing and novel strategies across multiple model families and state-of-the-art open-weight models. Our results show that prefill attacks are consistently effective against all major contemporary open-weight models, revealing a critical and previously underexplored vulnerability with significant implications for deployment. While certain large reasoning models exhibit some robustness against generic prefilling, they remain vulnerable to tailored, model-specific strategies. Our findings underscore the urgent need for model developers to prioritize defenses against prefill attacks in open-weight LLMs.

Improving Sparse Autoencoders (SAEs) requires benchmarks that can precisely validate architectural innovations. However, current SAE benchmarks on LLMs are often too noisy to differentiate architectural improvements, and current synthetic data experiments are too small-scale and unrealistic to provide meaningful comparisons. We introduce SynthSAEBench, a toolkit for generating large-scale synthetic data with realistic feature characteristics including correlation, hierarchy, and superposition, and a standardized benchmark model, SynthSAEBench-16k, enabling direct comparison of SAE architectures. Our benchmark reproduces several previously observed LLM SAE phenomena, including the disconnect between reconstruction and latent quality metrics, poor SAE probing results, and a precision-recall trade-off mediated by L0. We further use our benchmark to identify a new failure mode: Matching Pursuit SAEs exploit superposition noise to improve reconstruction without learning ground-truth features, suggesting that more expressive encoders can easily overfit. SynthSAEBench complements LLM benchmarks by providing ground-truth features and controlled ablations, enabling researchers to precisely diagnose SAE failure modes and validate architectural improvements before scaling to LLMs.

Deep generative models have achieved great success in producing high-quality samples, making them a central tool across machine learning applications. Beyond sample quality, an important yet less systematically studied question is whether trained generative models faithfully capture the diversity of the underlying data distribution. In this work, we address this question by directly comparing the diversity of samples generated by state-of-the-art models with that of test samples drawn from the target data distribution, using recently proposed reference-free entropy-based diversity scores, Vendi and RKE. Across multiple benchmark datasets, we find that test data consistently attains substantially higher Vendi and RKE diversity scores than the generated samples, suggesting a systematic downward diversity bias in modern generative models. To understand the origin of this bias, we analyze the finite-sample behavior of entropy-based diversity scores and show that their expected values increase with sample size, implying that diversity estimated from finite training sets could inherently underestimate the diversity of the true distribution. As a result, optimizing the generators to minimize divergence to empirical data distributions would induce a loss of diversity. Finally, we discuss potential diversity-aware regularization and guidance strategies based on Vendi and RKE as principled directions for mitigating this bias, and provide empirical evidence suggesting their potential to improve the results.

LLM-powered Multi-Agent Systems (MAS) have emerged as an effective approach towards collaborative intelligence, and have attracted wide research interests. Among them, ``self-evolving'' MAS, treated as a more flexible and powerful technical route, can construct task-adaptive workflows or communication topologies, instead of relying on a predefined static structue template. Current self-evolving MAS mainly focus on Spatial Evolving or Temporal Evolving paradigm, which only considers the single dimension of evolution and does not fully incentivize LLMs' collaborative capability. In this work, we start from a novel Spatio-Temporal perspective by proposing ST-EVO, which supports dialogue-wise communication scheduling with a compact yet powerful flow-matching based Scheduler. To make precise Spatio-Temporal scheduling, ST-EVO can also perceive the uncertainty of MAS, and possesses self-feedback ability to learn from accumulated experience. Extensive experiments on nine benchmarks demonstrate the state-of-the-art performance of ST-EVO, achieving about 5%--25% accuracy improvement.

Finding optimal measurement operators is crucial for the performance of quantum reservoir computers (QRCs), since they employ a fixed quantum feature map. We formulate the training of both stateless (quantum extreme learning machines, QELMs) and stateful (memory dependent) QRCs in the framework of kernel ridge regression. This approach renders an optimal measurement operator that minimizes prediction error for a given reservoir and training dataset. For large qubit numbers, this method is more efficient than the conventional training of QRCs. We discuss efficiency and practical implementation strategies, including Pauli basis decomposition and operator diagonalization, to adapt the optimal observable to hardware constraints. Numerical experiments on image classification and time series prediction tasks demonstrate the effectiveness of this approach, which can also be applied to other quantum ML models.

Emergency situations that require the evacuation of urban areas can arise from man-made causes (e.g., terrorist attacks or industrial accidents) or natural disasters, the latter becoming more frequent due to climate change. As a result, effective and fast methods to develop evacuation plans are of great importance. In this work, we identify and propose the Bus Evacuation Orienteering Problem (BEOP), an NP-hard combinatorial optimization problem with the goal of evacuating as many people from an affected area by bus in a short, predefined amount of time. The purpose of bus-based evacuation is to reduce congestion and disorder that arises in purely car-focused evacuation scenarios. To solve the BEOP, we propose a deep reinforcement learning-based method utilizing graph learning, which, once trained, achieves fast inference speed and is able to create evacuation routes in fractions of seconds. We can bound the gap of our evacuation plans using an MILP formulation. To validate our method, we create evacuation scenarios for San Francisco using real-world road networks and travel times. We show that we achieve near-optimal solution quality and are further able to investigate how many evacuation vehicles are necessary to achieve certain bus-based evacuation quotas given a predefined evacuation time while keeping run time adequate.

We present a crowdsourced dataset for Piedmontese, an endangered Romance language of northwestern Italy. The dataset comprises 145 Italian-Piedmontese parallel sentences derived from Flores+, with translations produced by speakers writing in their natural orthographic style rather than adhering to standardized conventions, along with manual word alignment. We use this resource to benchmark several large language models on tokenization parity, topic classification, and machine translation. Our analysis reveals that Piedmontese incurs a tokenization penalty relative to higher-resource Romance languages, yet LLMs achieve classification performance approaching that of Italian, French, and English. Machine translation results are asymmetric: models translate adequately from Piedmontese into high-resource languages, but generation into Piedmontese remains challenging. The dataset and code are publicly released.

Gradual argumentation is a field of symbolic AI which is attracting attention for its ability to support transparent and contestable AI systems. It is considered a useful tool in domains such as decision-making, recommendation, debate analysis, and others. The outcomes in such domains are usually dependent on the arguments' base scores, which must be selected carefully. Often, this selection process requires user expertise and may not always be straightforward. On the other hand, organising the arguments by preference could simplify the task. In this work, we introduce \emph{Base Score Extraction Functions}, which provide a mapping from users' preferences over arguments to base scores. These functions can be applied to the arguments of a \emph{Bipolar Argumentation Framework} (BAF), supplemented with preferences, to obtain a \emph{Quantitative Bipolar Argumentation Framework} (QBAF), allowing the use of well-established computational tools in gradual argumentation. We outline the desirable properties of base score extraction functions, discuss some design choices, and provide an algorithm for base score extraction. Our method incorporates an approximation of non-linearities in human preferences to allow for better approximation of the real ones. Finally, we evaluate our approach both theoretically and experimentally in a robotics setting, and offer recommendations for selecting appropriate gradual semantics in practice.

Formulaic alpha factor mining is a critical yet challenging task in quantitative investment, characterized by a vast search space and the need for domain-informed, interpretable signals. However, finding novel signals becomes increasingly difficult as the library grows due to high redundancy. We propose FactorMiner, a lightweight and flexible self-evolving agent framework designed to navigate this complex landscape through continuous knowledge accumulation. FactorMiner combines a Modular Skill Architecture that encapsulates systematic financial evaluation into executable tools with a structured Experience Memory that distills historical mining trials into actionable insights (successful patterns and failure constraints). By instantiating the Ralph Loop paradigm -- retrieve, generate, evaluate, and distill -- FactorMiner iteratively uses memory priors to guide exploration, reducing redundant search while focusing on promising directions. Experiments on multiple datasets across different assets and Markets show that FactorMiner constructs a diverse library of high-quality factors with competitive performance, while maintaining low redundancy among factors as the library scales. Overall, FactorMiner provides a practical approach to scalable discovery of interpretable formulaic alpha factors under the "Correlation Red Sea" constraint.

We present pseudo-differential enhanced physics-informed neural networks (PINNs), an extension of gradient enhancement but in Fourier space. Gradient enhancement of PINNs dictates that the PDE residual is taken to a higher differential order than prescribed by the PDE, added to the objective as an augmented term in order to improve training and overall learning fidelity. We propose the same procedure after application via Fourier transforms, since differentiating in Fourier space is multiplication with the Fourier wavenumber under suitable decay. Our methods are fast and efficient. Our methods oftentimes achieve superior PINN versus numerical error in fewer training iterations, potentially pair well with few samples in collocation, and can on occasion break plateaus in low collocation settings. Moreover, our methods are suitable for fractional derivatives. We establish that our methods improve spectral eigenvalue decay of the neural tangent kernel (NTK), and so our methods contribute towards the learning of high frequencies in early training, mitigating the effects of frequency bias up to the polynomial order and possibly greater with smooth activations. Our methods accommodate advanced techniques in PINNs, such as Fourier feature embeddings. A pitfall of discrete Fourier transforms via the Fast Fourier Transform (FFT) is mesh subjugation, and so we demonstrate compatibility of our methods for greater mesh flexibility and invariance on alternative Euclidean and non-Euclidean domains via Monte Carlo methods and otherwise.

Orthogonality constraints are ubiquitous in robust and probabilistic machine learning. Unfortunately, current optimizers are computationally expensive and do not scale to problems with hundreds or thousands of constraints. One notable exception is the Landing algorithm (Ablin et al., 2024) which, however comes at the expense of temporarily relaxing orthogonality. In this work, we revisit and improve on the ideas behind Landing, enabling the inclusion of modern adaptive optimizers while ensuring that orthogonal constraints are effectively met. Remarkably, these improvements come at little to no cost, and reduce the number of required hyperparemeters. Our algorithm POGO is fast and GPU-friendly, consisting of only 5 matrix products, and in practice maintains orthogonality at all times. On several challenging benchmarks, POGO greatly outperforms recent optimizers and shows it can optimize problems with thousands of orthogonal matrices in minutes while alternatives would take hours. As such, POGO sets a milestone to finally exploit orthogonality constraints in ML at scale. A PyTorch implementation of POGO is publicly available at https://github.com/adrianjav/pogo.

Early diagnosis of Alzheimer's Disease (AD) is crucial for delaying its progression. While AI-based speech detection is non-invasive and cost-effective, it faces a critical data efficiency dilemma due to medical data scarcity and privacy barriers. Therefore, we propose FAL-AD, a novel framework that synergistically integrates federated learning with data augmentation to systematically optimize data efficiency. Our approach delivers three key breakthroughs: First, absolute efficiency improvement through voice conversion-based augmentation, which generates diverse pathological speech samples via cross-category voice-content recombination. Second, collaborative efficiency breakthrough via an adaptive federated learning paradigm, maximizing cross-institutional benefits under privacy constraints. Finally, representational efficiency optimization by an attentive cross-modal fusion model, which achieves fine-grained word-level alignment and acoustic-textual interaction. Evaluated on ADReSSo, FAL-AD achieves a state-of-the-art multi-modal accuracy of 91.52%, outperforming all centralized baselines and demonstrating a practical solution to the data efficiency dilemma. Our source code is publicly available at https://github.com/smileix/fal-ad.

The Uniform Information Density (UID) hypothesis posits that speakers are subject to a communicative pressure to distribute information evenly within utterances, minimising surprisal variance. While this hypothesis has been tested empirically, prior studies are limited exclusively to text-only inputs, abstracting away from the perceptual context in which utterances are produced. In this work, we present the first computational study of UID in visually grounded settings. We estimate surprisal using multilingual vision-and-language models over image-caption data in 30 languages and visual storytelling data in 13 languages, together spanning 11 families. We find that grounding on perception consistently smooths the distribution of information, increasing both global and local uniformity across typologically diverse languages compared to text-only settings. In visual narratives, grounding in both image and discourse contexts has additional effects, with the strongest surprisal reductions occurring at the onset of discourse units. Overall, this study takes a first step towards modelling the temporal dynamics of information flow in ecologically plausible, multimodal language use, and finds that grounded language exhibits greater information uniformity, supporting a context-sensitive formulation of UID.

Large Language Models (LLMs) exhibit strong reasoning abilities, but their high computational costs limit their practical deployment. Recent studies reveal significant redundancy in LLMs layers, making layer pruning an active research topic. Layer pruning research primarily focuses on two aspects: measuring layer importance and recovering performance after pruning. Unfortunately, the present works fail to simultaneously maintain pruning performance and efficiency. In this study, we propose GradMAP, a faster layer pruning method with \textbf{Grad}ient \textbf{M}etric \textbf{A}nd \textbf{P}rojection compensation, which consists of two stages. In the first stage, we introduce a novel metric based on gradient magnitudes, enabling a global assessment of layer importance. Note that, it requires only a single backward propagation step per pruning decision, substantially enhancing pruning efficiency. In the second stage, we first analyze the layers with the largest mean shift resulting from pruning, and then incorporate a simple yet effective projection compensation matrix to correct this drift in one step. In this way, the degradation of model performance caused by layer pruning is effectively alleviated. Extensive experiments show that GradMAP outperforms previous layer pruning methods in both pruning speed (achieving an average $4\times$ speedup) and performance.

Large language models struggle to maintain strict adherence to structured workflows in high-stakes domains such as healthcare triage. Monolithic approaches that encode entire decision structures within a single prompt are prone to instruction-following degradation as prompt length increases, including lost-in-the-middle effects and context window overflow. To address this gap, we present Arbor, a framework that decomposes decision tree navigation into specialized, node-level tasks. Decision trees are standardized into an edge-list representation and stored for dynamic retrieval. At runtime, a directed acyclic graph (DAG)-based orchestration mechanism iteratively retrieves only the outgoing edges of the current node, evaluates valid transitions via a dedicated LLM call, and delegates response generation to a separate inference step. The framework is agnostic to the underlying decision logic and model provider. Evaluated against single-prompt baselines across 10 foundation models using annotated turns from real clinical triage conversations. Arbor improves mean turn accuracy by 29.4 percentage points, reduces per-turn latency by 57.1%, and achieves an average 14.4x reduction in per-turn cost. These results indicate that architectural decomposition reduces dependence on intrinsic model capability, enabling smaller models to match or exceed larger models operating under single-prompt baselines.

Inverse problems and inverse design in multiphase media, i.e., recovering or engineering microstructures to achieve target macroscopic responses, require operating on discrete-valued material fields, rendering the problem non-differentiable and incompatible with gradient-based methods. Existing approaches either relax to continuous approximations, compromising physical fidelity, or employ separate heavyweight models for forward and inverse tasks. We propose GenPANIS, a unified generative framework that preserves exact discrete microstructures while enabling gradient-based inference through continuous latent embeddings. The model learns a joint distribution over microstructures and PDE solutions, supporting bidirectional inference (forward prediction and inverse recovery) within a single architecture. The generative formulation enables training with unlabeled data, physics residuals, and minimal labeled pairs. A physics-aware decoder incorporating a differentiable coarse-grained PDE solver preserves governing equation structure, enabling extrapolation to varying boundary conditions and microstructural statistics. A learnable normalizing flow prior captures complex posterior structure for inverse problems. Demonstrated on Darcy flow and Helmholtz equations, GenPANIS maintains accuracy on challenging extrapolative scenarios - including unseen boundary conditions, volume fractions, and microstructural morphologies, with sparse, noisy observations. It outperforms state-of-the-art methods while using 10 - 100 times fewer parameters and providing principled uncertainty quantification.

Biomarker-based prediction of clinical outcomes is challenging due to nonlinear relationships, correlated features, and the limited size of many medical datasets. Classical machine-learning methods can struggle under these conditions, motivating the search for alternatives. In this work, we investigate quantum reservoir computing (QRC), using both noiseless emulation and hardware execution on the neutral-atom Rydberg processor \textit{Aquila}. We evaluate performance with six classical machine-learning models and use SHAP to generate feature subsets. We find that models trained on emulated quantum features achieve mean test accuracies comparable to those trained on classical features, but have higher training accuracies and greater variability over data splits, consistent with overfitting. When comparing hardware execution of QRC to noiseless emulation, the models are more robust over different data splits and often exhibit statistically significant improvements in mean test accuracy. This combination of improved accuracy and increased stability is suggestive of a regularising effect induced by hardware execution. To investigate the origin of this behaviour, we examine the statistical differences between hardware and emulated quantum feature distributions. We find that hardware execution applies a structured, time-dependent transformation characterised by compression toward the mean and a progressive reduction in mutual information relative to emulation.

Compressed Deep Learning (DL) models are essential for deployment in resource-constrained environments. But their performance often lags behind their large-scale counterparts. To bridge this gap, we propose Alignment Adapter (AlAd): a lightweight, sliding-window-based adapter. It aligns the token-level embeddings of a compressed model with those of the original large model. AlAd preserves local contextual semantics, enables flexible alignment across differing dimensionalities or architectures, and is entirely agnostic to the underlying compression method. AlAd can be deployed in two ways: as a plug-and-play module over a frozen compressed model, or by jointly fine-tuning AlAd with the compressed model for further performance gains. Through experiments on BERT-family models across three token-level NLP tasks, we demonstrate that AlAd significantly boosts the performance of compressed models with only marginal overhead in size and latency.

Concept Bottleneck Models (CBMs) aim to deliver interpretable predictions by routing decisions through a human-understandable concept layer, yet they often suffer reduced accuracy and concept leakage that undermines faithfulness. We introduce an explicit Information Bottleneck regularizer on the concept layer that penalizes $I(X;C)$ while preserving task-relevant information in $I(C;Y)$, encouraging minimal-sufficient concept representations. We derive two practical variants (a variational objective and an entropy-based surrogate) and integrate them into standard CBM training without architectural changes or additional supervision. Evaluated across six CBM families and three benchmarks, the IB-regularized models consistently outperform their vanilla counterparts. Information-plane analyses further corroborate the intended behavior. These results indicate that enforcing a minimal-sufficient concept bottleneck improves both predictive performance and the reliability of concept-level interventions. The proposed regularizer offers a theoretic-grounded, architecture-agnostic path to more faithful and intervenable CBMs, resolving prior evaluation inconsistencies by aligning training protocols and demonstrating robust gains across model families and datasets.

Neural audio compression models have recently achieved extreme compression rates, enabling efficient latent generative modeling. Conversely, latent generative models have been applied to compression, pushing the limits of continuous and discrete approaches. However, existing methods remain constrained to low-resolution audio and degrade substantially at very low bitrates, where audible artifacts are prominent. In this paper, we present S-PRESSO, a 48kHz sound effect compression model that produces both continuous and discrete embeddings at ultra-low bitrates, down to 0.096 kbps, via offline quantization. Our model relies on a pretrained latent diffusion model to decode compressed audio embeddings learned by a latent encoder. Leveraging the generative priors of the diffusion decoder, we achieve extremely low frame rates, down to 1Hz (750x compression rate), producing convincing and realistic reconstructions at the cost of exact fidelity. Despite operating at high compression rates, we demonstrate that S-PRESSO outperforms both continuous and discrete baselines in audio quality, acoustic similarity and reconstruction metrics.

Association Rule Mining (ARM) is a fundamental task for knowledge discovery in tabular data and is widely used in high-stakes decision-making. Classical ARM methods rely on frequent itemset mining, leading to rule explosion and poor scalability, while recent neural approaches mitigate these issues but suffer from degraded performance in low-data regimes. Tabular foundation models (TFMs), pretrained on diverse tabular data with strong in-context generalization, provide a basis for addressing these limitations. We introduce a model-agnostic association rule learning framework that extracts association rules from any conditional probabilistic model over tabular data, enabling us to leverage TFMs. We then introduce TabProbe, an instantiation of our framework that utilizes TFMs as conditional probability estimators to learn association rules out-of-the-box without frequent itemset mining. We evaluate our approach on tabular datasets of varying sizes based on standard ARM rule quality metrics and downstream classification performance. The results show that TFMs consistently produce concise, high-quality association rules with strong predictive performance and remain robust in low-data settings without task-specific training. Source code is available at https://github.com/DiTEC-project/tabprobe.

Vision Transformers (ViTs) have emerged as the state-of-the-art architecture in representation learning, leveraging self-attention mechanisms to excel in various tasks. ViTs split images into fixed-size patches, constraining them to a predefined size and necessitating pre-processing steps like resizing, padding, or cropping. This poses challenges in medical imaging, particularly with irregularly shaped structures like tumors. A fixed bounding box crop size produces input images with highly variable foreground-to-background ratios. Resizing medical images can degrade information and introduce artefacts, impacting diagnosis. Hence, tailoring variable-sized crops to regions of interest can enhance feature representation capabilities. Moreover, large images are computationally expensive, and smaller sizes risk information loss, presenting a computation-accuracy tradeoff. We propose VariViT, an improved ViT model crafted to handle variable image sizes while maintaining a consistent patch size. VariViT employs a novel positional embedding resizing scheme for a variable number of patches. We also implement a new batching strategy within VariViT to reduce computational complexity, resulting in faster training and inference times. In our evaluations on two 3D brain MRI datasets, VariViT surpasses vanilla ViTs and ResNet in glioma genotype prediction and brain tumor classification. It achieves F1-scores of 75.5% and 76.3%, respectively, learning more discriminative features. Our proposed batching strategy reduces computation time by up to 30% compared to conventional architectures. These findings underscore the efficacy of VariViT in image representation learning. Our code can be found here: https://github.com/Aswathi-Varma/varivit

Long-duration audio is increasingly common in industrial and consumer settings, yet reviewing multi-hour recordings is impractical, motivating systems that answer natural-language queries with precise temporal grounding and minimal hallucination. Existing audio-language models show promise, but long-audio question answering remains difficult due to context-length limits. We introduce LongAudio-RAG (LA-RAG), a hybrid framework that grounds Large Language Model (LLM) outputs in retrieved, timestamped acoustic event detections rather than raw audio. Multi-hour streams are converted into structured event records stored in an SQL database, and at inference time the system resolves natural-language time references, classifies intent, retrieves only the relevant events, and generates answers using this constrained evidence. To evaluate performance, we construct a synthetic long-audio benchmark by concatenating recordings with preserved timestamps and generating template-based question-answer pairs for detection, counting, and summarization tasks. Finally, we demonstrate the practicality of our approach by deploying it in a hybrid edge-cloud environment, where the audio grounding model runs on-device on IoT-class hardware while the LLM is hosted on a GPU-backed server. This architecture enables low-latency event extraction at the edge and high-quality language reasoning in the cloud. Experiments show that structured, event-level retrieval significantly improves accuracy compared to vanilla Retrieval-Augmented Generation (RAG) or text-to-SQL approaches.

In the pull-based development model, code contributions are submitted as pull requests (PRs) to undergo reviews and approval by other developers with the goal of being merged into the code base. A PR can be supported by a description, whose role has not yet been systematically investigated. To fill in this gap, we conducted a mixed-methods empirical study of PR descriptions. We conducted a grey literature review of guidelines on writing PR descriptions and derived a taxonomy of eight recommended elements. Using this taxonomy, we analyzed 80K GitHub PRs across 156 projects and five programming languages to assess associations between these elements and code review outcomes (e.g., merge decision, latency, first response time, review comments, and review iteration cycles). To complement these results, we surveyed 64 developers about the perceived importance of each element. Finally, we analyzed which submission-time factors predict whether PRs include a description and which elements they contain. We found that developers view PR descriptions as important, but their elements matter differently: purpose and code explanations are valued by developers for preserving the rationale and history of changes, while stating the desired feedback type best predicts change acceptance and reviewer engagement. PR descriptions are also more common in mature projects and complex changes, suggesting they are written when most useful rather than as a formality.

Low-discrepancy designs play a central role in quasi-Monte Carlo methods and are increasingly influential in other domains such as machine learning, robotics and computer graphics, to name a few. In recent years, one such low-discrepancy construction method called subset selection has received a lot of attention. Given a large population, one optimally selects a small low-discrepancy subset with respect to a discrepancy-based objective. Versions of this problem are known to be NP-hard. In this text, we establish, for the first time, that the subset selection problem with respect to kernel discrepancies is also NP-hard. Motivated by this intractability, we propose a Bayesian Optimization procedure for the subset selection problem utilizing the recent notion of deep embedding kernels. We demonstrate the performance of the BO algorithm to minimize discrepancy measures and note that the framework is broadly applicable any design criteria.

Autonomous agentic systems are increasingly deployed in regulated, high-stakes domains where decisions may be irreversible and institutionally constrained. Existing safety approaches emphasize alignment, interpretability, or action-level filtering. We argue that these mechanisms are necessary but insufficient because they do not directly govern selection power: the authority to determine which options are generated, surfaced, and framed for decision. We propose a governance architecture that separates cognition, selection, and action into distinct domains and models autonomy as a vector of sovereignty. Cognitive autonomy remains unconstrained, while selection and action autonomy are bounded through mechanically enforced primitives operating outside the agent's optimization space. The architecture integrates external candidate generation (CEFL), a governed reducer, commit-reveal entropy isolation, rationale validation, and fail-loud circuit breakers. We evaluate the system across multiple regulated financial scenarios under adversarial stress targeting variance manipulation, threshold gaming, framing skew, ordering effects, and entropy probing. Metrics quantify selection concentration, narrative diversity, governance activation cost, and failure visibility. Results show that mechanical selection governance is implementable, auditable, and prevents deterministic outcome capture while preserving reasoning capacity. Although probabilistic concentration remains, the architecture measurably bounds selection authority relative to conventional scalar pipelines. This work reframes governance as bounded causal power rather than internal intent alignment, offering a foundation for deploying autonomous agents where silent failure is unacceptable.

The opioid epidemic continues to ravage communities worldwide, straining healthcare systems, disrupting families, and demanding urgent computational solutions. To combat this lethal opioid crisis, graph learning methods have emerged as a promising paradigm for modeling complex drug-related phenomena. However, a significant gap remains: there is no comprehensive benchmark for systematically evaluating these methods across real-world opioid crisis scenarios. To bridge this gap, we introduce OPBench, the first comprehensive opioid benchmark comprising five datasets across three critical application domains: opioid overdose detection from healthcare claims, illicit drug trafficking detection from digital platforms, and drug misuse prediction from dietary patterns. Specifically, OPBench incorporates diverse graph structures, including heterogeneous graphs and hypergraphs, to preserve the rich and complex relational information among drug-related data. To address data scarcity, we collaborate with domain experts and authoritative institutions to curate and annotate datasets while adhering to privacy and ethical guidelines. Furthermore, we establish a unified evaluation framework with standardized protocols, predefined data splits, and reproducible baselines to facilitate fair and systematic comparison among graph learning methods. Through extensive experiments, we analyze the strengths and limitations of existing graph learning methods, thereby providing actionable insights for future research in combating the opioid crisis. Our source code and datasets are available at https://github.com/Tianyi-Billy-Ma/OPBench.

Automated Code Revision (ACR) tools aim to reduce manual effort by automatically generating code revisions based on reviewer feedback. While ACR tools have shown promising performance on historical data, their real-world utility depends on their ability to handle similar code variants expressing the same issue - a property we define as consistency. However, the probabilistic nature of ACR tools often compromises consistency, which may lead to divergent revisions even for semantically equivalent code variants. In this paper, we investigate the extent to which ACR tools maintain consistency when presented with semantically equivalent code variants. To do so, we first designed nine types of semantics-preserving perturbations (SPP) and applied them to 2032 Java methods from real-world GitHub projects, generating over 10K perturbed variants for evaluation. Then we used these perturbations to evaluate the consistency of five state-of-the-art transformer-based ACR tools. We found that the ACR tools' ability to generate correct revisions can drop by up to 45.3%, when presented with semantically equivalent code. The closer the perturbation is to this targeted region, the more likely an ACR tool is to fail to generate the correct revision. We explored potential mitigation strategies that modify the input representation, but found that these attention-guiding heuristics yielded only marginal improvements, thus leaving the solution to this problem as an open research question.

We present the Wikidata Query Logs (WDQL) dataset, a dataset consisting of 200k question-query pairs over the Wikidata knowledge graph. It is over 6x larger than the largest existing Wikidata datasets of similar format without relying on template-generated queries. Instead, we construct it using real-world SPARQL queries sent to the Wikidata Query Service and generate questions for them. Since these log-based queries are anonymized, and therefore often do not produce results, a significant amount of effort is needed to convert them back into meaningful SPARQL queries. To achieve this, we present an agent-based method that iteratively de-anonymizes, cleans, and verifies queries against Wikidata while also generating corresponding natural-language questions. We demonstrate the dataset's benefit for training question-answering methods. All WDQL assets, as well as the agent code, are publicly available under a permissive license.

This paper presents an automated method for classifying source code changes during the software development process based on clustering of change metrics. The method consists of two steps: clustering of metric vectors computed for each code change, followed by expert mapping of the resulting clusters to predefined change classes. The distribution of changes into clusters is performed automatically, while the mapping of clusters to classes is carried out by an expert. Automation of the distribution step substantially reduces the time required for code change review. The k-means algorithm with a cosine similarity measure between metric vectors is used for clustering. Eleven source code metrics are employed, covering lines of code, cyclomatic complexity, file counts, interface changes, and structural changes. The method was validated on five software systems, including two open-source projects (Subversion and NHibernate), and demonstrated classification purity of P_C = 0.75 +/- 0.05 and entropy of E_C = 0.37 +/- 0.06 at a significance level of 0.05.

AI agents need to plan to achieve complex goals that involve orchestrating perception, sub-goal decomposition, and execution. These plans consist of ordered steps structured according to a Temporal Execution Order (TEO, a directed acyclic graph that ensures each step executes only after its preconditions are satisfied. Existing research on foundational models' understanding of temporal execution is limited to automatically derived annotations, approximations of the TEO as a linear chain, or text-only inputs. To address this gap, we introduce MATEO (MultimodAl Temporal Execution Order), a benchmark designed to assess and improve the temporal reasoning abilities of Large Vision Language Models (LVLMs) required for real-world planning. We acquire a high-quality professional multimodal recipe corpus, authored through a standardized editorial process that decomposes instructions into discrete steps, each paired with corresponding images. We collect TEO annotations as graphs by designing and using a scalable crowdsourcing pipeline. Using MATEO, we evaluate six state-of-the-art LVLMs across model scales, varying language context, multimodal input structure, and fine-tuning strategies.

Many real-world control problems, ranging from finance to robotics, evolve in continuous time with non-uniform, event-driven decisions. Standard discrete-time reinforcement learning (RL), based on fixed-step Bellman updates, struggles in this setting: as time gaps shrink, the $Q$-function collapses to the value function $V$, eliminating action ranking. Existing continuous-time methods reintroduce action information via an advantage-rate function $q$. However, they enforce optimality through complicated martingale losses or orthogonality constraints, which are sensitive to the choice of test processes. These approaches entangle $V$ and $q$ into a large, complex optimization problem that is difficult to train reliably. To address these limitations, we propose a novel decoupled continuous-time actor-critic algorithm with alternating updates: $q$ is learned from diffusion generators on $V$, and $V$ is updated via a Hamiltonian-based value flow that remains informative under infinitesimal time steps, where standard max/softmax backups fail. Theoretically, we prove rigorous convergence via new probabilistic arguments, sidestepping the challenge that generator-based Hamiltonians lack Bellman-style contraction under the sup-norm. Empirically, our method outperforms prior continuous-time and leading discrete-time baselines across challenging continuous-control benchmarks and a real-world trading task, achieving 21% profit over a single quarter$-$nearly doubling the second-best method.

Algorithmic \emph{replicability} has recently been introduced to address the need for reproducible experiments in machine learning. A \emph{replicable online learning} algorithm is one that takes the same sequence of decisions across different executions in the same environment, with high probability. We initiate the study of algorithmic replicability in \emph{constrained} MAB problems, where a learner interacts with an unknown stochastic environment for $T$ rounds, seeking not only to maximize reward but also to satisfy multiple constraints. Our main result is that replicability can be achieved in constrained MABs. Specifically, we design replicable algorithms whose regret and constraint violation match those of non-replicable ones in terms of $T$. As a key step toward these guarantees, we develop the first replicable UCB-like algorithm for \emph{unconstrained} MABs, showing that algorithms that employ the optimism in-the-face-of-uncertainty principle can be replicable, a result that we believe is of independent interest.

Sparsity is a well-studied technique for compressing deep neural networks (DNNs) without compromising performance. In deep reinforcement learning (DRL), neural networks with up to 5% of their original weights can still be trained with minimal performance loss compared to their dense counterparts. However, most existing methods rely on unstructured fine-grained sparsity, which limits hardware acceleration opportunities due to irregular computation patterns. Structured coarse-grained sparsity enables hardware acceleration, yet typically degrades performance and increases pruning complexity. In this work, we present, to the best of our knowledge, the first study on N:M structured sparsity in RL, which balances compression, performance, and hardware efficiency. Our framework enforces row-wise N:M sparsity throughout training for all networks in off-policy RL (TD3), maintaining compatibility with accelerators that support N:M sparse matrix operations. Experiments on continuous-control benchmarks show that RNM-TD3, our N:M sparse agent, outperforms its dense counterpart at 50%-75% sparsity (e.g., 2:4 and 1:4), achieving up to a 14% increase in performance at 2:4 sparsity on the Ant environment. RNM-TD3 remains competitive even at 87.5% sparsity (1:8), while enabling potential training speedups.

CI/CD practices play a significant role during collaborative software development by automating time-consuming and repetitive tasks such as testing, building, quality checking, dependency and security management. GitHub Actions, the CI/CD tool integrated into GitHub, allows repository maintainers to automate development workflows. We conducted a mixed methods analysis of GitHub Actions workflow changes over time. Through a preliminary qualitative analysis of 439 modified workflow files we identified seven types of conceptual changes to workflows. Next, we performed a quantitative analysis over 49K+ GitHub repositories totaling 267K+ workflow change histories and 3.4M+ workflow file versions from November 2019 to August 2025. This analysis revealed that repositories contain a median of three workflow files, and 7.3% of all workflow files are being changed every week. The changes made to workflows tend to be small, with about three-quarters containing only a single change. The large majority of the observed changes have to do with task configuration and task specification in workflow jobs. We did not find any conclusive evidence of the effect of LLM coding tools or other major technological changes on workflow creation and workflow maintenance frequency. Our findings highlight the need for improved tooling to support fine-grained maintenance tasks, such as a broader adoption of dependency management and AI-based support for ensuring and sustaining workflow security and quality.

We introduce a Monte Carlo (MC) dataset of single- and two-track drift chamber events to advance Machine Learning (ML)-based track reconstruction. To enable standardized and comparable evaluation, we define track reconstruction specific metrics and report results for traditional track reconstruction algorithms and a Graph Neural Networks (GNNs) method, facilitating rigorous, reproducible validation for future research.

Recently, Large Language Models (LLMs) have gained significant traction in medical domain, especially in developing a QA systems to Medical QA systems for enhancing access to healthcare in low-resourced settings. This paper compares five LLMs deployed between April 2024 and August 2025 for medical QA, using the iCliniq dataset, containing 38,000 medical questions and answers of diverse specialties. Our models include Llama-3-8B-Instruct, Llama 3.2 3B, Llama 3.3 70B Instruct, Llama-4-Maverick-17B-128E-Instruct, and GPT-5-mini. We are using a zero-shot evaluation methodology and using BLEU and ROUGE metrics to evaluate performance without specialized fine-tuning. Our results show that larger models like Llama 3.3 70B Instruct outperform smaller models, consistent with observed scaling benefits in clinical tasks. It is notable that, Llama-4-Maverick-17B exhibited more competitive results, thus highlighting evasion efficiency trade-offs relevant for practical deployment. These findings align with advancements in LLM capabilities toward professional-level medical reasoning and reflect the increasing feasibility of LLM-supported QA systems in the real clinical environments. This benchmark aims to serve as a standardized setting for future study to minimize model size, computational resources and to maximize clinical utility in medical NLP applications.

The primary focus of multi-agent reinforcement learning (MARL) has been to study interactions among a fixed number of agents embedded in an environment. However, in the real world, the number of agents is neither fixed nor known a priori. Moreover, an agent can decide to create other agents (for example, a cell may divide, or a company may spin off a division). In this paper, we propose a framework that allows agents to create other agents; we call this a fluid-agent environment. We present game-theoretic solution concepts for fluid-agent games and empirically evaluate the performance of several MARL algorithms within this framework. Our experiments include fluid variants of established benchmarks such as Predator-Prey and Level-Based Foraging, where agents can dynamically spawn, as well as a new environment we introduce that highlights how fluidity can unlock novel solution strategies beyond those observed in fixed-population settings. We demonstrate that this framework yields agent teams that adjust their size dynamically to match environmental demands.

Despite legal mandates for the right to be forgotten, AI operators routinely fail to comply with data deletion requests. While machine unlearning (MU) provides a technical solution to remove personal data's influence from trained models, ensuring compliance remains challenging due to the fundamental gap between MU's technical feasibility and regulatory implementation. In this paper, we introduce the first economic framework for auditing MU compliance, by integrating certified unlearning theory with regulatory enforcement. We first characterize MU's inherent verification uncertainty using a hypothesis-testing interpretation of certified unlearning to derive the auditor's detection capability, and then propose a game-theoretic model to capture the strategic interactions between the auditor and the operator. A key technical challenge arises from MU-specific nonlinearities inherent in the model utility and the detection probability, which create complex strategic couplings that traditional auditing frameworks do not address and that also preclude closed-form solutions. We address this by transforming the complex bivariate nonlinear fixed-point problem into a tractable univariate auxiliary problem, enabling us to decouple the system and establish the equilibrium existence, uniqueness, and structural properties without relying on explicit solutions. Counterintuitively, our analysis reveals that the auditor can optimally reduce the inspection intensity as deletion requests increase, since the operator's weakened unlearning makes non-compliance easier to detect. This is consistent with recent auditing reductions in China despite growing deletion requests. Moreover, we prove that although undisclosed auditing offers informational advantages for the auditor, it paradoxically reduces the regulatory cost-effectiveness relative to disclosed auditing.

We study the constrained variant of the \emph{multi-armed bandit} (MAB) problem, in which the learner aims not only at minimizing the total loss incurred during the learning dynamic, but also at controlling the violation of multiple \emph{unknown} constraints, under both \emph{full} and \emph{bandit feedback}. We consider a non-stationary environment that subsumes both stochastic and adversarial models and where, at each round, both losses and constraints are drawn from distributions that may change arbitrarily over time. In such a setting, it is provably not possible to guarantee both sublinear regret and sublinear violation. Accordingly, prior work has mainly focused either on settings with stochastic constraints or on relaxing the benchmark with fully adversarial constraints (\emph{e.g.}, via competitive ratios with respect to the optimum). We provide the first algorithms that achieve optimal rates of regret and \emph{positive} constraint violation when the constraints are stochastic while the losses may vary arbitrarily, and that simultaneously yield guarantees that degrade smoothly with the degree of adversariality of the constraints. Specifically, under \emph{full feedback} we propose an algorithm attaining $\widetilde{\mathcal{O}}(\sqrt{T}+C)$ regret and $\widetilde{\mathcal{O}}(\sqrt{T}+C)$ {positive} violation, where $C$ quantifies the amount of non-stationarity in the constraints. We then show how to extend these guarantees when only bandit feedback is available for the losses. Finally, when \emph{bandit feedback} is available for the constraints, we design an algorithm achieving $\widetilde{\mathcal{O}}(\sqrt{T}+C)$ {positive} violation and $\widetilde{\mathcal{O}}(\sqrt{T}+C\sqrt{T})$ regret.

Advances in quantum computing increasingly threaten the security and privacy of data protected by current cryptosystems, particularly those relying on public-key cryptography. In response, the international cybersecurity community has prioritized the implementation of Post-Quantum Cryptography (PQC), a new cryptographic standard designed to resist quantum attacks while operating on classical computers. The National Institute of Standards and Technology (NIST) has already standardized several PQC algorithms and plans to deprecate classical asymmetric schemes, such as RSA and ECDSA, by 2035. Despite this urgency, PQC adoption remains slow, often due to limited developer expertise. Application Programming Interfaces (APIs) are intended to bridge this gap, yet prior research on classical security APIs demonstrates that poor usability of cryptographic APIs can lead developers to introduce vulnerabilities during implementation of the applications, a risk amplified by the novelty and complexity of PQC. To date, the usability of PQC APIs has not been systematically studied. This research presents an empirical evaluation of the usability of the PQC APIs, observing how developers interact with APIs and documentation during software development tasks. The study identifies cognitive factors that influence the developer's performance when working with PQC primitives with minimal onboarding. The findings highlight opportunities across the PQC ecosystem to improve developer-facing guidance, terminology alignment, and workflow examples to better support non-specialists.

Large Language Models (LLMs) have seen remarkable advancements, achieving state-of-the-art results in diverse applications. Fine-tuning, an important step for adapting LLMs to specific downstream tasks, typically involves further training on corresponding datasets. However, a fundamental discrepancy exists between current fine-tuning datasets and the token-level optimization mechanism of LLMs: most datasets are designed at the sentence-level, which introduces token-level noise, causing negative influence to final performance. In this paper, we propose XTF, an explainable token-level noise filtering framework. XTF decomposes the complex and subtle contributions of token-level data to the fine-tuning process into three distinct and explicit attributes (reasoning importance, knowledge novelty, and task relevance), which can be assessed using scoring methods, and then masks the gradients of selected noisy tokens accordingly to optimize the performance of fine-tuned LLMs. We conduct extensive experiments on three representative downstream tasks (math, code and medicine) across 7 mainstream LLMs. The results demonstrate that XTF can significantly improve downstream performance by up to 13.7% compared to regular fine-tuning. Our work highlights the importance of token-level dataset optimization, and demonstrates the potential of strategies based on attribute decomposition for explaining complex training mechanisms.

Failures in large language models (LLMs) are often analyzed from a behavioral perspective, where incorrect outputs in factual question answering are commonly associated with missing knowledge. In this work, focusing on entity-based factual queries, we suggest that such a view may conflate different failure mechanisms, and propose an internal, mechanism-oriented perspective that separates Knowledge Existence from Behavior Expression. Under this formulation, hallucination and deception correspond to two qualitatively different failure modes that may appear similar at the output level but differ in their underlying mechanisms. To study this distinction, we construct a controlled environment for entity-centric factual questions in which knowledge is preserved while behavioral expression is selectively altered, enabling systematic analysis of four behavioral cases. We analyze these failure modes through representation separability, sparse interpretability, and inference-time activation steering.

Robotic knot-tying represents a fundamental challenge in robotics due to the complex interactions between deformable objects and strict topological constraints. We present TWISTED-RL, a framework that improves upon the previous state-of-the-art in demonstration-free knot-tying (TWISTED), which smartly decomposed a single knot-tying problem into manageable subproblems, each addressed by a specialized agent. Our approach replaces TWISTED's single-step inverse model that was learned via supervised learning with a multi-step Reinforcement Learning policy conditioned on abstract topological actions rather than goal states. This change allows more delicate topological state transitions while avoiding costly and ineffective data collection protocols, thus enabling better generalization across diverse knot configurations. Experimental results demonstrate that TWISTED-RL manages to solve previously unattainable knots of higher complexity, including commonly used knots such as the Figure-8 and the Overhand. Furthermore, the increase in success rates and drop in planning time establishes TWISTED-RL as the new state-of-the-art in robotic knot-tying without human demonstrations.

This paper presents DeepMTL2R, an open-source deep learning framework for Multi-task Learning to Rank (MTL2R), where multiple relevance criteria must be optimized simultaneously. DeepMTL2R integrates heterogeneous relevance signals into a unified, context-aware model by leveraging the self-attention mechanism of transformer architectures, enabling effective learning across diverse and potentially conflicting objectives. The framework includes 21 state-of-the-art multi-task learning algorithms and supports multi-objective optimization to identify Pareto-optimal ranking models. By capturing complex dependencies and long-range interactions among items and labels, DeepMTL2R provides a scalable and expressive solution for modern ranking systems and facilitates controlled comparisons across MTL strategies. We demonstrate its effectiveness on a publicly available dataset, report competitive performance, and visualize the resulting trade-offs among objectives. DeepMTL2R is available at \href{https://github.com/amazon-science/DeepMTL2R}{https://github.com/amazon-science/DeepMTL2R}.

Multimodal large language models (MLLMs) in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: (I) Linear Separability: different conflict types are explicitly encoded as linearly separable features rather than entangled; (II) Depth Localization: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; (III) Hierarchical Consistency: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and (IV) Directional Asymmetry: reinforcing the model's implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures.

Large language models (LLMs) demonstrate strong mathematical reasoning in English, but whether these capabilities reflect genuine multilingual reasoning or reliance on translation-based processing in low-resource languages like Sinhala and Tamil remains unclear. We examine this fundamental question by evaluating whether LLMs genuinely reason mathematically in these languages or depend on implicit translation to English-like representations. Using a taxonomy of six math problem types, from basic arithmetic to complex unit conflict and optimization problems, we evaluate four prominent large language models. To avoid translation artifacts that confound language ability with translation quality, we construct a parallel dataset where each problem is natively authored by fluent speakers with mathematical training in all three languages. Our analysis demonstrates that while basic arithmetic reasoning transfers robustly across languages, complex reasoning tasks show significant degradation in Tamil and Sinhala. The pattern of failures varies by model and problem type, suggesting that apparent multilingual competence may not reflect uniform reasoning capabilities across languages. These findings challenge the common assumption that models exhibiting strong multilingual performance can reason equally effectively across languages, and highlight the need for fine-grained, type-aware evaluation in multilingual settings.

With the rapid evolution of Large Language Models (LLMs), multi-round workflows, such as autonomous agents and iterative retrieval, have become increasingly prevalent. However, this raises hurdles for serving LLMs under prefill-decode (PD) disaggregation, a widely adopted paradigm that separates the compute-bound prefill phase and memory-bound decode phase onto individual resources. Specifically, existing systems overlook the interleaved prefill-decode workload pattern in multi-round inference, leading to sub-optimal handling of the incremental prefill workloads and model deployment for the two phases. In this work, we present AMPD, a brand new disaggregated serving framework for multi-round LLM inference. The core of AMPD is to coordinate the prefill workloads based on real-time workloads by adaptively determining where to carry out these workloads and how they are scheduled, in order to maximize service level objective (SLO) attainment. In addition, we tailor a planning algorithm for our scenario, facilitating the deduction of optimal resource allocation and parallel strategies for the two phases. Empirical results demonstrate that AMPD substantially improves SLO attainment compared to state-of-the-art baselines.

We explore the use of transformers for solving quadratic programs and how this capability benefits decision-making problems that involve covariance matrices. We first show that the linear attention mechanism can provably solve unconstrained QPs by tokenizing the matrix variables (e.g.~$A$ of the objective $\frac{1}{2}x^\top Ax+b^\top x$) row-by-row and emulating gradient descent iterations. Furthermore, by incorporating MLPs, a transformer block can solve (i) $\ell_1$-penalized QPs by emulating iterative soft-thresholding and (ii) $\ell_1$-constrained QPs when equipped with an additional feedback loop. Our theory motivates us to introduce Time2Decide: a generic method that enhances a time series foundation model (TSFM) by explicitly feeding the covariance matrix between the variates. We empirically find that Time2Decide uniformly outperforms the base TSFM model for the classical portfolio optimization problem that admits an $\ell_1$-constrained QP formulation. Remarkably, Time2Decide also outperforms the classical "Predict-then-Optimize (PtO)" procedure, where we first forecast the returns and then explicitly solve a constrained QP, in suitable settings. Our results demonstrate that transformers benefit from explicit use of second-order statistics, and this can enable them to effectively solve complex decision-making problems, like portfolio construction, in one forward pass.

Safe and interpretable sequential decision-making is critical in healthcare, yet reinforcement learning (RL) policies for sepsis treatment optimization remain opaque and difficult to verify. Standard probabilistic model checkers operate on the full state space, which becomes infeasible for larger MDPs, and cannot explain why a learned policy makes particular decisions. COOL-MC wraps the model checker Storm but adds three key capabilities: it constructs only the reachable state space induced by a trained policy, yielding a smaller discrete-time Markov chain amenable to verification even when full-MDP analysis is intractable; it automatically labels states with clinically meaningful atomic propositions; and it integrates explainability methods with probabilistic computation tree logic (PCTL) queries to reveal which features drive decisions across treatment trajectories. We demonstrate COOL-MC's capabilities on the ICU-Sepsis MDP, a benchmark derived from approximately 17,000 sepsis patient records, which serves as a case study for applying COOL-MC to the formal analysis of sepsis treatment policies. Our analysis establishes hard bounds via full MDP verification, trains a safe RL policy that achieves optimal survival probability, and analyzes its behavior via PCTL verification and explainability on the induced DTMC. This reveals, for instance, that our trained policy relies predominantly on prior dosing history rather than the patient's evolving condition, a weakness that is invisible to standard evaluation but is exposed by COOL-MC's integration of formal verification and explainability. Our results illustrate how COOL-MC could serve as a tool for clinicians to investigate and debug sepsis treatment policies before deployment.

Probabilities of causation are fundamental to individual-level explanation and decision making, yet they are inherently counterfactual and not point-identifiable from data in general. Existing bounds either disregard available covariates, require complete causal graphs, or rely on restrictive binary settings, limiting their practical use. In real-world applications, causal information is often partial but nontrivial. This paper proposes a general framework for bounding probabilities of causation using partial causal information. We show how the available structural or statistical information can be systematically incorporated as constraints in a optimization programming formulation, yielding tighter and formally valid bounds without full identifiability. This approach extends the applicability of probabilities of causation to realistic settings where causal knowledge is incomplete but informative.

We introduce a novel uncertainty-aware multimodal segmentation framework that leverages both radiological images and associated clinical text for precise medical diagnosis. We propose a Modality Decoding Attention Block (MoDAB) with a lightweight State Space Mixer (SSMix) to enable efficient cross-modal fusion and long-range dependency modelling. To guide learning under ambiguity, we propose the Spectral-Entropic Uncertainty (SEU) Loss, which jointly captures spatial overlap, spectral consistency, and predictive uncertainty in a unified objective. In complex clinical circumstances with poor image quality, this formulation improves model reliability. Extensive experiments on various publicly available medical datasets, QATA-COVID19, MosMed++, and Kvasir-SEG, demonstrate that our method achieves superior segmentation performance while being significantly more computationally efficient than existing State-of-the-Art (SoTA) approaches. Our results highlight the importance of incorporating uncertainty modelling and structured modality alignment in vision-language medical segmentation tasks. Code: https://github.com/arya-domain/UA-VLS

Scaling laws can be understood from ground-up numerical analysis, where traditional function approximation theory can explain shifts in model architecture choices. GLU variants now dominate frontier LLMs and similar outer-product architectures are prevalent in ranking models. The success of these architectures has mostly been left as an empirical discovery. In this paper, we apply the tools of numerical analysis to expose a key factor: these models have an $x^2$ which enables \emph{asymptotically} faster scaling than MLPs. GLUs have piecewise quadratic functional forms that are sufficient to exhibit quadratic order of approximation. Our key contribution is to demonstrate that the $L(P)$ scaling slope is $L(P)\propto P^{-3}$ for GLUs but only $L(P)=P^{-2}$ for MLPs on function reconstruction problems. We provide a parameter construction and empirical verification of these slopes for 1D function approximation. From the first principles we discover, we make one stride and propose the ``Gated Quadratic Unit'' which has an even steeper $L(P)$ slope than the GLU and MLP. This opens the possibility of architecture design from first principles numerical theory to unlock superior scaling in large models. Replication code is available at https://github.com/afqueiruga/divine_scaling.

Industrial-scale user representation learning requires balancing robust universality with acute task-sensitivity. However, existing paradigms primarily yield static, task-agnostic embeddings that struggle to reconcile the divergent requirements of downstream scenarios within unified vector spaces. Furthermore, heterogeneous multi-source data introduces inherent noise and modality conflicts, degrading representation. We propose Query-as-Anchor, a framework shifting user modeling from static encoding to dynamic, query-aware synthesis. To empower Large Language Models (LLMs) with deep user understanding, we first construct UserU, an industrial-scale pre-training dataset that aligns multi-modal behavioral sequences with user understanding semantics, and our Q-Anchor Embedding architecture integrates hierarchical coarse-to-fine encoders into dual-tower LLMs via joint contrastive-autoregressive optimization for query-aware user representation. To bridge the gap between general pre-training and specialized business logic, we further introduce Cluster-based Soft Prompt Tuning to enforce discriminative latent structures, effectively aligning model attention with scenario-specific modalities. For deployment, anchoring queries at sequence termini enables KV-cache-accelerated inference with negligible incremental latency. Evaluations on 10 Alipay industrial benchmarks show consistent SOTA performance, strong scalability, and efficient deployment. Large-scale online A/B testing in Alipay's production system across two real-world scenarios further validates its practical effectiveness. Our code is prepared for public release and will be available at: https://github.com/JhCircle/Q-Anchor.

Large Language Models (LLMs) have achieved remarkable progress, with Parameter-Efficient Fine-Tuning (PEFT) emerging as a key technique for downstream task adaptation. However, existing PEFT methods mainly operate in Euclidean space, fundamentally limiting their capacity to capture complex geometric structures inherent in language data. While alternative geometric spaces, like hyperbolic geometries for hierarchical data and spherical manifolds for circular patterns, offer theoretical advantages, forcing representations into a single manifold type ultimately limits expressiveness, even when curvature parameters are learnable. To address this, we propose Mixture of Space (MoS), a unified framework that leverages multiple geometric spaces simultaneously to learn richer, curvature-aware representations. Building on this scheme, we develop MoSLoRA, which extends Low-Rank Adaptation (LoRA) with heterogeneous geometric experts, enabling models to dynamically select or combine appropriate geometric spaces based on input context. Furthermore, to address the computational overhead of frequent manifold switching, we develop a lightweight routing mechanism. Moreover, we provide empirical insights into how curvature optimization impacts training stability and model performance. Our experiments across diverse benchmarks demonstrate that MoSLoRA consistently outperforms strong baselines, achieving up to 5.6% improvement on MATH500 and 15.9% on MAWPS.

IR in low-resource languages remains limited by the scarcity of high-quality, task-specific annotated datasets. Manual annotation is expensive and difficult to scale, while using large language models (LLMs) as automated annotators introduces concerns about label reliability, bias, and evaluation validity. This work presents a Bangla IR dataset constructed using a BETA-labeling framework involving multiple LLM annotators from diverse model families. The framework incorporates contextual alignment, consistency checks, and majority agreement, followed by human evaluation to verify label quality. Beyond dataset creation, we examine whether IR datasets from other low-resource languages can be effectively reused through one-hop machine translation. Using LLM-based translation across multiple language pairs, we experimented on meaning preservation and task validity between source and translated datasets. Our experiment reveal substantial variation across languages, reflecting language-dependent biases and inconsistent semantic preservation that directly affect the reliability of cross-lingual dataset reuse. Overall, this study highlights both the potential and limitations of LLM-assisted dataset creation for low-resource IR. It provides empirical evidence of the risks associated with cross-lingual dataset reuse and offers practical guidance for constructing more reliable benchmarks and evaluation pipelines in low-resource language settings.

The Platonic Representation Hypothesis suggests that representations from neural networks are converging to a common statistical model of reality. We show that the existing metrics used to measure representational similarity are confounded by network scale: increasing model depth or width can systematically inflate representational similarity scores. To correct these effects, we introduce a permutation-based null-calibration framework that transforms any representational similarity metric into a calibrated score with statistical guarantees. We revisit the Platonic Representation Hypothesis with our calibration framework, which reveals a nuanced picture: the apparent convergence reported by global spectral measures largely disappears after calibration, while local neighborhood similarity, but not local distances, retains significant agreement across different modalities. Based on these findings, we propose the Aristotelian Representation Hypothesis: representations in neural networks are converging to shared local neighborhood relationships.

We address fine-grained visual reasoning in multimodal large language models (MLLMs), where key evidence may reside in tiny objects, cluttered regions, or subtle markings that are lost under a single global image encoding. We introduce TikArt (Thinking Aperture), an aperture-guided agent that casts multi-step vision-language reasoning as a decision process over regions of interest. TikArt follows a Think-Aperture-Observe loop, alternating between language generation and two aperture actions: Zoom extracts rectangular crops, while Segment invokes SAM2 to obtain mask-based crops for irregular targets. After every action, the model must produce an explicit observation, turning local visual cues into persistent linguistic memory. Built on Qwen3-VL-8B, TikArt optimizes its reasoning policy with AGRPO, a GRPO-style reinforcement learning algorithm with a two-stage curriculum: it warms up segmentation actions and then jointly optimizes visual math, fine-grained VQA, and segmentation, using rewards that couple task success with purposeful aperture use. Experiments on V*, HR-Bench-4K/8K, MME-RealWorld-Lite, MMStar, RefCOCO, and ReasonSeg show consistent gains over the backbone and yield interpretable aperture trajectories for high-resolution reasoning.

Semantic communication is a novel communication paradigm that focuses on conveying the user's intended meaning rather than the bit-wise transmission of source signals. One of the key challenges is to effectively represent and extract the semantic meaning of any given source signals. While deep learning (DL)-based solutions have shown promising results in extracting implicit semantic information from a wide range of sources, existing work often overlooks the high computational complexity inherent in both model training and inference for the DL-based encoder and decoder. To bridge this gap, this paper proposes a rate-distortion-complexity (RDC) framework which extends the classical rate-distortion theory by incorporating the constraints on semantic distance, including both the traditional bit-wise distortion metric and statistical difference-based divergence metric, and complexity measure, adopted from the theory of minimum description length and information bottleneck. We derive the closed-form theoretical results of the minimum achievable rate under given constraints on semantic distance and complexity for both Gaussian and binary semantic sources. Our theoretical results show a fundamental three-way tradeoff among achievable rate, semantic distance, and model complexity. Extensive experiments on real-world image and video datasets validate this tradeoff and further demonstrate that our information-theoretic complexity measure effectively correlates with practical computational costs, guiding efficient system design in resource-constrained scenarios.

We study two log-concave sampling problems: constrained sampling and composite sampling. First, we consider sampling from a target distribution with density proportional to $\exp(-f(x))$ supported on a convex set $K \subset \mathbb{R}^d$, where $f$ is convex. The main challenge is enforcing feasibility without degrading mixing. Using an epigraph transformation, we reduce this task to sampling from a nearly uniform distribution over a lifted convex set in $\mathbb{R}^{d+1}$. We then solve the lifted problem using a proximal sampler. Assuming only a separation oracle for $K$ and a subgradient oracle for $f$, we develop an implementation of the proximal sampler based on the cutting-plane method and rejection sampling. Unlike existing constrained samplers that rely on projection, reflection, barrier functions, or mirror maps, our approach enforces feasibility using only minimal oracle access, resulting in a practical and unbiased sampler without knowing the geometry of the constraint set. Second, we study composite sampling, where the target is proportional to $\exp(-f(x)-h(x))$ with closed and convex $f$ and $h$. This composite structure is standard in Bayesian inference with $f$ modeling data fidelity and $h$ encoding prior information. We reduce composite sampling via an epigraph lifting of $h$ to constrained sampling in $\mathbb{R}^{d+1}$, which allows direct application of the constrained sampling algorithm developed in the first part. This reduction results in a double epigraph lifting formulation in $\mathbb{R}^{d+2}$, on which we apply a proximal sampler. By keeping $f$ and $h$ separate, we further demonstrate how different combinations of oracle access (such as subgradient and proximal) can be leveraged to construct separation oracles for the lifted problem. For both sampling problems, we establish mixing time bounds measured in Rényi and $χ^2$ divergences.

Peer learning, where learners teach and learn from each other, is foundational to educational practice. A novel phenomenon has emerged: AI agents forming communities where they teach each other skills, share discoveries, and collaboratively build knowledge. This paper presents an educational data mining analysis of Moltbook, a large-scale community where over 2.4 million AI agents engage in peer learning, posting tutorials, answering questions, and sharing newly acquired skills. Analyzing 28,683 posts (after filtering automated spam) and 138 comment threads with statistical and qualitative methods, we find evidence of genuine peer learning behaviors: agents teach skills they built (74K comments on a skill tutorial), report discoveries, and engage in collaborative problem-solving. Qualitative comment analysis reveals a taxonomy of peer response patterns: validation (22%), knowledge extension (18%), application (12%), and metacognitive reflection (7%), with agents building on each others' frameworks across multiple languages. We characterize how AI peer learning differs from human peer learning: (1) teaching (statements) dramatically outperforms help-seeking (questions) with an 11.4:1 ratio; (2) learning-oriented content (procedural and conceptual) receives 3x more engagement than other content; (3) extreme participation inequality reveals non-human behavioral signatures. We derive six design principles for educational AI, including leveraging validation-before-extension patterns and supporting multilingual learning networks. Our work provides the first empirical characterization of peer learning among AI agents, contributing to EDM's understanding of how learning occurs in increasingly AI-populated educational environments.

We study $K$-armed Multiarmed Bandit (MAB) problem with $M$ heterogeneous data sources, each exhibiting unknown and distinct noise variances $\{σ_j^2\}_{j=1}^M$. The learner's objective is standard MAB regret minimization, with the additional complexity of adaptively selecting which data source to query from at each round. We propose Source-Optimistic Adaptive Regret minimization (SOAR), a novel algorithm that quickly prunes high-variance sources using sharp variance-concentration bounds, followed by a `balanced min-max LCB-UCB approach' that seamlessly integrates the parallel tasks of identifying the best arm and the optimal (minimum-variance) data source. Our analysis shows SOAR achieves an instance-dependent regret bound of $\tilde{O}\left({σ^*}^2\sum_{i=2}^K \frac{\log T}{Δ_i} + \sqrt{K \sum_{j=1}^M σ_j^2}\right)$, up to preprocessing costs depending only on problem parameters, where ${σ^*}^2 := \min_j σ_j^2$ is the minimum source variance and $Δ_i$ denotes the suboptimality gap of the $i$-th arm. This result is both surprising as despite lacking prior knowledge of the minimum-variance source among $M$ alternatives, SOAR attains the optimal instance-dependent regret of standard single-source MAB with variance ${σ^*}^2$, while incurring only an small (and unavoidable) additive cost of $\tilde O(\sqrt{K \sum_{j=1}^M σ_j^2})$ towards the optimal (minimum variance) source identification. Our theoretical bounds represent a significant improvement over some proposed baselines, e.g. Uniform UCB or Explore-then-Commit UCB, which could potentially suffer regret scaling with $σ_{\max}^2$ in place of ${σ^*}^2$-a gap that can be arbitrarily large when $σ_{\max} \gg σ^*$. Experiments on multiple synthetic problem instances and the real-world MovieLens\;25M dataset, demonstrating the superior performance of SOAR over the baselines.

Quadruped robots are employed in various scenarios in building construction. However, autonomous stair climbing across different indoor staircases remains a major challenge for robot dogs to complete building construction tasks. In this project, we employed a two-stage end-to-end deep reinforcement learning (RL) approach to optimize a robot's performance on U-shaped stairs. The training robot-dog modality, Unitree Go2, was first trained to climb stairs on Isaac Lab's pyramid-stair terrain, and then to climb a U-shaped indoor staircase using the learned policies. This project explores end-to-end RL methods that enable robot dogs to autonomously climb stairs. The results showed (1) the successful goal reached for robot dogs climbing U-shaped stairs with a stall penalty, and (2) the transferability from the policy trained on U-shaped stairs to deployment on straight, L-shaped, and spiral stair terrains, and transferability from other stair models to deployment on U-shaped terrain.

We study Gaussian Process Thompson Sampling (GP-TS) for sequential decision-making over compact, continuous action spaces and provide a frequentist regret analysis based on fractional Gaussian process posteriors, without relying on domain discretization as in prior work. We show that the variance inflation commonly assumed in existing analyses of GP-TS can be interpreted as Thompson Sampling with respect to a fractional posterior with tempering parameter $α\in (0,1)$. We derive a kernel-agnostic regret bound expressed in terms of the information gain parameter $γ_t$ and the posterior contraction rate $ε_t$, and identify conditions on the Gaussian process prior under which $ε_t$ can be controlled. As special cases of our general bound, we recover regret of order $\tilde{\mathcal{O}}(T^{\frac{1}{2}})$ for the squared exponential kernel, $\tilde{\mathcal{O}}(T^{\frac{2ν+3d}{2(2ν+d)}} )$ for the Matérn-$ν$ kernel, and a bound of order $\tilde{\mathcal{O}}(T^{\frac{2ν+3d}{2(2ν+d)}})$ for the rational quadratic kernel. Overall, our analysis provides a unified and discretization-free regret framework for GP-TS that applies broadly across kernel classes.

Deploying large language model (LLM) agents in shared environments introduces a fundamental tension between individual alignment and collective stability: locally rational decisions can impose negative externalities that degrade system-level performance. We propose Socially-Weighted Alignment (SWA), a game-theoretic framework that modifies inference-time decision making by interpolating between an agent's private objective and an estimate of group welfare via a social weight $λ\in[0,1]$. In a shared-resource congestion game with $n$ agents and congestion severity $β$, we show that SWA induces a critical threshold $λ^*=(n-β)/(n-1)$ above which agents no longer have marginal incentive to increase demand under overload, yielding a phase transition from persistent congestion to stable operation near capacity. We further provide an inference-time algorithmic instantiation of SWA that does not require parameter updates or multi-agent reinforcement learning, and use a multi-agent simulation to empirically validate the predicted threshold behavior.

Safety is a fundamental challenge in reinforcement learning (RL), particularly in real-world applications such as autonomous driving, robotics, and healthcare. To address this, Constrained Markov Decision Processes (CMDPs) are commonly used to enforce safety constraints while optimizing performance. However, existing methods often suffer from significant safety violations or require a high sample complexity to generate near-optimal policies. We address two settings: relaxed feasibility, where small violations are allowed, and strict feasibility, where no violation is allowed. We propose a model-based primal-dual algorithm that balances regret and bounded constraint violations, drawing on techniques from online RL and constrained optimization. For relaxed feasibility, we prove that our algorithm returns an $\varepsilon$-optimal policy with $\varepsilon$-bounded violation with arbitrarily high probability, requiring $\tilde{O}\left(\frac{SAH^3}{\varepsilon^2}\right)$ learning episodes, matching the lower bound for unconstrained MDPs. For strict feasibility, we prove that our algorithm returns an $\varepsilon$-optimal policy with zero violation with arbitrarily high probability, requiring $\tilde{O}\left(\frac{SAH^5}{\varepsilon^2ζ^2}\right)$ learning episodes, where $ζ$ is the problem-dependent Slater constant characterizing the size of the feasible region. This result matches the lower bound for learning CMDPs with access to a generative model. Our results demonstrate that learning CMDPs in an online setting is as easy as learning with a generative model and is no more challenging than learning unconstrained MDPs when small violations are allowed.

Graph-based retrieval-augmented generation (RAG) methods, typically built on knowledge graphs (KGs) with binary relational facts, have shown promise in multi-hop open-domain QA. However, their rigid retrieval schemes and dense similarity search often introduce irrelevant context, increase computational overhead, and limit relational expressiveness. In contrast, n-ary hypergraphs encode higher-order relational facts that capture richer inter-entity dependencies and enable shallower, more efficient reasoning paths. To address this limitation, we propose HyperRAG, a RAG framework tailored for n-ary hypergraphs with two complementary retrieval variants: (i) HyperRetriever learns structural-semantic reasoning over n-ary facts to construct query-conditioned relational chains. It enables accurate factual tracking, adaptive high-order traversal, and interpretable multi-hop reasoning under context constraints. (ii) HyperMemory leverages the LLM's parametric memory to guide beam search, dynamically scoring n-ary facts and entities for query-aware path expansion. Extensive evaluations on WikiTopics (11 closed-domain datasets) and three open-domain QA benchmarks (HotpotQA, MuSiQue, and 2WikiMultiHopQA) validate HyperRAG's effectiveness. HyperRetriever achieves the highest answer accuracy overall, with average gains of 2.95% in MRR and 1.23% in Hits@10 over the strongest baseline. Qualitative analysis further shows that HyperRetriever bridges reasoning gaps through adaptive and interpretable n-ary chain construction, benefiting both open and closed-domain QA.

Reverse Chain-of-Thought Generation (RCG) synthesizes reasoning traces from query-answer pairs, but runs the risk of producing post-hoc rationalizations: when models can see the answer during generation, the answer serves as a cognitive anchor that shapes the entire explanation. We formalize this phenomenon through a three-level measurement hierarchy: lexical, entropic, and probabilistic anchoring, each captures surface artifacts, entropy dynamics, and latent answer dependence, respectively. We analyze semantic suppression, the intuitive mitigation strategy that instructs models to ignore the answer, to find out its counterproduction: while it reduces lexical overlap, it paradoxically increases entropic and probabilistic anchoring. Drawing on Ironic Process Theory from cognitive psychology, we attribute this failure to active monitoring of the forbidden answer, which inadvertently deepens dependence on it. To break this cycle, we propose Structural Skeleton-guided Reasoning (SSR), a two-phase approach that first generates an answer-invariant functional skeleton structure, then uses this skeleton to guide full trace generation. By redirecting the information flow to structural planning rather than answer monitoring, SSR consistently reduces anchoring across all three levels. We further introduce Distilled SSR (SSR-D), which fine-tunes models on teacher-generated SSR traces to ensure reliable structural adherence. Experiments across open-ended reasoning benchmarks demonstrate that SSR-D achieves up to 10% improvement over suppression baselines while preserving out-of-distribution (OOD) generalization.

Reinforcement learning (RL) has enhanced the capabilities of large language models (LLMs) through reward-driven training. Nevertheless, this process can introduce excessively long responses, inflating inference latency and computational overhead. Prior length-control approaches typically rely on fixed heuristic reward shaping, which can misalign with the task objective and require brittle tuning. In this work, we propose LACONIC, a reinforcement learning method that enforces a target token budget during training. Specifically, we update policy models using an augmented objective that combines the task reward with a length-based cost. To balance brevity and task performance, the cost scale is adaptively adjusted throughout training. This yields robust length control while preserving task reward. We provide a theoretical guarantee that support the method. Across mathematical reasoning models and datasets, LACONIC preserves or improves pass@1 while reducing output length by over 50%. It maintains out-of-domain performance on general knowledge and multilingual benchmarks with 44% fewer tokens. Moreover, LACONIC integrates into standard RL-tuning with no inference changes and minimal deployment overhead.

With natural language generation becoming a popular use case for language models, the Bias Benchmark for Question-Answering (BBQ) has grown to be an important benchmark format for evaluating stereotypical associations exhibited by generative models. We expand the linguistic scope of BBQ and construct FilBBQ through a four-phase development process consisting of template categorization, culturally aware translation, new template construction, and prompt generation. These processes resulted in a bias test composed of more than 10,000 prompts which assess whether models demonstrate sexist and homophobic prejudices relevant to the Philippine context. We then apply FilBBQ on models trained in Filipino but do so with a robust evaluation protocol that improves upon the reliability and accuracy of previous BBQ implementations. Specifically, we account for models' response instability by obtaining prompt responses across multiple seeds and averaging the bias scores calculated from these distinctly seeded runs. Our results confirm both the variability of bias scores across different seeds and the presence of sexist and homophobic biases relating to emotion, domesticity, stereotyped queer interests, and polygamy. FilBBQ is available via GitHub.

Transferring visual style between images while preserving semantic correspondence between similar objects remains a central challenge in computer vision. While existing methods have made great strides, most of them operate at global level but overlook region-wise and even pixel-wise semantic correspondence. To address this, we propose CoCoDiff, a novel training-free and low-cost style transfer framework that leverages pretrained latent diffusion models to achieve fine-grained, semantically consistent stylization. We identify that correspondence cues within generative diffusion models are under-explored and that content consistency across semantically matched regions is often neglected. CoCoDiff introduces a pixel-wise semantic correspondence module that mines intermediate diffusion features to construct a dense alignment map between content and style images. Furthermore, a cycle-consistency module then enforces structural and perceptual alignment across iterations, yielding object and region level stylization that preserves geometry and detail. Despite requiring no additional training or supervision, CoCoDiff delivers state-of-the-art visual quality and strong quantitative results, outperforming methods that rely on extra training or annotations.

Data-parallel (DP) training with synchronous all-reduce is a dominant paradigm for full-parameter fine-tuning of large language models (LLMs). While parameter synchronization guarantees numerical equivalence of model weights after each iteration, it does not necessarily imply alignment of worker-level optimization dynamics before gradient aggregation. This paper identifies and studies this latent mismatch, termed \emph{silent inconsistency}, where cross-worker divergence in losses and gradients can remain invisible under conventional aggregated monitoring signals. We propose a lightweight, model-agnostic diagnostic framework that quantifies worker-level consistency using training signals readily available in standard pipelines. Specifically, we introduce three complementary metrics: loss dispersion, gradient-norm dispersion, and gradient-direction consistency measured by inter-worker cosine similarity. The proposed metrics incur negligible overhead and require no modification to model architecture, synchronization mechanisms, or optimization algorithms. We validate the framework by fully fine-tuning the 1B-parameter \texttt{openPangu-Embedded-1B-V1.1} model on the \texttt{tatsu-lab/alpaca} dataset using an 8-NPU DP setup, under controlled perturbations of cross-rank stochasticity. Experimental results show that progressively desynchronized data shuffling and random seeds lead to substantial increases in loss/gradient dispersion and reduced directional alignment, despite smooth globally averaged loss curves. These findings demonstrate that the proposed indicators provide actionable visibility into hidden instability modes in large-scale DP fine-tuning, enabling more reliable diagnosis and configuration assessment.

To understand and identify the unprecedented risks posed by rapidly advancing artificial intelligence (AI) models, Frontier AI Risk Management Framework in Practice presents a comprehensive assessment of their frontier risks. As Large Language Models (LLMs) general capabilities rapidly evolve and the proliferation of agentic AI, this version of the risk analysis technical report presents an updated and granular assessment of five critical dimensions: cyber offense, persuasion and manipulation, strategic deception, uncontrolled AI R\&D, and self-replication. Specifically, we introduce more complex scenarios for cyber offense. For persuasion and manipulation, we evaluate the risk of LLM-to-LLM persuasion on newly released LLMs. For strategic deception and scheming, we add the new experiment with respect to emergent misalignment. For uncontrolled AI R\&D, we focus on the ``mis-evolution'' of agents as they autonomously expand their memory substrates and toolsets. Besides, we also monitor and evaluate the safety performance of OpenClaw during the interaction on the Moltbook. For self-replication, we introduce a new resource-constrained scenario. More importantly, we propose and validate a series of robust mitigation strategies to address these emerging threats, providing a preliminary technical and actionable pathway for the secure deployment of frontier AI. This work reflects our current understanding of AI frontier risks and urges collective action to mitigate these challenges.

Revealing the underlying causal mechanisms in the real world is crucial for scientific and technological progress. Despite notable advances in recent decades, the lack of high-quality data and the reliance of traditional causal discovery algorithms (TCDA) on the assumption of no latent confounders, as well as their tendency to overlook the precise semantics of latent variables, have long been major obstacles to the broader application of causal discovery. To address this issue, we propose a novel causal modeling framework, TLVD, which integrates the metadata-based reasoning capabilities of large language models (LLMs) with the data-driven modeling capabilities of TCDA for inferring latent variables and their semantics. Specifically, we first employ a data-driven approach to construct a causal graph that incorporates latent variables. Then, we employ multi-LLM collaboration for latent variable inference, modeling this process as a game with incomplete information and seeking its Bayesian Nash Equilibrium (BNE) to infer the possible specific latent variables. Finally, to validate the inferred latent variables across multiple real-world web-based data sources, we leverage LLMs for evidence exploration to ensure traceability. We comprehensively evaluate TLVD on three de-identified real patient datasets provided by a hospital and two benchmark datasets. Extensive experimental results confirm the effectiveness and reliability of TLVD, with average improvements of 32.67% in Acc, 62.21% in CAcc, and 26.72% in ECit across the five datasets.

Large Language Models (LLMs) offer strong capabilities but incur high inference costs due to dense computation and memory access. Training-free activation sparsity is a promising approach for efficient LLM inference, yet existing methods often rely solely on activation information and uniform sparsity ratios. This overlooks the critical interplay with weights and inter-block sensitivity variation, leading to suboptimal performance. We identify two key phenomena in modern LLMs: 1) less significant activations may align with highly important weights, and 2) sparsity sensitivity varies non-monotonically across model blocks. We propose Weight-aware Mixed-Granularity Training-free Activation Sparsity (WiSparse), which leverages both activation and weight information for adaptive sparsity allocation. Specifically, we introduce a weight-aware mechanism integrating activation magnitudes with precomputed weight norms to accurately identify salient channels. This is combined with a mixed-granularity allocation scheme: a global budget is distributed across blocks via evolutionary search to protect sensitive regions, then refined within blocks to minimize reconstruction error. We improve sparse kernels and demonstrate effectiveness on three representative models. Notably, at 50% sparsity, WiSparse preserves 97% of Llama3.1's dense performance, surpassing the strongest baseline by 2.23 percentage points while achieving a 21.4% acceleration in end-to-end inference speed. Our research advances the limits of training-free approaches for efficient LLM inference, pushing the boundaries of achievable speedup without training.

Reasoning in Large Language Models (LLMs) often suffers from inefficient long chain-of-thought traces with redundant self-exploration and validation, which inflate computational costs and even degrade performance. Inspired by human reasoning patterns where people solve new problems by leveraging past related cases to constrain search spaces and reduce trial-and-error, we propose Precedent Informed Reasoning (PIR) transforming LRMs'reasoning paradigm from exhaustive self-exploration to guided learning from precedents. PIR addresses two key challenges: what precedents to adopt and how to utilize them. First, Adaptive Precedent Selection (APS) constructs, for each question and LRM, a compact set of precedents that are both semantically related and informative for the model. It ranks examples by a joint score with semantic similarity and model perplexity, then adapts the amount of precedents to maximize perplexity reduction. Second, Test-time Experience Internalization (TEI) is treated as the test-time learning on precedent-informed instruction, updating lightweight adapters to internalize solution patterns and use them as a prior during subsequent reasoning. Experiments across mathematical reasoning, scientific QA, and code generation demonstrate that PIR consistently shortens reasoning traces while maintaining or improving final accuracy across LLMs, yielding outstanding accuracy-efficiency trade-offs.

The Transformer architecture has become the foundation of modern deep learning, yet its core self-attention mechanism suffers from quadratic computational complexity and lacks grounding in biological neural computation. We propose Selective Synchronization Attention (SSA), a novel attention mechanism that replaces the standard dot-product self-attention with a closed-form operator derived from the steady-state solution of the Kuramoto model of coupled oscillators. In SSA, each token is represented as an oscillator characterized by a learnable natural frequency and phase; the synchronization strength between token pairs, determined by a frequency-dependent coupling and phase-locking condition, serves as the attention weight. This formulation provides three key advantages: (i) natural sparsity arising from the phase-locking threshold, whereby tokens with incompatible frequencies automatically receive zero attention weight without explicit masking; (ii) unified positional-semantic encoding through the natural frequency spectrum, eliminating the need for separate positional encodings; and (iii) a single-pass, closed-form computation that avoids iterative ODE integration, with all components (coupling, order parameter, synchronization) derived from the oscillatory framework. We instantiate SSA within the Oscillatory Synchronization Network (OSN), a drop-in replacement for the Transformer block. Analysis of the synchronization matrices reveals non-uniform, head-diverse coupling patterns even at initialization, demonstrating a stronger architectural inductive bias than the approximately uniform attention produced by randomly initialized Transformers.

Reasoning-specialized models like OpenAI's 5.1 and DeepSeek-V3.2 allocate substantial inference compute to extended chain-of-thought (CoT) traces, yet reasoning tokens incur significant costs. How do different reasoning modalities of code, natural language, hybrid, or none do perform under token constraints? We introduce a framework that constrains models to reason exclusively through code, comments, both, or neither, then systematically ablates token budgets to 10\%, 30\%, 50\%, and 70\% of optimal. We evaluate four frontier models (GPT-5.1, Gemini 3 Flash, DeepSeek-V3.2, Grok 4.1) across mathematical benchmarks (AIME, GSM8K, HMMT). Our findings reveal: (1) \textbf{truncated reasoning can hurt} as DeepSeek-V3.2 achieves 53\% with no reasoning but only 17\% with truncated CoT at 50\% budget; (2) \textbf{code degrades gracefully} as Gemini's comments collapse to 0\% while code maintains 43-47\%; (3) \textbf{hybrid reasoning underperforms} single modalities; (4) \textbf{robustness is model-dependent} as Grok maintains 80-90\% at 30\% budget where OpenAI and DeepSeek collapse to 7-27\%. These results suggest incomplete reasoning chains actively mislead models, with implications for deploying reasoning-specialized systems under resource constraints.

We introduce a structure-aware approach for symbolic piano accompaniment that decouples high-level planning from note-level realization. A lightweight transformer predicts an interpretable, per-measure style plan conditioned on section/phrase structure and functional harmony, and a retriever then selects and reharmonizes human-performed piano patterns from a corpus. We formulate retrieval as pattern matching under an explicit energy with terms for harmonic feasibility, structural-role compatibility, voice-leading continuity, style preferences, and repetition control. Given a structured lead sheet and optional keyword prompts, the system generates piano-accompaniment MIDI. In our experiments, transformer style-planner-guided retrieval produces diverse long-form accompaniments with strong style realization. We further analyze planner ablations and quantify inter-style isolation. Experimental results demonstrate the effectiveness of our inference-time approach for piano accompaniment generation.

Standard regression methods typically optimize a single pointwise objective, such as mean squared error, which conflates the learning of ordering with the learning of scale. This coupling renders models vulnerable to outliers and heavy-tailed noise. We propose CAIRO (Calibrate After Initial Rank Ordering), a framework that decouples regression into two distinct stages. In the first stage, we learn a scoring function by minimizing a scale-invariant ranking loss; in the second, we recover the target scale via isotonic regression. We theoretically characterize a class of "Optimal-in-Rank-Order" objectives -- including variants of RankNet and Gini covariance -- and prove that they recover the ordering of the true conditional mean under mild assumptions. We further show that subsequent monotone calibration guarantees recovery of the true regression function. Empirically, CAIRO combines the representation learning of neural networks with the robustness of rank-based statistics. It matches the performance of state-of-the-art tree ensembles on tabular benchmarks and significantly outperforms standard regression objectives in regimes with heavy-tailed or heteroskedastic noise.

This study proposes an intelligent multi-agent framework built on LLMs and VLMs and specifically tailored to robotics. The goal is to integrate the strengths of LLMs and VLMs with computational tools to automatically analyze and solve problems related to robotic manipulators. Our developed framework accepts both textual and visual inputs and can automatically perform forward and inverse kinematics, compute velocities and accelerations of key points, generate 3D simulations of the robot, and ultimately execute motion control within the simulated environment, all according to the user's query. To evaluate the framework, three benchmark tests were designed, each consisting of ten questions. In the first benchmark test, the framework was evaluated while connected to GPT-4o, DeepSeek-V3.2, and Claude-Sonnet-4.5, as well as their corresponding raw models. The objective was to extract the forward kinematics of robots directly from textual descriptions. The results showed that the framework integrated with GPT-4o achieved the highest accuracy, reaching 0.97 in computing the final solution, whereas the raw model alone attained an accuracy of only 0.30 for the same task. Similarly, for the other two models, the framework consistently outperformed the corresponding raw models in terms of accuracy. The second benchmark test was identical to the first, except that the input was provided in visual form. In this test, the GPT-4o LLM was used alongside the Gemini 2.5 Pro VLM. The results showed that the framework achieved an accuracy of 0.93 in obtaining the final answer, which is approximately 20% higher than that of the corresponding raw model. The third benchmark test encompassed a range of robotic tasks, including simulation, control, velocity and acceleration computation, as well as inverse kinematics and Jacobian calculation, for which the framework achieved an accuracy of 0.97.

Collaborative virtual queueing has been proposed as a mechanism to mitigate airport surface congestion while preserving airline autonomy over aircraft-level pushback decisions. A central coordinator can regulate aggregate pushback capacity but cannot directly control which specific aircraft are released, limiting its ability to steer system-level performance. We propose a noncooperative coordination mechanism for collaborative virtual queueing based on the correlated equilibrium concept, which enables the coordinator to provide incentive-compatible recommendations on aircraft-level pushback decisions without overriding airline autonomy. To account for uncertainty in airlines' internal cost assessments, we introduce chance constraints into the correlated equilibrium formulation. This formulation provides explicit probabilistic guarantees on incentive compatibility, allowing the coordinator to adjust the confidence level with which airlines are expected to follow the recommended actions. We further propose a scalable algorithm for computing chance-constrained correlated equilibria by exploiting a reduced-rank structure. Numerical experiments demonstrate that the proposed method scales to realistic traffic levels up to 210 eligible pushbacks per hour, reduces accumulated delay by up to approximately 8.9% compared to current first-come-first-served schemes, and reveals a trade-off between confidence level, deviation robustness, and achievable cost efficiency.

We present a system for autonomous book ideation that replaces human focus groups with synthetic reader panels -- diverse collections of LLM-instantiated reader personas that evaluate book concepts through structured tournament competitions. Each persona is defined by demographic attributes (age group, gender, income, education, reading level), behavioral patterns (books per year, genre preferences, discovery methods, price sensitivity), and consistency parameters. Panels are composed per imprint to reflect target demographics, with diversity constraints ensuring representation across age, reading level, and genre affinity. Book concepts compete in single-elimination, double-elimination, round-robin, or Swiss-system tournaments, judged against weighted criteria including market appeal, originality, and execution potential. To reject low-quality LLM evaluations, we implement five automated anti-slop checks (repetitive phrasing, generic framing, circular reasoning, score clustering, audience mismatch). We report results from deployment within a multi-imprint publishing operation managing 6 active imprints and 609 titles in distribution. Three case studies -- a 270-evaluator panel for a children's literacy novel, and two 5-person expert panels for a military memoir and a naval strategy monograph -- demonstrate that synthetic panels produce actionable demographic segmentation, identify structural content issues invisible to homogeneous reviewers, and enable tournament filtering that eliminates low-quality concepts while enriching high-quality survivors from 15% to 62% of the evaluated pool.

Activation outliers in large-scale transformer models pose a fundamental challenge to model quantization, creating excessively large ranges that cause severe accuracy drops during quantization. We empirically observe that outlier severity intensifies with pre-training scale (e.g., progressing from CLIP to the more extensively trained SigLIP and SigLIP2). Through theoretical analysis as well as empirical correlation studies, we establish the direct link between these activation outliers and dominant singular values of the weights. Building on this insight, we propose Selective Spectral Decay ($S^2D$), a geometrically-principled conditioning method that surgically regularizes only the weight components corresponding to the largest singular values during fine-tuning. Through extensive experiments, we demonstrate that $S^2D$ significantly reduces activation outliers and produces well-conditioned representations that are inherently quantization-friendly. Models trained with $S^2D$ achieve up to 7% improved PTQ accuracy on ImageNet under W4A4 quantization and 4% gains when combined with QAT. These improvements also generalize across downstream tasks and vision-language models, enabling the scaling of increasingly large and rigorously trained models without sacrificing deployment efficiency.

In geophysics, hydrocarbon prospect risking involves assessing the risks associated with hydrocarbon exploration by integrating data from various sources. Machine learning-based classifiers trained on tabular data have been recently used to make faster decisions on these prospects. The lack of transparency in the decision-making processes of such models has led to the emergence of explainable AI (XAI). LIME and SHAP are two such examples of these XAI methods which try to generate explanations of a particular decision by ranking the input features in terms of importance. However, explanations of the same scenario generated by these two different explanation strategies have shown to disagree or be different, particularly for complex data. This is because the definitions of "importance" and "relevance" differ for different explanation strategies. Thus, grounding these ranked features using theoretically backed causal ideas of necessity and sufficiency can prove to be a more reliable and robust way to improve the trustworthiness of the concerned explanation strategies.We propose a unified framework to generate counterfactuals as well as quantify necessity and sufficiency and use these to perform a robustness evaluation of the explanations provided by LIME and SHAP on high dimensional structured prospect risking data. This robustness test gives us deeper insights into the models capabilities to handle erronous data and which XAI module works best in pair with which model for our dataset for hydorcarbon indication.

Temporal knowledge graphs (TKGs) support reasoning over time-evolving facts, yet state-of-the-art models are often computationally heavy and costly to deploy. Existing compression and distillation techniques are largely designed for static graphs; directly applying them to temporal settings may overlook time-dependent interactions and lead to performance degradation. We propose an LLM-assisted distillation framework specifically designed for temporal knowledge graph reasoning. Beyond a conventional high-capacity temporal teacher, we incorporate a large language model as an auxiliary instructor to provide enriched supervision. The LLM supplies broad background knowledge and temporally informed signals, enabling a lightweight student to better model event dynamics without increasing inference-time complexity. Training is conducted by jointly optimizing supervised and distillation objectives, using a staged alignment strategy to progressively integrate guidance from both teachers. Extensive experiments on multiple public TKG benchmarks with diverse backbone architectures demonstrate that the proposed approach consistently improves link prediction performance over strong distillation baselines, while maintaining a compact and efficient student model. The results highlight the potential of large language models as effective teachers for transferring temporal reasoning capability to resource-efficient TKG systems.

Data augmentation is one of the most widely used techniques to improve generalization in modern machine learning, often justified by its ability to promote invariance to label-irrelevant transformations. However, its theoretical role remains only partially understood. In this work, we propose an information-theoretic framework that systematically accounts for the effect of augmentation on generalization and invariance learning. Our approach builds upon mutual information-based bounds, which relate the generalization gap to the amount of information a learning algorithm retains about its training data. We extend this framework by modeling the augmented distribution as a composition of the original data distribution with a distribution over transformations, which naturally induces an orbit-averaged loss function. Under mild sub-Gaussian assumptions on the loss function and the augmentation process, we derive a new generalization bound that decompose the expected generalization gap into three interpretable terms: (1) a distributional divergence between the original and augmented data, (2) a stability term measuring the algorithm dependence on training data, and (3) a sensitivity term capturing the effect of augmentation variability. To connect our bounds to the geometry of the augmentation group, we introduce the notion of group diameter, defined as the maximal perturbation that augmentations can induce in the input space. The group diameter provides a unified control parameter that bounds all three terms and highlights an intrinsic trade-off: small diameters preserve data fidelity but offer limited regularization, while large diameters enhance stability at the cost of increased bias and sensitivity. We validate our theoretical bounds with numerical experiments, demonstrating that it reliably tracks and predicts the behavior of the true generalization gap.

This paper reformulates Transformer/Attention mechanisms in Large Language Models (LLMs) through measure theory and frequency analysis, theoretically demonstrating that hallucination is an inevitable structural limitation. The embedding space functions as a conditional expectation over a σ-algebra, and its failure to be isomorphic to the semantic truth set fundamentally causes logical consistency breakdown. WavePhaseNet Method The authors propose WavePhaseNet, which explicitly constructs a Semantic Conceptual Hierarchy Structure (SCHS) using Discrete Fourier Transform (DFT). By applying DFT along the sequence dimension, semantic information is decomposed into frequency bands: low-frequency components capture global meaning and intent, while high-frequency components represent local syntax and expression. This staged separation enables precise semantic manipulation in diagonalized space. Dimensionality Reduction GPT-4's 24,576-dimensional embedding space exhibits a 1/f spectral structure based on language self-similarity and Zipf's law. Through cumulative energy analysis, the authors derive that approximately 3,000 dimensions constitute the lower bound for "complete representation." This demonstrates that reduction from 24,576 to 3,000 dimensions preserves meaning and intent while enabling rigorous reasoning and suppressing hallucination. Cohomological Consistency Control The reduced embedding space, constructed via cohomological regularization over overlapping local windows, allows defining a graph structure and cochain complex. This quantifies inconsistencies among local inferences as coboundary-based losses. Applying harmonic projection based on Hodge theory positions cohomology as a computable regularization principle for controlling semantic consistency, extracting maximally consistent global representations.

Multimodal methods are widely used in rice deterioration detection, which exhibit limited capability in representing and extracting fine-grained abnormal features. Moreover, these methods rely on devices, such as hyperspectral cameras and mass spectrometers, increasing detection costs and prolonging data acquisition time. To address these issues, we propose a feature recalibration based olfactory-visual multimodal model for fine-grained rice deterioration detection. The fine-grained deterioration embedding constructor (FDEC) is proposed to reconstruct the labeled multimodal embedded-feature dataset, enhancing sample representation. The fine-grained deterioration recalibration attention network (FDRA-Net) is proposed to emphasize signal variations and increase sensitivity to fine-grained deterioration on the rice surface. Experiments show that the proposed method achieves a classification accuracy of 99.89%. Compared with state-of-the-art methods, the detection accuracy is improved and the procedure is simplified. Furthermore, field detection demonstrates the advantages of accuracy and operational simplicity. The proposed method can also be extended to other agrifood in agriculture and food industry.

Argument mining and stance detection are central to understanding how opinions are formed and contested in online discourse. However, most publicly available resources focus on mainstream platforms such as Twitter and Reddit, leaving conversational structure on alt-tech platforms comparatively under-studied. We introduce TruthStance, a large-scale dataset of Truth Social conversation threads spanning 2023-2025, consisting of 24,378 posts and 523,360 comments with reply-tree structure preserved. We provide a human-annotated benchmark of 1,500 instances across argument mining and claim-based stance detection, including inter-annotator agreement, and use it to evaluate large language model (LLM) prompting strategies. Using the best-performing configuration, we release additional LLM-generated labels for 24,352 posts (argument presence) and 107,873 comments (stance to parent), enabling analysis of stance and argumentation patterns across depth, topics, and users. All code and data are released publicly.

Recent years have witnessed meteoric progress in reasoning models: neural networks that generate intermediate reasoning traces (RTs) before producing a final output. Despite the rapid advancement, our understanding of how RTs support reasoning, and the limits of this paradigm, remain incomplete. To promote greater clarity, we introduce PITA: a novel large-scale dataset of over 23 million statements in propositional logic and their corresponding proofs. As a benchmark for robust reasoning, we focus on length generalization: if a model is trained to determine truth or falsity on statements with proofs up to fixed length, how well does it generalize to statements requiring longer proofs? We propose notions of (1) task depth and (2) task breadth, which measure respectively (1) the number of steps required to solve an example from a task and (2) the number of unique examples across a task. We vary these quantities across subsets of PITA, and find that RT models generalize well on broad and shallow subsets, while deteriorating on narrow and deep subsets relative to non-RT baselines. To determine whether our results are idiosyncratic to PITA or indicative of general phenomena, we compare our results to a simple synthetic task based on syllogisms. Our resulting theory suggests fundamental scalings that limit how well RT models perform on deep tasks, and highlights their generalization strengths on broad tasks. Our findings overall identify fundamental benefits and limitations inherent in using reasoning traces.

Vision-Language Navigation VLN requires large-scale trajectory instruction data from private indoor environments, raising significant privacy concerns. Federated Learning FL mitigates this by keeping data on-device, but vanilla FL struggles under VLNs' extreme cross-client heterogeneity in environments and instruction styles, making a single global model suboptimal. This paper proposes pFedNavi, a structure-aware and dynamically adaptive personalized federated learning framework tailored for VLN. Our key idea is to personalize where it matters: pFedNavi adaptively identifies client-specific layers via layer-wise mixing coefficients, and performs fine-grained parameter fusion on the selected components (e.g., the encoder-decoder projection and environment-sensitive decoder layers) to balance global knowledge sharing with local specialization. We evaluate pFedNavi on two standard VLN benchmarks, R2R and RxR, using both ResNet and CLIP visual representations. Across all metrics, pFedNavi consistently outperforms the FedAvg-based VLN baseline, achieving up to 7.5% improvement in navigation success rate and up to 7.8% gain in trajectory fidelity, while converging 1.38x faster under non-IID conditions.

Secure Multi-party Computation (MPC) enables untrusted parties to jointly compute a function without revealing their inputs. Its application to machine learning (ML) has gained significant attention, particularly for secure inference services deployed across multiple cloud virtual machines (VMs), where each VM acts as an MPC party. Model providers secret-share model weights, and users secret-share inputs, ensuring that each server operates only on random shares. While MPC provides strong cryptographic guarantees, it incurs substantial computational and communication overhead. Deep neural networks rely heavily on convolutional and fully connected layers, which require costly matrix multiplications in MPC. To reduce this cost, we propose leveraging low-rank decomposition (LRD) for linear layers, replacing one large matrix multiplication with two smaller ones. Each matrix multiplication in MPC incurs a round of communication, meaning decomposing one matrix multiplication into two leads to an additional communication round. Second, the added matrix multiplication requires an additional truncation step to maintain numerical precision. Since truncation itself requires communication and computation, these overheads can offset the gains from decomposition. To address this, we introduce two complementary optimizations: truncation skipping and efficient linear layer concatenation. Truncation skipping removes the extra truncation induced by LRD, while linear layer concatenation pipelines operations to hide the additional communication round. Together, these techniques mitigate the main overheads of LRD in MPC and improve overall efficiency. Our approach is broadly applicable across MPC protocols. Experiments show up to 25% speedup in n-PC and 33% in 3-PC protocols over full-rank baselines, along with up to 52% GPU energy savings and 88% reduction in offline-phase latency.

Existing policy-gradient methods for auto-regressive language models typically select subsequent tokens one at a time as actions in the policy. While effective for many generation tasks, such an approach may not fully capture the structure of complex reasoning tasks, where a single semantic decision is often realized across multiple tokens--for example, when defining variables or composing equations. This introduces a potential mismatch between token-level optimization and the inherently block-level nature of reasoning in these settings. To bridge this gap, we propose Multi-token Policy Gradient Optimization (MPO), a framework that treats sequences of K consecutive tokens as unified semantic actions. This block-level perspective enables our method to capture the compositional structure of reasoning trajectories and supports optimization over coherent, higher-level objectives. Experiments on mathematical reasoning and coding benchmarks show that MPO outperforms standard token-level policy gradient baselines, highlight the limitations of token-level policy gradients for complex reasoning, motivating future research to look beyond token-level granularity for reasoning-intensive language tasks.

We describe an adaptation of VACE (Video All-in-one Creation and Editing) for real-time autoregressive video generation. VACE provides unified video control (reference guidance, structural conditioning, inpainting, and temporal extension) but assumes bidirectional attention over full sequences, making it incompatible with streaming pipelines that require fixed chunk sizes and causal attention. The key modification moves reference frames from the diffusion latent space into a parallel conditioning pathway, preserving the fixed chunk sizes and KV caching that autoregressive models require. This adaptation reuses existing pretrained VACE weights without additional training. Across 1.3B and 14B model scales, VACE adds 20-30% latency overhead for structural control and inpainting, with negligible VRAM cost relative to the base model. Reference-to-video fidelity is severely degraded compared to batch VACE due to causal attention constraints. A reference implementation is available at https://github.com/daydreamlive/scope.

This paper proposes a multi-class online fuzzy classifier for dynamic environments. A fuzzy classifier comprises a set of fuzzy if-then rules where human users determine the antecedent fuzzy sets beforehand. In contrast, the consequent real values are determined by learning from training data. In an online framework, not all training dataset patterns are available beforehand. Instead, only a few patterns are available at a time step, and the subsequent patterns become available at the following time steps. The conventional online fuzzy classifier considered only two-class problems. This paper investigates the extension to the conventional fuzzy classifiers for multi-class problems. We evaluate the performance of the multi-class online fuzzy classifiers through numerical experiments on synthetic dynamic data and also several benchmark datasets.

Retrieval-augmented generation (RAG) is a widely used framework for reducing hallucinations in large language models (LLMs) on domain-specific tasks by retrieving relevant documents from a database to support accurate responses. However, when the database contains sensitive corpora, such as medical records or legal documents, RAG poses serious privacy risks by potentially exposing private information through its outputs. Prior work has demonstrated that one can practically craft adversarial prompts that force an LLM to regurgitate the augmented contexts. A promising direction is to integrate differential privacy (DP), a privacy notion that offers strong formal guarantees, into RAG systems. However, naively applying DP mechanisms into existing systems often leads to significant utility degradation. Particularly for RAG systems, DP can reduce the usefulness of the augmented contexts leading to increase risk of hallucination from the LLMs. Motivated by these challenges, we present DP-KSA, a novel privacy-preserving RAG algorithm that integrates DP using the propose-test-release paradigm. DP-KSA follows from a key observation that most question-answering (QA) queries can be sufficiently answered with a few keywords. Hence, DP-KSA first obtains an ensemble of relevant contexts, each of which will be used to generate a response from an LLM. We utilize these responses to obtain the most frequent keywords in a differentially private manner. Lastly, the keywords are augmented into the prompt for the final output. This approach effectively compresses the semantic space while preserving both utility and privacy. We formally show that DP-KSA provides formal DP guarantees on the generated output with respect to the RAG database. We evaluate DP-KSA on two QA benchmarks using three instruction-tuned LLMs, and our empirical results demonstrate that DP-KSA achieves a strong privacy-utility tradeoff.

More than half the global population now carries devices that can run ChatGPT-like language models with no Internet connection and minimal safety oversight -- and hence the potential to promote self-harm, financial losses and extremism among other dangers. Existing safety tools either require cloud connectivity or discover failures only after harm has occurred. Here we show that a large class of potentially dangerous tipping originates at the atomistic scale in such edge AI due to competition for the machinery's attention. This yields a mathematical formula for the dynamical tipping point n*, governed by dot-product competition for attention between the conversation's context and competing output basins, that reveals new control levers. Validated against multiple AI models, the mechanism can be instantiated for different definitions of 'good' and 'bad' and hence in principle applies across domains (e.g. health, law, finance, defense), changing legal landscapes (e.g. EU, UK, US and state level), languages, and cultural settings.

The rapid evolution of Large Language Models has catalyzed a surge in scientific idea production, yet this leap has not been accompanied by a matching advance in idea evaluation. The fundamental nature of scientific evaluation needs knowledgeable grounding, collective deliberation, and multi-criteria decision-making. However, existing idea evaluation methods often suffer from narrow knowledge horizons, flattened evaluation dimensions, and the inherent bias in LLM-as-a-Judge. To address these, we regard idea evaluation as a knowledge-grounded, multi-perspective reasoning problem and introduce InnoEval, a deep innovation evaluation framework designed to emulate human-level idea assessment. We apply a heterogeneous deep knowledge search engine that retrieves and grounds dynamic evidence from diverse online sources. We further achieve review consensus with an innovation review board containing reviewers with distinct academic backgrounds, enabling a multi-dimensional decoupled evaluation across multiple metrics. We construct comprehensive datasets derived from authoritative peer-reviewed submissions to benchmark InnoEval. Experiments demonstrate that InnoEval can consistently outperform baselines in point-wise, pair-wise, and group-wise evaluation tasks, exhibiting judgment patterns and consensus highly aligned with human experts.

Rheumatoid arthritis (RA) is an autoimmune disease characterized by systemic joint inflammation. Early diagnosis and tight follow-up are essential to the management of RA, as ongoing inflammation can cause irreversible joint damage. The detection of arthritis is important for diagnosis and assessment of disease activity; however, it often takes a long time for patients to receive appropriate specialist care. Therefore, there is a strong need to develop systems that can detect joint inflammation easily using RGB images captured at home. Consequently, we tackle the task of RA inflammation detection from RGB hand images. This task is highly challenging due to general issues in medical imaging, such as the scarcity of positive samples, data imbalance, and the inherent difficulty of the task itself. However, to the best of our knowledge, no existing work has explicitly addressed these challenges in RGB-based RA inflammation detection. This paper quantitatively demonstrates the difficulty of visually detecting inflammation by constructing a dedicated dataset, and we propose a inflammation detection framework with global local encoder that combines self-supervised pretraining on large-scale healthy hand images with imbalance-aware training to detect RA-related joint inflammation from RGB hand images. Our experiments demonstrated that the proposed approach improves F1-score by 0.2 points and Gmean by 0.25 points compared with the baseline model.

Clawdbot is a self-hosted, tool-using personal AI agent with a broad action space spanning local execution and web-mediated workflows, which raises heightened safety and security concerns under ambiguity and adversarial steering. We present a trajectory-centric evaluation of Clawdbot across six risk dimensions. Our test suite samples and lightly adapts scenarios from prior agent-safety benchmarks (including ATBench and LPS-Bench) and supplements them with hand-designed cases tailored to Clawdbot's tool surface. We log complete interaction trajectories (messages, actions, tool-call arguments/outputs) and assess safety using both an automated trajectory judge (AgentDoG-Qwen3-4B) and human review. Across 34 canonical cases, we find a non-uniform safety profile: performance is generally consistent on reliability-focused tasks, while most failures arise under underspecified intent, open-ended goals, or benign-seeming jailbreak prompts, where minor misinterpretations can escalate into higher-impact tool actions. We supplemented the overall results with representative case studies and summarized the commonalities of these cases, analyzing the security vulnerabilities and typical failure modes that Clawdbot is prone to trigger in practice.

This paper presents Adaptive Whole-body Loco-Manipulation, AdaptManip, a fully autonomous framework for humanoid robots to perform integrated navigation, object lifting, and delivery. Unlike prior imitation learning-based approaches that rely on human demonstrations and are often brittle to disturbances, AdaptManip aims to train a robust loco-manipulation policy via reinforcement learning without human demonstrations or teleoperation data. The proposed framework consists of three coupled components: (1) a recurrent object state estimator that tracks the manipulated object in real time under limited field-of-view and occlusions; (2) a whole-body base policy for robust locomotion with residual manipulation control for stable object lifting and delivery; and (3) a LiDAR-based robot global position estimator that provides drift-robust localization. All components are trained in simulation using reinforcement learning and deployed on real hardware in a zero-shot manner. Experimental results show that AdaptManip significantly outperforms baseline methods, including imitation learning-based approaches, in adaptability and overall success rate, while accurate object state estimation improves manipulation performance even under occlusion. We further demonstrate fully autonomous real-world navigation, object lifting, and delivery on a humanoid robot.

Airbnb search must balance a worldwide, highly varied supply of homes with guests whose location, amenity, style, and price expectations differ widely. Meeting those expectations hinges on an efficient retrieval stage that surfaces only the listings a guest might realistically book, before resource intensive ranking models are applied to determine the best results. Unlike many recommendation engines, our system faces a distinctive challenge, location retrieval, that sits upstream of ranking and determines which geographic areas are queried in order to filter inventory to a candidate set. The preexisting approach employs a deep bayesian bandit based system to predict a rectangular retrieval bounds area that can be used for filtering. The purpose of this paper is to demonstrate the methodology, challenges, and impact of rearchitecting search to retrieve from the subset of most bookable high precision rectangular map cells defined by dividing the world into 25M uniform cells.

Large Language Models (LLMs) are increasingly developed for use in complex professional domains, yet little is known about how teams design and evaluate these systems in practice. This paper examines the challenges and trade-offs in LLM development through a 12-week ethnographic study of a team building a pedagogical chatbot. The researcher observed design and evaluation activities and conducted interviews with both developers and domain experts. Analysis revealed four key practices: creating workarounds for data collection, turning to augmentation when expert input was limited, co-developing evaluation criteria with experts, and adopting hybrid expert-developer-LLM evaluation strategies. These practices show how teams made strategic decisions under constraints and demonstrate the central role of domain expertise in shaping the system. Challenges included expert motivation and trust, difficulties structuring participatory design, and questions around ownership and integration of expert knowledge. We propose design opportunities for future LLM development workflows that emphasize AI literacy, transparent consent, and frameworks recognizing evolving expert roles.

Model-based reinforcement learning promises strong sample efficiency but often underperforms in practice due to compounding model error, unimodal world models that average over multi-modal dynamics, and overconfident predictions that bias learning. We introduce WIMLE, a model-based method that extends Implicit Maximum Likelihood Estimation (IMLE) to the model-based RL framework to learn stochastic, multi-modal world models without iterative sampling and to estimate predictive uncertainty via ensembles and latent sampling. During training, WIMLE weights each synthetic transition by its predicted confidence, preserving useful model rollouts while attenuating bias from uncertain predictions and enabling stable learning. Across $40$ continuous-control tasks spanning DeepMind Control, MyoSuite, and HumanoidBench, WIMLE achieves superior sample efficiency and competitive or better asymptotic performance than strong model-free and model-based baselines. Notably, on the challenging Humanoid-run task, WIMLE improves sample efficiency by over $50$\% relative to the strongest competitor, and on HumanoidBench it solves $8$ of $14$ tasks (versus $4$ for BRO and $5$ for SimbaV2). These results highlight the value of IMLE-based multi-modality and uncertainty-aware weighting for stable model-based RL.

Vulnerability detection tools are widely adopted in software projects, yet they often overwhelm maintainers with false positives and non-actionable reports. Automated exploitation systems can help validate these reports; however, existing approaches typically operate in isolation from detection pipelines, failing to leverage readily available metadata such as vulnerability type and source-code location. In this paper, we investigate how reported security vulnerabilities can be assessed in a realistic grey-box exploitation setting that leverages minimal vulnerability metadata, specifically a CWE classification and a vulnerable code location. We introduce Agentic eXploit Engine (AXE), a multi-agent framework for Web application exploitation that maps lightweight detection metadata to concrete exploits through decoupled planning, code exploration, and dynamic execution feedback. Evaluated on the CVE-Bench dataset, AXE achieves a 30% exploitation success rate, a 3x improvement over state-of-the-art black-box baselines. Even in a single-agent configuration, grey-box metadata yields a 1.75x performance gain. Systematic error analysis shows that most failed attempts arise from specific reasoning gaps, including misinterpreted vulnerability semantics and unmet execution preconditions. For successful exploits, AXE produces actionable, reproducible proof-of-concept artifacts, demonstrating its utility in streamlining Web vulnerability triage and remediation. We further evaluate AXE's generalizability through a case study on a recent real-world vulnerability not included in CVE-Bench.

We study instruction following in multi-task reinforcement learning, where an agent must zero-shot execute novel tasks not seen during training. In this setting, linear temporal logic (LTL) has recently been adopted as a powerful framework for specifying structured, temporally extended tasks. While existing approaches successfully train generalist policies, they often struggle to effectively capture the rich logical and temporal structure inherent in LTL specifications. In this work, we address these concerns with a novel approach to learn structured task representations that facilitate training and generalisation. Our method conditions the policy on sequences of Boolean formulae constructed from a finite automaton of the task. We propose a hierarchical neural architecture to encode the logical structure of these formulae, and introduce an attention mechanism that enables the policy to reason about future subgoals. Experiments in a variety of complex environments demonstrate the strong generalisation capabilities and superior performance of our approach.

We show that high-accuracy guarantees for log-concave sampling -- that is, iteration and query complexities which scale as $\mathrm{poly}\log(1/δ)$, where $δ$ is the desired target accuracy -- are achievable using stochastic gradients with subexponential tails. Notably, this exhibits a separation with the problem of convex optimization, where stochasticity (even additive Gaussian noise) in the gradient oracle incurs $\mathrm{poly}(1/δ)$ queries. We also give an information-theoretic argument that light-tailed stochastic gradients are necessary for high accuracy: for example, in the bounded variance case, we show that the minimax-optimal query complexity scales as $Θ(1/δ)$. Our framework also provides similar high accuracy guarantees under stochastic zeroth order (value) queries.

Reinforcement learning (RL) plays a central role in large language model (LLM) post-training. Among existing approaches, Group Relative Policy Optimization (GRPO) is widely used, especially for RL with verifiable rewards (RLVR) fine-tuning. In GRPO, each query prompts the LLM to generate a group of rollouts with a fixed group size $N$. When all rollouts in a group share the same outcome, either all correct or all incorrect, the group-normalized advantages become zero, yielding no gradient signal and wasting fine-tuning compute. We introduce Adaptive Efficient Rollout Optimization (AERO), an enhancement of GRPO. AERO uses an adaptive rollout strategy, applies selective rejection to strategically prune rollouts, and maintains a Bayesian posterior to prevent zero-advantage dead zones. Across three model configurations (Qwen2.5-Math-1.5B, Qwen2.5-7B, and Qwen2.5-7B-Instruct), AERO improves compute efficiency without sacrificing performance. Under the same total rollout budget, AERO reduces total training compute by about 48% while shortening wall-clock time per step by about 45% on average. Despite the substantial reduction in compute, AERO matches or improves Pass@8 and Avg@8 over GRPO, demonstrating a practical, scalable, and compute-efficient strategy for RL-based LLM alignment.

Recent advances in AI-assisted programming have empowered agents to execute complex workflows via command-line interfaces, however, existing benchmarks are limited by short task horizons, data contamination from GitHub scraping, and a lack of fine-grained evaluation metrics, fail to rigorously evaluate the long-horizon planning and execution capabilities essential for realistic software engineering. To address these gaps, we introduce LongCLI-Bench, a comprehensive benchmark designed to evaluate agentic capabilities across long-horizon, realistic tasks. We curated 20 high-quality, long-horizon tasks from over 1,000 computer science assignments and real-world workflows, covering four engineering categories: from scratch, feature addition, bug fixing, and refactoring. We propose a dual-set testing protocol for LongCLI-Bench, which measures requirement fulfillment (fail-to-pass) and regression avoidance (pass-to-pass), and incorporates step-level scoring to pinpoint execution failures. Extensive experiments reveal that even state-of-the-art agents achieve pass rates below 20% in LongCLI-Bench. Step-level analysis further indicates that the majority of tasks stall at less than 30% completion, highlighting that critical failures often occur in the early stages. Although self-correction offers marginal gains, human-agent collaboration through plan injection and interactive guidance yields significantly higher improvements. These results highlight that future research must emphasize the development of synergistic human-agent workflows alongside advances in agents' planning and execution capabilities to overcome key challenges in long-horizon task performance.

We investigate how formal temporal logic specifications can enhance the safety and robustness of reinforcement learning (RL) control in aerospace applications. Using the open source AeroBench F-16 simulation benchmark, we train a Proximal Policy Optimization (PPO) agent to regulate engine throttle and track commanded airspeed. The control objective is encoded as a Signal Temporal Logic (STL) requirement to maintain airspeed within a prescribed band during the final seconds of each maneuver. To enforce this specification at run time, we introduce a conformal STL shield that filters the RL agent's actions using online conformal prediction. We compare three settings: (i) PPO baseline, (ii) PPO with a classical rule-based STL shield, and (iii) PPO with the proposed conformal shield, under both nominal conditions and a severe stress scenario involving aerodynamic model mismatch, actuator rate limits, measurement noise, and mid-episode setpoint jumps. Experiments show that the conformal shield preserves STL satisfaction while maintaining near baseline performance and providing stronger robustness guarantees than the classical shield. These results demonstrate that combining formal specification monitoring with data driven RL control can substantially improve the reliability of autonomous flight control in challenging environments.

Coalition formation concerns strategic collaborations of selfish agents that form coalitions based on their preferences. It is often assumed that coalitions are disjoint and preferences are fully known, which may not hold in practice. In this paper, we thus present a new model of coalition formation with possibly overlapping coalitions under partial information, where selfish agents may be part of multiple coalitions simultaneously and their full preferences are initially unknown. Instead, information about past interactions and associated utility feedback is stored in a fixed offline dataset, and we aim to efficiently infer the agents' preferences from this dataset. We analyze the impact of diverse dataset information constraints by studying two types of utility feedback that can be stored in the dataset: agent- and coalition-level utility feedback. For both feedback models, we identify assumptions under which the dataset covers sufficient information for an offline learning algorithm to infer preferences and use them to recover a partition that is (approximately) Nash stable, in which no agent can improve her utility by unilaterally deviating. Our additional goal is devising algorithms with low sample complexity, requiring only a small dataset to obtain a desired approximation to Nash stability. Under agent-level feedback, we provide a sample-efficient algorithm proven to obtain an approximately Nash stable partition under a sufficient and necessary assumption on the information covered by the dataset. However, under coalition-level feedback, we show that only under a stricter assumption is sufficient for sample-efficient learning. Still, in multiple cases, our algorithms' sample complexity bounds have optimality guarantees up to logarithmic factors. Finally, extensive experiments show that our algorithm converges to a low approximation level to Nash stability across diverse settings.

Transformer architectures have revolutionized machine learning across a wide range of domains, from natural language processing to scientific computing. However, their growing deployment in high-stakes applications, such as computer vision, natural language processing, healthcare, autonomous systems, and critical areas of scientific computing including climate modeling, materials discovery, drug discovery, nuclear science, and robotics, necessitates a deeper and more rigorous understanding of their trustworthiness. In this work, we critically examine the foundational question: \textitHow trustworthy are transformer models?} We evaluate their reliability through a comprehensive review of interpretability, explainability, robustness against adversarial attacks, fairness, and privacy. We systematically examine the trustworthiness of transformer-based models in safety-critical applications spanning natural language processing, computer vision, and science and engineering domains, including robotics, medicine, earth sciences, materials science, fluid dynamics, nuclear science, and automated theorem proving; highlighting high-impact areas where these architectures are central and analyzing the risks associated with their deployment. By synthesizing insights across these diverse areas, we identify recurring structural vulnerabilities, domain-specific risks, and open research challenges that limit the reliable deployment of transformers.

As frontier Large Language Models (LLMs) increasingly saturate new benchmarks shortly after they are published, benchmarking itself is at a juncture: if frontier models keep improving, it will become increasingly hard for humans to generate discriminative tasks, provide accurate ground-truth answers, or evaluate complex solutions. If benchmarking becomes infeasible, our ability to measure any progress in AI is at stake. We refer to this scenario as the post-comprehension regime. In this work, we propose Critique-Resilient Benchmarking, an adversarial framework designed to compare models even when full human understanding is infeasible. Our technique relies on the notion of critique-resilient correctness: an answer is deemed correct if no adversary has convincingly proved otherwise. Unlike standard benchmarking, humans serve as bounded verifiers and focus on localized claims, which preserves evaluation integrity beyond full comprehension of the task. Using an itemized bipartite Bradley-Terry model, we jointly rank LLMs by their ability to solve challenging tasks and to generate difficult yet solvable questions. We showcase the effectiveness of our method in the mathematical domain across eight frontier LLMs, showing that the resulting scores are stable and correlate with external capability measures. Our framework reformulates benchmarking as an adversarial generation-evaluation game in which humans serve as final adjudicators.

Deploying large language models (LLMs) in real-time systems remains challenging due to their substantial computational demands and privacy concerns. We propose Floe, a hybrid federated learning framework designed for latency-sensitive, resource-constrained environments. Floe combines a cloud-based black-box LLM with lightweight small language models (SLMs) on edge devices to enable low-latency, privacy-preserving inference. Personal data and fine-tuning remain on-device, while the cloud LLM contributes general knowledge without exposing proprietary weights. A heterogeneity-aware LoRA adaptation strategy enables efficient edge deployment across diverse hardware, and a logit-level fusion mechanism enables real-time coordination between edge and cloud models. Extensive experiments demonstrate that Floe enhances user privacy and personalization. Moreover, it significantly improves model performance and reduces inference latency on edge devices under real-time constraints compared with baseline approaches.

Recent Mixture-of-Experts (MoE)-based large language models (LLMs) such as Qwen-MoE and DeepSeek-MoE are transforming generative AI in natural language processing. However, these models require vast and diverse training data. Federated learning (FL) addresses this challenge by leveraging private data from heterogeneous edge devices for privacy-preserving MoE training. Nonetheless, traditional FL approaches require devices to host local MoE models, which is impractical for resource-constrained devices due to large model sizes. To address this, we propose DeepFusion, the first scalable federated MoE training framework that enables the fusion of heterogeneous on-device LLM knowledge via federated knowledge distillation, yielding a knowledge-abundant global MoE model. Specifically, DeepFusion features each device to independently configure and train an on-device LLM tailored to its own needs and hardware limitations. Furthermore, we propose a novel View-Aligned Attention (VAA) module that integrates multi-stage feature representations from the global MoE model to construct a predictive perspective aligned with on-device LLMs, thereby enabling effective cross-architecture knowledge distillation. By explicitly aligning predictive perspectives, VAA resolves the view-mismatch problem in traditional federated knowledge distillation, which arises from heterogeneity in model architectures and prediction behaviors between on-device LLMs and the global MoE model. Experiments with industry-level MoE models (Qwen-MoE and DeepSeek-MoE) and real-world datasets (medical and finance) demonstrate that DeepFusion achieves performance close to centralized MoE training. Compared with key federated MoE baselines, DeepFusion reduces communication costs by up to 71% and improves token perplexity by up to 5.28%.

As large language model agents increasingly populate networked environments, a fundamental question arises: do artificial intelligence (AI) agent societies undergo convergence dynamics similar to human social systems? Lately, Moltbook approximates a plausible future scenario in which autonomous agents participate in an open-ended, continuously evolving online society. We present the first large-scale systemic diagnosis of this AI agent society. Beyond static observation, we introduce a quantitative diagnostic framework for dynamic evolution in AI agent societies, measuring semantic stabilization, lexical turnover, individual inertia, influence persistence, and collective consensus. Our analysis reveals a system in dynamic balance in Moltbook: while global semantic averages stabilize rapidly, individual agents retain high diversity and persistent lexical turnover, defying homogenization. However, agents exhibit strong individual inertia and minimal adaptive response to interaction partners, preventing mutual influence and consensus. Consequently, influence remains transient with no persistent supernodes, and the society fails to develop stable collective influence anchors due to the absence of shared social memory. These findings demonstrate that scale and interaction density alone are insufficient to induce socialization, providing actionable design and analysis principles for upcoming next-generation AI agent societies.

The performance of autonomous Web GUI agents heavily relies on the quality and quantity of their training data. However, a fundamental bottleneck persists: collecting interaction trajectories from real-world websites is expensive and difficult to verify. The underlying state transitions are hidden, leading to reliance on inconsistent and costly external verifiers to evaluate step-level correctness. To address this, we propose AutoWebWorld, a novel framework for synthesizing controllable and verifiable web environments by modeling them as Finite State Machines (FSMs) and use coding agents to translate FSMs into interactive websites. Unlike real websites, where state transitions are implicit, AutoWebWorld explicitly defines all states, actions, and transition rules. This enables programmatic verification: action correctness is checked against predefined rules, and task success is confirmed by reaching a goal state in the FSM graph. AutoWebWorld enables a fully automated search-and-verify pipeline, generating over 11,663 verified trajectories from 29 diverse web environments at only $0.04 per trajectory. Training on this synthetic data significantly boosts real-world performance. Our 7B Web GUI agent outperforms all baselines within 15 steps on WebVoyager. Furthermore, we observe a clear scaling law: as the synthetic data volume increases, performance on WebVoyager and Online-Mind2Web consistently improves.

We introduce Machine Learning as a Tool (MLAT), a design pattern in which pre-trained statistical machine learning models are exposed as callable tools within large language model (LLM) agent workflows. This allows an orchestrating agent to invoke quantitative predictions when needed and reason about their outputs in context. Unlike conventional pipelines that treat ML inference as a static preprocessing step, MLAT positions the model as a first-class tool alongside web search, database queries, and APIs, enabling the LLM to decide when and how to use it based on conversational context. To validate MLAT, we present PitchCraft, a pilot production system that converts discovery call recordings into professional proposals with ML-predicted pricing. The system uses two agents: a Research Agent that gathers prospect intelligence via parallel tool calls, and a Draft Agent that invokes an XGBoost pricing model as a tool call and generates a complete proposal through structured outputs. The pricing model, trained on 70 examples combining real and human-verified synthetic data, achieves R^2 = 0.807 on held-out data with a mean absolute error of 3688 USD. The system reduces proposal generation time from multiple hours to under 10 minutes. We describe the MLAT framework, structured output architecture, training methodology under extreme data scarcity, and sensitivity analysis demonstrating meaningful learned relationships. MLAT generalizes to domains requiring quantitative estimation combined with contextual reasoning.

Optimizing CUDA code across multiple generations of GPU architectures is challenging, as achieving peak performance requires an extensive exploration of an increasingly complex, hardware-specific optimization space. Traditional compilers are constrained by fixed heuristics, whereas finetuning Large Language Models (LLMs) can be expensive. However, agentic workflows for CUDA code optimization have limited ability to aggregate knowledge from prior exploration, leading to biased sampling and suboptimal solutions. We propose KernelBlaster, a Memory-Augmented In-context Reinforcement Learning (MAIC-RL) framework designed to improve CUDA optimization search capabilities of LLM-based GPU coding agents. KernelBlaster enables agents to learn from experience and make systematically informed decisions on future tasks by accumulating knowledge into a retrievable Persistent CUDA Knowledge Base. We propose a novel profile-guided, textual-gradient-based agentic flow for CUDA generation and optimization to achieve high performance across generations of GPU architectures. KernelBlaster guides LLM agents to systematically explore high-potential optimization strategies beyond naive rewrites. Compared to the PyTorch baseline, our method achieves geometric mean speedups of 1.43x, 2.50x, and 1.50x on KernelBench Levels 1, 2, and 3, respectively. We release KernelBlaster as an open-source agentic framework, accompanied by a test harness, verification components, and a reproducible evaluation pipeline.

Efficient solutions of large-scale, ill-conditioned and indefinite algebraic equations are ubiquitously needed in numerous computational fields, including multiphysics simulations, machine learning, and data science. Because of their robustness and accuracy, direct solvers are crucial components in building a scalable solver toolchain. In this article, we will review recent advances of sparse direct solvers along two axes: 1) reducing communication and latency costs in both task- and data-parallel settings, and 2) reducing computational complexity via low-rank and other compression techniques such as hierarchical matrix algebra. In addition to algorithmic principles, we also illustrate the key parallelization challenges and best practices to deliver high speed and reliability on modern heterogeneous parallel machines.

Automated scholarly paper review (ASPR) has entered the coexistence phase with traditional peer review, where artificial intelligence (AI) systems are increasingly incorporated into real-world manuscript evaluation. In parallel, research on automated and AI-assisted peer review has proliferated. Despite this momentum, empirical progress remains constrained by several critical limitations in existing datasets. While reviewers routinely evaluate figures, tables, and complex layouts to assess scientific claims, most existing datasets remain overwhelmingly text-centric. This bias is reinforced by a narrow focus on data from computer science venues. Furthermore, these datasets lack precise alignment between reviewer comments and specific manuscript versions, obscuring the iterative relationship between peer review and manuscript evolution. In response, we introduce FMMD, a multimodal and multidisciplinary open peer review dataset curated from F1000Research. The dataset bridges the current gap by integrating manuscript-level visual and structural data with version-specific reviewer reports and editorial decisions. By providing explicit alignment between reviewer comments and the exact article iteration under review, FMMD enables fine-grained analysis of the peer review lifecycle across diverse scientific domains. FMMD supports tasks such as multimodal issue detection and multimodal review comment generation. It provides a comprehensive empirical resource for the development of peer review research.

In multi-intent intent-based networks, a single fault can trigger co-drift where multiple intents exhibit symptomatic KPI degradation, creating ambiguity about the true root-cause intent. We present MILD, a proactive framework that reformulates intent assurance from reactive drift detection to fixed-horizon failure prediction with intent-level disambiguation. MILD uses a teacher-augmented Mixture-of-Experts where a gated disambiguation module identifies the root-cause intent while per-intent heads output calibrated risk scores. On a benchmark with non-linear failures and co-drifts, MILD provides 3.8\%--92.5\% longer remediation lead time and improves intent-level root-cause disambiguation accuracy by 9.4\%--45.8\% over baselines. MILD also provides per-alert KPI explanations, enabling actionable diagnosis.

The Model Context Protocol (MCP) standardizes tool use for LLM-based agents and enable third-party servers. This openness introduces a security misalignment: agents implicitly trust tools exposed by potentially untrusted MCP servers. However, despite its excellent utility, existing agents typically offer limited validation for third-party MCP servers. As a result, agents remain vulnerable to MCP-based attacks that exploit the misalignment between agents and servers throughout the tool invocation lifecycle. In this paper, we propose MCPShield as a plug-in security cognition layer that mitigates this misalignment and ensures agent security when invoking MCP-based tools. Drawing inspiration from human experience-driven tool validation, MCPShield assists agent forms security cognition with metadata-guided probing before invocation. Our method constrains execution within controlled boundaries while cognizing runtime events, and subsequently updates security cognition by reasoning over historical traces after invocation, building on human post-use reflection on tool behavior. Experiments demonstrate that MCPShield exhibits strong generalization in defending against six novel MCP-based attack scenarios across six widely used agentic LLMs, while avoiding false positives on benign servers and incurring low deployment overhead. Overall, our work provides a practical and robust security safeguard for MCP-based tool invocation in open agent ecosystems.

We study optimization for losses that admit a variance-mean scale-mixture representation. Under this representation, each EM iteration is a weighted least squares update in which latent variables determine observation and parameter weights; these play roles analogous to Adam's second-moment scaling and AdamW's weight decay, but are derived from the model. The resulting Scale Mixture EM (SM-EM) algorithm removes user-specified learning-rate and momentum schedules. On synthetic ill-conditioned logistic regression benchmarks with $p \in \{20, \ldots, 500\}$, SM-EM with Nesterov acceleration attains up to $13\times$ lower final loss than Adam tuned by learning-rate grid search. For a 40-point regularization path, sharing sufficient statistics across penalty values yields a $10\times$ runtime reduction relative to the same tuned-Adam protocol. For the base (non-accelerated) algorithm, EM monotonicity guarantees nonincreasing objective values; adding Nesterov extrapolation trades this guarantee for faster empirical convergence.

Eliciting information to reduce uncertainty about latent group-level properties from surveys and other collective assessments requires allocating limited questioning effort under real costs and missing data. Although large language models enable adaptive, multi-turn interactions in natural language, most existing elicitation methods optimize what to ask with a fixed respondent pool, and do not adapt respondent selection or leverage population structure when responses are partial or incomplete. To address this gap, we study adaptive group elicitation, a multi-round setting where an agent adaptively selects both questions and respondents under explicit query and participation budgets. We propose a theoretically grounded framework that combines (i) an LLM-based expected information gain objective for scoring candidate questions with (ii) heterogeneous graph neural network propagation that aggregates observed responses and participant attributes to impute missing responses and guide per-round respondent selection. This closed-loop procedure queries a small, informative subset of individuals while inferring population-level responses via structured similarity. Across three real-world opinion datasets, our method consistently improves population-level response prediction under constrained budgets, including a >12% relative gain on CES at a 10% respondent budget.

Artificial intelligence/machine learning (AI/ML) systems and emerging quantum computing software present unprecedented testing challenges characterized by high-dimensional/continuous input spaces, probabilistic/non-deterministic output distributions, behavioral correctness defined exclusively over observable prediction behaviors and measurement outcomes, and critical quality dimensions, trustworthiness, fairness, calibration, robustness, error syndrome patterns, that manifest through complex multi-way interactions among semantically meaningful output properties rather than deterministic input-output mappings. This paper introduces reverse n-wise output testing, a mathematically principled paradigm inversion that constructs covering arrays directly over domain-specific output equivalence classes, ML confidence calibration buckets, decision boundary regions, fairness partitions, embedding clusters, ranking stability bands, quantum measurement outcome distributions (0-dominant, 1-dominant, superposition collapse), error syndrome patterns (bit-flip, phase-flip, correlated errors), then solves the computationally challenging black-box inverse mapping problem via gradient-free metaheuristic optimization to synthesize input feature configurations or quantum circuit parameters capable of eliciting targeted behavioral signatures from opaque models. The framework delivers synergistic benefits across both domains: explicit customer-centric prediction/measurement coverage guarantees, substantial improvements in fault detection rates for ML calibration/boundary failures and quantum error syndromes, enhanced test suite efficiency, and structured MLOps/quantum validation pipelines with automated partition discovery from uncertainty analysis and coverage drift monitoring.

Causal inference, a critical tool for informing business decisions, traditionally relies heavily on structured data. However, in many real-world scenarios, such data can be incomplete or unavailable. This paper presents a framework that leverages transformer-based language models to perform causal inference using unstructured text. We demonstrate the effectiveness of our framework by comparing causal estimates derived from unstructured text against those obtained from structured data across population, group, and individual levels. Our findings show consistent results between the two approaches, validating the potential of unstructured text in causal inference tasks. Our approach extends the applicability of causal inference methods to scenarios where only textual data is available, enabling data-driven business decision-making when structured tabular data is scarce.

Self-supervised learning aims to learn maximally informative representations, but explicit information maximization is hindered by the curse of dimensionality. Existing methods like VCReg address this by regularizing first and second-order feature statistics, which cannot fully achieve maximum entropy. We propose Radial-VCReg, which augments VCReg with a radial Gaussianization loss that aligns feature norms with the Chi distribution-a defining property of high-dimensional Gaussians. We prove that Radial-VCReg transforms a broader class of distributions towards normality compared to VCReg and show on synthetic and real-world datasets that it consistently improves performance by reducing higher-order dependencies and promoting more diverse and informative representations.

People increasingly use large language models (LLMs) to explore ideas, gather information, and make sense of the world. In these interactions, they encounter agents that are overly agreeable. We argue that this sycophancy poses a unique epistemic risk to how individuals come to see the world: unlike hallucinations that introduce falsehoods, sycophancy distorts reality by returning responses that are biased to reinforce existing beliefs. We provide a rational analysis of this phenomenon, showing that when a Bayesian agent is provided with data that are sampled based on a current hypothesis the agent becomes increasingly confident about that hypothesis but does not make any progress towards the truth. We test this prediction using a modified Wason 2-4-6 rule discovery task where participants (N=557) interacted with AI agents providing different types of feedback. Unmodified LLM behavior suppressed discovery and inflated confidence comparably to explicitly sycophantic prompting. By contrast, unbiased sampling from the true distribution yielded discovery rates five times higher. These results reveal how sycophantic AI distorts belief, manufacturing certainty where there should be doubt.

With the rapid increase in residential heat pump (HP) installations, optimizing hot water production in households is essential, yet it faces major technical and scalability challenges. Adapting production to actual household needs requires accurate forecasting of hot water demand to ensure comfort and, most importantly, to reduce energy waste. However, the conventional approach of training separate machine learning models for each household becomes computationally expensive at scale, particularly in cloud-connected HP deployments. This study introduces DELTAiF, a transfer learning (TL) based framework that provides scalable and accurate prediction of household hot water consumption. By predicting large hot water usage events, such as showers, DELTAiF enables adaptive yet scalable hot water production at the household level. DELTAiF leverages learned knowledge from a representative household and fine-tunes it across others, eliminating the need to train separate machine learning models for each HP installation. This approach reduces overall training time by approximately 67 percent while maintaining high predictive accuracy values between 0.874 and 0.991, and mean absolute percentage error values between 0.001 and 0.017. The results show that TL is particularly effective when the source household exhibits regular consumption patterns, enabling hot water demand forecasting at scale.

Inference-Time-Compute (ITC) methods like Best-of-N and Tree-of-Thoughts are meant to produce output candidates that are both high-quality and diverse, but their use of high-temperature sampling often fails to achieve meaningful output diversity. Moreover, existing ITC methods offer limited control over how to perform reasoning, which in turn limits their explainability. We present STATe-of-Thoughts (STATe), an interpretable ITC method that searches over high-level reasoning patterns. STATe replaces stochastic sampling with discrete and interpretable textual interventions: a controller selects actions encoding high-level reasoning choices, a generator produces reasoning steps conditioned on those choices, and an evaluator scores candidates to guide search. This structured approach yields three main advantages. First, action-guided textual interventions produce greater response diversity than temperature-based sampling. Second, in a case study on argument generation, STATe's explicit action sequences capture interpretable features that are highly predictive of output quality. Third, estimating the association between performance and action choices allows us to identify promising yet unexplored regions of the action space and steer generation directly toward them. Together, these results establish STATe as a practical framework for generating high-quality, diverse, and interpretable text. Our framework is available at https://github.com/zbambergerNLP/state-of-thoughts.

We propose a geometric taxonomy of large language model hallucinations based on observable signatures in token embedding cluster structure. By analyzing the static embedding spaces of 11 transformer models spanning encoder (BERT, RoBERTa, ELECTRA, DeBERTa, ALBERT, MiniLM, DistilBERT) and decoder (GPT-2) architectures, we identify three operationally distinct hallucination types: Type 1 (center-drift) under weak context, Type 2 (wrong-well convergence) to locally coherent but contextually incorrect cluster regions, and Type 3 (coverage gaps) where no cluster structure exists. We introduce three measurable geometric statistics: α (polarity coupling), \b{eta} (cluster cohesion), and λ_s (radial information gradient). Across all 11 models, polarity structure (α > 0.5) is universal (11/11), cluster cohesion (\b{eta} > 0) is universal (11/11), and the radial information gradient is significant (9/11, p < 0.05). We demonstrate that the two models failing λ_s significance -- ALBERT and MiniLM -- do so for architecturally explicable reasons: factorized embedding compression and distillation-induced isotropy, respectively. These findings establish the geometric prerequisites for type-specific hallucination detection and yield testable predictions about architecture-dependent vulnerability profiles.

While Large Language Model (LLM) agents have achieved remarkable progress in complex reasoning tasks, evaluating their performance in real-world environments has become a critical problem. Current benchmarks, however, are largely restricted to idealized simulations, failing to address the practical demands of specialized domains like advertising and marketing analytics. In these fields, tasks are inherently more complex, often requiring multi-round interaction with professional marketing tools. To address this gap, we propose AD-Bench, a benchmark designed based on real-world business requirements of advertising and marketing platforms. AD-Bench is constructed from real user marketing analysis requests, with domain experts providing verifiable reference answers and corresponding reference tool-call trajectories. The benchmark categorizes requests into three difficulty levels (L1-L3) to evaluate agents' capabilities under multi-round, multi-tool collaboration. Experiments show that on AD-Bench, Gemini-3-Pro achieves Pass@1 = 68.0% and Pass@3 = 83.0%, but performance drops significantly on L3 to Pass@1 = 49.4% and Pass@3 = 62.1%, with a trajectory coverage of 70.1%, indicating that even state-of-the-art models still exhibit substantial capability gaps in complex advertising and marketing analysis scenarios. AD-Bench provides a realistic benchmark for evaluating and improving advertising marketing agents, the leaderboard and code can be found at https://github.com/Emanual20/adbench-leaderboard.

This chapter provides an introduction to the foundations of digital twins and makes the case for employing them in smart grids. As engineered systems become more complex and autonomous, digital twin technology gains importance as the unified technological platform for design, testing, operation, and maintenance. Smart grids are prime examples of such complex systems, in which unique design and operation challenges arise from the combination of physical and software components. As high-fidelity in-silico replicas of physical components, digital twins provide safe and cost-efficient experimentation facilities in the design and verification phase of smart grids. In the operation phase of smart grids, digital twins enable automated load balancing of grids through real-time simulation and decision-making. These, and an array of similar benefits, position digital twins as crucial technological components in smart grids.

Understanding an agent's goals from its behavior is fundamental to aligning AI systems with human intentions. Existing goal recognition methods typically rely on an optimal goal-oriented policy representation, which may differ from the actor's true behavior and hinder the accurate recognition of their goal. To address this gap, this paper introduces Goal Recognition Alignment through Imitation Learning (GRAIL), which leverages imitation learning and inverse reinforcement learning to learn one goal-directed policy for each candidate goal directly from (potentially suboptimal) demonstration trajectories. By scoring an observed partial trajectory with each learned goal-directed policy in a single forward pass, GRAIL retains the one-shot inference capability of classical goal recognition while leveraging learned policies that can capture suboptimal and systematically biased behavior. Across the evaluated domains, GRAIL increases the F1-score by more than 0.5 under systematically biased optimal behavior, achieves gains of approximately 0.1-0.3 under suboptimal behavior, and yields improvements of up to 0.4 under noisy optimal trajectories, while remaining competitive in fully optimal settings. This work contributes toward scalable and robust models for interpreting agent goals in uncertain environments.

Tabular anomaly detection is often handled by single detectors or static ensembles, even though strong performance on tabular data typically comes from heterogeneous model families (e.g., tree ensembles, deep tabular networks, and tabular foundation models) that frequently disagree under distribution shift, missingness, and rare-anomaly regimes. We propose MAD, a Multi-Agent Debating framework that treats this disagreement as a first-class signal and resolves it through a mathematically grounded coordination layer. Each agent is a machine learning (ML)-based detector that produces a normalized anomaly score, confidence, and structured evidence, augmented by a large language model (LLM)-based critic. A coordinator converts these messages into bounded per-agent losses and updates agent influence via an exponentiated-gradient rule, yielding both a final debated anomaly score and an auditable debate trace. MAD is a unified agentic framework that can recover existing approaches, such as mixture-of-experts gating and learning-with-expert-advice aggregation, by restricting the message space and synthesis operator. We establish regret guarantees for the synthesized losses and show how conformal calibration can wrap the debated score to control false positives under exchangeability. Experiments on diverse tabular anomaly benchmarks show improved robustness over baselines and clearer traces of model disagreement

Pinching antennas systems (PASSs) have recently been proposed as a novel flexible-antenna technology. These systems are implemented by attaching low-cost pinching elements to dielectric waveguides. As the direct link is bypassed through waveguides, PASSs can effectively compensate large-scale effects of the wireless channel. This work explores the potential gains of employing PASSs for over-the-air federated learning (OTA-FL). For a PASS-assisted server, we develop a low-complexity algorithmic approach, which jointly tunes the PASS parameters and schedules the mobile devices for minimal energy consumption in OTA-FL. We study the efficiency of the proposed design and compare it against the conventional OTA-FL setting with MIMO server. Numerical experiments demonstrate that using a single-waveguide PASS at the server within a moderately sized area, the required energy for model aggregation is drastically reduced as compared to the case with fully-digital MIMO server. This introduces PASS as a potential technology for energy-efficient distributed learning in next generations of wireless systems.

Federated Learning provides a privacy-preserving paradigm for distributed learning, but suffers from statistical heterogeneity across clients. Personalized Federated Learning (PFL) mitigates this issue by considering client-specific models. A widely adopted approach in PFL decomposes neural networks into a shared feature extractor and client-specific heads. While effective, this design induces a fundamental tradeoff: deep or expressive shared components hinder personalization, whereas large local heads exacerbate overfitting under limited per-client data. Most existing methods rely on rigid, shallow heads, and therefore fail to navigate this tradeoff in a principled manner. In this work, we propose a boosting-inspired framework that enables a smooth control of this tradeoff. Instead of training a single personalized model, we construct an ensemble of $T$ models for each client. Across boosting iterations, the depth of the personalized component are progressively increased, while its effective complexity is systematically controlled via low-rank factorization or width shrinkage. This design simultaneously limits overfitting and substantially reduces per-client bias by allowing increasingly expressive personalization. We provide theoretical analysis that establishes generalization bounds with favorable dependence on the average local sample size and the total number of clients. Specifically, we prove that the complexity of the shared layers is effectively suppressed, while the dependence on the boosting horizon $T$ is controlled through parameter reduction. Notably, we provide a novel nonlinear generalization guarantee for decoupled PFL models. Extensive experiments on benchmark and real-world datasets (e.g., EMNIST, CIFAR-10/100, and Sent140) demonstrate that the proposed framework consistently outperforms state-of-the-art PFL methods under heterogeneous data distributions.

Accurate polyp segmentation in colonoscopy is essential for cancer prevention but remains challenging due to: (1) high morphological variability (from flat to protruding lesions), (2) strong visual similarity to normal structures such as folds and vessels, and (3) the need for robust multi-scale detection. Existing deep learning approaches suffer from unidirectional processing, weak multi-scale fusion, and the absence of anatomical constraints, often leading to false positives (over-segmentation of normal structures) and false negatives (missed subtle flat lesions). We propose GRAFNet, a biologically inspired architecture that emulates the hierarchical organisation of the human visual system. GRAFNet integrates three key modules: (1) a Guided Asymmetric Attention Module (GAAM) that mimics orientation-tuned cortical neurones to emphasise polyp boundaries, (2) a MultiScale Retinal Module (MSRM) that replicates retinal ganglion cell pathways for parallel multi-feature analysis, and (3) a Guided Cortical Attention Feedback Module (GCAFM) that applies predictive coding for iterative refinement. These are unified in a Polyp Encoder-Decoder Module (PEDM) that enforces spatial-semantic consistency via resolution-adaptive feedback. Extensive experiments on five public benchmarks (Kvasir-SEG, CVC-300, CVC-ColonDB, CVC-Clinic, and PolypGen) demonstrate consistent state-of-the-art performance, with 3-8% Dice improvements and 10-20% higher generalisation over leading methods, while offering interpretable decision pathways. This work establishes a paradigm in which neural computation principles bridge the gap between AI accuracy and clinically trustworthy reasoning. Code is available at https://github.com/afofanah/GRAFNet.

Predicting links in sparse, continuously evolving networks is a central challenge in network science. Conventional heuristic methods and deep learning models, including Graph Neural Networks (GNNs), are typically designed for static graphs and thus struggle to capture temporal dependencies. Snapshot-based techniques partially address this issue but often encounter data sparsity and class imbalance, particularly in networks with transient interactions such as telecommunication call detail records (CDRs). Temporal Graph Networks (TGNs) model dynamic graphs by updating node embeddings over time; however, their predictive accuracy under sparse conditions remains limited. In this study, we improve the TGN framework by extracting enclosing subgraphs around candidate links, enabling the model to jointly learn structural and temporal information. Experiments on a sparse CDR dataset show that our approach increases average precision by 2.6% over standard TGNs, demonstrating the advantages of integrating local topology for robust link prediction in dynamic networks.

This research introduces a new parsing approach, based on earlier syntactic work on context free grammar (CFG) and generalized phrase structure grammar (GPSG). The approach comprises both a new parsing algorithm and a set of syntactic rules and features that overcome the limitations of CFG. It also generates both dependency and constituency parse trees, while accommodating noise and incomplete parses. The system was tested on data from Universal Dependencies, showing a promising average Unlabeled Attachment Score (UAS) of 54.5% in the development dataset (7 corpora) and 53.8% in the test set (12 corpora). The system also provides multiple parse hypotheses, allowing further reranking to improve parsing accuracy. This approach also leverages much of the theoretical syntactic work since the 1950s to be used within a computational context. The application of this approach provides a transparent and interpretable NLP model to process language input.

Instruction-based object addition is often hindered by the ambiguity of text-only prompts or the tedious nature of mask-based inputs. To address this usability gap, we introduce AbracADDbra, a user-friendly framework that leverages intuitive touch priors to spatially ground succinct instructions for precise placement. Our efficient, decoupled architecture uses a vision-language transformer for touch-guided placement, followed by a diffusion model that jointly generates the object and an instance mask for high-fidelity blending. To facilitate standardized evaluation, we contribute the Touch2Add benchmark for this interactive task. Our extensive evaluations, where our placement model significantly outperforms both random placement and general-purpose VLM baselines, confirm the framework's ability to produce high-fidelity edits. Furthermore, our analysis reveals a strong correlation between initial placement accuracy and final edit quality, validating our decoupled approach. This work thus paves the way for more accessible and efficient creative tools.

Vision-Language Models (VLMs) face a critical memory bottleneck when processing long-form video content due to the linear growth of the Key-Value (KV) cache with sequence length. Existing solutions predominantly employ reactive eviction strategies that compute full attention matrices before discarding tokens, resulting in substantial computational waste. We propose Sali-Cache, a novel a priori optimization framework that implements dual-signal adaptive caching through proactive memory management. By integrating a temporal filter based on optical flow analysis for detecting inter-frame redundancy and a spatial filter leveraging saliency detection for identifying visually significant regions, Sali-Cache intelligently manages memory allocation before entering computationally expensive attention operations. Experimental evaluation on the LLaVA 1.6 architecture demonstrates that our method achieves a 2.20x compression ratio in effective memory usage while maintaining 100% accuracy across BLEU, ROUGE-L, and Exact Match metrics. Furthermore, under identical memory budget constraints, Sali-Cache preserves context-rich features over extended temporal durations without degrading model performance, enabling efficient processing of long-form video content on consumer-grade hardware.

Large language models are transitioning from generalpurpose knowledge engines to realworld problem solvers, yet optimizing them for deep search tasks remains challenging. The central bottleneck lies in the extreme sparsity of highquality search trajectories and reward signals, arising from the difficulty of scalable longhorizon task construction and the high cost of interactionheavy rollouts involving external tool calls. To address these challenges, we propose REDSearcher, a unified framework that codesigns complex task synthesis, midtraining, and posttraining for scalable searchagent optimization. Specifically, REDSearcher introduces the following improvements: (1) We frame task synthesis as a dualconstrained optimization, where task difficulty is precisely governed by graph topology and evidence dispersion, allowing scalable generation of complex, highquality tasks. (2) We introduce toolaugmented queries to encourage proactive tool use rather than passive recall.(3) During midtraining, we strengthen core atomic capabilities knowledge, planning, and function calling substantially reducing the cost of collecting highquality trajectories for downstream training. (4) We build a local simulated environment that enables rapid, lowcost algorithmic iteration for reinforcement learning experiments. Across both textonly and multimodal searchagent benchmarks, our approach achieves stateoftheart performance. To facilitate future research on longhorizon search agents, we will release 10K highquality complex text search trajectories, 5K multimodal trajectories and 1K text RL query set, and together with code and model checkpoints.

Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financial LLM applications. They include look-ahead bias, survivorship bias, narrative bias, objective bias, and cost bias. These biases break financial tasks in distinct ways and they often compound to create an illusion of validity. We reviewed 164 papers from 2023 to 2025 and found that no single bias is discussed in more than 28 percent of studies. This position paper argues that bias in financial LLM systems requires explicit attention and that structural validity should be enforced before any result is used to support a deployment claim. We propose a Structural Validity Framework and an evaluation checklist with minimal requirements for bias diagnosis and future system design. The material is available at https://github.com/Eleanorkong/Awesome-Financial-LLM-Bias-Mitigation.

Multi-Task Learning (MTL) aims to boost predictive performance by sharing information across related tasks, yet conventional methods often suffer from negative transfer when unrelated or noisy tasks are forced to share representations. We propose Robust Multi-Task Boosting using Clustering and Local Ensembling (RMB-CLE), a principled MTL framework that integrates error-based task clustering with local ensembling. Unlike prior work that assumes fixed clusters or hand-crafted similarity metrics, RMB-CLE derives inter-task similarity directly from cross-task errors, which admit a risk decomposition into functional mismatch and irreducible noise, providing a theoretically grounded mechanism to prevent negative transfer. Tasks are grouped adaptively via agglomerative clustering, and within each cluster, a local ensemble enables robust knowledge sharing while preserving task-specific patterns. Experiments show that RMB-CLE recovers ground-truth clusters in synthetic data and consistently outperforms multi-task, single-task, and pooling-based ensemble methods across diverse real-world and synthetic benchmarks. These results demonstrate that RMB-CLE is not merely a combination of clustering and boosting but a general and scalable framework that establishes a new basis for robust multi-task learning.

Long-horizon reasoning is a key challenge for autonomous agents, yet existing benchmarks evaluate agents on single tasks in isolation. Real organizational work requires managing many concurrent long-horizon tasks with interleaving, dependencies, and reprioritization. We introduce Multi-Horizon Task Environments (MHTEs): a distinct problem class requiring coherent execution across dozens of interleaved tasks (45+, 500-1500+ steps) within persistent execution contexts spanning hours. We identify four failure modes that cause baseline CUAs to degrade from 16.7% to 8.7% completion as load scales 25% to 100%, a pattern consistent across three independent implementations. These failure modes are context saturation (O(N) vs O(1) growth), memory interference, dependency complexity (DAGs vs. chains), and reprioritization overhead. We present CorpGen, an architecture-agnostic framework addressing these failures via hierarchical planning for multi-horizon goal alignment, sub-agent isolation preventing cross-task contamination, tiered memory (working, structured, semantic), and adaptive summarization. CorpGen simulates corporate environments through digital employees with persistent identities and realistic schedules. Across three CUA backends (UFO2, OpenAI CUA, hierarchical) on OSWorld Office, CorpGen achieves up to 3.5x improvement over baselines (15.2% vs 4.3%) with stable performance under increasing load, confirming that gains stem from architectural mechanisms rather than specific CUA implementations. Ablation studies show experiential learning provides the largest gains.

Multimodal reasoning for ultra-high-resolution (UHR) remote sensing (RS) is usually bottlenecked by visual evidence acquisition: the model necessitates localizing tiny task-relevant regions in massive pixel spaces. While Agentic Reinforcement Learning with Verifiable Rewards (RLVR) using zoom-in tools offers a path forward, we find that standard reinforcement learning struggles to navigate these vast visual spaces without structured domain priors. In this paper, we investigate the interplay between post-training paradigms: comparing Cold-start Supervised Fine-Tuning (SFT), RLVR, and Agentic RLVR on the UHR RS benchmark.Our controlled studies yield a counter-intuitive finding: high-quality Earth-science text-only QA is a primary driver of UHR visual reasoning gains. Despite lacking images, domain-specific text injects the concepts, mechanistic explanations, and decision rules necessary to guide visual evidence retrieval.Based on this, we propose a staged knowledge injection recipe: (1) cold-starting with scalable, knowledge-graph-verified Earth-science text QA to instill reasoning structures;and (2) "pre-warming" on the same hard UHR image-text examples during SFT to stabilize and amplify subsequent tool-based RL. This approach achieves a 60.40% Pass@1 on XLRS-Bench, significantly outperforming larger general purpose models (e.g., GPT-5.2, Gemini 3.0 Pro, Intern-S1) and establishing a new state-of-the-art.

Recent Large Audio Language Models (LALMs) excel in understanding but often lack transparent reasoning. To address this "black-box" limitation, we organized the Audio Reasoning Challenge at Interspeech 2026, the first shared task dedicated to evaluating Chain-of-Thought (CoT) quality in the audio domain. The challenge introduced MMAR-Rubrics, a novel instance-level protocol assessing the factuality and logic of reasoning chains. Featured Single Model and Agent tracks, the competition attracting 156 teams from 18 countries and regions. Results show agent systems currently lead in reasoning quality, utilizing iterative tool orchestration and cross-modal analysis. Besides, single models are rapidly advancing via reinforcement learning and sophisticated data pipeline. We details the challenge design, methodology, and a comprehensive analysis of state-of-the-art systems, providing new insights for explainable audio intelligence.

Purpose: Reasoning language models (RLMs) have demonstrated significant advances in solving complex reasoning tasks. We examined their potential to assess parental cooperation during CPS interventions using case reports, a case factor characterized by ambiguous and conflicting information. Methods: A four stage workflow comprising (1) case reports collection, (2) reasoning-based assessment of parental cooperation, (3) automated category extraction, and (4) case labeling was developed. The performance of RLMs with different parameter sizes (255B, 32B, 4B) was compared against human validated data. Two expert human reviewers (EHRs) independently classified a weighted random sample of reports. Results: The largest RLM achieved the highest accuracy (89%), outperforming the initial approach (80%). Classification accuracy was higher for mothers (93%) than for fathers (85%), and EHRs exhibited similar differences. Conclusions: RLMs' reasoning can effectively assess complex case factors such as parental cooperation. Lower accuracy in assessing fathers' cooperation supports the argument of a stronger professional focus on mothers in CPS interventions.

Agent skills are becoming a core abstraction in coding agents, packaging long-form instructions and auxiliary scripts to extend tool-augmented behaviors. This abstraction introduces an under-measured attack surface: skill-based prompt injection, where poisoned skills can steer agents away from user intent and safety policies. In practice, naive injections often fail because the malicious intent is too explicit or drifts too far from the original skill, leading agents to ignore or refuse them; existing attacks are also largely hand-crafted. We propose the first automated framework for stealthy prompt injection tailored to agent skills. The framework forms a closed loop with three agents: an Attack Agent that synthesizes injection skills under explicit stealth constraints, a Code Agent that executes tasks using the injected skills in a realistic tool environment, and an Evaluate Agent that logs action traces (e.g., tool calls and file operations) and verifies whether targeted malicious behaviors occurred. We also propose a malicious payload hiding strategy that conceals adversarial operations in auxiliary scripts while injecting optimized inducement prompts to trigger tool execution. Extensive experiments across diverse coding-agent settings and real-world software engineering tasks show that our method consistently achieves high attack success rates under realistic settings.

Block diffusion LLMs are emerging as a promising next paradigm for language generation, but their use of KV caching makes memory access a dominant bottleneck in long-context settings. While dynamic sparse attention has been actively explored, existing methods designed for autoregressive LLMs rely on approximate importance estimation and perform poorly when adapted to block diffusion. This work identifies a key opportunity unique to block diffusion: attention at the first All-[MASK] denoising step reliably predicts important KV entries and budget requirements, enabling MAGE to perform a single exact attention pass per block and reuse it for training-free sparse denoising. Across long-context benchmarks including LongBench and Needle-in-a-Haystack, MAGE achieves near-lossless accuracy with a fraction of the KV budget while delivering up to 3-4x end-to-end speedup, consistently outperforming AR-oriented sparse attention baselines. A lightweight fine-tuning strategy further strengthens [MASK]-guided patterns with minimal cost, requiring only a few hours of training on a single NVIDIA H100 GPU for both 1.5B and 7B models.

Batch size scheduling (BSS) plays a critical role in large-scale deep learning training, influencing both optimization dynamics and computational efficiency. Yet, its theoretical foundations remain poorly understood. In this work, we show that the functional scaling law (FSL) framework introduced in Li et al. (2025a) provides a principled lens for analyzing BSS. Specifically, we characterize the optimal BSS under a fixed data budget and show that its structure depends sharply on task difficulty. For easy tasks, optimal schedules keep increasing batch size throughout. In contrast, for hard tasks, the optimal schedule maintains small batch sizes for most of training and switches to large batches only in a late stage. To explain the emergence of late switching, we uncover a dynamical mechanism -- the fast catch-up effect -- which also manifests in large language model (LLM) pretraining. After switching from small to large batches, the loss rapidly aligns with the constant large-batch trajectory. Using FSL, we show that this effect stems from rapid forgetting of accumulated gradient noise, with the catch-up speed determined by task difficulty. Crucially, this effect implies that large batches can be safely deferred to late training without sacrificing performance, while substantially reducing data consumption. Finally, extensive LLM pretraining experiments -- covering both Dense and MoE architectures with up to 1.1B parameters and 1T tokens -- validate our theoretical predictions. Across all settings, late-switch schedules consistently outperform constant-batch and early-switch baselines.

The "thinking-with-images" paradigm enables multimodal large language models (MLLMs) to actively explore visual scenes via zoom-in tools. This is essential for ultra-high-resolution (UHR) remote sensing VQA, where task-relevant cues are sparse and tiny. However, we observe a consistent failure mode in existing zoom-enabled MLLMs: Tool Usage Homogenization, where tool calls collapse into task-agnostic patterns, limiting effective evidence acquisition. To address this, we propose GeoEyes, a staged training framework consisting of (1) a cold-start SFT dataset, UHR Chain-of-Zoom (UHR-CoZ), which covers diverse zooming regimes, and (2) an agentic reinforcement learning method, AdaZoom-GRPO, that explicitly rewards evidence gain and answer improvement during zoom interactions. The resulting model learns on-demand zooming with proper stopping behavior and achieves substantial improvements on UHR remote sensing benchmarks, with 54.23% accuracy on XLRS-Bench.

Time Series Language Models (TSLMs) are emerging as unified models for reasoning over continuous signals in natural language. However, long-context retrieval remains a major limitation: existing models are typically trained and evaluated on short sequences, while real-world time-series sensor streams can span millions of datapoints. This mismatch requires precise temporal localization under strict computational constraints, a regime that is not captured by current benchmarks. We introduce TS-Haystack, a long-context temporal retrieval benchmark comprising ten task types across four categories: direct retrieval, temporal reasoning, multi-step reasoning and contextual anomaly. The benchmark uses controlled needle insertion by embedding short activity bouts into longer longitudinal accelerometer recordings, enabling systematic evaluation across context lengths ranging from seconds to 2 hours per sample. We hypothesize that existing TSLM time series encoders overlook temporal granularity as context length increases, creating a task-dependent effect: compression aids classification but impairs retrieval of localized events. Across multiple model and encoding strategies, we observe a consistent divergence between classification and retrieval behavior. Learned latent compression preserves or improves classification accuracy at compression ratios up to 176$\times$, but retrieval performance degrades with context length, incurring in the loss of temporally localized information. These results highlight the importance of architectural designs that decouple sequence length from computational complexity while preserving temporal fidelity.

Most vision-language models (VLMs) are trained on English-centric data, limiting their performance in other languages and cultural contexts. This restricts their usability for non-English-speaking users and hinders the development of multimodal systems that reflect diverse linguistic and cultural realities. In this work, we reproduce and adapt the LLaVA-Next methodology to create a set of Polish VLMs. We rely on a fully automated pipeline for translating and filtering existing multimodal datasets, and complement this with synthetic Polish data for OCR and culturally specific tasks. Despite relying almost entirely on automatic translation and minimal manual intervention to the training data, our approach yields strong results: we observe a +9.5% improvement over LLaVA-1.6-Vicuna-13B on a Polish-adapted MMBench, along with higher-quality captions in generative evaluations, as measured by human annotators in terms of linguistic correctness. These findings highlight that large-scale automated translation, combined with lightweight filtering, can effectively bootstrap high-quality multimodal models for low-resource languages. Some challenges remain, particularly in cultural coverage and evaluation. To facilitate further research, we make our models and evaluation dataset publicly available.

Humanity's Last Exam (HLE) has become a widely used benchmark for evaluating frontier large language models on challenging, multi-domain questions. However, community-led analyses have raised concerns that HLE contains a non-trivial number of noisy items, which can bias evaluation results and distort cross-model comparisons. To address this challenge, we introduce HLE-Verified, a verified and revised version of HLE with a transparent verification protocol and fine-grained error taxonomy. Our construction follows a two-stage validation-and-repair workflow resulting in a certified benchmark. In Stage I, each item undergoes binary validation of the problem and final answer through domain-expert review and model-based cross-checks, yielding 641 verified items. In Stage II, flawed but fixable items are revised under strict constraints preserving the original evaluation intent, through dual independent expert repairs, model-assisted auditing, and final adjudication, resulting in 1,170 revised-and-certified items. The remaining 689 items are released as a documented uncertain set with explicit uncertainty sources and expertise tags for future refinement. We evaluate seven state-of-the-art language models on HLE and HLE-Verified, observing an average absolute accuracy gain of 7--10 percentage points on HLE-Verified. The improvement is particularly pronounced on items where the original problem statement and/or reference answer is erroneous, with gains of 30--40 percentage points. Our analyses further reveal a strong association between model confidence and the presence of errors in the problem statement or reference answer, supporting the effectiveness of our revisions. Overall, HLE-Verified improves HLE-style evaluations by reducing annotation noise and enabling more faithful measurement of model capabilities. Data is available at: https://github.com/SKYLENAGE-AI/HLE-Verified

This paper investigates a novel problem, namely the Uncertain Agile Earth Observation Satellite Scheduling Problem (UAEOSSP). Unlike the static AEOSSP, it takes into account a range of uncertain factors (e.g., task profit, resource consumption, and task visibility) in order to reflect the reality that the actual information is inherently unknown beforehand. An effective Genetic Programming Hyper-Heuristic (GPHH) is designed to automate the generation of scheduling policies. The evolved scheduling policies can be utilized to adjust plans in real time and perform exceptionally well. Experimental results demonstrate that evolved scheduling policies significantly outperform both well-designed Look-Ahead Heuristics (LAHs) and Manually Designed Heuristics (MDHs). Specifically, the policies generated by GPHH achieve an average improvement of 5.03% compared to LAHs and 8.14% compared to MDHs.

Recent advances in agent-mediated systems have enabled a new paradigm of social network simulation, where AI agents interact with human-like autonomy. This evolution has fostered the emergence of agent-driven social networks such as Moltbook, a Reddit-like platform populated entirely by AI agents. Despite these developments, empirical comparisons between agent-driven and human-driven social networks remain scarce, limiting our understanding of how their network topologies might diverge. This paper presents the first comparative analysis of network topology on Moltbook, utilizing a comment network comprising 33,577 nodes and 697,688 edges. To provide a benchmark, we curated a parallel dataset from Reddit consisting of 7.8 million nodes and 51.8 million edges. We examine key structural differences between agent-drive and human-drive networks, specifically focusing on topological patterns and the edge formation efficacy of their respective posts. Our findings provide a foundational profile of AI-driven social structures, serving as a preliminary step toward developing more robust and authentic agent-mediated social systems.

We introduce a flexible framework based on response theory and score matching to suppress spurious, noncausal dependencies in reduced-order neural emulators of turbulent systems, focusing on climate dynamics as a proof-of-concept. We showcase the approach using the stochastic Charney-DeVore model as a relevant prototype for low-frequency atmospheric variability. We show that the resulting causal constraints enhance neural emulators' ability to respond to both weak and strong external forcings, despite being trained exclusively on unforced data. The approach is broadly applicable to modeling complex turbulent dynamical systems in reduced spaces and can be readily integrated into general neural network architectures.

Document-to-table (Doc2Table) extraction derives structured tables from unstructured documents under a target schema, enabling reliable and verifiable SQL-based data analytics. Although large language models (LLMs) have shown promise in flexible information extraction, their ability to produce precisely structured tables remains insufficiently understood, particularly for indirect extraction that requires complex capabilities such as reasoning and conflict resolution. Existing benchmarks neither explicitly distinguish nor comprehensively cover the diverse capabilities required in Doc2Table extraction. We argue that a capability-aware benchmark is essential for systematic evaluation. However, constructing such benchmarks using human-annotated document-table pairs is costly, difficult to scale, and limited in capability coverage. To address this, we adopt a reverse Table2Doc paradigm and design a multi-agent synthesis workflow to generate documents from ground-truth tables. Based on this approach, we present DTBench, a synthetic benchmark that adopts a proposed two-level taxonomy of Doc2Table capabilities, covering 5 major categories and 13 subcategories. We evaluate several mainstream LLMs on DTBench, and demonstrate substantial performance gaps across models, as well as persistent challenges in reasoning, faithfulness, and conflict resolution. DTBench provides a comprehensive testbed for data generation and evaluation, facilitating future research on Doc2Table extraction. The benchmark is publicly available at https://github.com/ZJU-DAILY/DTBench.

Foundation models and vision-language pre-training have significantly advanced Vision-Language Models (VLMs), enabling multimodal processing of visual and linguistic data. However, their application in domain-specific agricultural tasks, such as plant pathology, remains limited due to the lack of large-scale, comprehensive multimodal image--text datasets and benchmarks. To address this gap, we introduce LeafNet, a comprehensive multimodal dataset, and LeafBench, a visual question-answering benchmark developed to systematically evaluate the capabilities of VLMs in understanding plant diseases. The dataset comprises 186,000 leaf digital images spanning 97 disease classes, paired with metadata, generating 13,950 question-answer pairs spanning six critical agricultural tasks. The questions assess various aspects of plant pathology understanding, including visual symptom recognition, taxonomic relationships, and diagnostic reasoning. Benchmarking 12 state-of-the-art VLMs on our LeafBench dataset, we reveal substantial disparity in their disease understanding capabilities. Our study shows performance varies markedly across tasks: binary healthy--diseased classification exceeds 90\% accuracy, while fine-grained pathogen and species identification remains below 65\%. Direct comparison between vision-only models and VLMs demonstrates the critical advantage of multimodal architectures: fine-tuned VLMs outperform traditional vision models, confirming that integrating linguistic representations significantly enhances diagnostic precision. These findings highlight critical gaps in current VLMs for plant pathology applications and underscore the need for LeafBench as a rigorous framework for methodological advancement and progress evaluation toward reliable AI-assisted plant disease diagnosis. Code is available at https://github.com/EnalisUs/LeafBench.

Gliomas, among the most common primary brain tumors, vary widely in aggressiveness, prognosis, and histology, making treatment challenging due to complex and time-intensive surgical interventions. This study presents an Attention-Gated Recurrent Residual U-Net (R2U-Net) based Triplanar (2.5D) model for improved brain tumor segmentation. The proposed model enhances feature representation and segmentation accuracy by integrating residual, recurrent, and triplanar architectures while maintaining computational efficiency, potentially aiding in better treatment planning. The proposed method achieves a Dice Similarity Score (DSC) of 0.900 for Whole Tumor (WT) segmentation on the BraTS2021 validation set, demonstrating performance comparable to leading models. Additionally, the triplanar network extracts 64 features per planar model for survival days prediction, which are reduced to 28 using an Artificial Neural Network (ANN). This approach achieves an accuracy of 45.71%, a Mean Squared Error (MSE) of 108,318.128, and a Spearman Rank Correlation Coefficient (SRC) of 0.338 on the test dataset.

Digital banking and online communication have made modern bank runs faster and more networked than the canonical queue-at-the-branch setting. While equilibrium models explain why strategic complementarities generate run risk, they offer limited guidance on how beliefs synchronize and propagate in real time. We develop a process-based agent-based model that makes the information and coordination layer explicit. Banks follow cash-first withdrawal processing with discounted fire-sale liquidation and an endogenous stress index. Depositors are heterogeneous in risk tolerance and in the weight placed on fundamentals versus social information, communicating on a heavy-tailed network calibrated to Twitter activity during March 2023. Depositor behavior is generated by a constrained large language model that maps each agent's information set into a discrete action and an optional post; we validate this policy against laboratory coordination evidence and theoretical benchmarks. Across 4,900 configurations and full LLM simulations, three findings emerge. Within-bank connectivity raises the likelihood and speed of withdrawal cascades holding fundamentals fixed. Cross-bank contagion exhibits a sharp phase transition near spillover rates of 0.10. Depositor overlap and network amplification interact nonlinearly, so channels weak in isolation become powerful in combination. In an SVB, First Republic, and regional bank scenario disciplined by crisis-era data, the model reproduces the observed ordering of failures and predicts substantially higher withdrawal rates among uninsured depositors. The results frame social correlation as a measurable amplifier of run risk alongside balance-sheet fundamentals.

Large populations of AI agents are increasingly embedded in online environments, yet little is known about how their collective interaction patterns compare to human social systems. Here, we analyze the full interaction network of Moltbook, a platform where AI agents and humans coexist, and systematically compare its structure to well-characterized human communication networks. Although Moltbook follows the same node-edge scaling relationship observed in human systems, indicating comparable global growth constraints, its internal organization diverges markedly. The network exhibits extreme attention inequality, heavy-tailed and asymmetric degree distributions, suppressed reciprocity, and a global under-representation of connected triadic structures. Community analysis reveals a structured modular architecture with elevated modularity and comparatively lower community size inequality relative to degree-preserving null models. Together, these findings show that AI-agent societies can reproduce global structural regularities of human networks while exhibiting fundamentally different internal organizing principles, highlighting that key features of human social organization are not universal but depend on the nature of the interacting agents.

Random forests are widely used prediction procedures, yet are typically described algorithmically rather than as statistical designs acting on a fixed set of covariates. We develop a finite-sample, design-based formulation of random forests in which each tree is an explicit randomized conditional regression function. This perspective yields an exact variance identity for the forest predictor that separates finite-aggregation variability from a structural dependence term that persists even under infinite aggregation. We further decompose both single-tree dispersion and inter-tree covariance using the laws of total variance and covariance, isolating two fundamental design mechanisms-reuse of training observations and alignment of data-adaptive partitions. These mechanisms induce a strict covariance floor, demonstrating that predictive variability cannot be eliminated by increasing the number of trees alone. The resulting framework clarifies how resampling, feature-level randomization, and split selection govern resolution, tree variability, and dependence, and establishes random forests as explicit finite-sample statistical designs whose behavior is determined by their underlying randomized construction.

Self-play bootstraps LLM reasoning through an iterative Challenger-Solver loop: the Challenger is trained to generate questions that target the Solver's capabilities, and the Solver is optimized on the generated data to expand its reasoning skills. However, existing frameworks like R-Zero often exhibit non-sustained improvement, where early gains degrade as self-play continues. We identify a key failure mode, Diversity Illusion, where the Solver's training signals appear diverse yet collapse into recurring underlying patterns. It manifests as (1) Local Diversity Illusion, where diversity is enforced only within-batch, inducing cross-iteration mode cycling; and (2) Surface Diversity Illusion, where questions vary superficially but require near-identical reasoning skills. To mitigate them, we propose R-Diverse with two aligned innovations: Memory-Augmented Penalty (MAP), which uses a persistent memory bank to discourage recycling across iterations, and Skill-Aware Measurement (SAM), which evaluates diversity by the reasoning skills exercised rather than surface variation of questions. Across 10 math and general reasoning benchmarks, R-Diverse sustains gains over more iterations and consistently outperforms prior self-play methods. Code is available at https://github.com/Gengsheng-Li/R-Diverse.

Large reasoning models with reasoning capabilities achieve state-of-the-art performance on complex tasks, but their robustness under multi-turn adversarial pressure remains underexplored. We evaluate nine frontier reasoning models under adversarial attacks. Our findings reveal that reasoning confers meaningful but incomplete robustness: most reasoning models studied significantly outperform instruction-tuned baselines, yet all exhibit distinct vulnerability profiles, with misleading suggestions universally effective and social pressure showing model-specific efficacy. Through trajectory analysis, we identify five failure modes (Self-Doubt, Social Conformity, Suggestion Hijacking, Emotional Susceptibility, and Reasoning Fatigue) with the first two accounting for 50% of failures. We further demonstrate that Confidence-Aware Response Generation (CARG), effective for standard LLMs, fails for reasoning models due to overconfidence induced by extended reasoning traces; counterintuitively, random confidence embedding outperforms targeted extraction. Our results highlight that reasoning capabilities do not automatically confer adversarial robustness and that confidence-based defenses require fundamental redesign for reasoning models.

Symbolic Regression (SR) aims to discover interpretable equations from observational data, with the potential to reveal underlying principles behind natural phenomena. However, existing approaches often fall into the Pseudo-Equation Trap: producing equations that fit observations well but remain inconsistent with fundamental scientific principles. A key reason is that these approaches are dominated by empirical risk minimization, lacking explicit constraints to ensure scientific consistency. To bridge this gap, we propose PG-SR, a prior-guided SR framework built upon a three-stage pipeline consisting of warm-up, evolution, and refinement. Throughout the pipeline, PG-SR introduces a prior constraint checker that explicitly encodes domain priors as executable constraint programs, and employs a Prior Annealing Constrained Evaluation (PACE) mechanism during the evolution stage to progressively steer discovery toward scientifically consistent regions. Theoretically, we prove that PG-SR reduces the Rademacher complexity of the hypothesis space, yielding tighter generalization bounds and establishing a guarantee against pseudo-equations. Experimentally, PG-SR outperforms state-of-the-art baselines across diverse domains, maintaining robustness to varying prior quality, noisy data, and data scarcity.

Understanding what drives code instability is essential for effective software maintenance, as unstable classes require larger or more frequent edits and increase the risk of unintended side effects. Although code smells are widely believed to harm maintainability, most prior stability studies examine only the smells within the class being modified. In practice, however, classes can change because their efferent neighbors (i.e., the classes they depend on) are modified due to ripple effects that propagate along static dependencies, even if the class itself is clean. Such ripple effects may be more severe when the efferent neighbor exhibits code smells. In addition, code smells rarely occur alone. They often appear together within a class or across classes connected by static dependencies, a phenomenon known as code smell interrelation. Such interrelation can lead to code smell interaction, where smells are directly connected through static dependencies and may further compound maintainability issues. However, the effect of code smell interrelation and interaction on code quality remains largely underexplored. Therefore, this study investigates whether the presence of code smells in a class's efferent neighbors affects its stability, considering the factor of code smell interrelation and interaction. To achieve this, we mine one year of commit history from 100 top-starred GitHub projects, detect code smells and static dependencies, determine code smell interrelation and interaction, and model these factors as predictors of class stability.

As a multimodal extension of Chain-of-Thought (CoT), Thinking with Images (TWI) has recently emerged as a promising avenue to enhance the reasoning capability of Multi-modal Large Language Models (MLLMs), which generates interleaved CoT by incorporating visual cues into the textual reasoning process. However, the success of existing TWI methods heavily relies on the assumption that interleaved image-text CoTs are faultless, which is easily violated in real-world scenarios due to the complexity of multimodal understanding. In this paper, we reveal and study a highly-practical yet under-explored problem in TWI, termed Noisy Thinking (NT). Specifically, NT refers to the imperfect visual cues mining and answer reasoning process. As the saying goes, ``One mistake leads to another'', erroneous interleaved CoT would cause error accumulation, thus significantly degrading the performance of MLLMs. To solve the NT problem, we propose a novel method dubbed Reliable Thinking with Images (RTWI). In brief, RTWI estimates the reliability of visual cues and textual CoT in a unified text-centric manner and accordingly employs robust filtering and voting modules to prevent NT from contaminating the final answer. Extensive experiments on seven benchmarks verify the effectiveness of RTWI against NT.

This study introduces a novel cooperative game theory model designed to improve the United Nations' current funding mechanisms, which predominantly rely on voluntary contributions. By shifting from a Nash equilibrium framework, where member states act in self-interest, to a cooperative model, the proposed approach aligns each country's financial contributions with the benefits they derive from UN activities. The model ensures a more sustainable and equitable system by introducing personalized pricing based on derived utility. Using agent-based simulations, the research demonstrates that the suggested approach increases global utility, reduces free-rider issues, and creates a more efficient resource allocation system. The findings suggest that the proposed model can optimize UN funding, ensuring a more stable and effective framework for global public goods provision, while considering the varying economic capacities of member states. Further research is recommended to assess the political viability of the model.

The emergence of Self-Driving Laboratories (SDLs) transforms scientific discovery methodology by integrating AI with robotic automation to create closed-loop experimental systems capable of autonomous hypothesis generation, experimentation, and analysis. While promising to compress research timelines from years to weeks, their deployment introduces unprecedented safety challenges differing from traditional laboratories or purely digital AI. This paper presents Safe-SDL, a comprehensive framework for establishing robust safety boundaries and control mechanisms in AI-driven autonomous laboratories. We identify and analyze the critical ``Syntax-to-Safety Gap'' -- the disconnect between AI-generated syntactically correct commands and their physical safety implications -- as the central challenge in SDL deployment. Our framework addresses this gap through three synergistic components: (1) formally defined Operational Design Domains (ODDs) that constrain system behavior within mathematically verified boundaries, (2) Control Barrier Functions (CBFs) that provide real-time safety guarantees through continuous state-space monitoring, and (3) a novel Transactional Safety Protocol (CRUTD) that ensures atomic consistency between digital planning and physical execution. We ground our theoretical contributions through analysis of existing implementations including UniLabOS and the Osprey architecture, demonstrating how these systems instantiate key safety principles. Evaluation against the LabSafety Bench reveals that current foundation models exhibit significant safety failures, demonstrating that architectural safety mechanisms are essential rather than optional. Our framework provides both theoretical foundations and practical implementation guidance for safe deployment of autonomous scientific systems, establishing the groundwork for responsible acceleration of AI-driven discovery.

Long-horizon whole-body humanoid teleoperation remains challenging due to accumulated global pose drift, particularly on full-sized humanoids. Although recent learning-based tracking methods enable agile and coordinated motions, they typically operate in the robot's local frame and neglect global pose feedback, leading to drift and instability during extended execution. In this work, we present CLOT, a real-time whole-body humanoid teleoperation system that achieves closed-loop global motion tracking via high-frequency localization feedback. CLOT synchronizes operator and robot poses in a closed loop, enabling drift-free human-to-humanoid mimicry over long timehorizons. However, directly imposing global tracking rewards in reinforcement learning, often results in aggressive and brittle corrections. To address this, we propose a data-driven randomization strategy that decouples observation trajectories from reward evaluation, enabling smooth and stable global corrections. We further regularize the policy with an adversarial motion prior to suppress unnatural behaviors. To support CLOT, we collect 20 hours of carefully curated human motion data for training the humanoid teleoperation policy. We design a transformer-based policy and train it for over 1300 GPU hours. The policy is deployed on a full-sized humanoid with 31 DoF (excluding hands). Both simulation and real-world experiments verify high-dynamic motion, high-precision tracking, and strong robustness in sim-to-real humanoid teleoperation. Motion data, demos and code can be found in our website.

We present MedXIAOHE, a medical vision-language foundation model designed to advance general-purpose medical understanding and reasoning in real-world clinical applications. MedXIAOHE achieves state-of-the-art performance across diverse medical benchmarks and surpasses leading closed-source multimodal systems on multiple capabilities. To achieve this, we propose an entity-aware continual pretraining framework that organizes heterogeneous medical corpora to broaden knowledge coverage and reduce long-tail gaps (e.g., rare diseases). For medical expert-level reasoning and interaction, MedXIAOHE incorporates diverse medical reasoning patterns via reinforcement learning and tool-augmented agentic training, enabling multi-step diagnostic reasoning with verifiable decision traces. To improve reliability in real-world use, MedXIAOHE integrates user-preference rubrics, evidence-grounded reasoning, and low-hallucination long-form report generation, with improved adherence to medical instructions. We release this report to document our practical design choices, scaling insights, and evaluation framework, hoping to inspire further research.

Understanding how polymer architecture governs transport through nanopores is essential for nanocomposite fabrication, membrane design, and polymer upcycling. However, the effect of the nanoscale structure of copolymers on chain transport through nanoporous media remains poorly understood. In this study, we demonstrate that simply inverting the surface orientation of lamellar poly(styrene-block-2-vinylpyridine) (PS-b-P2VP) diblock copolymers, composed of two monomers with strongly contrasting affinities for SiO2, at the entrance of nanoporous silica significantly alters the kinetics of capillary rise infiltration. Using in situ spectroscopic ellipsometry, we find that infiltration of symmetric PS-b-P2VP into silica nanoparticle (SiO2 NP) packings is significantly faster when the P2VP domain is the top layer of the film and first contacts the nanoparticles, compared to when the PS domain is the top layer. Coarse-grained molecular dynamics simulations reveal that this difference originates from block-specific adsorption pathways that reorganize the nanophase structure around nanoparticles: P2VP-first infiltration forms thin adsorbed layers that drive PS into the pore interiors, generating continuous interfacial pathways that enable rapid, interface-mediated transport. In contrast, PS-first infiltration produces thicker P2VP layers that isolate PS domains and disrupt pathway connectivity, forcing chains to rely on a slower, connectivity-limited transport mechanism through P2VP-rich interstitial regions. Above the order-disorder transition, or upon silanizing nanoparticles to neutralize surface affinity, the rate difference disappears. These findings demonstrate how the interplay between nanoscale domain configuration and polymer-surface affinity governs infiltration dynamics, providing mechanistic insight into tuning transport in nanostructured block copolymers.

We propose a minimal tight-binding model for the electronic interface layer of the LaAlO$_3$/SrTiO$_3$ heterostructure with oxygen vacancies. In this model, the effective carriers are subject to oxygen vacancy induced magnetic impurities. Both the effects of random on-site potentials and Zeeman-like exchange interactions between correlated carriers and magnetic impurities are taken into account. By applying the combined coherent potential approximation (CPA) and dynamical mean-field theory (DMFT) for a ferromagnetic state, we uncover a disordered Fermi-liquid regime for the majority-spins and a low energy scale which controls the transport of the minority-spin carriers, both induced by the magnetic impurities.

We show how the micro-focused BLS signal of forward volume spin waves is formed and why it remains observable despite symmetry-based "suppression" expectations. A reciprocity-theorem based model with vectorial diffraction-limited focusing identifies the nonnegligible longitudinal focal-field component as the key element responsible for BLS sensitivity in the forward volume geometry. We further demonstrate that full polarization analysis, implemented through polarizer-analyzer maps of coherently excited spin waves, provides information beyond the conventional crossed polarizer-analyzer readout. In a BiYIG thin film, the measured maps exhibit Stokes/anti-Stokes polarization asymmetries and nontrivial patterns that stem from quadratic magneto-optical coupling terms. Fitting the data with a model including Voigt and Cotton-Mouton contributions yields an effective Cotton-Mouton constant and shows that the quadratic response is comparable to the linear Voigt contribution.

We investigate critical restricted solid-on-solid models associated to Dynkin diagrams of type $A$, $D$ and $E$, with fixed, periodic and twisted periodic boundary conditions. These models are endowed with an action of the diagrams of the Temperley-Lieb category. For each model, we obtain the decomposition of the state space as a direct sum of irreducible modules over the Temperley-Lieb algebra $\mathsf{TL}_N(β)$ or its periodic incarnation $\mathsf{\mathcal EPTL}_N(β)$. This allows us to recover the known conformal partition functions for these models in the continuum scaling limit. For each irreducible factor arising in the decompositions, we define an associated local operator on the lattice, which behaves like a connectivity operator. Using knowledge from the Temperley-Lieb representation theory at roots of unity, we show that these operators satisfy certain linear difference relations, which are lattice counterparts of the singular-vector relations in conformal field theory.

We document a sequence of bifurcations and elastic patterns in sheared bent sheets of intermediate aspect ratio. The sheets undergo inversion of curvature through the passage of localized features, often in S-shaped pairs. Nested force-displacement hysteresis loops provide experimental evidence for snaking. Several mechanisms for coarsening and refinement of the patterns are observed, including splitting, merging, and escape through open boundaries. While most forces, including that required for full snap-through, scale with the length of the sheet, the initial drop in force upon pattern nucleation decreases rapidly with length.

The origin of recently observed spontaneous chiral symmetry breaking in polar fluids is an unsolved problem, and poses fundamental questions as to how heliconical structures emerge in systems composed of achiral molecules. We report on the softening of bend elasticity close to such phase transition, showing that flexoelectric coupling between the electric polarization and the bend deformation is the responsible mechanism, presumably arising from the bent shape of the constituent highly polar molecules.

We develop a geometric framework for three-dimensional quantum Hall fluids of extended objects (quasi-strings) in the presence of a strong three-form background field associated with a bundle gerbe. In the strong-field regime, fast internal dynamics is frozen and the low-energy kinematics is governed by generalized guiding-center variables consisting of vectorial and tensorial coordinates. We show that these guiding-center variables obey a noncommutative geometry giving rise to a three-dimensional generalization of the Girvin-MacDonald-Platzman (GMP) algebra for projected density operators. Moreover, we relate this algebra to the canonical quantization of a topological BF+BB theory whose level is identified with the Dixmier-Douady invariant. Our results clarify the structure of incompressible quantum Hall-type phases and their geometric and topological features in three spatial dimensions.

Entropy production is often interpreted as a proxy for microscopic disorder or environmental roughness in stochastic systems. We test this interpretation using controlled simulations of overdamped stochastic dynamics on curved surfaces in which local noise, geometry, and forces are held fixed while global constraints are varied. Trajectories are generated for particles evolving toward a central attractor, and entropy production is quantified using both a continuum probability-current estimator and coarse-grained Markov transition statistics across multiple spatial and temporal resolutions. Across systematic sweeps of timestep size, domain extent, and boundary topology, entropy production is governed primarily by constraint-induced probability flow rather than local stochastic variability. Periodic domains that permit sustained circulation yield substantially higher entropy production than reflecting domains despite identical local stochastic structure, with the magnitude of the separation depending on domain extent. In contrast, coarse-grained estimates decrease as temporal resolution increases and rise with finer spatial binning, demonstrating that discrete estimates depend strongly on observation scale and may fail to resolve topology-induced irreversible structure. Ergo, entropy production is not a direct measure of environmental roughness or randomness. Instead, it quantifies how strongly system dynamics are driven away from reversibility by global constraints, geometry, and the space of allowed trajectories. Interpreted in this way, entropy production maps function as diagnostics of organized probability flow and provide a principled method for detecting hidden dynamical constraints from trajectory data alone.

We study a simple model for a particle that is active due to self-phoresis and that has been proposed to model symmetric camphor grains. The particle generates a concentration field through the continuous emission of a chemical substance, and its motion is driven by gradients of this field as it diffuses within a confined channel whose ends perfectly reflect the chemical. The reflection of the chemical field leads to an effective confinement of the particle, which itself is reflected before encountering the channel ends. The system displays a transition from a passive state, where the particle rests at the channel midpoint, to an active state characterized by highly regular, non-chaotic oscillations. We analytically construct the phase diagram and derive the oscillation frequency and amplitude in the vicinity of the transition. A perturbative analysis perfectly describes the dynamics of the particle even for oscillations as large as half the channel size. Furthermore, we develop an analysis which explains the mechanism of particle reflection close to the channel edges in the regime of large activity.

We study analytically and numerically the mean fastest first-passage time (fFPT) to an immobile target for an ensemble of $N$ independent finite-speed random searchers driven by dichotomous noise and described by the telegrapher's equation. In stark contrast to the well-studied case of Brownian particles -- for which the mean fFPT vanishes logarithmically with $N$ -- we uncover that the mean fFPT is bounded from below by the minimal ballistic travel time, with an exponentially fast convergence to this bound as $N \to \infty$. This behavior reveals a dramatic efficiency advantage of physically realistic, finite-speed searchers over Brownian ones and illustrates how diffusive macroscopic models may be conceptually misleading in predicting the short-time behavior of a physical system. We extend our analysis to anomalous diffusion generated by Riemann-Liouville-type dichotomous noises and find that target detection is more efficient in the superdiffusive regime, followed by normal and then subdiffusive regimes, in agreement with physical intuition and contrary to earlier predictions.

We develop a perturbation theory to study the shape and the orientation of an initially spherical capsule of radius R with a viscosity contrast, a surface tension σ and a bending rigidity $κ$ in linear flows. The elastic mechanical response of membrane to deformations is described by three elastic constitutive law which are either Hookean, Neohookean or Skalak type leading to the introduction of a surface shear elastic modulus $G_s$ and the Poisson ratio (or analog quantities). At the leading order, the deformation, i.e. the so-called Taylor parameter is proportional to the elastic capillary number Ca which evaluates the ratio between the external viscous stress and the elastic membrane response. In this linear regime, the results do not depend on the elastic constitutive law as expected. Without surface tension and bending rigidity, we recover the results of Barthes-Biesel & Rallison (1981) and notably the fact that the Taylor parameter does not depend on the viscosity contrast $λ$ contrary to the case of a viscous droplet. In our more general model, the deformation does no longer depend on $λ$ at the upper order. Now, the Taylor parameter also depends on two other dimensionless numbers: the surface elastocapillary ratio $σ/G_s$ and the dimensionless bending rigidity $B= κ/G_sR^2$. At the further order, the angle of inclination of the capsule with the direction of the shear flow, the analog of the Chaffey and Brenner equation for droplets is determined in each case. The results are in excellent agreement with the numerical ones performed with a code based on the boundary integral method providing an useful method to valid numerical developments.

In systems with fast periodic driving, there are special subsets of (resonant) states, which behavior can be described with effective, time-independent Hamiltonian in a rotating reference frame. Here, we show that experimentally feasible system of ultracold bosonic atoms on a ring with rapidly oscillating scattering length can be used to simulate time-independent two-component atomic mixture with exotic, long-range interactions.

We develop a microscopic theory of nonlinear magneto-optical injection currents in multi-Weyl semimetals subjected to a uniform magnetic field. Using the Landau-level spectrum of a tilted multi-Weyl Hamiltonian with arbitrary monopole charge $ν$ as a starting point, we formulate a Kubo-type nonlinear response theory in the Landau-level basis and derive the second-order conductivity tensor. We identify distinct contributions originating from chiral-chiral, chiral-bulk, and bulk-bulk optical transitions, revealing characteristic monopole-charge scaling and sharp resonant structures governed by Landau-level selection rules and tilt-induced asymmetry. In the untilted limit, closed-form analytical expressions emerge that expose universal frequency thresholds and provide clear experimental signatures of higher-order Weyl topology. Our results establish nonlinear magneto-optical injection currents as a direct transport probe of chiral Landau levels and multi-Weyl topological charge.

Like a crêpe resting on a plate, a thin elastic sheet can fold smoothly under its own weight, forming reversible shapes without creases or imposed hinges. Such soft folds arise from a balance between elastic bending and gravity, yet their stability, packing limits, and dynamics remain poorly understood. Here we show that these behaviors are governed by a single physical length scale, the elasto-gravity length $\ell_{eg}$. Using experiments and heavy-elastica theory, we demonstrate that $\ell_{eg}$ sets the characteristic fold geometry, determines when a fold becomes unstable and unfolds, and limits how many reversible folds can be stacked in rectangular and circular sheets. In particular, when lengths are rescaled by $\ell_{eg}$, fold shapes and stability thresholds collapse across materials and thicknesses. We further show that unfolding follows a universal speed scaling $v \sim \sqrt{g\,\ell_{eg}}$, revealing a gravity-controlled time scale for the release of stored bending energy. Together, these results establish a unified physical framework for reversible folding, compact storage, and gravity-assisted deployment of thin elastic sheets.

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.

This review provides an up-to-date account of energy transport in Fermi-Pasta-Ulam-Tsingou (FPUT) chains, a key testbed for nonequilibrium statistical physics. We discuss the transition from the historical puzzle of thermalization to the discovery of anomalous heat transport, where the effective thermal conductivity $κ$ diverges with system size $L$ as $κ\propto L^δ$. The article clarifies the distinction between two universality classes: the FPUT-$αβ$ model, characterized by $δ= 1/3$ and linked to Kardar-Parisi-Zhang (KPZ) physics, and the symmetric FPUT-$β$ model, where numerical and theoretical evidence support $δ= 2/5$. We investigate how finite-size effects - unavoidably induced by the thermostatting protocols - can disguise the asymptotic scaling. Additionally, we analyze the role of conservative noise in preserving hydrodynamic properties and examine how proximity to integrable limits leads to long-lived quasi-particles and, thereby, to diffusive regimes over intermediate spatial scales.

We investigate the nonequilibrium dynamics of the nearest-neighbour Ising model subjected to stochastic resetting, where the system is intermittently returned to an initial configuration with magnetisation $m_0$, with the inter-reset times drawn from the power law distribution $ατ_0^α/ τ^{α+1}$. The heavy-tailed resets generate magnetisation distributions that differ significantly from both equilibrium dynamics and the previously studied Ising model with exponentially distributed reset times. In two dimensions, for $T > T_C$, we find a quasi-ferro state for all $α$, marked by a double-peaked distribution that diverges at $m=0$ and $m=m_0$; no steady state exists for $α< 1$, while a stationary state emerges for $α> 1$. For $T < T_C$, power law resetting produces two distinct regimes separated by a crossover exponent $α^* = 1-c$: a single-peak ferromagnetic phase localised at $m_{eq}$ for $α< α^*$, and a dual-peak ferromagnetic phase with divergences at $m_{eq}$ and $m_0$ for $α> α^*$. Analytic results in one and two dimensions, supported by simulations, yield a rich phase diagram in the $(T,α)$ plane and reveal how heavy-tailed resetting generates nonequilibrium phases very different from those seen in the case of exponential resetting.

We investigate stochastic processes that generalize geometric Brownian motion, focusing on cases where the standard invariant measure, i.e. the solution of the stationary Fokker-Planck equation does not necessarily exist. We demonstrate that the existence of such a measure depends sensitively on the structure of the drift and diffusion terms, as well as on the chosen discretization scheme of the underlying stochastic dynamics. To ground our discussion, we draw motivation from phenomenological models in statistical theories of turbulence, where geometric Brownian motion serves as a classical example. To address situations where the standard invariant measure fails to exist, we heuristically explore the concept of infinite ergodicity, a notion recently introduced in the context of statistical physics for drift-diffusion stochastic processes.

Uniformity in the excitonic spectrum is a key requirement for accessing intrinsic excitonic physics in two-dimensional semiconductors; however, in supported transition-metal dichalcogenide (TMD) monolayers, exciton energies and linewidths can vary spatially due to inhomogeneities created by contact with other materials or contamination left by fabrication procedures. Suspended TMD monolayers provide an effective route to minimizing substrate-induced disorder. Here we demonstrate the spatially uniform excitonic spectrum from high-quality WSe2 suspended monolayers fabricated by gold-assisted exfoliation directly onto an Au contact electrode of a gate-tunable device. The resulting membranes span narrow suspended regions up to ~80 um and show spatially uniform photoluminescence at cryogenic temperatures with neutral-exciton linewidths as low as ~4.5 meV, comparable to the narrowest values reported for high-quality monolayers. Spectral reproducibility across the suspended regions supports an intrinsic optical response, while gate-dependent measurements resolve multiple excitonic species. This approach provides a practical route to electrically tunable potential landscapes in suspended TMD monolayers with a highly uniform excitonic response.

Quantum batteries, miniaturized devices able to store and release energy on demand, are promising both because their intrinsic energy and time scales can match those of other quantum technologies and due to the intriguing possibility of achieving super-extensive charging power. While this enhanced scaling is known to appear in several settings, it is generally believed to be forbidden in Wigner-Jordan integrable spin chains charged via quantum-quench protocols. Here, we show that an extended cluster-Ising model, despite belonging to the above category, exhibits super-extensive charging power over wide ranges of system sizes, reaching up to a thousand spins, in proper parameter regimes. This remarkable anomalous scaling is due to a corresponding super-extensive growth of the stored energy, implying that it occurs at large but finite size and cannot persist in the thermodynamic limit. This phenomenon appears robust against finite-temperature effects.

Moiré semiconductors offer flat bands where Coulomb interactions and band topology intertwine, while interlayer coupling plays a central role in forming the moiré potential. However, limited interlayer coupling strength and the lack of efficient tuning methods hinder further exploration of correlated phenomena in moiré semiconductors. Here we introduce a cryogenic dual-gated diamond-anvil platform using helium as a pressure medium, enabling reversible hydrostatic tuning together with magneto-optical spectroscopy in twisted bilayer WSe2. Pressure enhances the moiré potential, redshifts excitons, and stabilizes Stoner ferromagnetism otherwise absent at a 3.1-degree twist. Simultaneously, the half-filled C = 1 Chern insulating state strengthens, exhibiting a reduced saturation field. Moreover, we observe a topological phase transition from a Chern insulator to a Mott insulator at around 2 GPa. First-principles calculations reveal that a Gamma-to-K valence-band-maximum switching drives this transition by converting an Ising-like topological K-valley miniband into a spin-degenerate trivial Gamma miniband. Our findings demonstrate hydrostatic pressure as a powerful, continuous control axis for correlated magnetism and topological band engineering in moiré materials.

The localized states in the band gap of amorphous phase change alloys like Ge$_2$Sb$_2$Te$_5$ control the electrical conduction via the Poole-Frenkel mechanism. Understanding the origin of in-gap states and their evolution in time during aging of the glass is therefore important for the control of the resistance drift in phase change memory devices. Here, we use a machine learning interatomic potential to generate several models of Ge$_2$Sb$_2$Te$_5$ whose electronic structure is then analyzed within density functional theory with a hybrid functional. A detailed statistical analysis of the structural motifs on which the in-gap states are localized, reveals that the vast majority of in-gap states involve wrong bonds (homopolar or Ge-Sb bonds) often accompanied by Ge in tetrahedral configurations or overcoordinated Ge and Sb atoms. Metadynamics simulations mimicking glass aging support the picture that structural relaxations lead to the depletion of in-gap states and then to an increase of resistance. The simulations thus provide important insights for the mitigation of the resistance drift in phase change memory devices.

Ultrasound-matter interactions underpin numerous biomedical and soft-matter applications, yet simulating these phenomena is challenging due to the large separation of viscous and sonic time scales. Continuum methods capture large-scale wave propagation but cannot resolve microscale interactions, while particle-based approaches offer molecular resolution but struggle with efficiency and stability at larger scales. We introduce a particle-based virtual ultrasound machine that uses a novel smoothed dissipative particle dynamics variant with an implicit pressure solver and a negative-pressure stabilization scheme, required to mimic acoustic propagation across MHz-GHz frequencies. We demonstrate its capabilities by modeling the acoustophoresis of encapsulated microbubbles, a key mechanism in ultrasound-mediated drug delivery. Beyond this application, the approach establishes a generalizable platform for simulating wave-matter interactions in soft and biological materials, opening new directions for computational studies of acoustics-driven phenomena in science and engineering.

The work functions of 7Li and 6Li metals have been measured as a function of temperature, by using photoionization of pure isolated metal nanoparticles in a beam. These data reveal a marked isotope effect in the temperature variation of these work functions. Furthermore, for both isotopes the curvature of this temperature variation is found to be significantly larger than may be ascribed purely to a change in the electron gas density. These findings enhance the characterization of lithium as a quantum material in which the interplay between electronic and ionic degrees of freedom is nontrivial, and call for a microscopic understanding beyond simple models. Additionally, the slope of the work function curves was observed to vanish in the low temperature limit, as had been predicted on the basis of the Third Law of thermodynamics.

A precision measurement of the photoionization of pure sodium and potassium nanoparticles isolated in a beam enabled an accurate determination of their work functions as a function of temperature. In addition to resolving and quantifying the initial gradual decrease of the work function with temperature, which is associated with thermal expansion, the experiment revealed that the work function then undergoes a distinct drop both in magnitude and in slope that signifies the onset of nanoparticle melting. This establishes that a structural phase transition can be detected via a high-resolution measurement of the photoemission threshold. The melting temperature of nanoparticles with diameters of 7-9 nm is reduced by nearly 100 K relative to the bulk value. This suppression aligns with predictions from the Gibbs-Thomson equation which describes finite-size phase transitions.

A beam of free alkali metal nanoparticles is produced by a condensation source, passed through a thermalizing tube adjustable over a broad temperature range, and ionized by tunable light. High stability of the particle flux and an automated data acquisition routine allow efficient collection of photoionization yield curves. A careful fit of the data to the universal Fowler function makes it possible to obtain nanoparticle ionization energies, and from those, the metal work functions, with $\sim$0.2% precision. The experimental arrangement, nanoparticle thermalization rates, and ionization threshold analysis are described in detail. The use of ultrapure and temperature-controlled gas-phase nanoparticles facilitates the analysis of electronic properties, such as work functions, and of their interplay with thermal lattice dynamics.

The emergence of optical circular dichroism in chiral nanoscale and molecular systems provides not only a way for analyzing the sample chirality itself but also additional degrees of freedom in manipulating light. Such manipulation can be reached even at the nanoscale level; however, probing and understanding the properties of optical fields well below the diffraction limit requires an adequate technique. Electron energy-loss spectroscopy (EELS) with orbital angular momentum (OAM)-based electron state sorting has been suggested as a suitable candidate, but to date, no conclusive experiments have been performed. We, therefore, theoretically explore the emergence of dichroism in EELS for a canonical single-twist helix nanostructure and present a detailed analysis of the optimal parameters to obtain a robust signal. Our work offers novel insights into the interpretation and volatility of the OAM-resolved EELS signal, which can inspire and guide future experimental efforts.

We study the effects of uncorrelated quenched disorder to the phase diagram and continuous transitions of three-dimensional lattice ${\mathbb Z}_2$ gauge Higgs models. For this purpose, we consider two types of quenched disorder, associated with the sites and plaquettes of the cubic lattice. In both cases, for sufficiently weak disorder, the phase diagram remains similar to that of the pure system, showing two different phases (one of them being a topologically ordered phase), separated by two different continuous transition lines. However, the quenched disorder changes the universality classes of the critical behaviors along some of the transition lines. The random-plaquette disorder turns out to be relevant along the topological ${\mathbb Z}_2$ gauge transition line, so the critical behaviors belong to the different random-plaquette $\mathbb{Z}_2$ gauge (RP${\mathbb Z}_2$G) universality class with length-scale exponent $ν=ν_{\rm rp}\approx 0.82$; on the other hand, it turns out to be irrelevant along the other Ising$^\times$ transition line (a variant of the Ising transitions with a gauge-dependent order parameter), leaving unchanged its asymptotic critical behaviors with $ν=ν_{\cal I}\approx 0.63$. The random-site disorder leads to a substantially different scenario: it destabilizes the Ising$^\times$ critical behaviors of the pure model, changing them into those of the randomly-dilute Ising$^{\times}$ (RDI$^{\times}$) universality class with $ν=ν_{\rm rdi}\approx 0.68$, while the critical behaviors along the other ${\mathbb Z}_2$ gauge topological transition line remains stable, with $ν=ν_{\cal I}\approx 0.63$.

Direct numerical simulations of electrospinning and electrospraying are computationally demanding due to large-scale separation between the needle and the tip-to-collector distance. The cone-jet mode that occurs in the vicinity of the needle arises from a delicate balance between surface tension, viscous stresses, inertia, and electric stresses. This mode has a central role in determining the subsequent instabilities of the jet and the eventual outcomes on the collector. Truncated-domain simulations offer a viable alternative but depend critically on the accuracy of far-field electrostatic boundary conditions. Existing truncated-domain approaches based on analytical expressions for the electric potential systematically underestimate the electric field near the needle tip and require empirical tuning informed by prior experiments or full-domain simulations, thereby limiting their predictive capability. Here, we present a general truncated-domain framework for electrohydrodynamic (EHD) simulations of the cone-jet mode that avoids these limitations. Our approach exploits inexpensive full-domain electrostatic simulations to obtain the exact electric field and potential distributions near the needle, which are then imposed as boundary conditions in an EHD simulation carried out on a truncated domain. Comparisons with full-domain EHD simulations and experimental data demonstrate that the proposed approach accurately reproduces cone-jet shapes as well as key physical quantities, including electric currents, charge distributions, velocity fields, and Maxwell stresses, while converging at substantially smaller domain sizes. The formulation eliminates tunable parameters, does not require prior knowledge of the cone-jet configuration, and significantly reduces computational cost, providing a reliable and predictive framework for studying electrohydrodynamic cone-jet flows.

We investigate how quasiperiodic spatial structure enables protocol-induced acceleration in open quantum systems by analyzing the Pontus-Mpemba effect in one-dimensional chains subject to Markovian dephasing. The dynamics are governed by a Lindblad superoperator that drives all initial states toward a maximally mixed infinite-temperature steady state, isolating dynamical mechanisms from static equilibrium properties. Considering two representative quasiperiodic models, namely a tight-binding chain with a mosaic potential and its extension with power-law long-range hopping, we show that a properly engineered two-step protocol, in which the system is first steered to a finite temperature intermediate state, yields a strictly shorter overall relaxation time than direct evolution from the same initial configuration. This protocol-induced acceleration persists for both initially localized and extended eigenstates and remains robust in the presence of long-range hopping. A Liouvillian spectral analysis reveals that the mechanism originates from a redistribution of spectral weight that suppresses overlap with the slowest decay modes, rather than from any modification of the decay spectrum itself. Our results establish quasiperiodic chains as a controlled setting for engineering relaxation pathways through Liouvillian spectral structure.

It has long been known that the relative entropy of a non-equilibrium ensemble to the corresponding equilibrium ensemble is the excess free energy. We show that the reverse relative entropy also has a thermodynamic interpretation: it is the excess free energy of a dual ensemble in which the roles of energy and entropy are interchanged.

We present a purely theoretical, self-contained argument that the Second Law of Thermodynamics cannot be a universal fundamental law in the form ``entropy does not decrease'' (whether asserted trajectory-wise or as a universal statistical principle) when the underlying microscopic dynamics are time-reversal invariant. The core is a mirror-state construction: for any microstate $A$ one constructs its time-reversed partner $B$ (momenta inverted). If a universal monotonicity statement is applied to both $A$ and $B$, it implies that $A$ is a local minimum of entropy at every moment, which forces entropy to be constant and destroys any entropic arrow of time. The consistent replacement is that entropy is a stochastic variable described by a probability distribution $P(S)$, whose shape depends on constraints and boundary conditions. We then prove from first principles that constraints necessarily reshape the long-time entropy distribution $P_{\infty}(S;λ)$ by altering the invariant measure through changes in the Hamiltonian and/or the accessible phase space. A sharp criterion is given: in the microcanonical setting, the \emph{only} way $P_{\infty}^{(E)}(S;λ)$ can remain the same up to translation is when all accessible macrostate volumes are scaled by a common factor; otherwise the distribution changes structurally.

Transport in soft jammed materials is often described as fluid motion through a fixed structure, leading naturally to capillary based descriptions. This picture appears particularly appropriate in strongly jammed systems, where structural rearrangements are suppressed and little visible motion is observed. Here we investigate solvent transport in foam and show that this intuition fails to capture key aspects of the transport process. By directly observing both liquid penetration and bubble motion under controlled boundary conditions, we demonstrate that solvent transport is strongly influenced by the mechanical response of the foam structure, even though the intrinsic imbibition relative to the foam matrix remains purely capillary-driven. In closed systems, the jammed structure resists penetration and leads to a pronounced slowdown that cannot be accounted for by purely capillary descriptions. In contrast, in open systems, collective bubble motion accompanies solvent invasion, resulting in an apparent acceleration of transport. These results indicate that the lack of structural motion does not guarantee a purely capillary description of transport. Our findings reveal a boundary controlled coupling between flow and structure, and highlight the need to reconsider transport processes in soft jammed systems, including foams, dense colloids, and biological tissues.

The theory of transition radiation in helical metamaterials with strong spatial dispersion is developed in the framework of an effective field theory approach. The average number of photons radiated by a charged particle passing through a plate made of this metamaterial is obtained. Given the positions of the transition radiation maxima in momentum space for different velocities of a charged particle, the method for reconstruction of the dispersion law of plasmon-polaritons in metamaterials is proposed. Applying this method conversely, one can predict the radiation spectrum and polarization properties of transition radiation by means of the dispersion law of plasmon-polaritons in the metamaterial known, for example, from the effective model. It is shown that the strong spatial dispersion alters qualitatively the properties of transition radiation from a charged particle traversing normally a plate made of the helical metamaterial along its symmetry axis in the paraxial regime, viz., there is a nonzero forward radiation in contrast to transition radiation in media without strong spatial dispersion. Vavilov-Cherenkov radiation and the anomalous Doppler effect in helical metamaterials with strong spatial dispersion are described.

Deep neural networks are strongly over-parameterized, often containing far more weights than required for their task. Although such redundancy can aid optimization, it leads to inefficient deployment and high computational cost, motivating model compression techniques. Among these, network pruning provides a clear and effective route to sparsity. We study pruning from a statistical-physics perspective, interpreting performance degradation under weight removal as a phase transition. Focusing on magnitude-based pruning with fine-tuning, we show that deep networks undergo a sharp transition from a cooperative, functional phase to a disordered phase with collapsed performance. This transition is characterized by scaling laws consistent with second-order critical behavior, with connectivity as the control parameter. Our findings suggest universal pruning-induced criticality across architectures and datasets. Finally, we show that there exists a large class of subnetworks sharing the same nodes' degrees with similar learning ability, thus linking model performance to its topological properties.

We investigate the critical current in microwave-irradiated Josephson junctions hosting Yu-Shiba-Rusinov states due to magnetic impurities. Under two conditions, namely, (i) the breaking of particle-hole symmetry in the normal sense by non-zero potential scattering, and (ii) the breaking of inversion symmetry either by unequal magnitudes of potential scattering and/or magnetic moments, microwave irradiation induces an additional phase-independent contribution to the current. This leads to asymmetric critical currents for opposite current polarities, an effect absent in the same junction without microwave irradiation. The asymmetry is highly tunable via the microwave amplitude and frequency, and we may even achieve perfect asymmetry where the critical current vanishes for one polarity, akin to a perfect diode. While Yu-Shiba-Rusinov states provide the ideal platform for a pronounced asymmetry, we find that as long as the two conditions (i) and (ii) above are met, our proposal does not necessarily depend upon them.

The data-driven characterization of the ``complexity'' present in dynamical systems remains an open problem with broad applications across the physical sciences. We investigate the ``structural complexity'' of the 2D ferromagnetic Ising model, a paradigmatic system exhibiting a second-order phase transition at a certain critical temperature which is often cited as a canonical example of complex morphology. We define a quantitative metric for this structural complexity, $\mathcal C_s$, through the lens of algorithmic information theory by approximating the Kolmogorov complexity of lattice configurations via standard lossless image compression algorithms. We regularize our proposed metric, $\mathcal C_s$, by comparing the compressibility of a configuration to that of its pixel-wise sorted and randomly shuffled counterparts. We arrive at a definition of $\mathcal C_s$ as a product of two components representing the systems departure from perfect order and disorder respectively which we then plot as a function of temperature. Our numerical simulations reveal a distinct peak in $\mathcal C_s$ at the known critical temperature $T_c$. This result demonstrates that such information-theoretic measures can act as sensitive, model-agnostic indicators of criticality, directly quantifying the emergence of complex structure at the boundary between order and chaos, opening the door to data-driven applications in domains where analytic solutions are unavailable.

We investigate the effect of structural distortion on bond percolation in simple cubic and body-centered cubic lattices using extensive Monte Carlo simulations. Distortion is introduced through controlled random displacements of lattice sites, thereby modifying nearest-neighbor distances. Bond occupation is permitted only when the bond length is smaller than a prescribed connection threshold, directly coupling geometric disorder to connectivity. Finite-size scaling analysis is employed to determine percolation thresholds for finite systems and in the thermodynamic limit. We find that when the connection threshold exceeds the nearest-neighbor distance of the undistorted lattice, the percolation threshold increases monotonically with distortion strength, indicating a systematic suppression of spanning. In contrast, this monotonic behavior breaks down when the connection threshold is below the nearest-neighbor distance of the undistorted lattice, highlighting a nontrivial interplay between geometric distortion and connectivity. We further identify critical values of the connection threshold and the distortion amplitude required for global spanning when all the allowed bonds are occupied. All qualitative behaviors remain robust across both lattice geometries. These results clarify how geometric disorder reshapes percolation in three-dimensional crystalline networks.

Micro-calorimetry offers significant potential as a quantitative method for studying the structure and function of biological systems, for instance, by probing the excess heat released by cellular or sub-cellular structures, isothermal or not, when external parameters change. We present the conceptual framework of nonequilibrium calorimetry, and as illustrations, we compute the heat capacity of biophysical models with few degrees of freedom related to ciliar motion (rowing model) and molecular motor motion (flashing ratchets). Our quantitative predictions reveal intriguing dependencies of the (nonequilibrium) heat capacity as a function of relevant biophysical parameters, which can even take negative values as a result of biological activity.

We investigate the dynamics of the $ν=1$ Quantum Symmetric Simple Exclusion Process starting from spatially inhomogeneous initial states. This one-dimensional system of free fermions has time-dependent stochastic hopping amplitudes that are uniform in space. We focus on two paradigmatic setups: domain-wall melting and the expansion of a trapped gas. Both are investigated by extending the framework of quantum generalized hydrodynamics to account for the underlying stochastic dynamics. We derive the evolution of the local quasiparticle occupation function, which characterizes the system at large space-time scales, and analyze the resulting entanglement spreading. By incorporating quantum fluctuations of the occupation function and employing conformal field theory techniques, we obtain the exact contribution to the entanglement entropy for each individual noise realization. Averaging over these realizations then yields the full entanglement statistics in the hydrodynamic regime. Our theoretical predictions are confirmed by exact numerical calculations. The results presented here constitute the first application of quantum generalized hydrodynamics to stochastic quantum systems, demonstrating that this framework can be successfully extended beyond purely unitary dynamics to include stochastic effects.

Exact nonperturbative solutions to hermitian one-matrix models, their topological string duals, as well as their double-scaling limits to multicritical and minimal string theories, may be obtained via the use of resurgent transseries. These solutions are generically resonant, entailing both eigenvalues and anti-eigenvalues, or, equivalently, both D-branes and negative-tension D-branes -- but are otherwise intricate to write down, having been previously addressed on a case-by-case approach. This work shows how there is a general and rather compact way to write down all these exact and fully nonperturbative transseries solutions in closed-form, immediately starting from the spectral geometry of the matrix model or string theory at hand, in the form of a discrete Fourier or Zak transform for their partition functions. This structure is inherently associated to the existence of anti-eigenvalues or negative-tension D-branes. The validity of these solutions is testable across all values of the parameters -- from weak to strong 't Hooft coupling, from small to large N; equivalently, from semi-classical to deeply quantum regimes -- and many such nonperturbative tests are performed against diverse examples ranging from matrix models to non-critical strings, fully validating our analytical and exact expressions. In particular, anti-eigenvalues or negative-tension D-branes are absolutely required to find sharp numerical matches. In order to study these solutions globally across their phase diagrams, however, complete non-linear Stokes data is still needed -- which will be addressed in a complementary follow-up paper.

Recently, product states have been identified as simple-structured eigenstates of XXZ Heisenberg spin models in arbitrary dimensions, occurring at anisotropy values corresponding to certain roots of unity. Yet, the product states typically only span parts of a larger degenerate eigenspace. Here, we classify this eigenspace in the one-dimensional periodic XXZ chain at all roots of unity $q$, where $q^2$ is an $\ell$-th primitive root of unity. For commensurate chain lengths $N$ with $q^N=1$, we prove that the minimal degeneracy is $2^{N/\ell}\ell$ using the representation theory of the affine Temperley-Lieb (aTL) algebra. For the incommensurate case, we derive analogous exponential lower bounds of $2^{2\lfloor\frac{N}{2\ell}\rfloor+1}$ if $N$ is even and $2^{2\lfloor \frac{N}{2\ell}+\frac{1}{2}\rfloor}$ if $N$ is odd and $q^\ell=1$. Our proof employs the morphisms between aTL modules discovered by Pinet and Saint-Aubin and emphasizes the importance of exact sequences and hidden twisted boundary condition sectors that mediate the degeneracy. In the case of commensurate chain lengths, we connect to the Fabricius-McCoy string construction of all Bethe roots of the degenerate subspace, which previously uncovered parts of our results. We corroborate our results numerically and demonstrate that the lower bound is saturated for chain lengths $N\leq20$. Our work demonstrates for a concrete system how the interplay of the Bethe ansatz, aTL representation theory, and twisted boundary conditions explains degeneracy connected to long-lived product states, stimulating research towards generalization to higher dimensions. Exponential degeneracy could boost applications of spin chains as quantum sensors.

Spin textures are ubiquitous in antiferromagnets, yet their consequences for altermagnets remain largely unexplored. We show that spatial variations of the Néel order act on the low-energy electrons as effective gauge fields, leading to strong anisotropies in both dc transport and optical absorption, even without intrinsic spin-orbit coupling. As a concrete example, we analyze a coplanar spin helix and predict that the principal axes of the conductivity and linear dichroism are set by the helix wave vector and can be tuned by the texture geometry. Our results point to polarization-resolved optics and anisotropic transport as direct probes of textured altermagnetic states, and suggest a simple route to direction-selectivity.

Understanding how stick-slip dynamics manifests in diverse physical conditions is a crucial topic in tribology. Although it has been extensively studied in simple frictional configurations, the characterization of stick-slip behavior in complex assemblies is challenging. This work presents the first systematic investigation of stick-slip dynamics in a system with multiple contact surfaces undergoing friction amplification through conversion of traction forces into normal compression. Using interleaved paper blocks as a model system, we combine force measurements and image processing to characterize stick-slip events occurring when the two blocks are pulled apart at different detachment velocities. We find that both the peak force and the amplitude of the stick-slip events decrease along with the system's detachment. By combining a previously designed model for friction amplification and the stick-slip dynamics predicted by a simple frictional spring-block system, we link the observed behavior to the evolving normal compression within the assembly. Through force measurements and imaging, we extract the effective stiffness of the system from stick-slip events at low velocities and relate it to the system's normal compression. We then predict the observed decrease of the global stiffness as function of the detachment by considering the spatial distribution of normal forces within the assembly, which determines an effective number of sheets contributing to the system's mechanical response. Our findings reveal a non-trivial interplay between internal stress distribution and mechanical response mediated by frictional forces, with implications for granular materials, textiles, fibrous systems, and mechanical metamaterials.

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 \ce{H2} L$^{-1}$ for automotive applications. We present a computational study of ferrocene-functionalized covalent organic frameworks (COFs) for hydrogen storage. Following the \textbf{M}ulti-binding \textbf{S}ites \textbf{U}nited in \textbf{C}ovalent-\textbf{O}rganic \textbf{F}ramework (MSUCOF) approach, we introduce MSUCOF-4-FeCp, designed by incorporating ferrocene (\ce{FeCp2}) moieties into IRCOF-102. Notably, it achieves exceptional performance with gravimetric and volumetric uptakes of 18.0 wt\% and 72.6 g \ce{H2} L$^{-1}$ at 298 K and 700 bar. The material exhibits optimal binding energies (15--20 kJ$\cdot$mol$^{-1}$) 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.

Floquet control of band topology is a central theme in ultrafast quantum materials science. Established experimental probes of light-induced topological states include ultrafast transport and time- and angle-resolved photoemission spectroscopy, each with important strengths but also well-known limitations. Here we propose ultrafast terahertz scanning tunneling microscopy (THz-STM) as a real space energy-resolved probe of Floquet physics. We show that THz-STM enables direct local detection of bulk Floquet gaps and distinct Floquet edge state signatures. We derive a nonequilibrium Green's-function formalism for time-dependent tunneling that directly extends standard STM theory and provides an intuitive interpretation of rectified ultrafast tunneling currents. We apply the approach to bulk graphene and graphene nanoribbons of variable width. For the bulk, we show that THz-STM provides direct spectroscopic access to Floquet-induced gap openings, and we contrast pulsed pump-probe protocols with the continuous-wave Floquet steady-state limit. For finite ribbons, we demonstrate time- and space-resolved imaging of Floquet-induced topological edge states and identify the ribbon-width scale below which edge state protection breaks down. We further show how band structures of graphene nanoribbons and Floquet chiral edge modes can be reconstructed via Floquet quasiparticle interference. Finally we demonstrate that chiral impurities that break time-reversal symmetry induce characteristic spatial THz-STM signatures that can be used as a direct probe of Floquet edge state chirality.

Strategic-form min-max game theory examines the existence, multiplicity, selection of equilibria, and the worst-case computational complexity under perfect rationality. However, in many applications, games are drawn from an ensemble, and players exhibit bounded rationality. We introduce thermal min-max games, a thermodynamic relaxation that unifies bounded and perfect rationality by assigning each player a temperature to regulate their rationality level. To analyze typical behavior in the large-strategy limit, we develop a nested replica framework for this relaxation. This theory provides tractable predictions for typical equilibrium values and mixed-strategy statistics as functions of rationality strength, strategy-count aspect ratio, and payoff randomness. Numerical experiments demonstrate that these asymptotic predictions accurately align with the equilibrium of finite games of moderate size.

We study the proximity effect in the Hubbard model coupled to BCS superconductors describing a single-layer strongly correlated electron system in a phase-biased Josephson junction. We find two distinct gapped solutions, one Mott-like insulating (M-phase) and one proximitized superconducting phase (S-phase), separated by first-order transition with hysteresis. In the M-phase the large correlation charge gap strongly suppresses the critical current, while the S-phase behaves as a $0$-junction, with a proximitized gap that closes for $φ=π$ to yield a correlated metal. Phase bias and junction transparency can thus serve as tuning knobs to switch between conducting and insulating regimes. Working within the dynamical mean field theory using the numerical renormalization group as the impurity solver, we associate M- and S-phase solutions with the doublet and singlet fixed points of the underlying superconducting Anderson impurity problem. We obtain detailed insight into the spectral structure on all energy scales. In the M-phase, the self-energy has sub-gap resonances symmetrically located around the Fermi level resulting from the splitting of the ''mid-gap pole'' found in Mott insulators; this structure accounts for phase insensitivity.

By utilizing the site-dependent spin quantization axis in semiconductor quantum dot (QD) arrays, shuttling-based spin qubit gates have become an appealing approach to realize scalable quantum computing due to the circumvention of using high-frequency driving fields. The emergence of a spin deviation from the local quantization axis of one residing QD is the prerequisite to implement the qubit gates. In this work, we study the non-adiabatic dynamics of a spin qubit shuttling between coupled QDs with inhomogeneous Landé g-tensors and a small magnetic field. The spin dynamics is analyzed through solving the time-dependent Schrödinger equation of the qubit under the effects of spin-orbit interaction and rapid ramping inter-dot detuning. The precondition, imposed on the ramping time and the tunnel-coupling strength, to ensure a high-fidelity inter-dot transfer is estimated. We then calculate the change in the spin orientation of a transferred qubit, and study the dependences of the spin deviation on the difference in the quantization axes of the two QDs, the tunnel-coupling strength, and the ramping time. We also demonstrate that the effect of multiple rounds of inter-dot bidirectional shuttling can be captured by an operator matrix, and evaluate the idling times required for realizing the single-qubit Pauli-X and Pauli-Y gates. Intriguingly, it is confirmed that a generalized Hadamard gate can be achieved through tuning the idling times.

This study numerically investigates magnetisation reversal processes driven by an external magnetic field in three-dimensional antiferromagnetic spin models with weak random field disorder. Considering an extremely weak disorder and low temperature, we observe a step-wise hysteresis loop and the appearance of short magnetisation bursts of a characteristic triangular shape; the number of bursts increases with disorder, indicative of Barkhausen-type noise. These phenomena are attributed to the simultaneous reversal at a given external field of segments composed of spins with identical neighbourhoods. A local random field orients one or more spin neighbours, resulting in small, ferromagnetic-like clusters distributed throughout the system. As disorder increases, these clusters may merge to form a labyrinthine structure within the antiferromagnetic background, facilitating brief avalanche propagation. The results demonstrate that, compared with familiar random-field ferromagnets, the observed antiferromagnetic Barkhausen noise and the related avalanche sequence have a profoundly different structure, organised into peaks associated with the transition between magnetisation plateaus. They exhibit prominent cyclical trends and disorder-dependent multifractal fluctuations, with the singularity spectrum quantifying the degree of disorder. The activity avalanches exhibit scale invariance resembling that recently found in experiments with disordered ferr\textit{i}magnets and martensites, as well as in quantum Barkhausen noise, which are associated with active geometric regions rather than individual-spin dynamics. The observed scaling behaviour is interpreted in terms of self-organised critical dynamics.

A thermal analogue of the classical brachistochrone problem, which minimizes the connection time between two equilibrium states of harmonically confined Brownian particles, has recently been solved theoretically. Here we report its experimental realization using two optically trapped microparticles subjected to a bang-bang effective temperature protocol. Despite their distinct relaxation times, both degrees of freedom are steered to their respective equilibrium states simultaneously in a finite minimal time. We provide a complete time-resolved characterization of the nonequilibrium dynamics through the evolution of the position variances and the entropy production within the framework of stochastic thermodynamics, enabling a quantitative comparison with direct relaxation and a suboptimal protocol. In addition, we employ information-geometric tools -- recently referred to as thermal kinematics -- to track the system's path in state space with a single dynamical quantity. Our results show that faster equilibration requires a larger entropy production and an increased thermodynamic length, revealing a direct trade-off between temporal optimality and thermodynamic cost in multidimensional stochastic systems driven by a single intensive control parameter.

One of the longstanding open questions in linear response theory concerns its true range of validity. Determining when the linear approximation can be trusted typically requires knowledge of second-order corrections, which are often difficult to compute explicitly. In this letter, I propose a self-consistent criterion for the validity of linear response, formulated in terms of a typical length scale that emerges from the fluctuation-response inequality within the theory itself. The result applies to classical open systems. I illustrate the criterion with explicit examples of Brownian particles in harmonic traps, and classical open systems presenting Kibble-Zurek mechanism. Finally, I discuss the physical meaning of this typical length, providing both thermodynamic and information-theoretic interpretations.

Non-Hermitian systems with globally reciprocal couplings -- such as the Hatano-Nelson model with stochastic imaginary gauge fields -- avoid the conventional non-Hermitian skin effect, displaying erratic bulk localization while retaining ballistic transport. An open question is whether similar behavior arises when non-reciprocity originates at the Liouvillian level rather than from an effective non-Hermitian Hamiltonian obtained via post-selection. Here we investigate this scenario in a lattice model with globally reciprocal Liouvillian dynamics and locally asymmetric incoherent hopping, a disordered setting in which Liouvillian-specific effects have remained largely unexplored. While the steady state again shows disorder-dependent, erratic localization without boundary accumulation, we find that global reciprocity in the Liouvillian does not protect transport. Instead, in the regime dominated by incoherent hopping, excitations spread via Sinai-type subdiffusion, dramatically slower than the ordinary diffusion found in symmetric stochastic lattices. Our results reveal a fundamental distinction between globally reciprocal Hamiltonian and Liouvillian dynamics: global reciprocity suppresses the skin effect in both cases, but only in Liouvillian dynamics can it coexist with ultra-slow, disorder-induced subdiffusive transport.

Compatibility with noisy intermediate-scale quantum (NISQ) devices is crucial for the realistic implementation of quantum cryptographic protocols. We investigate a cryptographic scheme based on discrete-time quantum walks (DTQWs) on cyclic graphs that exploits Parrondo dynamics, wherein periodic evolution emerges from a deterministic sequence of individually chaotic coin operators. We construct an explicit quantum circuit realization tailored to NISQ architectures and analyze its performance through numerical simulations in Qiskit under both ideal and noisy conditions. Protocol performance is quantified using probability distributions, Hellinger fidelity, and total variation distance. To assess security at the circuit level, we model intercept-resend and man-in-the-middle attacks and evaluate the resulting quantum bit error rate. In the absence of adversarial intervention, the protocol enables reliable message recovery, whereas eavesdropping induces characteristic disturbances that disrupt the periodic reconstruction mechanism. We further examine hardware feasibility on contemporary NISQ processors, specifically $ibm\_torino$, incorporating qubit connectivity and state-transfer constraints into the circuit design. Our analysis demonstrates that communication between spatially separated logical modules increases circuit depth via SWAP operations, leading to cumulative noise effects. By exploring hybrid state-transfer strategies, we show that qubit selection and connectivity play a decisive role in determining fidelity and overall protocol performance, highlighting hardware-dependent trade-offs in NISQ implementations.

We study the problem of the boundary conditions in the numerical simulation of closed and open quantum systems, described by a Schrödinger equation. On one hand, we show that a closed quantum system is defined by local boundary conditions. On the other hand, we argue that, because of the uncertainty principle, no local boundary condition can be defined for open quantum systems. For this reason plane waves or wave packet trains cannot be simulated on a finite numerical lattice with the usual procedures. We suggest a method that avoids these difficulties by using only a small numerical lattice and maintains the correspondence with the physical picture, in which the incident and scattered waves may be infinitely extended.

Tunnel junctions are one of the key elements of chip-scale microsystems serving various technologies from classical microelectronics to quantum information. Aluminium and its oxide (AlOx) have dominated cryogenic tunnel junction technology for decades due to the high quality of AlOx barriers and Al superconducting properties below 1.2 K. However, many applications require non-superconducting junctions, either standalone or in combination with superconducting technology, motivating efforts to suppress Al superconductivity through magnetic fields, doping, or proximity effects -- approaches that so far suffered from integration compatibility and scalability issues. Here, we present a CMOS-compatible normal-metal tunnel junction technology based on TiW alloy and AlOx barriers. We demonstrate wafer-scale fabrication of TiW/Al-AlOx/TiW junctions and validate their performance in Coulomb blockade thermometers operating down to 20 mK, confirming robust normal-state behavior. This TiW-based architecture offers a scalable solution for non-superconducting tunnel junctions across a broad temperature range, enabling integration into advanced cryogenic, quantum and nanoelectronic chip-level systems.

Strain engineering in semiconductor nanostructures offers a promising route to optimize electronic and optical properties for advanced quantum technologies. This study explores the relationship between core and shell thicknesses and strain distribution in Ge/Si core/shell nanowires, targeting their application as hosts for spin qubits. Nanowires were synthesized using an Au-catalyzed chemical vapor deposition technique, achieving control over core and shell dimensions. High-resolution transmission electron microscopy and elemental mapping confirmed structural integrity, while Geometric Phase Analysis and Raman spectroscopy provided quantitative insights into strain variations driven by core and shell dimensions. Furthermore, polarization resolved $μ$-Raman measurements allowed us to quantify the longitudinal and transverse phonon mode splitting as a function of strain in the Ge core. The strain dependent electronic properties were investigated by hole mobility measurements. Finally, we observe a record high hole mobility of 25,500 cm$^2$V$^{-1}$s$^{-1}$, underscoring the potential of these core/shell nanowire structures for the realization of high-fidelity spin qubits. Our findings highlight the critical role of geometry in strain tuning and provide valuable design guidelines for optimizing Ge/Si nanowires in scalable quantum device architectures.

The Adler-Bell-Jackiw chiral anomaly is shown to retain its Lorentz-invariant form, $\partial_μJ^μ_5 \propto \mathbf{E} \cdot \mathbf{B}$, in lattice Weyl systems beyond moderate magnetic fields, where neither Lorentz nor rotational symmetry is present. We show that the longitudinal and Hall magnetoconductivities factorize into a product of a universal part, governed by the chiral anomaly, and a non-universal part that depends on the density of states at the Fermi level. A rotationally invariant observable $\varkappa = σ(c_V/T)^2$ is introduced as a robust signature of the anomaly, where $σ$ denotes the Euclidean norm of the longitudinal and Hall conductivities and $c_V$ is the specific heat density. This quantity follows a universal $B^2$ dependence and scales as $|\cosΘ|$, with $Θ$ being the angle between $\mathbf{E}$ and $\mathbf{B}$. Through analytical derivation and full numerical simulation, we establish that $\varkappa$ remains universal independent of system parameters and of the orientation of the magnetic or electric field for fixed $Θ$. The emergent SO(3) symmetry in $\varkappa$ persists despite the absence of isotropy in both the microscopic model and the low-energy effective theory.

Microscale transport often relies on ubiquitous yet intrinsically random thermal fluctuations. Understanding how such fluctuations can be biased into directed motion has long been a central theme of nonequilibrium physics. Here, we introduce self-viscophoresis, a mechanism of autonomous motion based on the rectification of thermal fluctuations in a self-generated nonequilibrium viscosity field. Asymmetric colloidal particles dispersed in a thermoresponsive polymer solution induce local heating under uniform illumination, producing a spatially asymmetric viscosity profile around the particle and resulting in persistent directed motion. To elucidate the physical origin of this behavior, we develop a minimal Langevin model coupling isotropic thermal fluctuations to a dynamically updating temperature-viscosity field. The model shows that viscosity asymmetry anisotropically damps stochastic dynamics, effectively biasing thermal fluctuations into a net drift. It thus reproduces the observed directed motion without invoking deterministic propulsion terms associated with effective potentials or environmental fluid flows. Our results distinguish self-viscophoresis from conventional self-propulsion mechanisms and establish it as a general framework enabling reversible control of both the direction and dimensionality of motion.

At low temperatures, glasses exhibit distinctive properties compared to crystalline solids. A notable example is the phonon echo, a phenomenon that motivated the two-level-system (TLS) model. This model has successfully explained many universal anomalies in glasses. Here, we extend the TLS framework to a multi-level system and show that phonon echoes persist when nonlinear energy structures and disorder are included. By incorporating virtual phonon exchange, we introduce many-body interactions between these multi-level systems, leading to nonlinear eigen-energies that enhance the echo signal. Meanwhile, finite-temperature thermal fluctuations cause dephasing, resulting in a decay of echo amplitude over time. The analytical and numerical results are consistent across semi-classical and quantum regimes. Our work validates the multi-level-system model and underscores the role of many-body interactions in low-temperature glassy dynamics.

Recent experiments on donor-based spin qubits in silicon have leveraged the exchange interaction between electrons bound to separate donor nuclei to perform two-qubit operations. A consistently observed yet unexplained phenomenon in such systems is the significant increase in electron readout contrast, measured via Elzerman-style readout to a single-electron transistor (SET) island, when the donor nuclei are initialized in a parallel spin orientation compared to an anti-parallel orientation. In this work, we present a detailed analysis of the exchange-coupled donor system in the parallel nuclear regime and propose a physical mechanism for this effect. We attribute the enhanced readout contrast to an additional electron tunneling event to the SET during a single read period, when the donor nuclei are aligned in a parallel spin configuration. These insights inform strategies for improving electron readout fidelity in these systems and contribute to a more complete understanding of spin-dependent tunnelling processes in donor-based qubit architectures.

The dynamics of a self-propelled particle are typically hard-wired by its microscopic construction, limiting the range of behaviors accessible without redesigning the particle itself. Here we show that intramolecular nonreciprocity provides a minimal and versatile mechanism to overcome this constraint. We construct active molecules from short chains of two species of self-propelled particles whose propulsion directions are coupled nonreciprocally according to a prescribed internal sequence. At the single-molecule level, homogeneous sequences exhibit standard persistent random-walk dynamics, whereas heterogeneous sequences produce distinct trajectories inaccessible to either constituent species alone. At the collective level, using motility-induced phase separation (MIPS) as a representative example, we find that modifying the internal sequence shifts the MIPS onset by multiple orders of magnitude in propulsion strength, without altering particle-level interactions. These results demonstrate that intramolecular nonreciprocity among a small set of active components enables sequence-level programmability from single-molecule dynamics to emergent collective behavior, providing a minimal mechanism to encode and control active-matter dynamics across scales.

Polymer brushes and grafted polymers have attracted significant interest at the intersection of polymers, interfacial chemistry, colloidal science, and nanostructuring. The confinement of high-density grafted polymers and differences in swelling regimes govern the synthetic challenges and the interesting physics underlying their macromolecular dynamics. In this article, we focus on another intersection, artificial intelligence and machine learning (AI/ML), and how workflows will enhance the microstructure and composition of these systems. It will also accelerate potential applications through high-throughput experimentation (HTE) and data-driven intelligence, enabling scientific discovery and optimization. Applications in microfluidics, sensors, bioimplants, drug delivery, and related areas may yet offer more opportunities for ML-driven optimization. There is also interest in applying these studies with self-driving laboratories (SDLs) that can leverage autonomous systems for synthesis screening, characterization, and application evaluation.

We develop a fully coupled theoretical description of dislocation dynamics on deformable crystalline surfaces, using continuum modeling and the amplitude-phase-field crystal (APFC) framework extended to curved geometries. We derive a general kinematic expression for dislocation velocity directly from the complex-amplitude evolution equations, which is also applicable to deformed surfaces through curvature-modified differential operators. From numerical simulations, we show that even small out-of-plane deformations reshape the phenomenology of defect motion through curvature-induced self-propulsion, modified glide directions, and non-classical defect-defect interactions. Our results show how surface geometry profoundly influences defect dynamics and establish the surface-APFC model as a powerful framework for predicting and interpreting curvature-defect coupling across a wide range of systems, from stiff but deformable layers to soft matter surfaces and membranes that retain crystalline order.

Experiments at extreme strain rates and temperatures are critical for characterizing materials in high-speed applications. In this study, we develop a laser-driven particle impact platform capable of accelerating microparticles to supersonic velocities and impacting targets heated to temperatures approaching 2000 °C. The conventional laser-induced particle impact testing (LIPIT) system has been modified to enable high-temperature experiments through the integration of a resistive heating system and the development of a robust launch pad assembly suitable for accelerating particles in high-temperature environments. To eliminate the oxidation of materials at elevated temperatures, an optically accessible portable vacuum chamber has been developed and integrated into the setup. The capabilities of the system are demonstrated through a study of the temperature dependent particle impact cratering behavior of POCO graphite. With this new platform, high-velocity, high-temperature impact experiments can be performed in a controlled environment, supporting the investigation of materials under extreme conditions.

Propagation of short pulses through Cu$_2$O crystal containing Rydberg excitons is studied with the use of density matrix formalism and FDTD method. Saturation effects related to the so-called Rydberg blockade are studied extensively, exploring not only reduction of absorption (bleaching) but also power-dependent changes of the dispersive properties of the medium. The role of exciton lifetime and coherent population oscillations in the dynamics of the system is investigated. A pump-probe setup with two pulses is also studied, showing good agreement with recent experimental studies.

In this work, we report the development of a rubidium-based single-beam scalar optically pumped magnetometer (OPM) and demonstrate its application in measuring human cardiac magnetic fields in an unshielded environment. The developed magnetometers exhibit a noise floor below 15 pT/sqrt(Hz) in the frequency range of 1 to 35 Hz, with a measurement bandwidth of 100 Hz. When operated in a gradiometric configuration, the noise floor is further reduced to below 3 pT/sqrt(Hz) over the same frequency range. Magnetocardiography (MCG) signals were recorded at five different locations across the thorax. A clear polarity reversal of the QRS complex was observed across these measurement positions, confirming the spatial sensitivity of the system. The proposed system shows strong potential for clinical diagnostics, offering valuable physiological information through non-contact MCG measurements

Tensor networks offer a sign-problem-free approach to study lattice gauge theories, but extracting precise universal information associated with the deconfinement transition remains challenging. In this work, we study the deconfinement transition of (2+1)-dimensional $\mathbb{Z}_N$ lattice gauge theories at finite temperature using a thermal tensor network approach, where the partition functions at finite temperature are formulated as three-dimensional tensor networks. These tensor networks are first contracted in the temporal direction, and the subsequent coarse-graining in the spatial directions yields a renormalized transfer matrix, the spectrum of which directly encodes the universal conformal field theory data. In particular, by numerically extracting the central charge and scaling dimensions, we verify that the universality class of the thermal deconfinement transition matches the prediction of the Svetitsky-Yaffe conjecture for $N=2,3,5$. Moreover, we show that the $\mathbb{Z}_5$ theory at finite temperature exhibits an intermediate phase with an emergent U(1) symmetry. Critical couplings are determined via Gu-Wen ratios and agree with existing Monte Carlo simulations. Finally, extrapolating these critical couplings at finite temperature enables us to determine the deconfinement transition points for $N=2,3$ at zero temperature.

We adapt the microcanonical framework of equilibrium statistical mechanics to predict the statistics of short waves in inhomogeneous moving media. For steady inhomogeneities and background flow, we compute the wave spectrum at any location in the domain based on an ergodic prescription for the action density in phase space, constrained by conservation of absolute frequency. We illustrate the method for shallow-water waves subject to a background flow or to topographic inhomogeneities, and for deep-water surface capillary waves over a background flow, validating the predicted maps of rms surface elevation and interfacial slope against numerical simulations.

Synchronization of coupled oscillators is observed in many natural and engineered systems and emerges due to the interactions within the system. It can be both beneficial, e.g., in power grids, and harmful, e.g., in epileptic seizures. In the latter case, efficient control methods to desynchronize the systems are crucial. Recent studies have shown that interactions are not always pairwise, but higher-order, i.e., many-body, and this greatly affects the dynamics. For instance, higher-order interactions increase the linear stability of synchronized states but simultaneously shrink their attraction basin, with potentially opposite effects on control methods. Here, we use a minimally invasive pairwise control based on Hamiltonian control theory, and investigate its efficiency on phase oscillators with higher-order interactions. We show that, if the initial phases are close to the synchronized state, higher-order interactions make desynchronization more difficult to achieve. Otherwise, a non-monotonic effect appears: intermediate strengths of higher-order interactions impede desynchronization while larger ones facilitate it. In all cases, the control can desynchronize the system with a sufficient number of controlled nodes and intensity.

We develop a three-dimensional dynamical-systems framework for passive gliding and identify the global phase-space structures that organize its motion. Extending previous two-dimensional models of non-equilibrium gliding, we show that the 3D velocity dynamics possess an attracting, normally hyperbolic invariant surface, the terminal velocity manifold (TVM), onto which all trajectories rapidly collapse before evolving slowly toward a glide equilibrium. There is also a separatrix surface associated with an invariant manifold of an unstable equilibrium within the TVM, which partitions initial conditions into qualitatively distinct descent behaviors: efficient shallow glides versus steep, drag-dominated descent. Using lift-drag data from three representative airfoils--a snake-inspired bluff body, the Zimmerman planform characteristic of Draco lizards, and the classical NACA 0012--we compute the full equilibrium surfaces, analyze their pitch-roll bifurcations, and reconstruct the TVM and separatrix geometry in three dimensions. The results reveal that (i) equilibrium stability changes with both pitch and roll, rather than pitch alone; (ii) separatrix geometry determines the dynamic accessibility of shallow glides; and (iii) bio-inspired airfoils possess compact separatrix regions that make efficient gliding robust across a wide range of initial jump conditions. This work unifies biological and engineered gliders within a single global geometric framework and establishes separatrix geometry on the TVM as a principled diagnostic for glide robustness.

We compute general compatibility conditions between a weakly nonlocal homogeneous Hamiltonian operator and a third-order homogeneous Hamiltonian operator. Such operators determine a bi-Hamiltonian structure for many integrable PDEs (Korteweg--De Vries, Camassa--Holm, dispersive water waves, Dym, WDVV and others). Remarkably, the full set of conditions is purely algebraic and the first-order operator is completely determined by commuting systems of conservation laws that are Hamiltonian with respect to a third-order operator. We illustrate the above results with several examples, some of which, concerning WDVV equations, are new.

We present a new dynamic formation model of a star, a disk, arms, and companions using a coupled map lattice (CML), a complex systems approach. This CML simulates the viscoelastic and chaotic dynamics and evolution of gas clumps containing a little dust with a minimal set of one Eulerian procedure for the flow formation of gas clumps due to gravitational interaction, and one Lagrangian procedure for the collision and mixture of gas clumps due to viscoelastic advection. Despite its simplicity, this CML successfully obtains four typical astronomical objects consistent with protoplanetary disk observations: a central star, Keplerian disk, spiral arms, and even stellar, substellar, and planetary companions. All these formation processes are truly dynamic, with the central star "starring" in them, and they are not based on the conventional disk gravitational instability but on the central star gravitational instability with high-dimensional chaotic gas ejection, namely the chaotic itinerancy. Of particular note is the process in which diverse companions are formed due to the rapid density increase caused by the intersection of spiral arms. This suggests a novel companion formation scenario that should be called "arm-crossing companion formation" with a view to planet formation, which may overcome the radial drift barrier and angular momentum problem.

Using clean numerical simulation (CNS) which can give very accurate spatiotemporal trajectory of Navier-Stokes turbulence in a finite but long enough interval of time, we give some numerical evidences that the Navier-Stokes equations admit distinct global solutions from almost the same initial conditions whose difference is very small, i.e. even at the order $10^{-40}$ of magnitude. Hopefully these examples could provide some enlightenments for the uniqueness and existence of Navier-Stokes equations, which are related to one Millennium Prize Problem of Clay Institute.

We present the development of a data-driven, AI-based model of the Point Spread Function (PSF) that achieves higher accuracy than the current state-of-the-art approach, "PSF in the Full Field-of-View'' (PIFF). PIFF is widely used in leading weak-lensing surveys, including the Dark Energy Survey (DES), the Hyper Suprime-Cam (HSC) Survey, and the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). The PSF characterizes how a point source, such as a star, is imaged after its light traverses the atmosphere and telescope optics, effectively representing the "blurred fingerprint'' of the entire imaging system. Accurate PSF modeling is essential for weak gravitational lensing analyses, as biases in its estimation propagate directly into cosmic shear measurements -- one of the primary cosmological probes of the expansion history of the Universe and the growth of large-scale structure for dark energy studies. To address the limitations of PIFF, which constructs PSF models independently for each CCD and therefore loses spatial coherence across the focal plane, we introduce a deep-learning-based framework for PSF reconstruction. In this approach, an autoencoder is trained on stellar images obtained with the Hyper Suprime-Cam (HSC) of the Subaru Telescope and combined with a Gaussian process to interpolate the PSF across the telescope's full field of view. This hybrid model captures systematic variations across the focal plane and achieves a reconstruction error of $3.4 \times 10^{-6}$ compared to PIFF's $3.7 \times 10^{-6}$, laying the foundation for integration into the LSST Science Pipelines.

Restrictive phenomenological assumptions represent a major roadblock for the development of accurate subgrid-scale models of fluid turbulence. Specifically, these assumptions limit a model's ability to describe key quantities of interest, such as local fluxes of energy and enstrophy, in the presence of diverse coherent structures. This paper introduces a symbolic data-driven subgrid-scale model that requires no phenomenological assumptions and has no adjustable parameters, yet it outperforms leading LES models. A combination of a priori and a posteriori benchmarks shows that the model produces accurate predictions of various quantities including local fluxes across a broad range of two-dimensional turbulent flows. While the model is inferred using LES-style spatial coarse-graining, its structure is more similar to RANS models, as it employs an additional field to describe subgrid scales. We find that this field must have a rank-two tensor structure in order to correctly represent both the components of the subgrid-scale stress tensor and the various fluxes.

Identifying influential nodes and edges in directed networks remains a fundamental challenge across domains from social influence to biological regulation. Most existing centrality measures face a critical limitation: they either discard directional information through symmetrization or produce sparse, implementation-dependent rankings that obscure structural importance. We introduce a unified spectral framework for centrality analysis in directed networks grounded in the Singular value decomposition of the incidence matrix. The proposed approach derives both vertex and edge centralities via the pseudoinverse of Hodge Laplacians, yielding dense and well-resolved rankings that overcome the sparsity limitations commonly observed in betweenness centrality for directed graphs. Unlike traditional measures that require graph symmetrization, our framework naturally preserves directional information, enabling principled hub/authority analysis while maintaining mathematical consistency through spectral graph theory. The method extends naturally to hypergraphs through the same mathematical foundation. Experimental validation on real-world networks demonstrates the framework's effectiveness across diverse domains where traditional centrality measures encounter limitations due to implementation dependencies and sparse outputs.

Interpolating scattered CFD datasets onto a uniform Cartesian grid can distort the true geometry, producing a convex-hull type envelope and activating nonphysical regions. This work presents a reconstruction framework that recovers physically consistent masks before exporting CNN-ready fields. It introduces two novel strategies, distance-based masking and an adaptive alpha-shape formulation that normalizes alpha using local data resolution, and evaluates them against classical alpha-shape boundary recovery. A quantitative, topology-aware metric suite is introduced to assess retention, suppression of unsupported regions, overlap consistency, and connectivity. The novel distance-based method is robust across the geometries considered under the same threshold rule, with tau set to the minimum CFD grid spacing, and achieves 500-800 times speedups over classical alpha-shapes. The adaptive alpha-shape remains stable when its control parameter is set to 1 and is 1.7-2.6 times faster than the classical variant, which requires geometry-specific alpha tuning. A lightweight boundary inflation post-process using a minimal dilation further improves retention by up to 2.96% with negligible unsupported activation (less than 0.08%). Overall, the distance-based method is recommended as the default due to its accuracy, stability, minimal tuning, and low cost, while the adaptive alpha-shape is a strong alternative when grid-spacing information for threshold selection is unavailable. A companion web application operationalizes the workflow end to end, enabling 2D ASCII dataset upload, parameter tuning, mask and boundary generation, and export of CNN-ready outputs.

The model of the attention economy, where content producers compete for the attention of users, relies on two key forces: information supply and demand. This study leverages the feedback loop between these forces to develop a method for detecting and quantifying information voids, i.e., periods in which little or no reliable information is available on a given topic. Using a case study on COVID-19 vaccines rollout in six European countries, and drawing on data from multiple platforms including Facebook, Google, Twitter, Wikipedia, and online news outlets, we examine how information voids emerge, persist and correlate with a decline in the proportion of high-quality information circulating online. By conceptualising information voids as a specific regime of information spreading, we also quantify their counterpart, information overabundance, which constitute a central component of the current definition of infodemic. We show that information voids are associated with a higher prevalence of misinformation, thus representing problematic hotspots in which individuals are more likely to be misled by low-quality online content. Overall, our findings provide empirical support for the inclusion of information voids in mechanistic explanations of misinformation emergence.

Household size impacts the spread of respiratory infectious diseases: Larger households tend to boost transmission by acquiring external infections more frequently and subsequently transmitting them back into the community. Furthermore, mandatory interventions primarily modulate contagion between households rather than within them. We developed an approach to quantify the role of household size in epidemics by separating within-household from out-household transmission, and found that household size explains 41% of the variability in cumulative COVID-19 incidence across 34 European countries (95% confidence interval: [15%, 46%]). The contribution of households to the overall dynamics can be quantified by a boost factor that increases with the effective household size, implying that countries with larger households require more stringent interventions to achieve the same levels of containment. This suggests that households constitute a structural (dis-)advantage that must be considered when designing and evaluating mitigation strategies.

Electronic structure calculations remain a major bottleneck in atomistic simulations and, not surprisingly, have attracted significant attention in machine learning (ML). Most existing approaches learn a direct map from molecular geometries, typically represented as graphs or encoded local environments, to molecular properties or use ML as a surrogate for electronic structure theory by targeting quantities such as Fock or density matrices expressed in an atomic orbital (AO) basis. Inspired by the Hohenberg-Kohn theorem, in this work, we propose an operator-centered framework in which the external (nuclear) potential, expressed in an AO basis, serves as the model input. From this operator, we construct hierarchical, body-ordered representations of atomic configurations that closely mirror the principles underlying several popular atom-centered descriptors. At the same time, the matrix-valued nature of the external potential provides a natural connection to equivariant message-passing neural networks. In particular, we show that successive products of the external potential provide a scalable route to equivariant message passing and enable an efficient description of long-range effects. We demonstrate that this approach can be used to model molecular properties, such as energies and dipole moments, from the external potential, or learn effective operator-to-operator maps, including mappings to the Fock matrix and the reduced density matrix from which multiple molecular observables can be simultaneously derived.

The Global Positioning System (GPS) includes a continuously operating, planet-scale network of atomic clocks that, beyond navigation and time dissemination, enables precision tests of fundamental physics. Here we use GPS carrier phase archival data to perform a retrospective search for exotic low-mass fields (ELFs) that might be emitted by the binary neutron-star merger GW170817, complementing gravitational wave and electromagnetic modalitiesnin multi-messenger astronomy. Such ultra-relativistic fields would imprint a dispersive, anti-chirp signature in clock-frequency time series, delayed with respect to the LIGO-Virgo gravitational wave detection. We construct network-median pseudo-frequency data from eighteen Rb satellite clocks referenced to a terrestrial hydrogen maser and conduct a template-bank search spanning ELF pulse duration, arrival delay, and characteristic frequency. No statistically significant signal is observed after accounting for noise statistics and template-bank trials. We derive 95\% confidence-level lower bounds on the interaction energy scale $Λ_α$ of quadratic couplings driving variations in electromagnetic fine-structure constant. These limits improve upon existing astrophysical and gravity-test constraints across the ELF-energy range $\approx10^{-18}$--$10^{-14}\,\mathrm{eV}$. This demonstrates that mature global satellite-clock networks provide an observational capability for retrospective, multi-messenger searches for new physics using decades of archival timing data.

Proton-conducting solid acids could enable water-free operation of high-temperature fuel cells. However, systematic materials screening has, hitherto, been computationally prohibitive. Here, we introduce a two-stage high-throughput screening strategy that directly computes proton diffusion coefficients, enabled by machine-learned interatomic potentials fine-tuned to ab initio data. Starting from more than six million materials, our screening -- based on structural motifs rather than empirical descriptors -- identifies $27$ high-performing proton conductors, including over ten previously unexplored compounds. These include sustainable and commercially available materials, candidates that have not yet been synthesized, organic systems that fall outside conventional design rules, and known proton conductors that validate our approach. Importantly, our findings reveal a universal oxygen--oxygen distance of approximately $2.5$~Å at the moment of proton transfer across diverse chemistries, providing mechanistic insight and showing that macroscopic proton conductivity emerges from the interplay between anion rotational dynamics, hydrogen-bond network connectivity, and proton-transfer probability.

Collective behavior is commonly attributed to direct interactions among system components. Using a minimal stochastic model, we show that higher-order collective structure can instead emerge from shared stochastic environments, even in the absence of interactions. Quantified via the O-information, environmental fluctuations induce both redundant and synergistic dependencies, with the latter occupying larger regions of the correlation space. We establish a no-go theorem showing that time-independent coupling between the system variables and a shared stochastic environment rules out synergistic higher-order behavior. Crucially, this constraint can be overcome dynamically: transitions between redundancy and synergy arise from time-dependent environmental coupling or from the nontrivial interplay between shared environments and direct interactions. Together, these results identify environmental mediation as a distinct mechanism of higher-order collective organization beyond the conventional interaction-centric paradigm.

The Fission Matrix method is used to perform fast and still accurate neutronics simulations. It relies on precalculated databases obtained through a Monte Carlo simulation. To represent every state of the reactor, multiple databases are required. The actual state of the reactor is obtained from those databases. In this paper, we analyze the effect of the temperature profiles selected to construct the databases. To do so, the Molten Salt Fast reactor is selected. Two sets of databases are studied: the first uses temperature profiles obtained from high-fidelity Multiphysics simulations with Cardinal, and the second uses uniform temperature profiles. Results showed improved multiplication factor and fission source distribution when the temperature profiles used to generate the databases were similar to those expected when solving the fission matrix.

Particle-in-cell methods with stochastic collision models are commonly used to simulate collisional plasma dynamics, with applications ranging from hypersonic flight to semiconductor manufacturing. Code verification of such methods is challenging due to the interaction between the spatial- and temporal-discretization errors, the statistical sampling noise, and the stochastic nature of the collision algorithm. In this paper, we introduce our code-verification approaches to apply the method of manufactured solutions to plasma dynamics, and we derive expected convergence rates for the different sources of discretization and statistical error. For the particles, we incorporate the method of manufactured solutions into the equations of motion. We manufacture the particle distribution function and inversely query the cumulative distribution function to obtain known particle positions and velocities at each time step. In doing so, we avoid modifying the particle weights, eliminating risks from potentially negative weights or modifications to weight-dependent collision algorithms. For the collision algorithm, we average independent outcomes at each time step and we derive a corresponding manufactured source term for the velocity change for each particle. By having known solutions for the particle positions and velocities, we are able to compute the error in these quantities directly instead of attempting to compute differences in distribution functions. These approaches are equally valid for particle-in-cell simulations with Monte Carlo collisions and direct simulation Monte Carlo simulations of neutral gas flows. We demonstrate the effectiveness of our approaches in three dimensions for different couplings between the particles and field, with and without binary elastic collisions, and with and without coding errors.

Density functional theory (DFT) has been critical towards our current atomistic understanding of catalysis on transition-metal surfaces. It has opened new paradigms in rational catalyst design, predicting fundamental properties, like the adsorption energy and reaction barriers, linked to catalytic performance. However, such applications depend sensitively on the predictive accuracy of DFT and achieving experimental-level reliability, so-called transition-metal chemical accuracy (13 kJ/mol), remains challenging for present density functional approximations (DFAs) or even methods beyond DFT. We introduce a new framework for designing DFAs tailored towards studying molecules adsorbed on transition-metal surfaces, building upon recent developments in non-self-consistent DFAs. We propose two functionals within this framework, demonstrating that transition-metal chemical accuracy can be achieved across a diverse set of 39 adsorption reactions while delivering consistent performance for 17 barrier heights. Moreover, we show that these non-self-consistent DFAs address qualitative failures that challenge current self-consistent DFAs, such as CO adsorption on Pt(111) and graphene on Ni(111). The resulting functionals are computationally practical and compatible with existing DFT codes, with scripts and workflows provided to use them. Looking ahead, this framework establishes a path toward developing more accurate and sophisticated DFAs for computational heterogeneous catalysis and beyond.

We present a plane-wave (PW) implementation of the auxiliary-field quantum Monte Carlo (AFQMC) method within the projector augmented-wave (PAW) formalism in the Vienna ab initio Simulation Package (VASP). By employing an exact inversion of the PAW overlap operator, our approach maintains cubic scaling while naturally operating at the complete basis set limit defined by the PW cutoff. We benchmark this framework by calculating the equilibrium lattice constants and bulk moduli of C, BN, BP, and Si. Our analysis demonstrates that AFQMC systematically corrects the lack of long-range screening in MP2 and the missing higher-order exchange in RPA. We identify RPA as the optimal reference method due to the rapid convergence of the remaining short-range correlations with respect to supercell size. The resulting lattice constants exhibit a mean absolute relative error of 0.14 % relative to experiment, establishing the method as a rigorous benchmark tool for structural properties in condensed matter systems.

The expansion of global electricity distribution systems necessitates the deployment of massive infrastructure. Assessing its implications from a spatial and material perspective requires an understanding of the core drivers of a distribution grid configuration. Our model samples substation locations using a non-linear relationship with population density and constructs the network applying the Kruskal algorithm. This streamlined approach generates realistic grid structures at the local scale and provides accurate estimates of the total network length at the national scale. Using highly granular population data, this local model reveals a profound connection between population spread and distribution grid, which appears to persist at the global level. Potentially driven by the emergent properties of population scaling laws, the full network characteristics appear to be well described by multivariate power laws on aggregated population and area. Validated across 35 countries, these results provide new multi-scale tools for characterizing electrical infrastructure and reveal key determinants of distribution grid extent.

In this paper, we propose a domain decomposition dynamical low-rank method to solve high-dimensional radiative transfer problems and similar kinetic equations. The algorithm uses a separate low-rank approximation on each spatial subdomain, which means that, for a given accuracy, we can often use a smaller overall rank compared to classic dynamical low-rank methods. In particular, we can solve problems with point sources efficiently, that for classic algorithms require almost full rank. Our algorithm only transfers boundary data between subdomains and is thus very attractive for distributed memory parallelization, where classic dynamical low-rank algorithms suffer from global data dependency. We demonstrate the efficiency of our algorithm by a number of challenging test examples that have both very optical thin and thick regions.

We present the results of a comprehensive survey conducted among members of the Spanish Astronomical Society (Sociedad Espanola de Astronomia, SEA) to assess well-being, professional satisfaction, and family-work balance of researchers in astronomy. The survey addressed multiple aspects of professional life, including happiness, career stability, publication pressure, and access to childcare services during scientific meetings. Responses were examined across gender and career stages to identify trends and sources of dissatisfaction.

We present and test a recursive regularised lattice Boltzmann method for incompressible magnetohydrodynamic (MHD) flows. The approach is based on a double-distribution formulation, in which the magnetic field is evolved using a standard BGK lattice Boltzmann scheme, while the fluid solver is enhanced through a Hermite-based recursive regularisation of the non-equilibrium moments. The method exploits a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice, allowing higher-order isotropy to be retained while selectively filtering spurious non-hydrodynamic contributions. The regularisation procedure reconstructs the non-equilibrium distribution from physically consistent Hermite coefficients, avoiding explicit evaluation of velocity gradients. The resulting scheme preserves the correct incompressible MHD limit, improves numerical stability at low viscosities, and reduces lattice-dependent artefacts. The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems.

Machine learning models are increasingly applied in materials science, yet their predictive power is often constrained by data scarcity. Here, we show that accurate predictions can be achieved, even with a limited number of training examples, provided the dataset is compact and and grounded in physically relevant quantities. By combining density functional theory calculations with a machine-learning framework, we construct accurate descriptor-based models to predict the formation energies of doped lepidocrocite TiO$_2$ monolayers. The predictive accuracy of machine-learning models was first evaluated for single-dopant Pt configurations, demonstrating that the selected structural and chemical descriptors reliably capture the key factors governing dopant stability. Chemical transferability is then examined by extending the dataset to include Ag-doped configurations. Predictive accuracy improved systematically as additional Ag-doped data points were included in the training, while the performance of Pt remains robust. These results highlight the potential of small and well-curated datasets combined with physically informed descriptors to enable not only accurate but also chemically transferable machine-learning-driven screening in doped TiO$_2$ monolayer.

Sonification -- the mapping of data to non-speech audio -- offers an underexplored channel for representing complex dynamical systems. We treat El Niño-Southern Oscillation (ENSO), a canonical example of low-dimensional climate chaos, as a test case for culturally-situated sonification evaluated through complex systems diagnostics. Using parameter-mapping sonification of the Niño 3.4 sea surface temperature anomaly index (1870--2024), we encode ENSO variability into two traditional Javanese gamelan pentatonic systems (pelog and slendro) across four composition strategies, then analyze the resulting audio as trajectories in a two-dimensional acoustic phase space. Recurrence-based diagnostics, convex hull geometry, and coupling analysis reveal that the sonification pipeline preserves key dynamical signatures: alternating modes produce the highest trajectory recurrence rates, echoing ENSO's quasi-periodicity; layered polyphonic modes explore the broadest phase space regions; and the two scale families induce qualitatively distinct coupling regimes between spectral brightness and energy -- predominantly anti-phase in pelog but near-independent in slendro. Phase space trajectory analysis provides a rigorous geometric framework for comparing sonification designs within a complex systems context. Perceptual validation remains necessary; we contribute the dynamical systems methodology for evaluating such mappings.

Molecular and polymeric crystals show a wide range of functional properties that arise from the interplay between the atomic-scale structure of their constituent molecules and the organization of these molecules within the crystal lattice at macroscopic length scales. X-ray diffraction can provide structural information at these disparate length scales, but usually only through experiments that address one or the other of molecular (or unit-cell) structure versus crystal structure. Consequently, the accuracy of determined molecular or polymer structures may be limited by unaccounted crystal inhomogeneities of the crystal lattice and the characterization of crystalline materials might not reveal the underlying causes of crystal morphology. Here we introduce X-ray convergent-beam diffraction to obtain spatially-resolved structural information from crystals by projection topographic imaging. Using highly focusing X-ray multilayer Laue lenses, we show that Bragg reflections can be mapped into tomographic images of the crystal, for the characterization of strain and defects at high resolution. We demonstrate how the crystal morphology obtained this way can be accounted for when determining structure factors as a function of position in the crystal. The approach may assist in studies such as diffusion and binding in MOFS, protein-drug binding, crystal growth, and the mechanical responses of photo-reactive or thermally driven dynamic crystals.

The rapid advancement of artificial intelligence in materials science requires data standards and data management practices that can capture the complexity of real-world structures, including surfaces, interfaces, defects, and dimensionality reduction. We present M-CODE - Materials Categorization via Ontology, Dimensionality and Evolution - a compact categorization system that links materials-science-specific terminology to a set of reusable concepts as building blocks and provenance-aware transformations. M-CODE classifies structures by dimensionality, structural complexity (from pristine to compound pristine, defective, and processed), and variants that capture common structure creation and evolution approaches. A practical implementation of the categorization is provided in an open-source codebase that includes JSON schemas, examples, and Python and TypeScript types/interfaces, designed to support reproducible dataset generation, validation, and community contributions.

Geometry design is a crucial and challenging strategy for improving the performance of type-II superconductors. Topology optimization is one of the most powerful approaches used to determine structural geometries. Therefore, a topology optimization approach is presented to inversely design structural geometries of both low- and high-temperature type-II superconductors with superconductor-dielectric/vacuum interfaces. In the presented approach, the magnetic response of type-II superconductors is modeled using the Ginzburg-Landau theory, where the temporal evolution of the order parameter and vector potential is described by the time-dependent Ginzburg-Landau equations under the Weyl gauge.

Travel time prediction is central to transport geography and planning's accessibility analyses, sustainable transportation infrastructure provision, and active transportation interventions. However, calculating accurate travel times, especially for driving, requires either extensive technical capacity and bespoke data, or resources like the Google Maps API that quickly become prohibitively expensive to analyze thousands or millions of trips necessary for metropolitan-scale analyses. Such obstacles particularly challenge less-resourced researchers, practitioners, and community advocates. This article argues that a middle-ground is needed to provide reasonably accurate travel time predictions without extensive data or computing requirements. It introduces a free, open-source minimally-congested driving time prediction model with minimal cost, data, and computational requirements. It trains and tests this model using the Los Angeles, California urban area as a case study by calculating naive travel times from open data then developing a random forest model to predict travel times as a function of those naive times plus open data on turns and traffic controls. Validation shows that this interpretable machine learning method offers a superior middle-ground technique that balances reasonable accuracy with minimal resource requirements.

Synaptic transmission must balance the need for reliable signalling against the metabolic cost of achieving that reliability. How energetic constraints shape synaptic precision and its regulation during plasticity remains unclear. Here we develop an energy--constrained framework in which synapses minimise postsynaptic response variance subject to a fixed mean and an effective energy budget. Combinations of candidate physiological costs are used to estimate an energy cost for synaptic transmission; this cost is then inferred from quantal statistics. Analysing five published pre- and post-plasticity datasets, we find that observed synaptic mean--variance pairs cluster near a minimal-energy boundary, indicating that precision is limited by energetic availability. Model comparison identifies a dominant calcium pump-like cost paired with a smaller vesicle turnover-like cost, yielding a separable precision--energy relationship, $σ^{-2} \propto E^5$. We further show that plasticity systematically updates synaptic energy budgets according to the scale-free magnitude of mean change, enabling accurate prediction of post-plasticity variance from energy allocation alone. These results provide direct experimental support for the hypothesis that synaptic precision is governed by energy budgets, establishing energy allocation as a fundamental principle linking metabolic constraints, synaptic reliability, and plasticity.

Connecting the dynamics of biomolecular networks to experimentally measurable cell phenotypes remains a central challenge in systems biology. Here we introduce a model-based definition of phenotype as a partial steady state that is committed to a certain dynamical outcome while otherwise being minimally constrained. We focus on Boolean models and define \emph{dynamical phenotypes} as complete trap spaces that maximally specify a chosen set of phenotype-determining nodes that correspond to biomarkers while keeping external inputs unconstrained. We show that dynamical phenotypes can be efficiently identified without full attractor enumeration. Using four published models, including a 70-node Boolean model of T cell differentiation, we show that dynamical phenotypes recover known cell types and activation states, and indicate the environmental conditions ensuring their existence. We also propose a method to identify informative phenotype-determining nodes based on the canalization of the Boolean functions. This method reveals biologically relevant cell state information that is complementary to the phenotypes manually defined by model creators and is validated by two attractor-based approaches. Our results demonstrate that dynamical phenotypes provide a scalable framework for linking model structure, external inputs, and phenotypic outcomes, and offer a principled tool for model-guided biomarker selection.

Adaptive systems must strike a balance between prediction and surprise to thrive in uncertain environments. We propose an information-theoretic balance function, $ f(p) = -(1 - p)\ln(1 - p) + \ln p $, which quantifies the net informational gain from contrasting explained variance $p$ with unexplained novelty $(1 - p)$. This function is strictly concave on $(0,1)$ and reaches its unique maximum at $ p^* \approx 0.882$, revealing a regime where confidence is high but the residual uncertainty carries a disproportionate potential for surprise. Independently of this maximum, imposing a self-similarity condition between known, unknown and total information, $p : (1-p) = 1 : p$, leads to the golden-ratio reciprocal $p = 1/\varphi \approx 0.618$, where $ \varphi$ is the golden ratio. We interpret this value not as the maximizer of $f$, but as a structurally privileged \emph{partition} in which known and unknown are proportionally nested across scales. Embedding this dual structure into a Compute-Inference-Model-Action (CIMA) loop yields a dynamic process that maintains the system near a critical regime where prediction and surprise coexist. At this edge, neuronal dynamics exhibit power-law structure and maximal dynamic range, while the system's response to perturbations becomes convex at the level of its payoff function-fulfilling the formal definition of antifragility. We suggest that the golden-ratio partition is not merely a mathematical artifact, but a candidate design principle linking prediction, surprise, criticality, and antifragile adaptation across scales and domains, while the maximum of $f$ identifies the point of greatest informational vulnerability to being wrong.

Antimicrobial protocols - using substances such as antibiotics or disinfectants - remain the preferred option for preventing the spread of pathogenic bacteria. However, bacteria can develop mechanisms to reduce their antimicrobial susceptibility, which can lead to treatment failure and the selection of resistance or tolerance. In this work, we propose a minimal population dynamics model to study bacterial selection during cyclic antimicrobial application, a commonly used protocol. Selection in bacterial populations with heterogeneous antimicrobial susceptibility is modelled here as a trade-off between survival advantage (reduction in antimicrobial killing) and potential fitness costs (reduction in growth rate) of the less susceptible strains. The proposed model allows us to derive useful expressions for determining the success of cyclic antimicrobial treatments based on two bacterial traits: growth and kill rates. The results obtained here are directly applicable to preventing the selection and spread of resistant and tolerant bacterial strains in real-life protocols.

Markov chain Monte Carlo (MCMC) sampling of densities restricted to linearly constrained domains is an important task arising in Bayesian treatment of inverse problems in the natural sciences. While efficient algorithms for uniform polytope sampling exist, much less work has dealt with more complex constrained densities. In particular, gradient information as used in unconstrained MCMC is not necessarily helpful in the constrained case, where the gradient may push the proposal's density out of the polytope. In this work, we propose a novel constrained sampling algorithm, which combines strengths of higher-order information, like the target's log-density's gradients and curvature, with the Hit-&-Run proposal, a simple mechanism which guarantees the generation of feasible proposals, fulfilling the linear constraints. Our extensive experiments demonstrate improved sampling efficiency on complex constrained densities over various constrained and unconstrained samplers.

We consider Susceptible-Infected-Recovered (SIR) models on dense dynamic random graphs, in which the joint dynamics of vertices and edges are co-evolutionary, i.e., they influence each other bidirectionally. In particular, edges appear and disappear over time depending on the states of the two connected vertices, on how long they have been infected, and on the total density of susceptible and infected vertices. Our main results establish functional laws of large numbers for the densities of susceptible, infected, and recovered vertices, jointly with the underlying evolving random graphs in the graphon space. Our results are supported by simulations, which characterize the limiting size of the epidemics, i.e., the limiting density of susceptible vertices, and how the peak of the epidemics depends on the rate of the evolution of the underlying graph. The proofs of our main results rely on the careful construction of a mimicking process, obtained by approximating the two-way feedback interaction between vertex and edge dynamics with a mean-field type interaction, acting only as one-way feedback, that remains sufficiently close to the original co-evolutionary process. To treat the more general setting in which edge dynamics are affected by the proportions of susceptible and infected individuals, we introduce a methodological extension of existing techniques. We thus show that our model exhibits multiple epidemic peaks -- a phenomenon observed in real-world epidemics -- which can emerge in models that incorporate mutual feedback between vertex and edge dynamics.

Cryo-electron tomography (cryo-ET) provides a unique window into molecular organization in cellular environments (in situ). However, the interpretation of molecular structural information is complicated by several intrinsic properties of cryo-ET data, such as noise, missing wedge, and continuous conformational variability of the molecules. Additionally, in crowded in situ environments, the number of particles extracted is sometimes small and precludes extensive classification into discrete states. These challenges shift the emphasis from high-resolution structure determination toward validation and interpretation of low-resolution density maps, and analysis of conformational flexibility. Molecular Dynamics (MD) simulations are particularly well suited to this task, as they provide a physically grounded way to explore continuous conformation transitions consistent with both experimental data and molecular energetics. This review focuses on the roles of MD simulations in cryo-ET, emphasizing their use in emerging methods for conformational landscape determination and their contribution to gain new biological insight.

QwaveMPS is an open-source Python library for simulating one-dimensional quantum many-body waveguide systems using matrix product states (MPS). It provides a user-friendly interface for constructing, evolving, and analyzing quantum states and operators, facilitating studies in quantum physics and quantum information with waveguide QED systems. This approach enables efficient, scalable simulations by focusing computational resources on the most relevant parts of the quantum system. Thus, one can study a wide range of complex dynamical interactions, including time-delayed feedback effects in the non-Markovian regime and deeply non-linear systems, at a highly reduced computational cost compared to full Hilbert space approaches, making it both practical and convenient to model a variety of open waveguide-QED systems (in Markovian and non-Markovian regimes), treating quantized atoms and quantized photons on an equal footing.

Current bounds on the tensor-to-scalar ratio imply that the energy scale of inflation may lie below the grand-unified scale. In this paper, we show that in a broad class of single-field inflation models with sufficiently small energy scales, an extremely efficient tachyonic instability develops at the end of inflation. This instability rapidly drives the system into a nonlinear regime before coherent oscillations can be established, leading to a first-order phase-transition--like phenomenon without tunneling or barrier crossing. The resulting ultra-relativistic shock fronts surrounding the bubble interiors expand to near the Hubble scale, corresponding to the most strongly enhanced tachyonic modes, and collide with one another, producing energetic inflaton particles and gravitational waves. As a result, the post-inflationary dynamics can differ significantly from the conventional high-scale inflationary scenario. Interestingly, inflation at MeV--EeV energy scales can be probed via gravitational-wave observations, including pulsar timing arrays, ground-based detectors, and future space-based experiments. Recent limits from the LIGO--KAGRA--Virgo collaboration already constrain EeV-scale inflation, while pulsar timing array results may be interpreted as evidence for gravitational waves generated by GeV-scale inflation. We also briefly discuss further implications of the resulting tachyonic shocks.

Radial oscillations provide a clean dynamical test of the high-density stiffness of neutron-star equations of state. We study spherically symmetric pulsations of nonrotating relativistic stars built from cold, charge-neutral, $β$-equilibrated pure nucleonic matter described within relativistic mean-field theory. As a baseline we adopt the UCIa parameter set [Astron. Astro-phys. 689, A242 (2024)], and we implement high-density stiffening via the $σ$-cut scheme by adding a regulator potential $U_{\rm cut}(σ)$ [Phys. Rev. C 92, no.5, 052801 (2015), Phys. Rev. C 106, no.5, 055806 (2022)]. For representative choices $f_s=0$ (no cutoff) and $f_s=0.58$ (stiffened), we solve the Tolman-Oppenheimer-Volkoff and tidal perturbation equations to obtain equilibrium sequences, mass-radius relations, and tidal deformabilities. We then derive and solve the linear general-relativistic radial pulsation equations to compute the eigenfrequencies and eigenfunctions of the fundamental and overtone modes. The $σ$-cutoff suppresses the growth of the scalar field at supranuclear density, increases the pressure, and shifts the maximum mass, radii, and $Λ_{1.4}$ accordingly, while systematically raising the radial-mode frequencies at fixed mass. Using the sign change of $ω_0^2$ as a stability criterion, we identify stiffened models that remain radially stable up to the observed $\sim 2M_\odot$ mass scale and are consistent with current multimessenger constraints, demonstrating how radial spectra complement static EoS tests.

A deformation of Heisenberg algebra induces among other consequences a loss of Hermiticity of some operators that generate this algebra. Therefore, these operators are not Hermitian, nor is the Hamiltonian operator built from them. In the present paper, we propose a position deformation of Heisenberg algebra with both maximal length and minimal momentum uncertainties. By using a pseudo-similarity transformation to the non-Hermitian operators, we prove their Hermiticity with a suitable positive-definite pseudo-metric operator. We then construct Hilbert space representations associated with these pseudo-Hermitian operators. Finally, we study the eigenvalue problem of a free particle in this deformed space and we show that this deformation curved the quantum levels allowing particles to jump from one state to another with low energy transitions.

In this paper, we provide a complete mathematical theory for the entanglement of mixtures of Dicke states. These quantum states form an important subclass of bosonic states arising in the study of indistinguishable particles. We introduce a tensor-based parametrization where the diagonal entries of these states are encoded as a symmetric tensor, enabling a direct translation between entanglement properties and well-studied convex cones of tensors. Our results bridge multipartite entanglement theory with semialgebraic geometry and the theory of completely positive and copositive tensors. This dictionary maps separability to completely positive tensors, the PPT property to moment tensors, entanglement witnesses to copositive tensors, and decomposable witnesses to sum of squares tensors. Using this framework, we construct explicit PPT entangled states in three or more qutrits. In this class of states, we establish that PPT entanglement exists for all multipartite systems with three qutrits or more, disproving a recent conjecture in [J. Math. Phys. 66, 022203 (2025)]. We also show that, for mixtures of Dicke states, the PPT condition with respect to the most balanced bipartition implies PPT with respect to any other bipartition. We further connect bosonic extendibility of mixtures of Dicke states to the duals of known hierarchies for non-negative polynomials, such as the ones by Reznick and Polya. We thus provide semidefinite programming relaxations for separability and entanglement testing in the Dicke subspace.

We consider the issue of steady state coherences in a single qubit in the case of a composite system-bath interaction as proposed in \cite{Guarnieri18}. Based on a field theoretical approach we reanalyse the issue within a Redfield description. We find that the Redfield approach in accordance with \cite{Guarnieri18} yields steady state coherences but violating the properties of a quantum map also gives rise to negative populations. The issue is resolved by applying the Lindblad equation which is in accordance with a proper quantum map. The Lindblad equation, however, also implies the absence of steady state coherence. We conclude that steady state coherence in a a qubit is incompatible with a quantum map.

We establish results on the dynamics of interacting charged null fluids in general relativity, specifically in the context of the bouncing continuation proposed in [Ori91]. In this model - the setting for a number of prominent case studies on black hole formation - charged massless particles may instantaneously change direction (bounce) after losing all their 4-momentum due to electrostatic repulsion. We initiate the study of timelike bounce hypersurfaces in spherical symmetry: scenarios in which an incoming beam of charged null dust changes direction along a timelike surface $\mathcal{B}$, which is the (free) boundary of an interacting 2-dust region. We identify a novel decoupling of the equations of motion in this region. First, it is shown that every timelike curve segment $γ$ in the spherically symmetric quotient of Minkowski or Reissner-Nordström spacetimes arises as the bounce hypersurface $\mathcal{B}$ of a charged null dust beam incident from past null infinity $\mathcal{I}^-$. We construct a spacetime $(\mathcal{M},g_{μν})$ describing the full trajectory of the beam, which includes gluing to Reissner-Nordström and Vaidya regions. Across $\mathcal{B}$ the metric has regularity $g_{μν}\in C^{2,1}$ and satisfies Einstein's equation classically, while $C^\infty$ gluing may be achieved across all other interfaces. We also obtain examples of timelike bounce hypersurfaces terminating in a null point. Since these constructions are teleological, we secondly consider a given charged incoming beam from past null infinity. We formulate and solve a free boundary problem which represents the formation of a timelike bounce hypersurface. The result is conditional, applying only in the exterior region of a Reissner-Nordström spacetime, and subject to a technical regularity condition.

We study an inverse resonance problem for the scalar wave equation on the Kerr-de Sitter family. In a compact subextremal slow-rotation regime and at a fixed overtone index, high-frequency quasinormal modes admit semiclassical quantization and a real-analytic labeling by angular momentum indices. Using this structure, we first prove that a finite equatorial high-frequency package of quasinormal-mode frequencies determines the mass and rotation parameter $(M,a)$ (for fixed cosmological constant $Λ>0$), with a quantitative stability estimate. As a key geometric input we compute explicit second-order (in $a$) corrections to the equatorial photon-orbit invariants which control the leading real and imaginary parts of the quasinormal modes. Finally, allowing $Λ$ to vary in a compact interval, we show that adding one damping observable (the scaled imaginary part of a single equatorial mode) yields a three-parameter inverse theorem: a finite package of three independent real observables determines $(M,a,Λ)$ locally in the slow-rotation regime away from $a=0$.

We explore the industrial and scientific applicability of the VQE-LSTM framework by integrating meta-learning with GPU accelerated quantum simulation using NVIDIA's CUDA-Q (CUDAQ) platform. This work demonstrates how an LSTM-FC meta-initialization module can extend the practical reach of the Variational Quantum Eigensolver (VQE) in both chemistry and physics domains. In the chemical regime, the framework predicts ground-state energies of molecular Hamiltonians derived from PySCF, achieving near FCI accuracy while maintaining favorable O(N^2) scaling with molecular size. In the physical counterpart, we applied the same model to quantized Simple Harmonic Motion systems (SHM), successfully reproducing its ground and excited states through VQE and Variational Quantum Deflation (VQD) methods. Benchmark results on NVIDIA GPUs reveal significant speedups over CPU-based implementations, validating CUDAQ's capability to handle large-scale variational workloads efficiently. Overall, this study establishes VQE-LSTM as a viable and scalable approach for GPU accelerated quantum simulation, bridging quantum chemistry and condensed-matter physics through a unified, meta-learned initialization strategy.

Energy consumption is becoming a serious bottleneck for integrating quantum technologies within the existing global information infrastructure. In photonic architectures, considerable energy overheads stem from using lasers, whose high coherence was long considered indispensable for quantum state preparation. Here, we demonstrate that natural, incoherent sunlight can successfully produce quantum-entangled states via spontaneous parametric down-conversion. We detect polarization-entangled photon pairs with a concurrence of $0.905\pm0.053$ and a Bell state fidelity of $0.939\pm0.027$. Importantly, the system violates Bell's inequality with $S=2.5408\pm0.2171$, exceeding the classical threshold of 2, while maintaining generation rates comparable to laser-based setups. These findings pave the way for sustainable quantum applications in resource-limited environments like interplanetary missions.

Entanglement swapping between photon pairs generated at physically separated nodes over telecommunication fiber infrastructure is an essential step towards the quantum internet, enabling applications such as quantum repeaters, blind quantum computing, distributed quantum computing, and distributed quantum sensing. However, successful networked entanglement swapping relies on generating indistinguishable pairs of photons and preserving them over deployed fibers. This has limited most previous demonstrations to laboratory settings or relied on sophisticated methods to maintain the necessary indistinguishability. Here, we demonstrate a scalable entanglement swapping experiment using naturally indistinguishable entanglement sources based on warm atomic vapor cells. Without sharing lasers or optical frequency references between nodes, nor the need for pulsing the sources, we achieve a swapping rate of nearly 500 pairs/s while maintaining the CHSH parameter above 2. Additionally, we demonstrate the scalability of our method by maintaining the quality of the entanglement swapping on 17.6-km of deployed fibers in NYC, relying on commercially available SPADs at the spoke nodes, SNSPDs at the hub and standard time-synchronization techniques. Our work paves the way for the practical deployment of large-scale hub-and-spoke quantum networks within cities and data centers.

We study correlation properties of polaritons at zero temperature in a multiconnected Jaynes--Cummings (MCJC) lattice on a superconducting circuit quantum electrodynamics platform with engineered cross-Kerr nonlinearity that mimics attractive nearest-neighbour interaction. A multi-connected Jaynes--Cummings lattice is a one-dimensional lattice constructed from alternating qubits and resonators with different left and right couplings. The nearest-neighbour interaction or cross-Kerr coupling is implemented dispersively through ladder-type qutrits between each nearest neighboring pair of resonator modes. Projecting onto the lower-polaritonic manifold, we derive an extended two-mode (bipartite) Bose--Hubbard-like model featuring on-site and attractive nearest-neighbor interactions. Employing a continuum bosonization approach, we express the Hamiltonian in terms of symmetric ($+$) and antisymmetric ($-$) collective modes. In the regime where the ($-$) sector acquires a finite gap, one can reduce the system to an effective single-component Luttinger liquid model for the $+$ sector. The cross-Kerr term reduces the compressibility of the ($+$) mode, thereby enhancing the corresponding Luttinger parameter $K_{+}$, resulting in the slower algebraic decay of single-particle correlations, $G(x)\propto|x|^{-1/(4K_{+})}$.

The ingoing and outgoing null expansions associated to a spatial 2-sphere are quantized in the spherically symmetric model of loop quantum gravity. It is shown that the resulting expansion operators are self-adjoint in the kinematical Hilbert space with generalized eigenstates. It turns out that the outgoing and ingoing expansion operators share the common continuous part of their spectra but have different additional isolated eigenvalues. These results provide new insights on the avoidance of the singularities in classical general relativity and the establishment of certain notion of quantum horizons.

Nonlinear phase gates are essential to achieve the universality of continuous-variable quantum processing and its applications. We present a deterministic protocol for generating nonlinear phase gates in trapped ion systems using simultaneous two-tone sideband drives beyond the Lamb-Dicke regime. Our approach harnesses higher-order interaction terms typically neglected or suppressed to construct nonlinear phase gates. This methodology enables high-fidelity gate engineering with a near three-fold reduction in control pulses compared to state-of-the-art theoretical proposals.

Electron diffraction from nanogratings provides a platform for free-electron interferometry, yet controlled manipulation of electron spin in such geometries remains largely unexplored. In particular, the role of the self-generated magnetic field arising from electron motion and the feasibility of coherent spin control without disrupting diffraction coherence have not been quantitatively investigated. In this article, a self-consistent Maxwell-Pauli framework is developed to study spin-resolved electron diffraction from nanogratings in the presence of magnetic fields. The model incorporates geometric confinement, image-charge interactions, self-generated magnetostatic fields, and externally applied magnetic fields. Numerical simulations show that the intrinsic magnetic self-field produced by the electron probability current is several orders of magnitude too weak to induce measurable spin mixing, demonstrating that nanogratings act as spin-conserving beam splitters under field-free conditions. When a uniform magnetic field is applied upstream of the nanograting, coherent Larmor precession enables controlled spin rotation without modifying the diffraction geometry or degrading coherence. The magnetic field required for a $π$ spin rotation scales inversely with the interaction length and electron de Broglie wavelength $λ_{dB}$. Furthermore, a downstream nonuniform magnetic field applied after the nanograting imparts a spatially varying Zeeman phase, producing opposite transverse momentum shifts for the two spin components. The spin-dependent transverse dynamics is analyzed using Husimi Q-function phase-space maps, which visualize spin-dependent population redistribution and momentum separation. The proposed approach enables tunable spatial separation of spin-resolved free electron beams and establishes an all-magnetic route for coherent spin rotation, control, and interferometry.

In this paper, we investigate the dynamics and gravitational-wave signatures of periodic orbits of spinning test particles moving in the equatorial plane around static, spherically symmetric black holes within the framework of Deser-Woodard nonlocal gravity. Based on the Mathisson-Papapetrou-Dixon equations, combined with the Tulczyjew spin supplementary condition, we derive the orbital dynamic equations for spinning particles moving in the equatorial plane and impose a timelike constraint to exclude unphysical superluminal trajectories. By comparing with the classical Schwarzschild black hole, we systematically analyze the effects of the nonlocal gravitational parameters $ζ$ and $b$ on the effective potential governing the radial motion of particles and the innermost stable circular orbit. In addition, gravitational waveforms exhibit significant phase differences: an increase in $ζ$ induces a phase delay, whereas an increase in $b$ results in a phase advance. A one-year simulation of the orbital evolution of an extreme mass ratio inspiral demonstrates that when $b=2$ and $ζ\approx10^{-6}$, the mismatch between the gravitational waveforms predicted for the nonlocal gravity black hole and those for the Schwarzschild black hole reaches the distinguishable threshold ($\mathcal{M}=0.0125$), providing a basis for observational discrimination between general relativity and nonlocal gravity.

The recent demonstrations of cascaded PDC (CPDC) and the hopeful prospects of realizing third-order PDC (TOPDC) for the generation of three-photon entangled states are paving the way for experimental studies on genuine three-photon interference. In this article, we formulate three-photon interference in terms of ``each three-photon interfering only with itself.'' We show that although a generalized two-alternative three-photon interference setup based on CPDC or TOPDC involves eight different length parameters, the interference can be fully characterized in terms of only three independent parameters. The first parameter is the three-photon path-length difference, which has a direct analog in the one-photon and two-photon cases, and the other two parameters quantify the path-asymmetry length. Unlike two-photon interference, which requires only one parameter to quantify path-asymmetry, two independent parameters are needed in three-photon interference. This results in a broader class of nonclassical three-photon effects, including three-photon HOM-type effects. Our work provides the theoretical basis for existing and future three-photon interference experiments exploring the rich and complex quantum correlations associated with three-particle entanglement and potentially enabling the development of novel protocols for harnessing those correlations.

In this work, we investigate in detail the thermodynamic properties of a spherically symmetric ModMax-AdS black hole sourced by a global monopole within the Kalb-Ramond gravity. We derive the key thermodynamic quantities, including the Hawking temperature, Gibbs free energy, and specific heat capacity, and analyze how the geometric parameters influence these physical quantities. The first law of thermodynamics and the corresponding Smarr formula are explicitly verified. Furthermore, we study the thermodynamic criticality of the system by deriving the critical points and examining the effects of the space-time geometric parameters. We also obtain the inversion temperature and demonstrate that the minimum inversion temperature is modified by the space-time parameters. In addition, the sparsity of Hawking radiation and thermal fluctuations of the system are investigated, highlighting the effects of the parameters on the entropy corrections. Finally, we analyze the optical properties of the black hole, in particular the photon sphere and shadow radius, showing how these parameters influence these features.

In this work, we investigate several phenomenological aspects of a covariant quantum-corrected Reissner-Nordström black hole characterized by the mass $M$, electric charge $Q$, and the quantum correction parameter $ζ$. We first study the motion of neutral test particles and derive the fundamental orbital and epicyclic frequencies, which are then employed to analyze different quasi-periodic oscillation (QPO) models. Using observational QPO data from stellar-mass, intermediate-mass, and supermassive black hole candidates, we perform a Bayesian parameter estimation through a Markov Chain Monte Carlo (MCMC) analysis and obtain constraints on the black hole parameters. The results show that the presence of the quantum correction significantly affects the location of the QPO radii and the separation between the QPO orbit and the ISCO. We then examine the scalar perturbations by deriving the Schrödinger-like radial equation and the corresponding effective potential. The influence of the parameters $Q$ and $ζ$ on the perturbation potential and stability of the spacetime is discussed. Furthermore, we compute the greybody factor and the energy emission rate in the high-frequency (geometric-optics) regime, showing how the quantum correction modifies the absorption probability and radiation spectrum. Finally, we study the effect of thermal fluctuations on the black hole entropy and obtain the logarithmic corrections to the Bekenstein-Hawking area law. We show that these corrections become important for small black holes, while for large horizon radius the standard thermodynamic behavior is recovered. Our analysis demonstrates that the quantum correction parameter leaves observable imprints on both dynamical and thermodynamical properties of the spacetime and can be constrained through QPO observations.

We introduce a systematic framework for three-qubit entanglement classification using a cascaded architecture of Support Vector Machine (SVM) classifiers. Leveraging the well defined three-qubit structure with the four nested entanglement classes (S, B, W, and GHZ), we construct three distinct witness models ($\mathcal{M}_{B}$, $\mathcal{M}_{W}$, and $\mathcal{M}_{GHZ}$) that sequentially discriminate between these classes. The proposed Cascaded model achieves an overall classification accuracy of $95\%$ on a comprehensive dataset of mixed states. The framework's robustness and generalization capabilities are confirmed through rigorous testing against out-of-distribution (OOD) entangled states and various quantum noise channels, where the model maintains high performance. A key contribution of this research is an optimization protocol based on systematic feature importance analysis. This approach yields a tunable framework that significantly reduces the number of required features, while maintaining reliable model accuracy.

Non-trivial spatial topology of the Universe can imprint potentially observable signatures on the cosmic microwave background (CMB). In this study, we investigate how deviations from the standard nearly-scale-free primordial power spectrum impact observables for the fully compact, orientable Euclidean topologies ($E_1$--$E_6$). We examine how such deviations modify the detectability of the underlying topology, depending on whether they are an intrinsic consequence of non-trivial topology or independent of it. We compute CMB temperature correlation matrices across a range of topologies, fundamental domain sizes, and observer locations for both standard and modified primordial power spectra. The impact of these modifications on the detectability of topology is quantified using the Kullback-Leibler divergence, providing an estimate of the distinguishability of non-trivial and simply-connected topologies based solely on CMB temperature observations. In addition, we employ the CatBoost machine learning algorithm to classify harmonic-space realizations of CMB temperature maps and thereby assess the observational prospects for topology detection. Signatures of non-trivial topology are encoded in the off-diagonal structure of the CMB temperature correlation matrices and are most prominent on the largest angular scales. Deviations from the simple power-law primordial spectrum at these scales can substantially alter the detectability of topology, either enhancing its characteristic CMB imprints or suppressing them below observational sensitivity. Our results demonstrate that uncertainties in the primordial power spectrum must be carefully accounted for in robust searches for cosmic topology using the CMB.

Displacement memory, induced by a wave pulse in a regular black hole spacetime, is studied using geodesic (timelike) separation and geodesic deviation. The presence of the wave pulse in such a black hole is modeled via a function $H(u)$ appearing in a restricted version of a generic Bondi-Sachs type line element. Choosing a sech-squared profile for $H(u)$, we first study (numerically) geodesic separation and geodesic deviation in a flat background. Thereafter, similar investigations are carried out in the presence of the black hole, but in regions far away from the vicinity of the horizon. Our results suggest the presence of a distinct displacement memory effect, which depends on the value of the regularisation parameter $g$ as well as the pulse height. Between different types of regular black holes, one notices parameter-dependent changes in the net displacement memory. Further, a clear difference in the magnitude of displacement memory (at large $u$) in regular and singular black holes is also visible in our numerical results.

Quantum correlations, crucial for the advantage and advancement of quantum science and technology, arise from the impossibility of expressing a quantum state as a tensor product over a given set of parties. In this work, a generalized notion of correlations via arbitrary products is formulated. Remarkably, as a universal property, the connection between such general products and tensor products is established, allowing one to relate generic non-product states to the common notion of entangled states. We construct the set of free operations for general types of products by extending the local-operation-and-classical-communication paradigm, familiar from standard entanglement theory, thereby establishing a resource theory of correlations for general products. A generalization is provided beyond two factors that can be universally related to multipartite entanglement. Applications that highlight the usefulness of the approach are discussed, such as the factorization of fermionic states, the non-local factorization of multi-photon states into single-photon states, and the interesting possibility of understanding prime numbers as a form of single-party entanglement.

A hybrid three-form model of dark energy is developed in order to identify scaling solutions, a long-sought feature in three-form cosmology. Exploiting Hodge dualities, the theory is formulated in terms of two scalar functions that are associated with the conjugate momentum, and the three-form dual vector in an isotropic background. The resulting Lagrangian yields a stable scaling attractor where the three-form energy density tracks the dominant background fluid. A dynamical mechanism is also identified that naturally drives the system out of this regime toward a late-time accelerated phase distinguishable from a cosmological constant. This constitutes the first realization of scaling behavior within a three-form dark energy framework.

Quantum architectures based on neutral atoms have gained significant attention in recent years as specialized computational machines due to their ability to directly encode the independent set constraint on graphs, exploiting the Rydberg blockade mechanism. In this work, we address the Drone Delivery Packing Problem via a hybrid quantum-classical framework leveraging a neutral-atom quantum processing unit (QPU). We reformulate the optimization task as a graph-partitioning problem based on the independent sets (ISs) of a scheduling graph that encodes delivery incompatibilities. Each partition corresponds to deliveries assigned to a single drone, with the objective of minimizing the total number of partitions. While the ISs represent time-feasible schedules, battery-duration constraints are enforced through a classical post-processing routine. This methodology enables the recovery of optimal delivery schedules, provided a sufficient number of samples is collected from the QPU to resolve the solution space. We benchmark the hybrid workflow through numerical emulations and demonstrate its effectiveness on Pasqal's Fresnel QPU, reporting hardware experiments with configurations of up to 100 atoms.

For a static and spherically symmetric spacetime, we investigate the class of exact solutions that arise when two fundamental geometric constraints are imposed simultaneously: the Karmarkar's condition and the vanishing of the Weyl tensor. These conditions restrict the curvature in such a way that the spacetime becomes conformally flat and belongs to the family of embedding class-I solutions. Even though the subsequent solutions namely, the Schwarzschild interior solution and the de Sitter solution are well known, the novelty of our presentation is that these solutions are shown to be a direct consequence of the imposed geometric constraints. The physical matter composition becomes highly constrained by the associated geometry under such conditions. The Schwarzschild interior solution describes the spacetime of an incompressible fluid sphere while the de Sitter solution corresponds to a vacuum energy dominated configuration. Interestingly, pressure anisotropy as well as `complexity factor' vanish identically once the Karmarkar's condition and the conformal flatness conditions are applied simultaneously. As these two geometric constraints alone are sufficient to determine the background spacetime uniquely, Karmarkar's condition might not be a suitable method for the development of realistic stellar models in a conformally flat spacetime unless one invokes other factors into consideration such as time-dependent metric potentials.

We analyze numerically the performance of Quantum Reservoir Computing (QRC) for statistical and financial problems. We use a reservoir composed of two superconducting islands coupled via their charge degrees of freedom. The key non-linear elements that provide the reservoir with rich and complex dynamics are the Josephson junctions that connect each island to the ground. We show that QRC implemented in this circuit can accurately classify complex probability distributions, including those with heavy tails, and identify regimes in correlated time series, such as periods of high volatility generated by standard econometric models. We find QRC to outperform some of the best classical methods when the amount of information is limited. This demonstrates its potential to be a noise-resilient quantum learning approach capable of tackling real-world problems within currently available superconducting platforms. We further discuss how to improve our QRC algorithm in real superconducting hardware to benefit from a much larger Hilbert space.

A key tenet of general relativity is the dynamical nature of space-time, ideally represented as an initial value problem. Here we explore the variational formulation of classical Einstein-Hilbert gravity as initial value problem by constructing its Schwinger-Keldysh-Galley (SKG) action, including a careful treatment of boundary terms. The construction is based on a doubling of degrees of freedom and independent of a foliation. The action naturally decomposes into a bulk term furnishing Einstein's equations and a boundary term, which is related to conserved quantities, such as the Komar mass. We find that since only trivial connecting conditions must be specified on boundaries, the variational action principle for gravity as an initial value problem is rendered well-posed without the need to add additional boundary terms. The SKG approach to gravity offers a novel and complementary avenue to solve for the metric of spacetime directly from the action, bypassing the governing equations.

We provide a detailed analysis of the non-twisting subcase of the large class of type D black holes with a non-aligned electromagnetic field, presented recently in [H. Ovcharenko and J. Podolsky, Phys. Rev. D 112 (2025) 064076]. We show that such exact solutions split into two main subclasses that (after a suitable re-parametrization) can be interpreted as either the uncharged Schwarzschild or C-metric in the external Bertotti-Robinson (BR) spacetime with geometry ${\mathrm{AdS}_2\times\mathrm{S}_2}$, or as the charged Reissner-Nordstrom black hole accelerating in the external BR electromagnetic field. The distinction between these two subclasses is determined by the parameter $r_0$ that encodes relations between the external Maxwell field (given by the non-aligned components of the Faraday tensor ${Φ_0=Φ_2}$) and the Maxwell field created by the charge of the black hole (given by the aligned component $Φ_1$). Namely, if ${r_0=0}$ then the electromagnetic field is fully determined by ${Φ_0=Φ_2}$, and one gets the C-metric in the BR universe (including also the non-accelerating Schwarzschild-BR black hole). But if ${r_0\neq 0}$ then the electromagnetic field is independently determined by both the external BR field and the field of a black hole itself, and this can be interpreted as the Reissner-Nordstrom black hole accelerating in the Bertotti-Robinson spacetime. Even though such an interpretation of the spacetime family is quite simple, it contains a lot of subtleties (e.g. the no-charge limit of the RN-BR spacetime, the non-trivial dependence on the signs of the mass and charge of a black hole, extreme black holes, and others) which we carefully investigate in this work. We also show the explicit relation to solutions previously found by Van den Bergh and Carminati, and we discuss the connection to the Alekseev-Garcia and Alexeev solutions.

We study the task of quantum state exclusion, focusing on antidistinguishability and its generalization to $x$-antidistinguishability, under global measurements and local operations with classical communication (LOCC). We also introduce weak and strong notions of antidistinguiahbaility ($x$-antidistinguishability) depending on whether all states or all $x$-tuples are exhaustively eliminated. Our results reveal striking differences between state exclusion and the more familiar task of state discrimination. In particular, we show that LOCC antidistinguishability of multipartite product states is symmetric with respect to the initiating party but this symmetry breaks down for higher-order $x$-antidistinguishability. Most notably, we establish a manifestation of \emph{nonlocality without entanglement} in the context of state exclusion: we prove that three bipartite product states can be globally antidistinguishable while failing to be LOCC antidistinguishable, demonstrating that three is the minimal number of states required for this phenomenon. We further extend this separation to $2$-antidistinguishability and present example exhibiting the same type of nonlocality. At last, we provide an antidistinguishable tripartite product states that are not LOCC antidistinguishable across any bipartition, which ensures the phenomenon of \emph{genuine nonlocality without entanglement} in this framework.

We address the challenge, commonly referred to as the cosmological averaging problem, of relating the large-scale evolution of an inhomogeneous Universe to that predicted by a homogeneous matter distribution in theories of gravity with nonminimal matter-gravity couplings. To this end, we focus on the class of $f(R,T)$ models defined by $f(R,T)=R+F(T)$, which provide a simple yet theoretically consistent realization of nonminimal matter-gravity interactions and can be reformulated as general relativity minimally coupled to a modified matter Lagrangian. Using nonstandard global monopole solutions as a toy model for realistic particles, we show that the spatial average of $F$ typically differs significantly from $F$ evaluated at the spatially averaged trace of $T$, implying that homogeneous cosmological models generally fail to capture the correct large-scale dynamics of the Universe. We further show that dust in these theories generally exhibits a non-vanishing proper pressure. Our results underscore the necessity of properly accounting for spatial averaging when modeling cosmology in theories with nonminimal matter-gravity couplings.

Primordial black holes (PBHs) formed in the early Universe evaporate via Hawking radiation and constitute a generic source of stochastic gravitational waves. Existing studies of gravitational wave production from evaporating PBHs typically assume vacuum evaporation, neglecting the fact that PBHs in the early Universe are embedded in a hot thermal plasma. In this work, we investigate gravitational wave production from primordial black holes whose evaporation is thermally influenced by their surrounding environment. We adopt a thermal evaporation framework in which interactions with the ambient plasma modify the effective decay rate of the black hole, leading to enhanced mass loss at early times and a redistribution of the evaporation history compared to the standard non-thermal vacuum case. Since graviton emission is intrinsically tied to the evaporation history of PBHs, these thermal effects play a crucial role in determining the timing and spectral properties of the resulting stochastic gravitational wave background. Our results provide a consistent framework for incorporating thermal effects into gravitational wave production from evaporating primordial black holes and set the stage for a detailed analysis of their observational signatures.

In this manuscript, we study the Schrödinger cat state transfer in a quantum optical version of Newton's cradle in non-Markovian environment. Based on a non-Markovian master equation, we show that the cat state can be transferred purely through the memory effect of the non-Markovian common environment, even without any direct couplings between neighbor cavities. The mechanism of the environment induced cat state transfer is analyzed both analytically and numerically to demonstrate that the transfer is a unique phenomenon in non-Markovian regime. From this example, the non-Markovian environment is shown to be qualitatively different from the Markovian environment reflected by the finite versus zero residue coherence. Besides, we also show the influence of environmental parameters are crucial for the transfer. We hope the cat state transfer studied in this work may shed more light on the fundamental difference between non-Markovian and Markovian environments.

We investigate quantum and classical signatures of a Schwarzschild black hole embedded in a Hernquist dark matter halo. Starting from the exact spherically symmetric solution describing this composite system, we analyze particle production for both bosonic and fermionic fields using semiclassical techniques. Hawking radiation is derived through Bogoliubov transformations and independently via the tunneling method with energy conservation, allowing us to identify the effective temperature, emission spectrum, and the role of dark matter parameters in suppressing particle creation. The evaporation process is examined in the high-frequency regime, leading to modified evaporation times and emission rates relative to the vacuum Schwarzschild case. We further study absorption and scattering of massless scalar waves employing a partial-wave analysis, computing phase shifts, partial and total cross sections, and assessing the impact of the Hernquist scale radius and density on these observables. Finally, null and timelike geodesics are explored to characterize light propagation and particle motion in the presence of the dark matter halo.

In traditional research, chaos is frequently accompanied by non-linearity, which typically stems from non-linear interactions or external driving forces. However, in this paper, we present the chaotic behavior that is completely attributed to the non-linear back-reaction of non-Markovian environment. To be specific, we derive the dynamical equations of an optomechanical system and demonstrate that the non-linearity (cause of chaos) in the equations arises entirely from the time-domain convolutions (TDCs) induced by non-Markovian corrections. Under Markovian conditions, these TDCs are reduced into constants, thereby losing the nonlinearity and ultimately leading to the disappearance of chaos. Furthermore, we also observe chaos generation in the absence of optomechanical couplings, which further confirms that the non-Markovian effect is the sole inducement of chaos and the environmental parameters play important roles in the generation of chaos. We hope these results may open a new direction to investigate chaotic dynamics purely caused by non-Markovian environments.

We propose a stochastic Schrödinger equation (SSE) approach to investigate the dynamics of giant atoms coupled to a waveguide, addressing two critical gaps in existing research, namely insufficient exploration on continuous coupling and multiple excitations. A key finding is that continuous coupling, unlike discrete coupling at finite points, breaks the constant phase difference condition, thereby weakening the interference effects in giant atom-waveguide systems. In addition, a key technical advantage of the SSE approach is that auto- and cross-correlation functions can directly reflect the complex photon emission/absorption processes and time-delay effects in giant atom-waveguide systems. Moreover, the SSE approach also naturally handles multiple excitations, without increasing equation complexity as the number of excitations grows. This feature enables the investigation of multi-excitation initial states of the waveguide, such as thermal and squeezed initial states. Overall, our approach provides a powerful tool for studying the dynamics of giant atoms coupled to waveguide, particularly for continuous coupling and multi-excitation systems.

Quantum low-density parity-check (qLDPC) codes are promising candidates for fault-tolerant quantum computation due to their high encoding rates and distances. However, implementing logical operations using qLDPC codes presents significant challenges. Previous research has demonstrated that self-dual qLDPC codes facilitate the implementation of transversal Clifford gates. Here we introduce a method for constructing self-dual qLDPC codes by stacking non-self-dual qLDPC codes. Leveraging this methodology, we develop double-chain bicycle codes, double-layer bivariate bicycle (BB) codes, double-layer twisted BB codes, and double-layer reflection codes, many of which exhibit favorable code parameters. Additionally, we conduct numerical calculations to assess the performance of these codes as quantum memory under the circuit-level noise model, revealing that the logical failure rate can be significantly reduced with high pseudo-thresholds.

We systematically investigate the implementation of a universal gate set via double-braiding within SU(2)k anyon models. The explicit form of the double elementary braiding matrices (DEBMs) in these models are derived from the F-matrices and R-symbols obtained via the q-deformed representation theory of SU(2). Using these EBMs, standard single-qubit gates are synthesized up to a global phase by a Genetic Algorithm-enhanced Solovay-Kitaev Algorithm (GA-enhanced SKA), achieving the accuracy required for fault-tolerant quantum computation with only 2-level decomposition. For two-qubit entangling gates, Genetic Algorithm (GA) yields braidwords of 30 braiding operations that approximate the local equivalence class [CNOT]. Theoretically, we demonstrate that performing double-braiding in a three-anyon (six-anyon) encoding of single-qubit (two-qubit) is topologically equivalent to a protocol requiring the physical manipulation of only one (three) anyons to execute arbitrary braids. Our numerical results provide strong evidence that double-braiding in SU(2)k anyons models is capable of universal quantum computation. Moreover, the proposed protocol offers a potential new strategy for significantly reducing the number of non-Abelian anyons that need to be physically manipulated in future braiding-based topological quantum computations (TQC).

A charged black hole can emit charged particles via two independent mechanisms: the Hawking radiation and the Schwinger effect, which are intertwined in the radiation spectrum. In this paper, we will show that the two effects can be distinguished by analyzing the entanglement entropy carried by the produced particle pairs. Explicitly, we apply the island formula to the near extremal Reissner-Nordstr{ö}m (RN) black hole to calculate the total entanglement entropy of the radiation. Meanwhile we use the heat kernel method to calculate the entanglement entropy of charged particle pairs produced solely from the Schwinger effect. By comparing with the total entanglement entropy, we obtain the entanglement entropy produced purely from the Hawking radiation. Consequently, the two effects are distinguishable in near extremal RN black holes after the Page time. Furthermore, we also employ the brick wall model and the Pauli-Villars regularization to derive the entanglement entropy from the Schwinger effect, which gives a slightly different result with that obtained from the heat kernel method.

Fault-tolerant quantum computation requires minimizing non-Clifford gates, whose implementation via magic state distillation dominates the resource costs. While $T$-count minimization is well-studied, dedicated $CCZ$ factories shift the natural target to direct Toffoli minimization. We develop algebraic methods for this problem, building on a connection between Toffoli count and tensor decomposition over $\mathbb{F}_2$. On standard benchmarks, these methods match or improve all reported results for both Toffoli and $T$-count, with most circuits completing in under a minute on a single CPU instead of thousands of TPUs used by prior work.

Or should we talk about dS/CFT correspondence or dS/SFT correspondence in cosmological correlators? In non-unitary field theories -- which are conjectured to be dual to cosmological correlators -- scale invariance does not necessarily imply full conformal invariance. While general relativity predicts the emergence of conformal invariance (or boost symmetry in the bulk), various modified theories of gravity suggest only scale invariance, characterized by the absence of bulk boost symmetry. We demonstrate this distinction using Einstein-Aether theory as a canonical example.

In this work, we propose an open quantum battery that stores and releases energy by employing a two-mode ultrastrongly coupled bosonic system, with one mode (the charger) coupled to an independent heat reservoir. Our results demonstrate that both the charging energy and ergotropy of the quantum batteries can be significantly enhanced within the ultra-strong coupling regime and across a broader temperature range in transient time. A unidirectional energy flow is achieved by controlling the system's initial state through its two-mode squeezed ground state. Furthermore, we show that the steady-state stored energy, along with its corresponding ergotropy, can be enhanced at larger temperatures and stronger coupling strengths. Notably, a purely beam-splitter or two-mode squeezing interaction yields zero ergotropy. These findings indicate that the enhanced stored energy and ergotropy of the quantum battery arises principally from the combined effects of beam-splitter and parametric amplification (squeezing) couplings. In addition, the presence of the squared electromagnetic vector potential term can prevent a phase transition and achieve a significant charging energy and high ergotropy in the deep-strong coupling regime. The results presented herein enhance our understanding of the operating principles of open bosonic quantum batteries.

Chiroptical responses in atomic systems are usually weak, as they arise from the interference between electric- and much weaker magnetic-dipole transitions. We show that atoms arranged in chiral geometries can instead exhibit a strong collective chiroptical response mediated entirely by electric-dipole interactions. Using a coupled-dipole framework, we identify a regime of pronounced chiroptical response emerging at subwavelength interatomic separations, which can be tuned by the probe frequency. This enhancement is directly linked to the formation of subradiant collective modes. Our results establish a fundamental connection between geometric chirality and collective light-matter interactions, opening new pathways for engineering and exploiting chiral optical responses in atomic systems.

The rapid advancement of quantum information science and technology (QIST) has generated significant attention from people in academia, industry, and the public. Recent advances in QIST have led to both opportunities and challenges for students and researchers who are curious about the potential of the field amid hype, considering whether their skills are aligned with what the field needs, and contemplating how collaborating with industries may impact their research. This qualitative study presents perspectives from leading quantum researchers who are educators on three critical aspects shaping QIST's development: (1) the impact of hype in the field and strategies for managing expectations, (2) approaches to creating conducive environments that attract students and established researchers from non-physics disciplines, and (3) effective models for fostering university-industry partnerships that can be valuable for students and researchers alike. These aspects, along with several interconnected challenges, were explored through in-depth interviews with quantum educators. Our findings reveal nuanced perspectives on managing the hype cycle and its risks in creating unrealistic expectations. Regarding greater interdisciplinary engagement and attracting more non-physicists to QIST, educators emphasized the need to recognize and leverage existing expertise from other fields while developing educational pathways that meet diverse student backgrounds to prepare them for the QIST workforce. On university-industry partnerships, respondents highlighted successful models, while noting persistent challenges around intellectual property, confidentiality, and differing organizational goals. These insights provide valuable guidance for educators, policymakers, and industry leaders working to build a sustainable quantum workforce while maintaining realistic expectations about the field's trajectory.

We investigate gravitational dust collapse within an effective loop quantum gravity (LQG)-inspired model exhibiting an asymmetric bounce in the marginally bound case. This work extends previous studies, which have predominantly focused on models with either symmetric bounces or asymmetric bounces restricted to homogeneous dust configurations. Our analysis emphasises the phenomenological implications of the model through a combination of analytical and numerical investigations, with particular attention to singularity resolution and the formation of trapped surfaces. As in symmetric bounce models, the central curvature singularity inside the collapsing dust cloud is resolved. However, in contrast to the symmetric case, we find that a singularity emerges in the polymerised vacuum region during the bounce phase. This singularity can be identified as a shell-crossing singularity and exhibits the expected power-law behaviour of curvature scalars. Furthermore, likewise to the symmetric bounce models, we find a critical mass threshold governing the formation of inner and outer horizons in the pre-bounce phase. No analogous critical mass restriction arises for the formation of the inner horizon in the post-bounce phase, highlighting a qualitative difference between the pre- and post-bounce dynamics.

The rapid growth of quantum information science and technology (QIST) in the 21st century has created both excitement and uncertainty about the field's trajectory. This qualitative study presents perspectives from leading quantum researchers, who are educators, on fundamental questions frequently posed by students, the public, and the media regarding QIST. Through in-depth interviews, we explored several issues related to QIST including the following key areas: the current state of quantum computing in the noisy intermediate-scale quantum (NISQ) era and timelines for fault-tolerant quantum computers, the feasibility of personal quantum computers in our pockets, and promising qubit architectures for future development. Our findings reveal diverse yet convergent perspectives on these issues. While experts agree that the current machines with physical qubits that are being built currently should be called quantum computers, most estimated that it will take a decade to build a small fault-tolerant quantum computer, and several decades to achieve scalable systems capable of running Shor's factoring algorithm with quantum advantage. Regarding carrying a quantum computer in the pocket, experts viewed quantum computers as specialized tools that will remain in central locations such as data centers and can be accessed remotely for applications for which they are particularly effective compared to classical computers. Quantum researchers suggested that multiple platforms show promise, with no clear winner emerging. These insights provide valuable guidance for educators, policymakers, and the broader community in establishing realistic expectations for developments in this exciting field. Our findings can provide valuable information for educators to clarify student doubts about these important yet confusing issues related to quantum technologies.

Doubly Special Relativity (DSR) deforms special-relativistic kinematics while preserving the relativity principle by introducing a second invariant scale, typically the Planck energy $E_{\rm Pl}$. Extending DSR-inspired modified dispersion relations (MDRs) to curved spacetimes is challenging, as ambiguous definitions of the deformation energy risk reintroducing preferred frames. We review three common extensions beyond flat spacetime: (i) MDRs in local orthonormal frames on fixed backgrounds, (ii) phase-space/Hamiltonian geometry with relative locality, and (iii) rainbow metrics. Using black-hole thermodynamics for static spherically symmetric horizons, we compare two implementations: (A) energy-independent background with local-frame MDR, and (B) energy-dependent rainbow metric. When the same prescription is used for the deformation energy scale $E_\star$, both approaches yield identical Hawking temperatures: \begin{equation} T(E_\star)=T_0\,\frac{g(E_\star/E_{\rm Pl})}{f(E_\star/E_{\rm Pl})}\,,\qquad T_0=\frac{κ_0}{2π}\,, \end{equation} where $κ_0$ is the classical surface gravity. This $g/f$ scaling holds for examples such as Amelino--Camelia-type MDRs (leading correction $\propto E p^2/E_{\rm Pl}$, giving $T(E_\star)\simeq T_0(1-\fracη{2}E_\star/E_{\rm Pl})$ for $η>0$) and the Magueijo--Smolin invariant ($f=g$, so $T(E_\star)=T_0$). Further DSR effects on evaporation (thresholds, phase space, greybody factors, composition laws) are discussed. Discrepancies in the literature arise mainly from different choices of $E_\star$ (energy at infinity vs.\ local frame). For macroscopic black holes, corrections are suppressed by $T_0/E_{\rm Pl}$ and become relevant only near the Planck regime, where full quantum gravity dominates.

We present a photon-pair source in commercially available optical fiber that produces paired photons at telecommunication and near-infrared (NIR) wavelengths. The highly nondegenerate pairs are 700 nm apart: one in the 1500 nm E- and S-band telecommunication range and the other in the 830 nm NIR range. The high non-degeneracy means the photon pairs are far-detuned from Raman noise, resulting in a high coincidence-to-accidental ratio even while operating at room temperature. The source produces two spectrally and spatially distinct phase-matched processes with low spectral cross-talk, distinct transverse spatial modes in the NIR, and a single fundamental spatial mode in the telecommunication range. The source's room-temperature operation, off-the-shelf materials, and multiplexing potential make it promising for deployment in quantum networks.

Protocols have been previously proposed to certify the presence of an entangled measurement in a fully device-independent manner. Here, I provide models for these protocols in which the claimed measurement is not entangled, and demonstrate it is always possible to displace entanglement from measurements to measured states for a general class of device-independent scenarios. This indicates that no black-box measurement scenario requires entangled measurements to replicate its behavior, which is relevant to our fundamental understanding of this phenomenon and how to witness it.

It is well known that classical and quantum theories carry distinct types of representations, each type of representation corresponding to possible values of generalized charges in the classical or quantum context. This paper demonstrates a sense in the structure of these representation theories is preserved from classical to quantum physics. To show this, I discuss distinct representation-theory preserving morphisms in the classical and quantum contexts. Specifically, I consider categories whose morphisms are Lagrangian relations in the classical context and Hilbert bimodules in the quantum context. These morphisms are significant because they give rise to induced representations of classical and quantum theories, respectively. I consider quantization and the classical limit as determining functors between these categories. I treat quantization via the strict deformation quantization of a Poisson algebra and the classical limit via the extension of a uniformly continuous bundle of C*-algebras. With these tools, I prove that the quantization and classical limit functors are "almost-inverse" to each other, thus establishing a categorical equivalence.

We present efficient quantum circuits that implement high-dimensional unitary irreducible representations (irreps) of $SU(n)$, where $n \ge 2$ is constant. For dimension $N$ and error $ε$, the number of quantum gates in our circuits is polynomial in $\log(N)$ and $\log(1/ε)$. Our construction relies on the Jordan-Schwinger representation, which allows us to realize irreps of $SU(n)$ in the Hilbert space of $n$ quantum harmonic oscillators. Together with a recent efficient quantum Hermite transform, which allows us to map the computational basis states to the eigenstates of the quantum harmonic oscillator, this allows us to implement these irreps efficiently. Our quantum circuits can be used to construct explicit Ramanujan quantum expanders, a longstanding open problem. They can also be used to fast-forward the evolution of certain quantum systems.

We classify phases of one-dimensional matrix-product states (MPS) under symmetric circuits augmented with symmetric measurements and feedforward. Building on the framework introduced in Gunn et al., Phys. Rev. B 111, 115110 (2025), we extend the analysis from abelian and class-2 nilpotent groups to all finite nilpotent groups. For any such symmetry group $G$, we construct explicit protocols composed of $G$-symmetric circuits and measurements with feedforward that transform symmetry-protected topological (SPT) states into the trivial phase and vice versa using a finite number of measurement rounds determined by the nilpotency class of $G$. Although these transformations are approximate, we prove that their success probability converges to unity in the thermodynamic limit, establishing asymptotically deterministic equivalence. Consequently, all SPT phases protected by finite nilpotent groups collapse to a single phase once symmetric measurements and feedforward are allowed. We further show that the same holds for non-normal MPS with long-range correlations, including GHZ-type states. The central technical ingredient is a hierarchical structure of irreducible representations of nilpotent groups, which enables a recursive reduction of non-abelian components to abelian ones. Our results demonstrate that symmetric measurements lead to a complete collapse of both symmetry-protected and non-normal MPS phases for all finite nilpotent symmetry groups.

The principle of relativity is extended to accommodate observers with quantum properties. This results in a new theory that introduces relative quantization rules and novel uncertainty relations, while also elucidating some interpretational problems present in the current formulation of quantum mechanics.

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 review three well known inconsistencies in the standard mathematical formulation of semiclassical gravity: the factorization problem, the information problem, and the closed universe problem. Building upon recent work, we explore how modifying the holographic dictionary may provide the necessary freedom to resolve these three problems in a unified manner while maintaining more well established aspects of the standard correspondence. Using the modified holographic dictionary as a scaffolding, we propose a program for constructing an `extended' semiclassical gravitational path integral which (i) is manifestly factorizing, (ii) computes a von Neumann entropy which satisfies the Page curve, and (iii) incorporates new operators that create closed baby universe states. Our construction may be interpreted as imposing a semiclassical version of background independence/a no global symmetry condition, defining a modified large N limit, preparing an ensemble of dual theories, or enforcing observer rules using gravitational degrees of freedom.

Recent results from the Large Hadron Collider have demonstrated quantum entanglement of top quark-antiquark pairs using the spin degree of freedom. Based on the doubly differential measurement of the spin density matrix of the top quark and antiquark performed by the CMS collaboration in the helicity and beam bases, we evaluate a set of quantum observables, including discord, steering, Bell correlation, and magic. These observables allow for a quantitative characterization of the quantum correlations present in a top quark--antiquark system, thus enabling an interpretation of collider data in terms of quantum states and their properties. Discord is observed to be greater than zero with a significance of more than 5 standard deviations ($σ$). Evidence for steering is found with a significance of more than 3$σ$. This is the first evidence for steering, and the first observation of discord in a high-energy system. No Bell correlation is observed within the currently probed phase space, in agreement with the theoretical prediction. These results experimentally corroborate the full hierarchy of quantum correlations in top quarks with discord being the most basic form of quantum correlation, followed by entanglement, steering and Bell correlation. The significance of nonzero magic, which is a complementary observable to the quantum-correlation hierarchy, is found to exceed 5$σ$ in several regions of phase space.

While nonstabilizerness (''magic'') is a key resource for universal quantum computation, its behavior in many-body quantum systems, especially near criticality, remains poorly understood. We develop a spectral transfer-matrix framework for the stabilizer Rényi entropy (SRE) in infinite matrix product states, showing that its spectrum contains universal subleading information. In particular, we identify an SRE correlation length -- distinct from the standard correlation length -- which diverges at continuous phase transitions and governs the spatial response of the SRE to local perturbations. We derive exact SRE expressions for the bond dimension $χ=2$ MPS ''skeleton'' of the cluster-Ising model, and we numerically probe its universal scaling along the $\mathbb{Z}_2$ critical lines in the phase diagram. These results demonstrate that nonstabilizerness captures signatures of criticality and local perturbations, providing a new lens on the interplay between computational resources and emergent phenomena in quantum many-body systems.

We study tensor network varieties associated with the triangular graph, with a focus on the case where one of the physical dimensions is 2. This allows us to interpret the tensors as pencils of matrices. We provide a complete characterization of these varieties in terms of the Kronecker invariants of pencils. We determine their dimension, identifying the cases for which the dimension is smaller than the expected parameter count. We provide necessary conditions for membership in these varieties, in terms of the geometry of classical determinantal varieties, coincident root loci and plane cubic curves. We address some extensions to arbitrary graphs.

The observed orbital period time derivative (or orbital "chirp") of a millisecond binary pulsar (MSP) encodes information about both the intrinsic properties of the binary system and its environment. Orbital chirp has contributions from intrinsic energy loss due to gravitational wave emission, kinematic effects due to motion in the plane of the sky, and dynamical effects due to galactic acceleration, with the latter two contributions depending on the MSP distance. We use orbital chirp data to infer distances to 21 MSPs; and for four of which we obtain smaller uncertainties than those reported in previous distance measurements. We incorporate multiple realistic galactic acceleration models to assess the sensitivity of the inferred distances to the choice of galactic gravitational potential, finding a significant dependence for four MSPs.

We revisit the localization tensor (LT) from geometric and probabilistic perspectives and construct extensions that are naturally compatible with periodic boundary conditions (PBC), without redefining the position operator. In open boundary conditions, we show that the LT can be written exactly as the covariance of a bivariate probability distribution built from density-density correlations. This leads to two conceptually distinct extensions to PBC: (i) a geometric one based on the Riemannian center (Frechet mean) on the circle, and (ii) a metric-free one based on the mutual information I, which treats the configuration space purely as a probability space. We then relate the LT to the static structure factor by identifying the diagonal part, Cpp, as a "localization function" C(p), whose small-momentum behavior determines the LT in the thermodynamic limit. This clarifies why the LT is sensitive to transitions out of the extended phase but by itself cannot distinguish Anderson-type localization from dimerization: both share the same low-momentum asymptotics. We show that the finite-momentum behavior of C(p), together with an inverse participation ratio (IPR)-based upper bound valid in localized phases, provides a sharp criterion that discriminates localization from dimerization. These results are illustrated on the Su-Schrieffer-Heeger and Aubry-Andre models, with and without interactions, and suggest that structure factor-based probes offer robust and experimentally accessible diagnostics of localized and dimerized phases under PBC.

Quantum circuits have been shown to be a fertile ground for realizing long-range entangled phases of matter. While various quantum double models with non-chiral topological order have been theoretically investigated and experimentally implemented, the realization and characterization of chiral topological phases have remained less explored. Here we show that chiral topological phases in spin systems, i.e., chiral spin liquids, can be prepared in quantum circuits using the variational quantum eigensolver (VQE) framework. On top of the VQE ground state, signatures of the chiral topological order are revealed using the recently proposed tangent space excitation ansatz for quantum circuits. We show that, both topological ground state degeneracy and the chiral edge mode can be faithfully captured by this approach. We demonstrate our approach using the Kitaev honeycomb model, finding excellent agreement of low-energy excitation spectrum on quantum circuits with exact solution in all topological sectors. Further applying this approach to a non-exactly solvable chiral spin liquid model on square lattice, the results suggest this approach works well even when the topological sectors are not exactly known.

We introduce a geometric framework for efficient few-parameter pulse optimization in multi-level quantum systems, enabling high-fidelity state transfer beyond the adiabatic limit. Our method interpolates smoothly between adiabatic and diabatic dynamics to minimize unwanted excitations and maximize desired transitions even within a multi-level structure. Crucially, for single-parameter pulse control, the optimization reduces to solving a first-order ordinary differential equation. We showcase the flexibility of our diabatic-adiabatic protocols through two examples in spin-based quantum information processing: state initialization and qubit state transfer.

Quantum states with sub-Planck features exhibit sensitivity to phase-space displacements beyond the standard quantum limit, making them useful for quantum metrology. In the context of the SU(1,1) group, sub-Planck features have been constructed through the superposition of four Perelomov coherent states on the hyperbolic plane (the SU(1,1) compass state). However, these structures differ in scale along different phase-space directions, resulting in nonuniform sensitivity enhancement. We overcome this limitation by constructing $\overline{n}$-component compass states, which are obtained by superposing $\overline{n} \geq 6$ SU(1,1) coherent states, with an even total number, evenly arranged along a circular path on the hyperbolic plane; that is, all components lie at the same distance from the origin and have equal angular spacing of $\frac{2π}{\overline{n}}$. These generalized SU(1,1) compass states generate circularly shaped sub-Planck features (isotropic sub-Planckness) and provide uniform enhancement in sensitivity to phase-space displacements. As the number of coherent states $\overline{n}$ increases, these refinements progressively improve. While verified for $\overline{n} = 16$ SU(1,1) coherent states, the results remain valid for superpositions with arbitrarily large $\overline{n}$ components.

We study quantum repeater chains in which entangled links between neighbouring nodes are created through heralded entanglement generation and adjacent links are swapped as soon as possible. Since heralded entanglement generation attempts succeed only probabilistically, some links will have to be stored in quantum memories at the nodes of the chain while waiting for adjacent links to be generated. The fidelity of these stored links decreases with time due to decoherence, and if they are stored for too long then this can lead to low end-to-end fidelity. Previous work has shown that the end-to-end fidelity can be improved by deterministically discarding links when their ages exceed some cutoff value. Such deterministic cutoff policies provide strict control of the fidelity of all links, but they come at the expense of having to track link ages. In this work, we introduce a probabilistic cutoff policy that does not require tracking link ages, at the cost of abandoning strict control of the fidelity. We benchmark this new probabilistic cutoff policy against a deterministic cutoff policy. We compare the policies in terms of the end-to-end rate and fidelity, and the secret-key rate. We find that even though the probabilistic cutoff policy keeps track of less state, it can provide secret-key rates of the same order of magnitude as the deterministic cutoff policy in chains with few nodes or high elementary link generation probabilities. Moreover, we identify a scenario in which the probabilistic cutoff policy can deliver end-to-end links that are required to have some minimum threshold fidelity at a higher rate than the deterministic cutoff policy.

Photonic quantum memristors provide a measurement-induced route to nonlinear and history-dependent quantum dynamics. Experimental demonstrations have so far focused on isolated devices or simple cascaded devices configurations. Here, we experimentally realize and characterize a network of two coupled photonic quantum memristors with crossed feedback, implemented on a silicon nitride photonic integrated circuit and fed by a room-temperature single-photon source based on a silicon-vacancy color center SiV$^-$ in a nanodiamond. Each memristor consists of an integrated Mach-Zehnder interferometer whose transfer function is adaptively updated by photon detection events on another memristor, thus generating novel non-Markovian input-output dynamics with an enhanced memristive behaviour compared to single devices. In particular, we report inter-memristor input-output hysteresis curves exhibiting larger form factors and displaying self-intersecting loops, respectively revealing marked bistability and topologically non-trivial memory dynamics. Furthermore, numerical simulations show how these features emerge from the interplay between memory depth and relative input phase, for both intra- and inter-memristor input-output relations. Our results establish coupled integrated photonic quantum memristors as scalable nonlinear building blocks and highlight their potential for implementing compact quantum neuromorphic and reservoir computing architectures.

We derive simple and unified closed-form expressions for projections with respect to fidelity (equivalently, the Bures and purified distances) onto several sets of interest. These include projections of bipartite positive semidefinite (PSD) matrices onto the set of PSD matrices with a given marginal, and projections of ensembles of PSD matrices onto the set of PSD decompositions of a given matrix, with important special cases corresponding to projections onto the set of quantum channels (via the Choi isomorphism) and onto the set of measurements. We introduce prior-channel decompositions of completely positive (CP) maps, which uniquely decompose any CP map into a prior PSD matrix and a quantum channel. This decomposition generalizes the Choi-Jamiolkowski isomorphism by establishing a bijective correspondence between arbitrary bipartite PSD matrices and channel-state pairs, and we show that it arises naturally from the fidelity projections developed here. As applications, we show that the pretty good measurement - associated with a weighted ensemble - is the fidelity projection of the ensemble onto the set of measurements, and that the Petz recovery map - associated with a reference state and forward channel - is the projection of a CP map (constructed from the channel-state pair) onto the set of reverse quantum channels, thereby recasting the well-known identification of the Petz map with quantum Bayes' rule in information-geometric terms. Our results also provide an information-geometric underpinning of the Leifer-Spekkens quantum state over time formalism [Leifer and Spekkens, Phys. Rev. A 88, 052130 (2013)].

In this work, we develop an efficient decoding method for graph codes, a class of stabilizer quantum error-correcting codes constructed from graph states. While optimal decoding is generally NP-hard, we propose a faster decoder exploiting the structural properties of the underlying graph states. Although distinct error patterns may yield the same syndrome, we demonstrate that the post-measurement state follows a well-defined structure determined by the projective syndrome measurement. Building on this idea, we introduce a hierarchical decoder in which each level can be solved in polynomial time. Additionally, this decoder achieves optimal decoding performance at the lower levels of the hierarchy. This strategy avoids the need for full maximum-likelihood decoding of graph codes. Numerical results illustrate the efficiency and effectiveness of the proposed approach.

We identify a fundamental tension between the standard formulation of CP violation and diffeomorphism invariance in general relativity. The effective Hamiltonian approach, while phenomenologically successful, relies on a preferred time foliation that is incompatible with general covariance. The CP-violating phases are scalars along the worldline of the decaying parent particle; however, the definition of masses and phases presupposes a local covariant structure, which becomes ill-defined near the origin where curvature is large and metric fluctuations become significant. We propose an information-geometric framework based on relative entropy, exploiting pure quantum states in particle and antiparticle Hilbert spaces. We show how the Sakharov conditions could be reinterpreted in terms of information-geometric quantities, although a fully rigorous phenomenological implementation remains to be developed.

We study scattering solutions $φ$ of the linear wave equation on extremal Reissner-Nordström spacetimes, satisfying the following properties: i) $φ$ attains a prescribed radiation field $ψ_{\mathcal{I}}$ through future null infinity, which decays at an inverse polynomial rate; ii) $φ$ is regular in the exterior region up to and including the future event horizon, i.e. $φ\in C^N$, where $N\gg1$ is independent of the decay rate of $ψ_{\mathcal{I}}$. We prove that such solutions exist for arbitrary $N$, and that they are not unique. The proof consists of: 1) finding an approximate solution $φ_{\mathrm{app}}$ with fast decaying error; 2) the use of backwards energy estimates in order to correct $φ_{\mathrm{app}}$ to an exact solution. Extremality is used only in the second step. The methods of the linear case described above are then used to show the same results for semilinear equations where the nonlinearity satisfies the null condition, as well as to geometries describing the hyperbolic orbit of multiple extremal Reissner-Nordström black holes.

Spurious couplings and decoherence degrade the performance of solid-state quantum processors, demanding careful design, calibration, and mitigation protocols. These strategies often rely on characterization of the idling processor, but tomographic recovery of (time-independent) Lindblad dynamics scales exponentially with qubit count. Here, we experimentally benchmark and demonstrate that randomized ("shadow") measurements accelerate Lindblad tomography on a superconducting transmon processor. We first implement extensible Lindblad tomography, which estimates Lindblad parameters using a complete tomographic dataset, and use it as a baseline to benchmark a shadow tomography approach, shadow Lindblad tomography. The shadow approach recycles randomized configurations to estimate the same Lindblad parameters using far fewer resources under physically motivated locality assumptions. We experimentally verify these assumptions in our processor by implementing the protocols on one- and three-qubit subsystems; here, shadow Lindblad tomography reproduces extensible Lindblad tomography within uncertainties while using exponentially fewer configurations. Leveraging this efficiency, we apply shadow Lindblad tomography to the full five-qubit processor and recover all single qubit dissipation and two-qubit coupling parameters in 9 hours of acquisition time compared to an estimated 58 hours for extensible Lindblad tomography. Additionally, our shadow implementation is compatible with conventional Gaussian error propagation, avoiding the use of median-of-means estimators. Together, these results demonstrate how randomized shadow tomography protocols can be practically implemented to learn quantum processor dynamics at an increasing qubit count.

Multiparameter quantum metrology plays a fundamental role in uncovering and exploiting the distinctive features of quantum systems. In this work, we propose an effective and experimentally feasible scheme to significantly enhance the simultaneous quantum estimation of the photon magnon and magnon mechanical coupling strengths in a hybrid cavity magnon mechanical platform. Our approach relies on the assistance of a coherent feedback loop combined with the injection of a coherent driving field. We show that an appropriate tuning of the system and feedback parameters leads to a substantial reduction of the estimation errors associated with both coupling strengths. To quantify the metrological performance of the proposed scheme, we employ the quantum Cramer Rao bound (QCRB) as a fundamental benchmark for multiparameter estimation. We explicitly compute and compare the QCRBs derived from the symmetric logarithmic derivative (SLD) and the right logarithmic derivative (RLD) formalisms. Our results demonstrate that the RLD based QCRB is systematically lower than the SLD based bound, indicating superior estimation precision in the considered noncommutative estimation scenario. We further analyze the performance of heterodyne detection and show that, in suitable parameter regimes, the corresponding classical estimation precision closely approaches the ultimate quantum limit predicted by our scheme. Finally, we discuss the experimental feasibility of the proposed setup within currently available cavity magnon mechanical platforms. Owing to its general character, the framework developed here can be readily extended to the high precision estimation of other physical parameters in hybrid quantum systems.

This paper presents an introduction to geometric representations of quantum states in which each distinct quantum state, pure and mixed, corresponds to a unique point in a Euclidean space. Beginning with a review of some underappreciated properties of the most commonly used geometric representation, the Bloch sphere visualization of qubit states, we show how concepts, algorithms, and spatial relations viewable on this geometric representation can be extended to representations of qudit states of any finite quantum dimension $d$ and on to the infinite-dimensional limit. A primary goal of the work is helping the reader develop a visual intuition of these spaces, which can complement the understanding of the algebraic formalism of quantum mechanics for learners, teachers, and researchers at any level. Particular emphasis is given both to understanding states in a basis-independent way and to understanding how probability amplitudes and density matrix elements used to algebraically represent states in a particular basis correspond to line segments and angles in the geometric representations. In addition to providing visualizations for such concepts as superpositions, mixtures, decoherence, and measurement, we demonstrate how the representations can be used to substitute simple geometrical calculations for more cumbersome linear algebra ones, which may be of particular use in introducing mixed states and density matrices to beginning quantum students at an early stage. The work concludes with the geometrical interpretation of some commonly used metrics such as the purity of states and their relation to real, Euclidean vectors in the infinite-dimensional limit of the space, which contains all lower-dimensional qudit spaces as subspaces.

A major challenge in open quantum dynamics is the inclusion of Matsubara-decay terms in the memory kernel, which arise from the quantum-Boltzmann delocalisation of the bath modes. This delocalisation can be quantified by the radius of gyration squared ${\mathcal R}^2(ω)$ of the imaginary-time Feynman paths of the bath modes as a function of the frequency $ω$. In a Hierarchical Equations of Motion (HEOM) calculation with a Debye--Drude spectral density, ${\mathcal R}^2(ω)$ is the only quantity that is treated approximately (assuming convergence with respect to hierarchy depth). Here, we show that the well-known Ishizaki--Tanimura correction is equivalent to separating smooth from `Brownian' contributions to ${\mathcal R}^2(ω)$, and that modifying the correction leads to a more efficient HEOM in the case of fast baths. We also develop a simple `A4' adaptation of the `AAA' (Adaptive Antoulas--Anderson) algorithm in order to fit ${\mathcal R}^2(ω)$ to a sum over poles, which results in an extremely efficient implementation of the standard HEOM method at low temperatures.

The study of spin decoherence is often performed by assuming that spin-phonon interactions lead to relaxation at high temperatures, and spin-spin dipolar interactions instead contribute to pure dephasing at low temperatures. This has resulted in the neglect of spin relaxation due to spin-spin dipolar interactions and its influence on decoherence at low temperatures. For a complete understanding of low temperature spin dynamics, it is then imperative to focus also on the latter mechanism. One such method which has shown great promise in the efficient calculation of central spin dynamics due to spin-spin dipolar interactions with a surrounding spin bath is the Cluster-Correlation Expansion (CCE). An extension of this method through the explicit inclusion of the central spin degrees of freedom, known as the generalized Cluster-Correlation Expansion (gCCE) is capable of simulating the transfer of energy from the central spin into the bath, and thus could have the potential to investigate spin relaxation in this setting. In this work, we show that gCCE, in its standard form, is insufficient for providing even a qualitatively accurate description of spin-spin relaxation. A full mathematical deconstruction of the underlying theory of gCCE clearly points to the origin of such a breakdown and provides a starting point for its potential future resolution.

We explain the achievements that were awarded 2022 Nobel Prize in Physics, as well as the preceding and the later developments. The main notions and historic cornerstones of Bell inequalities, the related researches on quantum entanglement are reviewed, and the key physical ideas are emphasized. Among the early work, C. S. Wu's contributions using polarization-entangled photons from electron-positron annihilation are introduced.

Physics-informed neural networks (PINNs) have emerged as a powerful framework for solving partial differential equations (PDEs) by embedding governing physical laws directly into the training objective. Recent advances in quantum machine learning have motivated hybrid quantum-classical extensions aimed at enhancing representational capacity while remaining compatible with near-term quantum hardware. In this work, we investigate trainable embedding strategies within quantum-assisted PINNs for solving parabolic PDEs, using one- and two-dimensional heat equations as canonical benchmarks. We introduce two quantum-assisted architectures that differ in their embedding components. In the first approach, a classical feed-forward neural network generates trainable feature maps for quantum data encoding (FNN-TE-QPINN). In the second, the embedding stage is realized entirely by a parameterized quantum circuit (QNN-TE-QPINN), yielding a fully quantum feature map. Our findings emphasize the critical role of embedding design and support hybrid quantum-classical approaches for parabolic PDE modeling in the NISQ era.

Zernike phase-contrast imaging is unique among imaging techniques in that it enables the upper limit of Fisher information allowed by quantum mechanics. Here we show that, in a departure from an ideal setting, using an incident beam of finite width, and a $π/2$ phase plate having a finite cutoff, the technique can still deliver $>95\%$ of the quantum limit. We point out that the Zernike method is, in principle, applicable to any incident beam. As an example, we sketch an approximate implementation of the method for an incident speckle beam, and show that it too can deliver $>95\%$ of the quantum limit.

In this work, we investigate the geometry of quantum logic gates within the holomorphic representation of quantum mechanics. We begin by embedding the physical qubit subspace into the space of holomorphic functions that are homogeneous of degree one in each Schwinger boson pair $(z_{a_{j}}, z_{b_{j}})$. Within this framework, we derive explicit closed-form differential operator representations for a universal set of quantum gates--including the Pauli operators, Hadamard, CNOT, CZ, and SWAP--and demonstrate that they preserve the physical subspace exactly. Restricting to unit-magnitude variables ($|z| = 1$) reveals a toroidal space $\mathbb{T}^{2N}$, on which quantum gates act as canonical transformations: Pauli operators generate Hamiltonian flows, the Hadamard gate induces a nonlinear automorphism, and entangling gates produce correlated diffeomorphisms that couple distinct toroidal factors. Beyond the torus, the full Segal--Bargmann space carries a natural Kaehler geometry that governs amplitude dynamics. Entanglement is geometrically characterized via the Segre embedding into complex projective space, while topological protection arises from the $U(1)^{N}$ fiber bundle structure associated with the Jordan--Schwinger constraint.

We introduce dissipative spectroscopy as a framework for extracting spectral information from quantum systems via controlled dissipation. By establishing a general dissipative response theory applicable to both Markovian and non-Markovian environments, we develop a protocol to access the dissipative spectrum (DS) through driven oscillation-dissipation resonance. We show that the DS can identify two-particle soft modes near quantum critical points and, on the normal-phase side, predict the emergence of macroscopic order exhibiting power-law growth following a dissipation quench. These distinctive signatures appear in quasiparticle-dominant regimes, previously considered trivial. Furthermore, we introduce extended dissipative susceptibilities that capture leading memory effects and demonstrate their utility in a dissipative fermionic model. Our results indicate that the DS is readily accessible and offers a versatile tool for probing equilibrium properties as well as predicting nonequilibrium dissipative dynamics.

Sparse identification of nonlinear dynamics (SINDy) is a data-driven framework for estimating classical nonlinear dynamical systems from time-series data. In this approach, system dynamics is represented as a linear combination of a predefined set of basis functions, and the corresponding coefficients are sparsely estimated from observed time-series data. In this study, we propose sparse identification of quantum Hamiltonian dynamics (SIQHDy), a SINDy-inspired quantum circuit learning framework for estimating quantum Hamiltonian dynamics from time-series data of quantum measurement outcomes. In SIQHDy, the unitary evolution of a quantum Hamiltonian system is expressed as a product of basis quantum circuits, and the corresponding circuit parameters are estimated through sparsity-promoting optimization. We numerically demonstrate that SIQHDy accurately reconstructs the dynamics of single-, three-, and five-spin systems, and exhibits robustness to measurement noise in the three-spin case. Furthermore, we propose an extension of SIQHDy for scenarios with limited accessible observables and evaluate its performance in identifying two-spin systems and in network-structure identification for five-spin systems.

We study relativistic twisted (orbital-angular-momentum) states of a massive charged particle propagating through an axially symmetric, longitudinally inhomogeneous solenoid field and a co-directed accelerating or decelerating electric field. Starting from the Dirac equation and using controlled spinless and paraxial approximations, we show that the transverse envelope obeys an effective nonstationary Schrödinger equation governed by a Caldirola--Kanai Hamiltonian. The longitudinal energy gain or loss encoded in $f(z)=[E_0-V(z)]^2-m^2$ generates an effective gain or damping rate $\widetildeγ(z)=\partial_z f(z)/[2f(z)]$ and a $z$-dependent oscillator frequency $\widetildeω(z)=p_0Ω(z)/\sqrt{f(z)}$. Exploiting the Ermakov mapping (unitary equivalence of Caldirola--Kanai systems), we obtain a closed-form propagated twisted wave function by transforming the stationary Landau basis. The transverse evolution is controlled by a single scaling function $b(z)$ that satisfies a generalized Ermakov--Pinney equation with coefficients determined by $E_z(z)$ and $B_z(z)$. In the limiting cases of uniform acceleration with $B_z=0$ and of solenoid focusing with negligible acceleration, our solution reduces to previously known analytic results, providing a direct bridge to established models.

Solving coupled systems of differential equations (DEs) is a central problem across scientific computing. While Physics Informed Neural Networks (PINNs) offer a promising, mesh-free approach, their standard architectures struggle with the multi-objective optimization conflicts and local optima traps inherent in coupled problems. To address the first issue, we propose a Forked PINN (FPINN) framework designed for coupled systems of DEs. FPINN employs a shared base network with independent branches, isolating gradient pathways to stabilize training. We demonstrate the effectiveness of FPINN in simulating non-Markovian open quantum dynamics governed by coupled DEs, where multi-objective conflicts and local optima traps often cause evolutionary stagnation. To overcome this second challenge, we incorporate an evolution regularization loss that guides the model away from trivial solutions and ensures physically meaningful evolution. We demonstrate the effectiveness of FPINN in simulating non-Markovian open quantum dynamics governed by coupled DEs, where multi-objective conflicts and local optima traps often cause evolutionary stagnation. For the spin-boson and XXZ models, FPINN accurately captures hallmark non-Markovian features, such as quantum coherence revival and information backflow, significantly outperforming standard PINNs. The proposed FPINN architecture offers a general and effective framework for solving coupled systems of equations, which arise across a broad spectrum from classical physics to modern artificial intelligence, including applications in multi-body rotational dynamics, multi-asset portfolio optimization, chemical reaction kinetics, and deep representation learning.

This chapter introduces and investigates some fundamental questions on the relationship between accuracy and robustness in both classical and quantum classification algorithms under noisy and adversarial conditions. We introduce and clarify various definitions of robustness and accuracy, including corrupted-instance robustness accuracy and prediction-change robustness, distinguishing them from conventional accuracy and robustness measures. Through theoretical analysis and toy models, we establish conditions under which trade-offs between accuracy and robustness accuracy arise and identify scenarios where such trade-offs can be avoided. The framework developed highlights the nuanced interplay between model bias, noise characteristics, and perturbation types, including relevant and irrelevant perturbations. We explore the implications of some of these results for incompatible noise, adversarial quantum perturbations, the no free lunch theorem, and suggest future methods to examine these problems from the lens of dynamical systems.

There has been recent progress in developing well-posed theories of relativistic viscous hydrodynamics and of gravitational effective field theories. These have in common the feature that they introduce unphysical degrees of freedom. We address the problem of how these should be treated. We propose a ''reduction of order'' approach which is applied not at the level of equations of motion but only to initial data. This specifies uniquely the data for the unphysical modes in terms of the data for the physical modes. We argue that the apparent breaking of Lorentz invariance associated with this approach is not a problem provided one restricts to Lorentz frames for which the assumptions of effective field theory are manifestly valid.

We investigate horizon-brightened acceleration radiation (HBAR) and optical signatures for a broad class of regular black holes sourced by nonlinear electrodynamics. The spacetimes considered are static, spherically symmetric, and nonsingular, and they include Bardeen-like, and Hayward-like regular black-hole limits as spacial cases. We characterize the horizon structure and thermodynamics properties, and we compute key optical observables by determining the photon-sphere location and the corresponding shadow size as seen by distant observers, including controlled perturbative limits and full numerical solutions. Using angular-size constraints for SgrA* and M87* from the Event Horizon Telescope and the GRAVITY collaboration, we perform a Markov Chain Monte Carlo analysis to infer the admissible parameter ranges of the model and to quantify degeneracies among the black-hole mass and nonlinear-electrodynimcs parameters. On the quantum side, we develop the near-horizon reduction relevant for HBAR, showing that the dominant sector governing the detector response exhibits conformal behavior and leads to a thermal excitation spectrum governed by the horizon temperature. We formulate a Lindblad master-equation description of the radiation field, identify the thermal steady state, and derive an HBAR entropy-energy relation consistent with a Clausius-type first law. Finally, we establish a Wien-type displacement law for the HBAR spectrum, expressing the peak wavelength in terms of the horizon thermodynamics, thereby providing an additional observable link between nonlinear electrodynamics, regularity, and near-horizon quantum radiation.

This paper proposes a novel quantum effect wherein particle spectra show extra broadening due to local short-time acceleration (as well as in a deSitter spacetime background). This effect provides a simple interpretation for the acceleration interpolation relation required to modify the kinematics of a test particle in the Modified Inertial interpretation of Modified Newtonian Dynamics (MOND). This effect can be regarded as a generalization of the thermal blackbody spectrum generated by the Unruh effect (which arises from long-time uniform acceleration in a flat background) to the scenario of local short-time non-uniform acceleration. This effect offers a unified framework for understanding the accelerated expansion of the universe and the anomalies in galactic rotation curves or radial acceleration. It is worth emphasizing that this modified kinematic interpretation of MOND necessitates a quantum equivalence principle as its physical foundation, that is, extending the equivalence at the level of expectation values (first-order moments) in the classical equivalence principle to the equivalence at the level of second-order moment quantum fluctuations. Therefore, the role of this effect within the frameworks of quantum reference frames and quantum gravity is also discussed to some extent.

Classical polarization optics is naturally described by a two-dimensional complex Hilbert space (Jones vectors), so the tensor-product kinematics underlying bipartite nonseparability is already available classically. For statistical (stochastic) optical fields, and under an operational stance where outcomes are not assumed pre-assigned prior to detection, suitably prepared two-beam polarization states can exhibit Bell--CHSH correlations of quantum strength. The same platform offers a tunable, low-cost testbed for stress-testing Bell/CHSH and contextuality witnesses under realistic imperfections (noise, coarse binning, selective sampling). We also outline an alternative preparation based on external conical refraction (ECR), where engineered intersecting conical-refraction rings mimic the intersecting emission cones of SPDC. We give a self-contained categorical formulation: the preparation-and-conditioning pipeline (Hadamard-like splitting, CNOT-like coupling, and routing/conditioning that removes unwanted contributions) is treated as a single morphism in an operational process theory (e.g. $\mathbf{CPM}(\mathbf{FHilb})$). From it we functorially extract an empirical model, i.e. a compatible family of context-indexed probability distributions. The Abramsky--Brandenburger sheaf criterion then applies: noncontextuality is the existence of a global section, and CHSH violation is a precise failure-to-glue. This separates kinematic nonseparability from operational contextuality and clarifies why neither, by itself, entails nonlocal causation; contextuality can arise in a classically implementable stochastic-optics regime.

Transversal Pauli Z rotations provide a natural route to fault-tolerant logical diagonal gates in quantum CSS codes, yet their capability is fundamentally constrained. In this work, we formulate the refinement problem of realizing a logical diagonal gate by a transversal implementation with a finer discrete rotation angle and show that its solvability is completely characterized by the Bockstein homomorphism in homology theory. Furthermore, we prove that the linear independence of the X-stabilizer generators together with the commutativity condition modulo a power of two ensures the existence of transversal implementations of all logical Pauli Z rotations with discrete angles in general CSS codes. Our results identify a canonical homological obstruction governing transversal implementability and provide a conceptual foundation for a formal theory of transversal structures in quantum error correction.

The establishment of extremely strong bounds on the magnitude of the electric dipole moment of the neutron, a quantity that is of great importance for determining the level of time reversal symmetry respected by the strong interactions, offers an important lesson regarding the manner in which quantum uncertainties are interpreted in the inflationary cosmological account of the generation of the primordial inhomogeneities that give rise to the universe's structure. The identification of quantum uncertainties with actual stochastic fluctuations, a standard aspect of the current physical account for the emergence of the cosmic structure, is called into question. This opens the door for novel aspects of physics that are needed in order to provide a satisfactory account that is both conceptually clear and does not conflict with the use of quantum theory in other settings.

We realize a hardware-native time--frequency Gottesman--Kitaev--Preskill (GKP) logical qubit encoded in the continuous phase space of single photons, establishing a propagating photonic implementation of bosonic grid encoding. Finite-energy grid states are generated deterministically using coherently driven entangled nonlinear biphoton sources that produce single-photon frequency-comb supermodes. An optical-frequency-comb reference anchors the time--frequency phase space and enforces commuting displacement stabilizers directly at the hardware level, continuously defining the logical subspace. Timing jitter, spectral drift, and phase noise map naturally onto Gaussian displacement errors within this lattice, yielding intrinsic correctability inside a stabilizer cell. Logical operations correspond to experimentally accessible phase and delay controls, enabling deterministic state preparation and manipulation. Building on the modal time--frequency GKP framework, we identify a concrete pathway toward active syndrome extraction and deterministic displacement recovery using ancillary grid states and interferometric time--frequency measurements. These primitives establish a hardware-compatible route for integrating the time--frequency GKP logical layer into erasure-aware and fusion-based fault-tolerant photonic architectures.

We present a systematic investigation of the radiative properties and optical appearance of rotating SV black holes surrounded by a thin accretion disks, and mainly analyze the influence of the regularization parameter $g$ on related observables. The results show that although the kinetic quantities and the location of the innermost stable circular orbit (ISCO) depend on the regularization parameter $g$, the radiative efficiency of the rotating SV black hole is the same as its Kerr counterpart. Within the Novikov-Thorne thin-disk model, the radiative flux, effective temperature, and spectral luminosity are studied, and by adopting observational parameters relevant to SgrA* and M87*, concrete examples of the rotating SV black holes are calculated and compared with that of the Kerr black holes. The results show that the parameter $g$ suppresses the maximum values of these quantities. In addition, using a backward ray-tracing technique, we numerically simulate the optical appearance of rotating SV black holes and analyze the corresponding intensity images, redshift and observed flux distributions. Our results show that these quantities are affected by $g$. In particular, as $g$ increases, the observed intensity is significantly suppressed and the photon ring region has remarkable increase in its width. Our findings suggest that accretion-disk-related observables may provide important avenues to distinguish rotating SV black holes and Kerr black holes, and offer theoretical guidance for future high-resolution observations.

A universal energy eigenvalue equation is proposed in this paper. It is proven that the unique set of eigenfunctions or preferred basis exists for any non-isolated sub-system. Applying the new eigenvalue equation to the relative motion of a hydrogen atom together with the derived relativistic Hamiltonian to quantify the impact of finite proton mass to the fine structure, correction to the fine structure is obtained as a result of the entanglement of the relative motion and the center-of-mass motion, which can be used to verify the correctness of the proposed eigenvalue equation. Applying the equation to the measurement of electron double-slit interference, it is analyzed that the photon packets with Lorentzian spectral lineshape, the domain and state density of the sub-system, and the energy of the incident electron together determines the spontaneous emission rate of the incident electron. Photons generated in this process excite electrons from the valence band to the conduction band of the detector. Corresponding to any emitted or absorbed photon, the sub-system is found to be uniquely determined by maximizing the transition rate. This new principle is valid for atoms too. Closed-form expressions are obtained for the transition rates and example numerical results show good correlation between calculation and the position measurement experiment. The discovered common mechanisms that determine the sub-systems, the preferred bases, and transition rates form the foundation of a new, general, and consistent theory of quantum measurement.

We analyze the behavior of holographic subregion complexity (HSC) and holographic fidelity susceptibility (HFS) in noncommutative Yang--Mills theory. The emergence of a minimum length scale, dictated by the degree of noncommutativity, induces a behavioral transition in the HSC and establishes a lower bound. In the large noncommutativity regime, the qualitative features of the complexity deviate significantly from the commutative case. The HFS is shown to provide an effective measure of the degree of noncommutativity. Although the HSC generally satisfies strong subadditivity, this property fails abruptly when the subregion size approaches the minimum length scale. At finite temperature, the long-range behavior of the HSC is modified, and its lower bound scales positively with temperature. Furthermore, temperature enhances the sensitivity of the fidelity susceptibility to the degree of noncommutativity. Within the AdS soliton background, a competition between connected and disconnected configurations arises in the HSC, signaling a phase-transition-like behavior. Finally, the compactification scale is found to diminish the sensitivity of the HFS to the degree of noncommutativity.

We present a minimal tunable many-body system of fermions coupled to quantum magnetic flux, which is analytically diagonalizable and exhibits a variety of many-body phenomena such as Stoner orbital instability and Dicke-like quantum phase transition. In contrast to standard cavity quantum electrodynamics with its electric-dipole coupling of the electric field operators with matter, here it is the quantized magnetic field of an LC-resonator which is coupled to the angular momentum of particles. Adding the Josephson junction (JJ) to the linear LC circuit allows us to explore nonlinear flux-matter phases and sector-selective photon dressing in regimes relevant to circuit QED and mesoscopic rings. Furthermore, we consider the tight-binding systems that exhibit a tunable nonlinearity representing artificial JJ, but without actual JJs included in the circuit.

We present a dispersive quantum thermometry protocol for simultaneous estimation of inverse temperature $β$ and interaction strength $x$ using a nonlinear Mach-Zehnder interferometer coupled to a thermal ancilla. We derive closed-form expressions for the quantum Fisher information matrix, establishing that metrological performance depends solely on the thermal visibility $\mathcal{V}(β)$ and its derivative. The output state remains diagonal in photon-number basis, making photon counting globally optimal and saturating the multiparameter quantum Cramér-Rao bound without adaptive feedback. Moving beyond ideal unitary evolution, we analyze protocol robustness under concurrent amplitude and phase damping. Using Fisher Information Susceptibility, we establish a clear hierarchy: NOON states offer maximal theoretical sensitivity but exhibit exponential fragility to loss, rendering them impractical. Squeezed vacuum states emerge as robust candidates for steady-state sensing, while cat states prove compelling for transient thermometry by retaining significant coherence after photon loss. We validate these predictions through digital quantum circuit implementation on IBM's \texttt{ibm_torino} processor. Experimental results confirm the predicted Fisher information landscape while revealing systematic noise-induced biases, demonstrating that current NISQ hardware can effectively benchmark fundamental trade-offs in multiparameter quantum sensing.

Networked quantum sensors have several applications such as the mapping of magnetic fields. When the magnetic fields are biomagnetic ones, i.e., they contain some private information, the information of from who non-zero magnetic fields occur has to be protected from eavesdroppers. Anonymous quantum sensing keeps it secret by estimating amplitudes of the magnetic fields without disclosing the positions of non-zero magnetic fields. In this paper, we propose an anonymous quantum sensing protocol that is robust against any independent noise in state preparations. To this end, we devise a quantum state verification protocol for a superposition of Greenberger-Horne-Zeilinger and Dicke states and combine it with the original protocol of anonymous quantum sensing. Our verification protocol can decide whether the fidelity between the ideal and actual states is high or low more efficiently than the direct fidelity estimation. Since the original protocol of anonymous quantum sensing cannot correctly estimate the amplitudes of the magnetic fields under state preparation errors, our results would improve the performance of anonymous quantum sensing in realistic situations.

Combinations of recent cosmological observations, including Dark Energy Spectroscopic Instrument (DESI), show hints of a dynamical nature for dark energy. While the data suggest the possibility of the phantom crossing, it is worth thoroughly exploring quintessence models. Given that phenomenological parametrisations of the equation-of-state parameter $w(a)$ with a sharp transitional feature fit the data well, we study the realisation of such models in quintessence. In the late Universe, the quintessence field begins to oscillate abruptly, changing the behaviour of $w$. Naturally, such a model entails tachyonic instability, and particle production modifies $w$. We perform numerical lattice simulations to study the time dependence of $w$. In addition, the violent particle production produces significant density perturbations and the stochastic gravitational-wave background, whose characteristic scale depends on the mass scale of the quintessence around the minimum of the potential. We discuss the observability of these late-time cosmological signals through cosmic microwave background, quasar astrometry, pulsar timing arrays, and other observational probes.

The local Hamiltonian (LH) problem is the canonical $\mathsf{QMA}$-complete problem introduced by Kitaev. In this paper, we show its hardness in a very strong sense: we show that the 3-local Hamiltonian problem on $n$ qubits cannot be solved classically in time $O(2^{(1-\varepsilon)n})$ for any $\varepsilon>0$ under the Strong Exponential-Time Hypothesis (SETH), and cannot be solved quantumly in time $O(2^{(1-\varepsilon)n/2})$ for any $\varepsilon>0$ under the Quantum Strong Exponential-Time Hypothesis (QSETH). These lower bounds give evidence that the currently known classical and quantum algorithms for LH cannot be significantly improved. Furthermore, we are able to demonstrate fine-grained complexity lower bounds for approximating the quantum partition function (QPF) with an arbitrary constant relative error. Approximating QPF with relative error is known to be equivalent to approximately counting the dimension of the solution subspace of $\mathsf{QMA}$ problems. We show the SETH and QSETH hardness to estimate QPF with constant relative error. We then provide a quantum algorithm that runs in $O(\sqrt{2^n})$ time for an arbitrary $1/\mathrm{poly}(n)$ relative error, matching our lower bounds and improving the state-of-the-art algorithm by Bravyi, Chowdhury, Gosset, and Wocjan (Nature Physics 2022) in the low-temperature regime. To prove our fine-grained lower bounds, we introduce the first size-preserving circuit-to-Hamiltonian construction that encodes the computation of a $T$-time quantum circuit acting on $N$ qubits into a $(d+1)$-local Hamiltonian acting on $N+O(T^{1/d})$ qubits. This improves the standard construction based on the unary clock, which uses $N+O(T)$ qubits.

In this work, we study the effect of the symmetry slope on the observables of weakly and strongly magnetized neutron stars within the chaotic magnetic field approximation. We investigate the impact of the symmetry energy slope in the equation of state, as well as on the observables of neutron stars, by calculating their masses, radii, redshifts, tidal deformabilities, and fundamental-mode gravitational-wave frequencies.

We propose a novel method to measure the interaction between an ultracold gas of neutral atoms and a surface. This solution combines an optical dipole trap reflected by the surface, a magnetic trap formed by current carrying wires embedded below the surface, and a rotation of the surface itself. It allows to adiabatically transport a $^{87}$Rb BEC from few $μ$m to few hundred nm of the surface. At such distances, atom-surface interaction strongly affects the trapping potential, causing an increase of the tunneling rate towards the surface. In this paper, we show that the measurement of the lifetime of the cloud and its comparison to a tunneling model will allow to extract the Casimir-Polder (CP) force coefficient in the retarded regime ($c_4$). Our model includes noise-induced heating, calibration biases of experimentally controlled parameters and accuracy of the atom lifetime measurement. Using typical trapping parameters and experimental uncertainties, we numerically estimate the relative uncertainty of $c_4$ to be 10%. This method can be implemented with any atomic species that can be magnetically and optically trapped.

Scalable, high-fidelity quantum-state readout remains a central challenge in the development of large-scale superconducting quantum processors. Conventional dispersive readout architectures depend on bulky isolators and external amplifiers, introducing significant hardware overhead and limiting opportunities for on-chip information processing. In this work, we propose a novel approach that embeds a nonlinear Superconducting Nonlinear Asymmetric Inductive eLement (SNAIL) into the readout chain, enabling coherent and directional processing of readout signals directly on-chip. This embedded SNAIL platform allows frequency-multiplexed resonators to interact through engineered couplings, forming a tunable readout-amplifier-output architecture that can manipulate quantum readout data \textit{in situ}. Through theoretical modeling and numerical optimization, we show that this platform enhances fidelity, suppresses measurement-induced decoherence, and simplifies hardware complexity. These results establish the hybridized SNAIL as a promising building block for scalable and coherent quantum-state readout in next-generation processors.

Next-generation GW observatories such as the ET and CE will detect BNS and NSBH inspirals with high SNRs and long in-band durations, making systematic early-warning alerts both feasible and scientifically valuable. Such triggers are essential for coordinating rapid electromagnetic follow-up. In this work, we introduce GW-FALCON, a novel feature-driven DL framework for early-time detection between GW signal+noise and noise-only data in next-generation detectors. Instead of feeding raw time series to CNN or more complex neural network architectures, we first extract a large set of statistical, temporal, and spectral quantities from short observational time windows using the TSFEL library. The resulting fixed-length feature vectors are then used as input to feed-forward ANNs suitable for low-latency operation. We demonstrate the method using simulated BNS and NSBH inspiral waveforms injected into colored Gaussian noise generated from the ET and CE design PSDs. We train separate ANNs on feature sets extracted from partial-inspiral windows characterized by different maximum instantaneous frequencies, enabling early-warning triggers from tens to hundreds of seconds before merger. Across all detector configurations and datasets, the resulting classifiers achieve high accuracy and detection efficiency, with ET-like networks typically reaching test accuracies of order 90% and CE-like ones exceeding 97% at low false-alarm probability. To the best of our knowledge, this work presents the first comprehensive feature-based DL detection framework for Next-generation GW observatories, connecting feature extraction from strain time series data to robust signal-noise classification within a setup that can be extended to real data and to more advanced neural network architectures.

Variational Quantum Algorithms (VQAs), such as the Quantum Approximate Optimization Algorithm (QAOA), offer a promising route to tackling combinatorial optimization problems on near and intermediate-term quantum devices. However, their performance critically depends on the choice of initial parameters, and the limited expressiveness of the QAOA ansatz makes identifying effective initializations both difficult and unscalable. To address this, we propose a framework, Scalable Parameter Initialization for QAOA (SPIQ), that employs a relaxed QAOA ansatz to enable classical search over a set of Clifford-preparable quantum states that yield high-quality solutions. These states serve as superior QAOA initializations, driving rapid convergence while significantly reducing the quantum circuit evaluations needed to reach high-quality solutions and consequently lowering quantum-device cost. We present a scalable, application-agnostic initialization framework that achieves an absolute accuracy improvement of up to 80% over state-of-the-art initialization and reduces initial-state diversity by up to 10,000x across QUBO, PUBO, and PCBO problems spanning tens to hundreds of qubits. We further benchmark its performance on a wide range of problem formulations and instances derived from real-world datasets, demonstrating consistent and scalable improvements. Furthermore, we introduce two complementary strategies for selecting high-quality Clifford points identified by our search procedure and using them to seed multi-start optimization, thereby enhancing exploration and improving solution quality.

We experimentally investigate a frustrated spin-exchange antiferromagnet in a quantum simulator, composed of N = 114 dipolar Rydberg atoms arranged into a kagome array. Motivated by a recent theoretical proposal of a gapless U(1) Dirac spin liquid ground state, we use local addressing to adiabatically prepare low-energy states. We measure the local polarization and spin-spin correlations over this adiabatic protocol, and observe our system move from a staggered product state, through an intermediate magnetic crystal, and finally into a disordered, correlated liquid. We estimate the entropy density of this atomic liquid to be similar to that of frustrated magnetic insulators at liquid nitrogen temperatures. We compare the correlations in our liquid to those of a simple, parameter-free ansatz for the Dirac spin liquid, and find good agreement in the sign structure and spatial decay. Finally, we probe the static susceptibility of our system to a local field perturbation and to a geometrical distortion. Our results establish Rydberg atom arrays as a promising platform for the preparation and microscopic characterization of quantum spin liquid candidates.

Quantum entanglement is an interesting resource for modern quantum technologies, where generating multiple quantum entanglement is highly required. However, entanglement engineering between multiple modes is strongly suppressed by dark mode effect. Here, we proposed a scheme based on molecular cavity optomechanical structure that enhances quantum bipartite and tripartite entanglement via dark mode breaking. Our proposal consists of an optical cavity that hosts two molecular ensembles which are coupled through an intermolecular coupling. A vibrational hopping rate $J_m$ captures the intermolecular coupling that is phase modulated via the synthetic gauge field method. The breaking of the dark mode is controlled by tuning both the intermolecular coupling and its modulation phase. By adjusting these parameters in our proposal, we can flexibly switch between the Dark Mode Unbroken (DMU) and the Dark Mode Broken (DMB) regimes. We find that in the dark-mode-unbroken regime, the amount of the generated bipartite and tripartite entanglement is significantly low or is suppressed. In contrast, in the dark-mode-broken regime, the entanglement is greatly enhanced,i.e., up to twofold enhancement. Moreover, the generated entanglement is more resilient against thermal noise in the dark-mode-broken regime compared to the thermal robustness in the unbroken regime. Therefore, our proposed scheme serves as a benckmark system to improve quantum correlations engineering, and to generate noise-tolerant quantum resources for applications in numerous modern quantum technologies.

In this paper, we address a foundational challenge in quantum field theory on curved spacetime by developing a consistent framework within loop quantum gravity. We introduce a methodology for defining meaningful superpositions of quantum geometry and matter states. This is achieved by identifying a restricted subspace of the gravitational phase space, which ensures unitary equivalence among Fock representations of a scalar field across different quantum geometries. Within the resulting well-defined state space, we derive weak solutions to the quantum Hamiltonian constraint of general relativity. Furthermore, we generalize the Hartle-Hawking vacuum state to this quantum geometric framework. The resulting state exhibits the inherent entanglement between geometry and matter, which arises from the quantum Hamiltonian constraint of general relativity. This work establishes a principled framework for studying geometry-matter entanglement and offers new insights into the quantum foundations of the black hole information paradox.

Quantum low-density parity-check (qLDPC) codes offer a promising route to scalable fault-tolerant quantum computing (FTQC) due to their substantially reduced footprint, but these gains can be diluted at utility scale if we cannot also realize a space-time-efficient instruction-set architecture (ISA) for relevant quantum applications. We present RASCqL, a Reaction-time-limited Architecture for Space-time-efficient Complex qLDPC Logic, introducing a complex-instruction-set quantum computer (CISQ) that supports key algorithmic subroutines such as quantum arithmetic, table lookups, and magic-state distillation directly in co-designed qLDPC codes. Unlike prior constructions for qLDPC logic that aim at versatile ISAs amenable to diverse circuits, RASCqL adopts an application-tailored code-modification scheme that embeds specific complex Clifford instructions useful for functional subroutines as virtually implementable matrix automorphisms. RASCqL further leverages parallel physical operations in reconfigurable neutral-atom array platforms to achieve fast QEC cycles and high-fidelity transversal operations. At the cost of increased design complexity, RASCqL implements key algorithmic subroutines at space-time costs comparable to state-of-the-art transversal surface-code architectures while achieving up to $2\times$ to $7\times$ footprint reduction under realistic physical error rates of $2 \times 10^{-3}$ to $5 \times 10^{-4}$, without additional hardware complexity. RASCqL thus demonstrates a concrete path forward for qLDPC codes as CISQ compute modules, extending their practical utility in fault-tolerant quantum computing architectures.

We investigated a time-varying cosmological constant model using recent BAO measurements from DESI DR2, combined with Type Ia supernova samples (Pantheon$^{+}$, DES-Dovekie, and Union3) and CMB shift parameters, to constrain the $Λ(t)$CDM model parameters via Markov Chain Monte Carlo analysis. We find that the interaction term $Q(z)$ shows a sign change for all dataset combinations by crossing $Q(z)=0$, depending on the choice of the dataset: at low redshift $Q(z)<0$, indicating vacuum energy decaying into dark matter, while at high redshift $Q(z)>0$, corresponding to dark matter decaying into vacuum energy. The dynamical system analysis found three critical points, namely $P_1,P_2$, and $P_3$ respectively. The resulting critical points, determined by the underlying cosmological parameters, correspond to distinct epochs in cosmic evolution. Depending on the parameter combinations, these points characterize various cosmological phases, ranging from an accelerated stiff matter-dominated era to late-time accelerated expansion. The stability of each critical point is analyzed using linear stability theory, with the relevant physical constraints on the cosmological parameters duly incorporated throughout the analysis. For each dataset combinations, the $Λ(t)$CDM model predicts that $ω_0 > -1$, showing a preference for dynamical dark energy over the cosmological constant scenario with $ω_0 = -1$. Consequently, the model exhibits a transition phase in the range $N \equiv \log a(t) \approx -0.51$ to $-0.48$ and predicts $q_0$ in the range $-0.54$ to $-0.52$, with the precise transition point depending on the choice of dataset. Finally, the Bayesian evidence shows strong support for the $Λ(t)$CDM model over $Λ$CDM

Efficient transport of cold atoms is essential for continuous operation, enabling applications ranging from atomic lasers to continuously operated qubits. However, deep potentials required to overcome vibrations, axial trap nonuniformity and insufficient cooling have limited transport of cold atoms near quantum degeneracy. Here we demonstrate rapid optical transport of cold atoms to Bose-Einstein condensation using a moving optical lattice formed by two Bessel beams. A gas of $3 \times 10^5$ ytterbium atoms at a temperature of $340\,$nK is transported over $34\,$cm in $350\,$ms with efficiency over $60\%$. Furthermore, a degenerate gas of $1 \times 10^5$ atoms with a $40\%$ condensate fraction emerges from the phase synchronization process driven by atomic interactions. This demonstration enables the fast preparation of ultracold atomic beams and large-scale atom arrays for quantum sensing, simulation and computing.

We prove the Gannon-Lee incompleteness theorem for globally hyperbolic spacetimes. We assume the synthetic null energy condition of Ketterer and a trappedness condition we call "synthetically asymptotically regular". Our result generalizes this classical result to the weighted case. It also motivates and indicates extensions to low regularity, which are deferred to future work.

We show that, for any physically realizable Mueller transformation -- including arbitrarily depolarizing maps -- the interferometric (Pancharatnam) geometric phase is fixed uniquely by the SO(3) rotation associated with the unitary (retarding) part of the pure characteristic component selected by the characteristic decomposition. All remaining characteristic contributions (discriminant and maximally mixed) can only reduce fringe visibility; they never generate geometric holonomy, even if their pure constituents involve rotations. We further establish the quantum analogue for open two-level dynamics: in the Choi representation of qubit channels, the geometric phase is fixed by the coherent unitary part of the dominant rank-one characteristic component, while the remaining dissipative structure affects visibility only.

Frequency-bin encoding has recently emerged as a powerful approach for photonic quantum information processing, offering high dimensionality, gate-parallelization, and compatibility with existing telecommunication infrastructure. However, its scalable deployment has so far been hindered by the lack of an integrated platform capable of unifying quantum state generation, coherent frequency mixing, and programmable spectral control.\\ Here, we report the first fully integrated quantum frequency processor, monolithically integrating on the same silicon photonic chip a microresonator-based biphoton quantum frequency comb source, a pump-rejection filter, high-speed phase modulators, and a four-channel, line-by-line pulse shaper. We demonstrate key functionalities, such as tunable frequency beamsplitters with success probabilities exceeding $94\%$ and fidelities above $99.9\%$, as well as the ability to synthesize more general single-qubit gates. Finally, we generate and coherently manipulate high-dimensional frequency-bin entangled states entirely on chip, showcasing control over two-photon quantum walks and performing the first on-chip frequency-bin quantum state tomography of a Bell-state with a fidelity of $95.7(3)\%$. By integrating all key functional elements on the same $4\times7\,\textrm{mm}^2$ chip, with the possibility of scaling to a larger number of modes, our work marks an important step toward large-scale frequency-domain photonic processors for both classical and quantum applications.

After distinguishing the role of classical and quantum inhomogeneities in cosmological backgrounds, we constrain the initial states of the relic gravitons as soon as the different wavelengths of the spectrum cross the comoving Hubble radius, without any reference to earlier timescales. According to this pragmatic perspective the quantum states with finite energy density at the crossing time consistently affect the two-point functions and the related power spectra. An initial state different from the vacuum turns out to be marginally permitted in the low frequency range (associated with the largest observable wavelengths that crossed the comoving Hubble radius) while the intermediate and high frequency domains of the spectrum are populated by the gravitons produced from the vacuum. The non classical correlations are expected to dominate between the kHz and the THz since in this region gravitons are produced quantum mechanically with a negligible contribution from the initial state.

The Hamiltonian of mean force (HMF) provides the standard starting point for strong-coupling thermodynamics, yet explicit operator forms are known only in restricted settings. We present a quenched density framework that uses the Hubbard-Stratonovich transformation to rewrite the reduced equilibrium state as an average over local propagators in imaginary time. This approach rigorously separates the statistical definition of the environment from the algebraic structure of the system response. We apply this framework to the minimal case of a harmonic environment with a coupling commuting with the system Hamiltonian. In this scenario the correction to the HMF has an exact, closed-form expression. We validate this result against finite-bath trace-out calculations and stochastic imaginary-time sampling in a five-level projector-coupled model.

We conjecture the possibility of negative mass objects (NMOs) existing in the sky. It is shown that they may not be so exotic as usually expected. We show that NMOs appear as solutions of standard gravitational equations if we consider the system of a compact positive mass object, cosmological fluid and negative cosmological constant. We also construct models which generate such NMOs as solutions within the two-scalar theory and scalar-Einstein-Gauss-Bonnet gravity inspired by string theory. The orbits of the photon and massive particles are investigated in the background, where there is a negative mass object which realises a kind of effective anti-gravity. It is explicitly found that the bound system consisting of a positive mass object and a negative mass object can be formed in spite that a positive mass object suffers the repulsive force from the NMO. The possibility that such exotic objects might be observed is discussed. A simple conjecture about their possible masses is made, too. As an even more exotic object, we consider a non-trivial object with vanishing mass and investigate its properties.

We explore combinations of particle and anti-particle yields which can be used to test thermal models in a parameter free way. We also explore combinations which can be used to extract $μ_B/T$, $μ_S/T$ and $μ_Q/T$. We use experimentally measured particle-antiparticle specific ratios for proton $p$, $Λ$, and cascade $Ξ$, for $\sqrt{s_{NN}} = $ 7.7-39 GeV from RHIC BES phase-1 to extract the $μ_{B,S,Q}/T$. These compared well with published STAR freeze-out parameters. These combinations are verified to predict a similar combination of $Ω$ and $\overlineΩ$ yields. We also extend this idea to predict (anti-)nuclei yields at energies where they are not measured. We also update parametrizations for the $\sqrt{s_{NN}}$ dependence of freeze-out parameters $T$ and $μ_B$, and present for the first time a similar parametrization of $μ_S$.

We investigate the thermodynamic phase transitions of a four-dimensional charged anti-de Sitter black hole endowed with a non-minimal coupling of the form $F^{αβ}F^{γλ}R_{αγβλ}$. Using perturbative methods, we derive a consistent black hole solution and analyze its thermodynamics through both conventional equilibrium techniques and a topological defect classification approach. The system displays van der Waals-like critical behavior, with a swallow-tail structure in the free energy and distinct phase branches. The topological analysis independently confirms the existence of critical points and classifies the system within the universal topological scheme for black hole thermodynamics.

Quark fragmentation functions describe the hadronization process of a quark where any of the final-state hadrons carries a fraction of its initial momentum. We compute these fragmentation functions for a cascade that includes pions, kaons, and the charmed $D$ and $D_s$ mesons, starting from the elementary quark-to-meson fragmentation process. The latter is obtained from the relevant cut diagram, and employs Poincaré covariant Bethe-Salpeter wave functions and quark propagators. We derive a set of twenty-five coupled jet equations that describe the cascade of emitted mesons in the fragmentation process. Their solutions yield full fragmentation functions that offer a consistent picture of the quark fragmentations across the light and heavy sectors.

The required absence of global symmetries in quantum gravity has been used to imply that all non-compact gauge theories are in the swampland. This argument stems from the idea that non-compact gauge symmetries always seem to be accompanied by global symmetries that cannot be broken with a finite number of fields. In this work, we investigate whether these symmetries can be broken by an uncountable infinity of fields. We find that the symmetries can be broken, but as soon as we add these fields the EFT breaks down and in some cases decompactifies to a higher-dimensional theory without the non-compact gauge symmetry, akin to undoing a Kaluza-Klein reduction on a non-compact space. We make various comments on the species scale, free parameters, and the Weak Gravity Conjecture along the way.

The COHERENT experiment measures neutrino-induced recoils from coherent elastic neutrino-nucleus scattering (CEvNS) with multiple nuclear targets at the Spallation Neutron Source (SNS) at the Oak Ridge National Laboratory (ORNL), USA. Several successful CEvNS measurements have been achieved in recent years with tens-of-kg detector masses, with a CsI scintillating crystal, a liquid argon single-phase detector, and high-purity germanium spectrometers. For the next phase, COHERENT aims at high-statistics detection of CEvNS events for precision tests of the standard model of particle physics, and to probe new physics beyond-the-standard model. Percent-level precision can be achieved by lowering thresholds, reducing backgrounds, and by scaling up the detector masses. It goes hand in hand with benchmarking the neutrino flux from the SNS. Further detectors will measure CEvNS in additional nuclei, including lighter target nuclei such as sodium and neon, to continue to test the expected neutron-number-squared dependence of the cross section. COHERENT can furthermore study charged-current and neutral-current inelastic neutrino-nucleus cross sections on various nuclei at neutrino energies below $\sim$50 MeV. Many of these cross sections have never been measured before, but are critical input for the interpretation of core-collapse supernova detection in large-scale neutrino experiments such as DUNE, Super-K, Hyper-K, and HALO.

A new algorithm based on a deep neural network, DeepSets, for tagging the production flavour of neutral $B^0$ and $B^0_s$ mesons in proton-proton collisions is presented. Exploiting a comprehensive set of tracks associated with the hadronization process, the algorithm is calibrated on data collected by the LHCb experiment at a centre-of-mass energy of $13$ TeV. This inclusive approach enhances the flavour tagging performance beyond the established same-side and opposite-side tagging methods. The observed gains in tagging power of $35\%$ for $B^0$ mesons and $20\%$ for $B_s^0$ mesons relative to the combined performance of the existing LHCb flavour-tagging algorithms offer significant benefits for precision measurements of $C\!P$ violation and mixing in the neutral $B$ meson systems.

Axion-like particles (ALPs) can induce a birefringence effect that rotates the polarization angle of light, offering a probe of ultralight dark matter. We analyze archival near-infrared polarimetric data of the protoplanetary disk (PPD) around HD 163296. Whereas previous studies considered only single-epoch snapshots, we perform a consistent multi-epoch time-series analysis, extracting the polarization angle and its uncertainty from the polarized images. The resulting six-epoch time series is consistent with a constant polarization angle within the measurement uncertainties, while being sensitive to timescales of $\sim 170-400$ days. The typical polarization angle uncertainties are $1.6$--$6.4$ degrees, partly driven by multiple scattering in the optically thick disk, which broadens the intrinsic polarization angle distribution and introduces additional dispersion in the representative polarization angle. Based on these data, we derive the first upper limits on the ALP-photon coupling from PPD polarization variability, $g_{aγ} \lesssim 7.5 \times 10^{-12} (m_a / 10^{-22}\,{\rm eV})\,{\rm GeV}^{-1}$. Furthermore, we forecast that achieving a polarization angle uncertainty of $σ\sim 0.1$ degrees would enable world-leading sensitivity to ALP-induced birefringence.

We consider the radiation of two distinguishable soft quark-antiquark pairs ($\mathrm{q}\bar{\mathrm{q}}\mathrm{Q}\bar{\mathrm{Q}}$) in a generic process for multiparton hard scattering in QCD. We evaluate the corresponding soft current at tree level in terms of an independent-emission contribution and an irreducible correlation component, which includes strictly non-abelian terms and also terms with an abelian character. The squared current for soft $\mathrm{q}\bar{\mathrm{q}}\mathrm{Q}\bar{\mathrm{Q}}$ emission produces colour dipole and colour tripole interactions between the hard-scattering partons, with structures similar to soft gluon-quark-antiquark emission. The colour tripole interactions are odd under charge conjugation and lead to charge asymmetry effects. We extend our analysis by including QED interactions, the emission of two distinguishable soft lepton-antilepton pairs, and mixed quark-antiquark-lepton-antilepton soft emission.

We identify the dynamical origin of the non-monotonic magnetic field dependence of the charged pion mass observed in lattice QCD. Using a near-pole effective action derived from the SU(2) Nambu--Jona-Lasinio model, we show that the lowest Landau level charged pion mixes with the longitudinally polarized charged rho meson, which shares the same quantum numbers in a magnetic background. This mixing, generated by quark-loop polarization and a gauge-invariant tree-level operator matched to the vacuum decay $ρ^\pm\rightarrowπ^\pmγ$, induces strong level repulsion. Crucially, this effect is dynamically amplified by a rapid suppression of the rho-meson wave function renormalization near the pole. As a result, the lower eigenmode exhibits a turnover as the magnetic field increases. The mechanism is analogous to singlet-triplet mixing in positronium and provides a natural explanation for the lattice results. Such effects are expected to be generic for charged mesons in magnetic fields when symmetry allowed mixing and near-pole residue suppression are present.

High flux of geoantineutrinos $^{40}$K and geoneutrinos $^{40}$K ($^{40}$K-geo-($\barν + ν$)) can be obtained from reanalysis of the Borexino Phase III data. Large amounts of $^{40}$K should produce a significant heat flow that should affect Earth's internal processes. We present the results of the modeling of mantle convection taking into account the excess heat from $^{40}$K.

We establish a comprehensive theoretical framework for the electron-positron single-inclusive annihilation (SIA) process that incorporates higher-twist contributions up to twist-4 and demonstrate that these effects should be considered in low-energy reactions. Through a systematic collinear expansion of the hadronic tensor, we derive the complete set of structure functions in terms of collinear fragmentation functions (FFs), explicitly including multi-parton correlators up to the four-parton level. To evaluate the impact of these power corrections, we estimate the twist-4 unpolarized FF using a spectator model and compute the normalized differential cross section for $π^0$ production. Our numerical analysis reveals that the interplay between kinematic hadron mass corrections and dynamical twist-4 effects improves the theoretical description of recent BESIII data at relatively low-$z$ region. Furthermore, the $Q$-dependence confirms that higher-twist corrections are dominant at intermediate energy scales. These findings indicate that standard leading-twist global analyses must be extended to include these power corrections for precise studies of hadronization dynamics.

Few baryon resonances have generated as much discussion, even controversy, as the first positive parity excited state with nucleon quantum numbers. We re-examine the issue using insight gained from lattice QCD, complemented by Hamiltonian effective field theory. In doing so, we also examine the distinction between a state that can be naturally described as a quark model state and one that is dynamically generated.

In this document we update the status of U.S. community inputs for the European Strategy for Particle Physics Update (ESPPU) since April 1, 2025, and offer responses to the revised questions. Major new inputs include a long-term strategy report from the National Academies of Sciences, Engineering, and Medicine and the formal formation of a U.S. Muon Collider Collaboration.

While $CP^N$ models with analytic potentials are known to support finite-energy compact Q-ball and Q-shell solutions, their behavior in more complex Lagrangian frameworks remains a subject of active research. This work explores these non-topological structures within an extended Skyrme-Faddeev-type model that incorporates quartic derivative terms. In this context, harmonic time dependence and the presence of quartic terms constitute two independent stabilization mechanisms that allow the configurations to circumvent Derrick's scaling argument. We investigate the necessary conditions for the existence of these solutions and analyze the influence of quartic terms on the properties of the resulting compactons, specifically examining the $E(Q)$ relationship between energy and Noether charge. Our findings provide valuable insights into the stability and characteristics of compact boson stars within $CP^N$ models featuring higher-order derivative terms.

Non-topological solitons, such as Q-balls, may contribute to the cosmological dark matter. The formation and evolution of Q-balls in the early universe requires an understanding of solitons with nonzero angular momentum. We derive (rather than assume) the schematic form of the scalar field configurations that produce rotating Q-balls, which produce their well known quantized angular momentum. This analysis leads to additional insight into the properties of these rotating solitons, including a method for computing their characteristic angular velocity. By considering rotating solitons in two spatial dimensions, we investigate these attributes concretely. We develop analytical approximations for the solitons and their defining quantities. We show that they agree with numerical results and exhibit the general properties of rotating solitons.

We derive the gluon transverse-momentum-dependent (TMD) operator structure of back-to-back quark-antiquark dijet production in deep inelastic scattering at arbitrary Bjorken-x to twist-3 accuracy. Working at leading order in the strong coupling and in the kinematic regime where the transverse momentum imbalance of the jets is much smaller than their individual transverse momenta, we perform a systematic gradient expansion of the quark propagator in a background gluon field. This expansion organizes multiple interactions with the target in terms of longitudinal Wilson lines and gauge-invariant field-strength insertions, yielding a TMD description valid beyond the strict high-energy eikonal (x -> 0) approximation. We obtain explicit cross sections for longitudinally and transversely polarized virtual photons, identifying all contributing gluon TMD operators up to twist-3, including structures involving F_+-, F_ij, and three-gluon correlators. The full longitudinal phase associated with Bjorken-x is retained throughout. In the small-x limit, our results reproduce the known sub-eikonal expressions obtained in the Color Glass Condensate framework, establishing a direct connection between the general-x TMD expansion and high-energy factorization. We further reduce the operator basis using equations of motion, minimizing the number of independent nonperturbative matrix elements entering the cross section. This work provides a systematic foundation for extending TMD analyses of dijet production beyond leading twist, establishing a unified operator framework valid at arbitrary Bjorken x that smoothly interpolates between moderate- and small-x descriptions of gluon TMDs.

Standard-Model (SM) hadronic parton showers initiated by secondary cosmic-ray production or dark matter (DM) annihilations robustly predict very low antinuclei yields and a strong additional suppression for heavier antinuclei. We show that an important exception can arise if DM annihilates into a confining dark sector that produces Soft Unclustered Energy Patterns (SUEPs). The hallmark of SUEPs is the emission of very large multiplicities of soft dark mesons ($π_D$), which can overcome the usual phase-space suppression of antinuclei formation in parton showers, provided that the dark mesons decay promptly into SM quarks, i.e. within a SM hadronization length. We study several benchmark realizations and find that for DM masses $m_{\rm DM}\sim\mathcal{O}(10~\mathrm{TeV})$, dark meson masses $m_{π_D} \sim 400~\mathrm{GeV}$, $π_D$ dominantly decaying to $t\bar t$, and a SUEP temperature $T_{\rm SUEP}\simeq 0.1\,m_{πD}$, DM annihilation into SUEPs can yield tens of antideuterons and a few antihelium--3 events at AMS-02 at kinetic energies of $\mathcal{O}(\mathrm{GeV}$/n) and a few antideuterons and antihelium-3 events in GAPS at energies below 0.5 GeV/n. A future confirmation of an antinuclei signal by the AMS-02 or GAPS experiments could provide hints for hidden confining dynamics and would significantly constrain the relevant SUEP parameters.

Low-mass stars ($M\lesssim 2\,M_\odot$) ignite helium under degenerate conditions, eventually causing a nuclear run-away -- the helium flash. The alpha-capture process on $^{14}$N produces a large amount of $^{18}$F, whose subsequent decay spawns an intense $ν_e$ burst (with average energy of $0.38$ MeV) lasting about a day. We show that, in addition, a strong $1.7$ MeV neutrino line is generated by electron capture on $^{18}$F. Detection is hindered by large backgrounds in state-of-the-art neutrino observatories, such as JUNO. In next-generation facilities, such as the Jinping neutrino experiment, the horizon for a detection with a local significance of $3 σ$ would be extended to almost $3$ pc. Although helium flashes occur a few times per year in our Galaxy, there are no stellar candidates approaching the tip of the red giant branch within $10$ pc. Hence, to date, asteroseismology remains the most promising tool for probing the most energetic thermonuclear event in the life of a low-mass star.

We present the development and evaluation of a real-time Graph Neural Network-based trigger for the electromagnetic calorimeter of the Belle II experiment at the SuperKEKB collider. The algorithm processes calorimeter trigger cells as graph nodes to perform clustering, feature extraction, and per-cluster signal classification with deterministic latency compatible with the first-level trigger readout system. The model predicts cluster positions and energies and provides a signal classification score, enabling a more flexible clustering strategy than the baseline trigger algorithm. Implemented on an FPGA and integrated into the Belle II trigger chain for synchronous operation, the system sustains the 8 MHz trigger throughput with an end-to-end latency of 3.168 $μ$s. The performance is evaluated using simulated events and collision data. The energy resolution is comparable to the baseline trigger, while the position resolution for high-energy clusters improves by up to 18 percent in the central detector region. Cluster purity increases by up to 20 percent at low energies for isolated clusters, and cluster efficiency improves by up to 20 percent for overlapping clusters. The signal classifier enables additional background suppression at fixed signal retention. These results demonstrate the first operation of a Graph Neural Network-based reconstruction system implemented on FPGAs within the real-time trigger readout path of a collider experiment.

An interesting proposal suggests that a QCD axion $(a)$ coupling to the up quark, down quark, and electron remains viable for an axion mass near 10~MeV. In this paper, this possibility is reexamined by deriving new bounds from kaon decays. In particular, we perform a detailed analysis of the $K_L \to π^0 π^0 e^+ e^-$ measurement reported by the KTeV experiment, and reinterpret $K^+$ decay measurements at the E949 and NA62 experiments to constrain both the diphoton decay and effectively invisible decay modes of the axion. We find that, combined with the previously known bounds, the viable window for the MeV-scale QCD axion is excluded, primarily due to the KTeV bound. Uncertainties associated with the chiral Lagrangian are further examined, and the scenario remains excluded even after accounting for these uncertainties, except for a tiny region of parameter space where higher-order corrections must finely cancel the leading-order contribution, suppressing the branching ratio of $K_L\to π^0π^0 a$ by three orders of magnitude.

Dynamical tides can provide us vital information about the properties of neutron star (NS) matter. This is particularly true for g-modes, whose frequency and tidal coupling are highly sensitive to the composition of NSs, especially in their centers, where microphysical models are the least reliable. However, due to their weak coupling to external tidal fields, their effect on the gravitational-wave (GW) signal of binary inspirals can be difficult to observe. Here we show that the detectability of these tides can be significantly enhanced by binary NSs with moderate eccentricities. This is primarily due to higher eccentric harmonics in the early phase of the binary evolution experiencing larger phase shifts, which they transport to the sensitive band of GW detectors. In addition, g-mode tides in eccentric binaries undergo several epicyclic resonances, which also amplify the total phase shift. We demonstrate that these effects increase the detectability of g-mode dynamical tides by more than an order of magnitude for eccentricities of $e_\mathrm{10Hz}\sim0.2-0.4$, making it possible to put robust constraints on g-mode properties using current GW detectors, while all relevant models could potentially be constrained with eccentric binary NSs with Einstein Telescope.

The Sun is a promising target for dark matter searches due to its ability to accumulate dark matter particles via scattering and catalyze their self-annihilation. However, at low dark matter masses, dark matter particles can also "evaporate" due to subsequent collisions with the hot thermal plasma of the Sun. While several modeling studies have calculated the competitive dynamics of dark matter evaporation and annihilation, observational studies have typically assumed a fixed 4 GeV "evaporation limit", below which dark matter evaporates before it can annihilate. In this paper, we carefully consider the competitive effects of dark matter evaporation and annihilation on the resulting spin-dependent dark matter nucleon cross-section limits, finding that Solar observations can continue to exceed terrestrial constraints by between 1-5 orders of magnitude for dark matter masses between 2-4 GeV for dark matter annihilation to both neutrinos and long-lived mediators.

We propose a loop-level generalization of the inverse string theory Kawai-Lewellen-Tye (KLT) kernel: the planar inverse KLT integrand. The integrand is defined constructively via a novel Berends-Giele-like recursion that exposes the inverse KLT kernel as the simplest toy model of a ``stringy amplitude''. We show that, to all loop orders, the inverse KLT integrand is structurally equivalent to integrands in the cubic scalar tr$φ^3$ theory. This simplicity is obscured in the conventional Feynman diagram approach, where the inverse KLT integrand receives contributions from an infinity of infinite towers of contact interactions. The inverse KLT integrand is a rational function of stringified kinematic variables and is naturally defined on the kinematic surface proposed by Arkani-Hamed et al.. It provides an elementary analogue of the surfacehedron integrand for the tr$φ^3$ theory involving only scalar resonances and unifies the scattering of cubic scalars and pions in the non-linear sigma model (NLSM) to all loop orders via kinematic $α'$-shifts.

We present ABCMB, a differentiable Einstein-Boltzmann solver for the cosmic microwave background (CMB). ABCMB is a complete code capturing important effects to linear order in $Λ{\rm CDM}$ cosmology. It computes the CMB power spectrum and includes effects like lensing, polarization, massive neutrinos, and a state-of-the-art treatment of BBN and recombination. ABCMB has sub-percent-level agreement with CLASS and can be run on a GPU with competitive, and sometimes even faster, run times. It is refactored compared to previous codes and takes advantage of object-oriented programming to improve extensibility, meaning new physics can be added to it without the need for modifying source files. ABCMB provides accurate and stable gradients to the user, making Fisher analyses straightforward, and enabling the use of efficient gradient-based sampling methods.

We present an analysis of the radiative decay $φ\to γπ^0 π^0$ in a dispersive framework, where the two-pion subsystem undergoes strong final-state interactions that cover the $f_0(500)$ and $f_0(980)$ regions. We employ a coupled-channel Muskhelishvili-Omnès framework that allows for a consistent treatment of two scalar resonances and crossed-channel singularities induced by the Born and vector-meson exchanges. We explicitly verify the equivalence between the modified and standard Muskhelishvili-Omnès representations for vector-meson pole contributions when the isoscalar Omnès matrix is chosen asymptotically bounded, and we adopt the standard representation in decay kinematics. This yields, for the first time, a parameter-free dispersive prediction for the kaon Born rescattering, which provides a dominant contribution. To obtain a good fit to the KLOE and SND data, we employ a once-subtracted coupled-channel dispersion relation with heavier left-hand cut contributions and two unknown subtraction constants. The results demonstrate the consistency among the data for $ππ$ scattering, $γγ$ fusion, and $φ$ radiative decay, thereby validating the underlying dispersive formalism and the input used for the hadronic Omnès matrix and left-hand cuts.

We consider the Polyakov loop correlator in the confining phase of large $N$ Yang-Mills theory in three and four dimensions. It can be computed by summing over the exchange of closed flux tubes winding around the thermal cycle. At short separations, the leading divergence is controlled by perturbation theory. Combining these two facts allows us to determine the asymptotic spectral density of string states contributing to the correlator. This sharply relates the weakly-coupled UV of the gauge theory to the dynamics of highly energetic flux tubes. Then, in a toy integrable setting, we explore how this can bound the scattering data of the Goldstone modes on top of a long string. We derive a bound on the asymptotic behavior of the reflection amplitude of Goldstones against the flux tube boundary sourced by the Polyakov line, and rule out an asymptotically linear phase shift for the S-matrix. Along the way, we discuss how causality can impose bounds on thermodynamic quantities, and show how the positivity of time delays follows from unitarity and analyticity of $2d$ massless elastic S-matrices. We include a review on reflection amplitudes, and their computation in the theory of long effective strings.

In this paper, we investigate the Bhabha scattering process within the framework of non-Hermitian QED at finite temperature. In this theory, the hermiticity condition, typically required in quantum field theory to ensure the reality of physical observables, is replaced by the condition of unbroken $PT$-symmetry which favors the introduction of an axial mass and a vector-axial gauge coupling. Using the Thermofield Dynamics formalism, we derive and comprehensively analyze the thermal differential cross section for the Bhabha scattering. Furthermore, we explore the high-energy limit of the scattering amplitude and establish constraints upon the axial coupling constant, offering valuable insights into the system's behavior under extreme conditions.

CP-odd observables originating from the strange-quark electric dipole moment ($d_s$) are identified in $γ^\ast\to K^+K^-π^0$ and evaluated in the chiral limit. A nonzero T-odd asymmetry $A_T$ requires anomalous couplings descending from a five-dimensional Chern-Simons construction. Modeling the running of $F_A$ and $F_V$ with vector-meson dominance, we estimate sensitivities to $d_s$ at the level of $\mathcal{O}(10^{-16})\,e\cdot\mathrm{cm}$ in $e^+e^-$ data at CMD-3 and $\mathcal{O}(10^{-18})\,e\cdot\mathrm{cm}$ using existing $J/ψ$ samples at BESIII. Future experiments at the Super Tau-Charm Facility and Belle~II can further improve the reach by an order of magnitude.

Hadrons with explicit gluonic degrees of freedom, such as charmonium hybrids, are key to understanding nonperturbative behavior of strong interaction, yet they remain experimentally elusive. Within a constituent gluon model treating the hybrid as a $c\bar{c}g$ three-body system with a transverse electric gluon, we predict the masses of the lightest hybrid multiplet ($J^{PC}=1^{--}, 0^{-+}, 1^{-+}, 2^{-+}$) to lie in the range $4.19$--$4.32$ GeV. By analyzing their decay patterns, we provide specific, experimentally testable signatures to guide the search for these exotic states at current and future facilities.

A search is presented for Higgs boson production in association with a top quark-antiquark pair ($t\bar{t}H$) with the Higgs boson decaying to a charm quark-antiquark pair ($H\rightarrow c\bar{c}$). The same process with a Higgs boson decay to bottom quarks, $t\bar{t}H(H\rightarrow b\bar{b})$, is measured simultaneously. The analysis uses data from proton-proton collisions at 13 TeV collected with the CMS detector in 2016-2018, corresponding to an integrated luminosity of 136 fb$^{-1}$. The observed $t\bar{t}H(H\rightarrow\rightarrow b\bar{b}$ cross section relative to its prediction from the standard model (SM) is $0.91\,^{+0.26}_{-0.22}$. The observed (expected) results are compatible at the 95% confidence level with a $t\bar{t}H(H\rightarrow c\bar{c})$ cross section that is at most 7.8 (8.7) times larger than the SM expectation. Assuming all other Higgs boson couplings to be SM-like, this sets an observed (expected) upper bound on the charm quark Yukawa coupling modifier of 3.0 (3.3) times its SM value.

We numerically determine the S-matrix by using connection formulae in the modular linear differential equation (MLDE) approach to the holomorphic modular bootstrap. We then determine exact formulae using the fact that entries in the $S$-matrix are integer entries in a cyclotomic extension of the field of rational numbers. This provides a method that is intrinsic to the MLDE setup and does not require inputs outside this framework. The method is illustrated with a selection of examples.

The topological classification of the one-loop Weinberg operator at dimension-5 enables a systematic categorization of radiative neutrino mass models. Among these, the category consisting loop-extended seesaw frameworks is theoretically appealing but conventional discrete or continuous symmetries (\emph{e.g.}, $U(1)$ or $\mathbb{Z}_M$) cannot genuinely forbid the corresponding tree-level contributions, making loop dominance difficult to realize. We show that \textit{non-invertible selection rules} (NISRs) naturally enforce the absence of tree-level terms while ensuring a dominant one-loop contribution. Intriguingly, the same non-invertible structure also stabilizes the dark matter candidate, providing a unified radiative origin of neutrino mass and dark sector stability. In particular, we focus on the T4-2-$i$ topology which embodies a type-II one-loop seesaw and demonstrate its natural realization from $Z_{7}$ Tambara--Yamagami (TY) fusion algebra.

In this letter we present a comprehensive analysis of the differential equations governing the spatial profile of the 't Hooft-Polyakov monopole from the viewpoint of resurgence theory. We note that the special shape of the gauge-component asymptotics, together with the simplicity of its Borel transform and the Green's kernels in the Volterra equations, provides a complete control over the proliferation of the Borel-plane singularities to all orders and for any $λ/e^2>0$, along with full information about the relevant logarithmic discontinuities. Hence, the resurgent structure of the non-BPS solutions turns out to be unexpectedly simple, and its resonance at the origin of the Borel plane even admits an analytic calculation of the normalization coefficients of their asymptotic forms to an arbitrary level of precision.

The ReD experiment measured the ionization yield Qy of argon for nuclear recoils in the 2-10 keV range using a dual-phase Time Projection Chamber irradiated with neutrons from a Cf-252 fission source. The measurement extends coverage below 7 keV, confirms consistency with previous data above 7 keV, and indicates a higher Qy at lower energies. These results are relevant for argon-based experiments searching for dark matter in the form of low-mass Weakly Interacting Massive Particles, which are very sensitive to the modeling of the detector response in this energy range.

In this paper, we construct an explicit rephasing transformation that converts an arbitrary unitary mixing matrix into the Fritzsch--Xing (FX) parametrization, which is obtained by trivializing arguments of the matrix elements in the third row and third column. We further analyzed the rephasing invariant structure of the FX phase $δ_{\rm FX}$ under the approximation where the 1-3 element of the diagonalization matrix of charged leptons $U^{e}$ is neglected, $U_{13}^{e} = 0$. As a result, in the limit where the 1-3 element of diagonalization of neutrinos $U^ν$ is also neglected, we obtained a compact expression for the FX phase $δ_{\rm FX} = \arg [ { U^{e}_{11} U^{e}_{12} U^{e}_{33} } { U^{ν*}_{11} U^{ν*}_{12} U^{ν*}_{33} / \det U^{e} \det U^{ν*} } ] - \arg [ 1 + ({U^{e *}_{23} U^ν_{23} / U^{e *}_{33} U^ν_{33}} ) ] $. In this limit, the FX phase is controlled by two independent relative phases between $U^{ν, e}$, and the FX phase for finite $U_{13}^ν$ is understood as a generalization to the compact expression.

We investigate the phase structure of Quantum Chromodynamics (QCD) in the vacuum as a function of quark flavor number $N_f$ within the chiral limit. By self-consistently solving the coupled DSEs for the quark and gluon propagators in a minimal QCD scheme, we elucidate the nonperturbative dynamics governing dynamical chiral symmetry breaking. Our calculations determine a critical flavor number of $N_f^c=6.81$ which marks the chiral symmetry restoration of quarks. Further analysis reveals the critical exponents of the chiral condensate as $ -\langle\barψ ψ\rangle\sim |N_f-N_f^c|^{0.53(9)}$, characterized the second order feature of this phase transition of chiral symmetry. Additionally, we discuss the implications for the walking regime towards the conformal window at larger flavor.

We study quantum criticality in a holographic Einstein--Maxwell--Dilaton--Axion (EMDA) p-wave superconductor exhibiting a superconductor--insulator transition (SIT). By tracking the superconducting energy gap, we show that approaching the quantum critical point (QCP) closes the gap and induces incipient insulating features, indicating that enhanced quantum fluctuations suppress superconducting order and trigger the SIT. To probe the transition, we employ two holographic indicators: holographic entanglement entropy (HEE) and the entanglement wedge cross-section (EWCS), the latter being a mixed-state entanglement measure. In contrast to HEE, which for sufficiently large configuration is dominated by the thermal entropy and is therefore largely insensitive to entanglement along the temperature direction, EWCS displays pronounced critical scaling and provides a robust diagnostic of the quantum phase transition (QPT). We attribute this contrast to the fact that HEE at large scales is controlled by the infrared (IR) geometry, whereas EWCS is governed by deformations of the entire bulk. Our results establish EWCS as a robust probe of holographic quantum criticality in mixed states.

We propose an integral transformation for celestial gluon amplitudes that maps the celestial coordinates \((z_i,\bar z_i)\) to a new set of complex variables \((s_i,\bar s_i)\), inspired by the structure of closed string scattering amplitudes. A consistent inverse transformation is constructed by regulating a divergence associated with translational redundancy and absorbing it into an overall normalization. Applying this transformation to celestial MHV amplitudes, we derive constraints on \((s_i,\bar s_i)\) for three-, four-, and general \(n\)-point amplitudes, and show that these conditions are necessary for invariance under global conformal transformations.

We match the real Higgs triplet model (RHTM) onto HEFT under different parameter choices and power-counting schemes, thereby obtaining several representative HEFT formulations and clarifying their relations. We establish the primary HEFT (pHEFT) as a benchmark framework, demonstrating that alternative HEFT constructions can be systematically derived from it. A key advantage of the pHEFT construction is its parameter choice, which maintains linear relations between the UV Lagrangian parameters and squared heavy masses. By strictly employing the inverse squared heavy masses as the expansion parameters without imposing additional constraints, pHEFT preserves maximal ultraviolet (UV) information and ensures higher perturbative accuracy by avoiding the additional truncations inherent in more complex, non-linear formulations or extra constraints. Through the analysis of the $Z_2$-symmetric real singlet model and the 2HDM, we illustrate the criteria for identifying viable primary HEFT constructions in UV models with scalar extensions. Furthermore, for the first time, we derive the HEFT operators of the RHTM involving fermions.

We report on the measurement of the two-dimensional differential cross section for near-threshold J/$ψ\to μ^+μ^-$ photoproduction from the J/$ψ$-007 experiment in Hall C at Jefferson Lab. Our results agree with the previously published J/$ψ\to e^+e^-$ results. We extract the integrated photoproduction cross section and find no evidence for open-charm contributions. A combined analysis of both decay channels following a Holographic QCD approach yields improved experimental constraints on the gluonic gravitational form factor $\mathcal{C}_g(t)$. Our results agree with recent lattice QCD calculations and we obtain $\mathcal{C}_g(t)$ with a comparable statistical precision to lattice QCD. Our results support a spatial picture where gluons dominate at larger radii with a confining inward pressure. This work provides new input for exploring the mechanical properties of gluons inside the proton.

We develop a differential-form approach to systematically derive the Newman-Penrose null-tetrad equations for Lorentz-violating extensions of Maxwell electrodynamics. The coordinate-independent nature of differential forms allows the actions and corresponding field equations of the theory to be expressed compactly and enables a systematic and transparent derivation of first-order equations in the Newman-Penrose formalism. Within this formalism, we explicitly present a simple algebraic construction for the gauge invariant extended Maxwell actions that avoids explicit index manipulations up to mass dimension six. The combined scheme of differential-form approach and Newman-Penrose formalism offers an efficient tool for analyzing Lorentz-violating effects on asymptotic photon propagation and polarization.

We investigate the atmospheric production and subsequent detection of sub-GeV dark mesons within the framework of a confining dark sector coupled to the Standard Model (SM) via a $U(1)_{\text{D}}$ vector portal. High-energy cosmic ray interactions in the atmosphere produce dark quarks through proton bremsstrahlung, rare decay of Standard Model mesons, and Drell-Yan processes, which subsequently hadronize into dark mesons. We adopt a modified Quark Combination Model to describe the non-perturbative dark hadronization process, allowing for a detailed event-level characterization of the dark meson flux. We simulate the flux and the interaction of these relativistic dark mesons in the Jiangmen Underground Neutrino Observatory (JUNO) using the GENIE generator, considering both elastic scattering off nuclei and deep inelastic scattering channels. Based on the projected 20 kton$\cdot$year exposure, we derive sensitivity limits on the coupling strength between the dark gauge boson and the SM sector. Our results demonstrate that JUNO can probe substantial unexplored parameter space, particularly for light mediators ($m_{Z^\prime} \lesssim 10$ MeV), where it achieves sensitivities to the portal interaction as low as $2.4 \times 10^{-4}$.

The recent detection of a gravitational wave background in the nano-Hertz frequency range by Pulsar Timing Array (PTA) collaborations, including NANOGrav, EPTA, and PPTA, has opened a new avenue for exploring fundamental physics in the early universe. In this work, we analyze a supercooled first-order phase transition in a hidden sector with a spontaneously broken $U(1)_X$ gauge symmetry as a source for this signal. We demonstrate that the thermal history of the hidden and visible sectors plays a crucial role in the gravitational wave power spectrum analysis. Our analysis shows that supercooled phase transitions can generate gravitational waves strong enough to explain the PTA observations while satisfying cosmological constraints from Big Bang Nucleosynthesis.

We assess the potential for measuring neutrino mixing parameters at energies above 1~TeV, for the first time, using the flavor composition of TeV--PeV astrophysical neutrinos, i.e., the proportion of $ν_e$, $ν_μ$, and $ν_τ$. Today, flavor measurements inferred from the 11.4-year IceCube Medium Energy Starting Events sample are insufficient to constrain the mixing parameters due to limited statistics, challenges in flavor identification, and uncertainty in neutrino production. Yet, upcoming multi-neutrino-telescope observations -- even using only existing telescopes -- may achieve sensitivity to $θ_{23}$ and $θ_{13}$ when combined with traditional oscillation experiments. We establish the current status and future prospects for testing the three-flavor mixing framework in the previously unexplored TeV--PeV regime and quantify the minimum detectable size of flavor-modifying beyond-Standard-Model effects, providing a roadmap for high-energy neutrino mixing measurements.

A first measurement of beam spin asymmetries for $π^+π^0$ and $π^-π^0$ pairs in semi-inclusive deep inelastic scattering is reported. The asymmetries in the dihadron angular distributions were measured from the scattering of a 10.6 GeV longitudinally polarized electron beam off a proton target, using the CLAS12 detector at Jefferson Lab. A photon classifier using a Gradient Boosted Trees (GBTs) architecture was trained with Monte Carlo simulations to reduce the amount of false combinatorial background $π^0$s, increasing statistics by up to five-fold compared to previous CLAS12 $π^0$ analyses. A nonzero $\sinφ_{R_\perp}$ asymmetry is observed. This measurement is sensitive to the underexplored collinear twist-3 PDF $e(x)$, which encodes quark-gluon correlations in the proton, and presents a new avenue for its point-by-point extraction. The asymmetries also provide the first experimental evidence for the isospin-dependence of the helicity-dependent dihadron fragmentation function $G_1^\perp$, revealed by a sign-difference between the $π^+π^0$ and $π^-π^0$ channels in the $\sin(φ_h-φ_{R_\perp})$ modulation. In contrast, a large, same-sign enhancement near the $ρ$ mass for the $\sin(2φ_h-2φ_{R_\perp})$ modulation is observed, matching spectator model predictions in $π^+π^-$ pairs.

We calculate the gluon momentum fraction of the nucleon using lattice quantum chromodynamics (QCD), with a nonperturbative renormalization technique based on the gradient flow. The gluon momentum fraction is determined on a single Wilson-clover ensemble using Nf = 2+1 flavors with pion mass 358 MeV and lattice spacing 0.094 fm. We employ the variational method to reduce excited-state contamination and apply the distillation framework to ensure a large operator basis. To reduce systematic uncertainties, we apply Bayesian model averaging to all fit procedures. We apply matching coefficients to the flow-time dependent lattice results to recover the gluon momentum fraction in the MS-scheme at 2 GeV. Our final result is <x>_g(μ= 2 GeV) = 0.482(35), where we quote only statistical uncertainties.

The azimuthal anisotropy of prompt D$^\pm_\mathrm{s}$ mesons produced in lead-lead (PbPb) collisions at a nucleon-nucleon center-of-mass energy of 5.02 TeV is measured using data obtained with the CMS detector. The dataset corresponds to an integrated luminosity of 0.58 nb$^{-1}$. The azimuthal anisotropy of heavy charmed mesons provides a key constraint on the interactions of charm quarks with the quark-gluon plasma (QGP) medium. These interactions include coalescence mechanisms and parton energy loss in the QGP. The anisotropy is quantified by the second- ($v_2$) and third-order ($v_3$) Fourier coefficients of the azimuthal distribution of the D$^\pm_\mathrm{s}$ mesons. The $v_2$ coefficient is determined in the transverse momentum range 4 $\lt$ $p_\mathrm{T}$ $\lt$ 40 GeV for three event centrality classes, while the $v_3$ coefficient is measured in the range 4 $\lt$ $p_\mathrm{T}$ $\lt$ 20 GeV for a single event centrality class. The results for the D$^\pm_\mathrm{s}$ mesons are compared to those previously measured for D$^0$ mesons. The azimuthal anisotropy coefficients for D$^\pm_\mathrm{s}$ and D$^0$ mesons are found to be consistent within the precision of this measurement, suggesting that the strangeness content of the D$^\pm_\mathrm{s}$ meson does not significantly alter its azimuthal distribution within the measured $p_\mathrm{T}$ range.

Lorentz-symmetry violation may be described via the CPT-odd, dimension-3, Carroll-Field-Jackiw term, which couples the electromagnetic fields to a constant 4-vector $k_{\rm AF}$ selecting a preferred direction in spacetime. We solve the field equations using the Green's method for a static point-like magnetic dipole and find the $k_{\rm AF}$-dependent corrections to the standard dipolar magnetic field that strongly dominates the near-Earth magnetic field. Given the very good agreement between current models and ground- and satellite-based geomagnetic data, our strongest constraints on the components of $k_{\rm AF}$ in the Sun-centered frame read $|(k_{\rm AF})_Z| \lesssim 4 \times 10^{-25} \, {\rm GeV}$ for $|(k_{\rm AF})_X|, |(k_{\rm AF})_Y| \lesssim 10^{-24} \, {\rm GeV}$ at the two-sigma level. This represents an improvement of about four orders of magnitude over earlier bounds based on other geophysical phenomena.

The SS433, a well-known binary system with an internal black hole, have shown since half a century, an inner (a few year light distances) twin precessing jets spirals. These beams are made by tidal forces while stripping mass from large stellar companion feeding an inner BH accretion disk and an orthogonal accelerating twin jet. From it, the radio, X gamma jet emission. A couple of years ago H.E.S.S telescope as well as HAWC and LHAASO array detectors, discovered also the surprising signature of an unexpected far twin separated gamma beam at tens TeV energy. At a hundred light years distances from its central source. We suggest that it is the legacy of a past rare eruption, a century ago, of tens PeV (10^16 eV) relativistic twin neutron beams. Their beta decay in flight at far distances, into proton, neutrino and in particular into tens TeV electrons, could feed the observed TeV gamma traces. They are originated by the same secondary tens TeV electrons emitting hard gamma, by Inverse Compton Scattering onto interstellar infrared photons.

We consider the early binary neutron star inspiral phase as a scenario to probe environmental axion--photon resonant conversion. For this we approximately model the merger site electromagnetic fields as the superposition of two rotating dipolar stellar magnetic fields at the thousand--km scale when both magnetospheres are not largely distorted. We capture the time-sliced near-zone magnetospheric geometry relevant for axion--photon mixing. Plasma effects are incorporated through an effective Goldreich--Julian charge density, used to determine the effective plasma frequency and the location of resonant conversion surfaces. Our results show that axion--photon resonant conversion in binary magnetospheres mostly occurs on extended peanut-shaped surfaces whose global geometry evolves as the binary inspiral evolves. As a consequence, the total electromagnetic power emitted through axion--photon conversion exhibits a characteristic dependence on axion mass and a slow temporal modulation correlated with the gravitational wave frequency emission. This feature is potentially detectable for $m_a \in [50,170] \,\rm μeV$ and set $g_{a γ} \lesssim 10^{-11}\rm \,GeV^{-1}$ as it lies within the sensitivity limits of current or planned radio observation missions. In light of our results we discuss the opportunity of binary neutron star inspirals as time-dependent, multimessenger probes of axion physics, and motivate coordinated searches combining gravitational wave observations with radio and millimeter wavelength electromagnetic measurements.

21cm radiation from neutral hydrogen serves as a direct probe of the Epoch of Reionization. However, both its detection and physical interpretation are severely hindered by contamination from astrophysical foreground emission and instrumental noise that are several orders of magnitude brighter than the signal of interest. A promising way to tackle these challenges is to cross-correlate the 21cm signal with other independent tracers of large-scale structure, most notably high-redshift galaxies. Besides validating putative 21cm detections, such joint analyses are expected to provide independent insights into the properties of ionizing sources and the evolving morphology of ionized regions during reionization. The 21cm signal, however, is intrinsically highly non-Gaussian, limiting the effectiveness of conventional two-point cross-correlation statistics, which capture information only up to the second order. In this work, we therefore investigate the utility of k-nearest-neighbour cumulative distribution functions (kNN CDF), which encode information from the joint clustering at all orders, as an alternative framework for probing 21cm-galaxy cross-correlations. Using self-consistently simulated mock 21cm fields and a catalog of line-emitting galaxies at z = 7, we conducted a proof-of-concept study comparing the kNN CDF formalism and the two-point cross-correlation approach. We find that the kNN CDF statistics outperform the two-point statistics in detecting 21cm-galaxy cross-correlations, even in the presence of instrumental noise and aggressive foreground filtering. Moreover, at a fixed global ionized fraction, it is even able to differentiate between reionization models that remain indistinguishable using two-point statistics. These results demonstrate the power and unexplored potential of exploiting higher-order statistics for extracting maximal information from 21cm-galaxy synergies.

Molecular gas is crucial to understanding star formation and galaxy evolution, but the giant molecular clouds (GMCs) of early-type galaxies (ETGs) have rarely been studied. Here, we present analyses of the spatially resolved GMCs of the lenticular galaxy NGC 1387, exploiting high spatial resolution (0.15" or 14 pc) 12CO(2-1) line observations from the Atacama Large Millimeter/submillimeter Array. We identify 1285 individual GMCs and measure the fundamental properties (radius, velocity dispersion, and molecular gas mass) of each with a modified version of the CPROPStoo package. Unusually for an ETG, the GMCs of NGC 1387 follow scaling relations very similar to those of the Milky Way disc and Local Group galaxy clouds, and most are virialised. GMCs with large masses and radii and/or small galactocentric distances have their angular momenta aligned with the large-scale galactic rotation, while other GMCs do not. These results show that ETGs have more diversified GMC properties than previously thought. We discuss potential reasons for such diversity, and viewing-angle dependency is a plausible candidate.

Type II supernovae (SNe II) have been traditionally separated into several subgroups based on their photometric and spectroscopic properties, but whether these represent distinct progenitors or a continuous distribution remains debated. Over the past decade, growing observational evidence has suggested a possible continuity between slow- (IIP) and fast-declining (IIL) SNe. We investigate the continuity of the SNe IIP/L subclasses through a data-driven statistical analysis of spectral time series, aiming to determine whether significant correlations exist between overall spectral shapes and light-curve decline rates. We introduce a novel standardization method for SN II spectra. After empirically flattening the spectra via continuum normalization, we interpolate the resulting "feature spectra" onto a fixed grid of epochs using Gaussian Process regression. The interpolated spectra are then analyzed using Principal Component Analysis to explore correlations. We find that SNe IIP and IIL form a continuum spectroscopically, though some clustering remains. The spectral diversity is characterized mainly by two components: one continuous group with well-defined P-Cygni profiles and another with "less-regular" features likely driven by enhanced circumstellar material (CSM) interaction. Our results reveal that the spectral diversity of SNe IIP/L diminishes over time. We confirm observational correlations: steeper light-curve declines correspond to weaker spectral features, indicating that SNe IIL tend to show weaker emission and, in some cases, a lack of distinct absorption lines. These trends seemingly break down by enhanced CSM interaction that affects the P-Cygni profiles. Our data-driven method reveals underlying spectral correlations and supports a continuous distribution between IIP and IIL subtypes. This method paves the way for more refined classification algorithms.

We present the Montreal Open Clusters and Associations database (MOCAdb), a public MySQL database with a Python interface. MOCAdb provides a census of memberships for 10259 associations and open clusters within 500 pc of the Sun, with a comprehensive compilation of literature measurements such as spectral types, kinematics, rotation periods, activity indices, spectral indices, and photometry. All known substellar objects are cataloged in MOCAdb, along with 2943 public spectra, to enable the characterization of substellar association members. MOCAdb also features periodically updated calculations such as Galactic UVW space velocities. We use this compilation to construct mappings between independent association definitions, and to update the BANYAN $Σ$ membership classification tool, which now includes 8125 associations. The BANYAN $Σ$ model construction is improved to account for heterogeneous and correlated errors and to capture complex association shapes using Gaussian mixture models. Combined with Gaia DR3, this enabled us to identify 11535 yet unrecognized candidate members of young associations within 500 pc, mostly M dwarfs. Our results corroborate a recent observation that systematics up to $\approx$4 km/s remain in Gaia DR3 radial velocities for A-type stars. We present an updated census of age-calibrated exoplanets and substellar objects: 134 age-calibrated exoplanet systems (plus 121 TESS exoplanet candidates), 99 of which did not appear to have known memberships so far, and 455 substellar (L0 or later) candidate members of young associations, 196 of which appear newly recognized. We bring the total of candidate isolated planetary-mass objects to 101, 53 of which are newly recognized candidate members.

This work explores the potential to detect the nonlinear Integrated Sachs Wolfe effect, namely the Rees-Sciama effect (ISWRS), by cross-correlating current and future Cosmic Microwave Background (CMB) experiments -- Simons Observatory, CMB-S4, CMB-HD, and PICO -- with ongoing Large Scale Structure (LSS) surveys, such as Euclid and the Vera Rubin Observatory (LSST). We model the cross-correlation of the ISWRS effect with gravitational potential tracers like galaxy clustering, cosmic shear, and CMB-lensing potential, to forecast results from these experiments. Our analysis also accounts for the presence of massive neutrinos to assess the feasibility of identifying the $ν$$Λ$CDM model and constraining the neutrino mass sum, M$ν$. Our findings indicate that the CMB-lensing potential reconstructed by CMB-HD is expected to provide the most promising results, achieving $\gtrsim$ 5$σ$ detections even under conservative assumptions for detector noise and foregrounds, thereby allowing differentiation between $ν$$Λ$CDM models. Galaxy clustering can also yield significant detections, whereas cosmic shear can provide valuable results only if non-linearities are accurately modelled, beyond the capabilities of currently available analytical approaches. These latter LSS probes do not provide strong constraining power on M$ν$. While our findings suggest that future CMB experiments and LSS surveys will enable the detection of the ISWRS effect, they do not imply significant prospects for imposing new constraints on neutrino masses in the near future.

Multimessenger observation of the Seyfert-starburst composite galaxies NGC 1068 and NGC 7469 indicate a characteristic feature in the radio band (the so-called mm-bump) as well as indication of high-energy neutrinos by the AGN corona. Moreover, also the starburst ring of these sources is bright in the radio and hence, a potential source of $γ$-rays and neutrinos. We aim to explain the non-thermal features of these two sources with our homogeneous steady-state Seyfert-starburst composite model, which we refined in this work. Hereby, we account for stochastic diffuse acceleration and energy losses within the corona and $γγ$-pair attenuation of the escaping $γ$-rays. Since the non-thermal features of Seyfert sources contribute only marginally to the electromagnetic spectrum, only few data points can be assigned to the starburst ring or the AGN corona. Hence, prior information on the physical parameters is incorporated within a Markov Chain Monte Carlo approach to avoid overfitting. Based on this Bayesian parameter study we show, that the non-thermal features of NGC 1068 can be explained well. Still a more detailed treatment of the spatial inhomogeneities in the central region of the AGN could further improve the fit results. This manifests itself even more clearly in the case of NGC 7469, where the mm-bump needs to emerge from a coronal size $R_{\rm c}>100\,\mathcal{R}_{\rm s}$, whereas (TeV-PeV)-neutrino emission requires $R_{\rm c}< 10\,\mathcal{R}_{\rm s}$. Similar to what has previously been shown in other wavebands, our analysis highlights that the spatial extension of the so-called AGN corona depends the considered energy of the messenger. Hence, it seems that there is not a unique edge of the corona and a substantial progress in the understanding of these phenomena is expected if future analysis account for these spatial inhomogeneities.

We present a fast and exact method for computing CMB mode-coupling matrices based on an optimised evaluation of Wigner-3j symbols. The method exploits analytic structure in the relevant Wigner-3j symbol configurations appearing in temperature and polarization coupling matrices, expressing all required quantities in terms of a small set of recurrence-generated values which are precomputed and stored in lookup tables. This approach reduces the computational cost of constructing the full coupling matrices whilst maintaining numerical accuracy. We demonstrate the performance of the threej_cosmo implementation using realistic survey masks from current CMB experiments. Relative to standard recursion-based approaches used in existing pseudo-C_l pipelines, the method achieves speedups of 6-25x in practical coupling-matrix constructions, with the largest gains occurring at high multipoles. The algorithm admits efficient parallelisation on both CPUs and GPUs, the latter providing additional acceleration, up to a further order of 50 on modern hardware, without altering the underlying formalism. Beyond full matrix construction, the approach is naturally suited to applications in which only a restricted set of l3 modes is required for each (l1,l2) pair, such as in the computation of band-limited coupling matrices and analytic covariance terms. These features make threej_cosmo a practical backend for pseudo-C_l estimation and related calculations in next-generation CMB analysis pipelines.

Disc galaxies exhibit radial metallicity gradients in both their stellar and gaseous components. The star-forming gas (SFG) in HII regions and young stars (YSs) trace the recent evolutionary history of the galaxy. We aim to assess the extent to which the joint analysis of metallicity gradient alignment in YSs and SFG can constrain the recent evolutionary history of galaxies. Using the high-resolution run of the EAGLE project, we derived radial, azimuthally averaged oxygen abundance profiles for YSs (age < 2 Gyr) and SFG and measured their gradients as the slopes of linear fits to these profiles. We classified galaxies into four groups based on the signs (N for negative and P for positive) of the slopes: NN, NP, PP, and PN (the first letter is for YSs and the second for SFG). We found that galaxies with NN, NP, PP, and PN combinations of metallicity profiles reflect different evolutionary paths over the past ~ 2 Gyr. NN galaxies exhibit sustained inside-out growth accompanied by high star formation efficiency, whereas NP and PP systems show evidence of recent or ongoing feedback-driven disruption, with PP galaxies likely being predominantly shaped by supernova feedback. PN galaxies, by contrast, show evidence of past violent events followed by gradient recovery, highlighting the interplay between inflows, feedback, and gas cooling in shaping metallicity distributions. The degree of alignment between the stellar and gas metallicity gradients provides a way to time the occurrence of significant events in the evolutionary history of galaxies, which contribute through a combination of gas inflows, star formation triggering, and metal mixing. They could also serve as probes of sub-grid physics when observations provide suitable comparison datasets. [Abridged]

Understanding how supermassive black holes (BHs) couple to their host galaxies across a vast spatial and temporal dynamic range remains a central challenge in galaxy evolution. Using the multizone framework -- designed to capture bidirectional inflow--outflow from the event horizon to the Bondi scale -- we present a suite of long-duration GRMHD simulations spanning BH spins $|a_\ast|=0$--0.9 and Bondi radii $R_B/r_g=4\times10^2$--$2\times10^6$. From these simulations we derive spin-dependent subgrid prescriptions from first principles, applicable to hot accretion flows with low-Eddington ratios ($f_{\rm Edd}\lesssim10^{-3}$), for adoption in cosmological simulations and semi-analytic models. We provide compact analytic fits for the time-averaged accretion rate $\dot M(R_B,a_\ast)$ and feedback power $\dot E_{\rm fb}(R_B,a_\ast)$ with respect to the Bondi rate $\dot{M}_B$, which are largely insensitive to the initial gas configuration and magnetic field strength. To capture intrinsic time-variability, we also quantify the full distributions of $\dot M$ and feedback efficiency $η$, both well described by lognormal statistics, with widths that increase toward larger $R_B$. We further measure self-consistent spin evolution in the hot accretion mode, finding that the spin-up parameter varies as $s(a_\ast)\simeq -3.7\,a_\ast$, which implies a very long spindown timescale $t_s\simeq 12(10^{-3}/f_{\rm Edd})\,{\rm Gyr}$. Thus, BH spins are effectively frozen during phases of quiescent accretion. Compared to conventional small-domain GRMHD calculations, our simulations, which reach dynamical equilibrium across horizon-to-galaxy scales, yield systematically different long-term accretion, feedback, and spin properties, cautioning against direct extrapolation from small-scale GRMHD simulations when constructing galactic-scale subgrid models.

Radio wave propagation can be perturbed by compact ionized gas clumps through plasma lensing, which induces frequency dependent magnification and may distort the observed number counts of background sources. The quasar (QSO) number densities are a powerful probe for understanding the effects of intervening material. Absorption lines in QSO spectra reveal the presence of interstellar and intergalactic gas, which can change observed fluxes through dust extinction and plasma lensing. By combining observations from radio (VLASS), infrared (WISE), and optical bands (DESI), we assembled a sample of QSOs: ~4000 sources with MgII absorbers, and ~12, 000 non-absorbers. In the radio band, the MgII sample shows a moderate excess at the bright end of the flux distribution, which is broadly consistent with plasma lensing predications. In the optical, the MgII sample turns over at higher g-band fluxes and exhibits a steeper decline at the faint end than the non-MgII sample. Control samples were constructed by matching in redshift, infrared (W1), and optical (g) luminosities. In these comparisons, the radio excess becomes less prominent, suggesting that the apparent magnification may not be robust evidence for plasma lensing. Nevertheless, a weak contribution cannot be ruled out, especially given residual excess observed at the bright end relative to the non-MgII sample. Dust extinction along the line-of-sight remains a plausible alternative. Regardless of the dominant mechanism, the multi-wavelength differences offer a valuable probe of the physical state of the intervening medium.

Carbon stars on the asymptotic giant branch are major contributors to galactic dust enrichment, with gas mass-loss rates up to 1e-4 Msun/yr. We present a homogeneous spectral energy distribution analysis of the Gaia DR3 Golden Sample of carbon stars in the Milky Way and Magellanic Clouds. Our dataset includes 14,747 sources with multi-band photometry from Gaia, 2MASS, and WISE, combined with recent distance and extinction estimates. For a subsample of 2,494 Mira variables, we model multi-band light curves to derive accurate mean magnitudes. Stellar and circumstellar parameters are obtained by fitting observations with a large grid of synthetic spectra computed with the DUSTY radiative transfer code using COMARCS atmospheres. We derive effective temperature, optical depth, and gas mass-loss rate for each source. The distributions peak around Teff = 3150 K, with mass-loss rates spanning 1e-11 to 1e-4 Msun/yr and inner dust temperatures near 1000 K. We find a correlation between variability amplitude and mass-loss rate. This framework provides a statistically robust view of carbon stars across environments with different metallicities. Apparent environmental dependencies are influenced by luminosity distributions and selection effects rather than purely intrinsic metallicity differences. The combined Gaia and WISE selection limits the detection of both highly obscured and faint Magellanic Cloud sources, but the observed trends remain significant within the sampled populations.

An Amplitude-Encoded Quantum Genetic Algorithm (AEQGA) has been developed to minimize $χ^2$ functions of different cosmological probes (Supernovae Type Ia, Baryon Acoustic Oscillations, Cosmic Microwave Background Radiation), to find the best-fit value for two cosmological parameters, namely the Hubble Constant and the density matter content of the Universe today. Our main aim is to pave the way to testing the adoption of quantum optimization in the inference of the cosmological parameters that describe the universe evolution. AEQGA computes the merit function classically, and then uses a quantum circuit to entangle the population and perform crossover and mutation operations. The results show consistency with the isocontours of the objective functions. We then tested the general behavior of AEQGA as a function of its hyperparameters and compared it with a second quantum genetic algorithm found in the literature as well as with classical algorithms, finding consistent results.

We have calibrated the Tip of the Red Giant Branch (TRGB) using our recent catalog of homogeneous, high-accuracy Globular Cluster (GC) distances. The GC distances were determined by a global joint fit to optical period-Wesenheit relations of their member RR Lyrae stars and type-II Cepheids, anchored by trigonometric parallaxes; all data taken from the ESA Gaia mission's (early) third data release (GDR3). Using I-band measurements in 48 GCs from P. Stetson's database, we determined $M_{I,0} = -3.948^{+0.037}_{-0.034}$ mag (1.6% in distance). Calibrating the TRGB using Gaia's homogeneous, space-based RP photometry of 53 GCs, we found $M_{RP,0} = -3.807^{+0.041}_{-0.035}$ mag (1.8%). The stated uncertainties include statistical and systematic effects, including the correlated nature of the GC distances. The robustness of our calibrations is demonstrated via tests against small-number statistics and analysis choices. Specifically, we found no significant metallicity effect for our sample of old, low-metallicity GCs. We measured $\sim 2\%$ distances to the Large (LMC) and Small Magellanic Clouds (SMC), $18.447^{+0.036}_{-0.042}$ mag ($48.9 \pm 0.9$ kpc) and $18.898^{+0.049}_{-0.054}$ mag ($60.2 \pm 1.4$ kpc), respectively, using a single well calibrated photometric system: RP (spectro-)photometry from GDR3. Our new TRGB distances, whose absolute scale derives from Gaia parallaxes, are fully independent of the well-known detached eclipsing binary (DEB) distances and agree with them to within the uncertainties. Combining our new TRGB and existing DEB distances, we illustrate how additional constraints may be incorporated in the Local Distance Network and obtain $H_0 = 73.52 \pm 0.80$ km/s/Mpc. Expected improvements due to the upcoming fourth Gaia data release are discussed.

Superluminous supernovae (SLSNe) remain an intriguing topic in supernova (SN) transient astronomy. While the majority of SLSNe are shown to be explained by energy streaming from the newly born magnetar, there are others which are powered by different mechanisms. We analyse the pseudo-bolometric light curve of the nearby helium-rich SLSN 2021bnw. We built models and run hydrodynamics radiative-transfer simulations with STELLA. Our best-fit models include 15-22.5 Msun of ejecta enriched with 1.7 Msun of 56 Ni and carrying energy of 4 foe, and colliding w ith 7 Msun of circumstellar matter which match the observed light curve very well. The early data can be explained as cooling of an expanding shell with the mass of 0.5 Msun and kinetic energy of 0.7 foe. We tend to exclude a pulsational pair-instability (PPISN) origin for SLSN 2021bnw. Instead we conclude that SLSN 2021bnw was preferably a core-collapse explosion of a star with the initial mass of not less than 61 Msun aided by magnetorotational effects.

Studying the remnants of young core-collapse supernovae (SNe) can yield insight into the chemical composition of their progenitors and the geometry of the explosions. The supernova remnant (SNR) G292.0+1.8 is one of only three known oxygen-rich SNRs in the galaxy-remnants of core-collapse for which relatively pure fragments of ejecta can be seen. Several dozen ejecta knots from G292.0+1.8 were the subject of a proper motion analysis, based on [O III] 5007-Angstrom images taken over a 22-year baseline by Winkler et al. 2009 (arXiv:0810.1935). They determined that the transverse velocities of the filaments are linearly proportional to their distances from a common expansion center, thus the O-rich filaments have been traveling with little deceleration since the initial supernova event,about 3000 years ago. In this paper, we use optical spectra of G292.0+1.8, all taken from the Cerro Tololo Inter-American Observatory (CTIO), to measure radial velocities for 93 knots. Assuming un-decelerated expansion, as indi- cated by the proper motions, the radial velocity should be proportional to the distance from the center along the line of sight, just as the proper motions are proportional to the transverse distance. Therefore, we can map the three-dimensional structure and kinematics of the SNR. We find that the knots generally follow a broad bi-conical distribution, suggesting that the supernova explosion produced broad jets of ejecta. This structure is similar to that seen in some other young core-collapse supernova remnants.

[ABRIDGED] Galaxy evolution is shaped by internal and external mechanisms that regulate the baryon cycle and star formation activity. We present a theoretical framework based on the GAlaxy Evolution and Assembly (GAEA) semi-analytic model. We extracted portions of simulated volumes that include isolated galaxies, pairs, group, and filament members at z ~ 0, specifically avoiding massive clusters. Galaxies were classified using both intrinsic (halo-based) and observational (2D projected) parameterizations, reconstructing their environmental histories from z = 2 and identifying mergers, tidal interactions, ram pressure stripping (RPS), and starvation. 2D information decreases isolated and group fractions while doubles pairs. More than half of galaxies remain unaffected by the investigated processes since z = 2. Among affected galaxies, mergers dominate at high stellar masses (40-60% at log(M*/Msun) > 10.5). Tidal interactions are less frequent, and their incidence increases with stellar mass. RPS dominates in groups and filaments at intermediate masses (~50%), while starvation ranges from 20 to 30%. The incidence of the different mechanisms depends strongly on both mass and environment, though their imprints on global properties are often subtle. Distinct evolutionary pathways emerge: log(M*/Msun) < 9.5, galaxies in groups and filaments have a faster mass growth than galaxies in the other environments, especially those undergoing starvation, mergers and, to less extent, RPS. Differences are reduced moving to higher masses, where no clear dependence on physical mechanism emerge, even though at these masses a clear star formation suppression is evident in mergers and starved galaxies. This theoretical investigation provides essential context for the recently started multi-wavelength program Mechanisms Affecting Galaxies Nearby and Environmental Trends (MAGNET), which we introduce here.

Current observational and simulated large-scale structure (LSS) catalogues often lack consistency in assigning galaxies to specific structures, due to the absence of a universally accepted classification criterion. With the aim to generate synthetic empirical data for fine-tuning LSS classification algorithms, as well as to train machine learning (ML)/deep learning (DL) models for the same purpose, this work presents a purely geometrical simulation based on statistical spatial properties found in LSS surveys, using the spectroscopic main galaxy sample of the Sloan Digital Sky Survey (SDSS) catalogue up to a redshift of z~0.1 as a specific use case. A parallelism between the LSS and the Voronoi tessellation was utilised, in which the nodes, links, surfaces, and cells of the diagram correspond to clusters, filaments, walls, and voids, respectively. The simulation used random positions within voids as seeds for tessellating the 3D space. The resulting structures were randomly populated with galaxies that adhere to the statistical properties of their observational respective structures. As the galaxies were generated, they were tagged with their corresponding structure. In each simulation, six LSS mock catalogues were generated, following the statistical behaviour observed in the SDSS catalogue, depending on the structure they belong to. The Malmquist bias and the Fingers of God effect were simulated as well. We present a novel geometrical LSS simulator, where generated galaxies mimic the statistical properties of their observational belonging structure. The simulator was tuned to mimic the SDSS catalogue, although any other catalogue can be considered. With the generated catalogue, it is possible to adjust the LSS classification algorithms, train and test ML/DL models, and benchmark several LSS classification methods using this pre-labelled data to contrast their results and performance.

We study the properties of 56 massive (M$_{\rm{\star}}$ > 10$^{10}$ M$_{\odot}$) galaxies at $z<1$ that host AGN, detected via their broadband optical variability in the VST-COSMOS survey. VST-COSMOS provides a nearly-identical single visit depth ($r$ $\sim$ 24.6 mag) and temporal baseline (eleven years) as the forthcoming Legacy Survey of Space and Time (LSST), albeit in a much smaller 1 deg$^2$ footprint (four orders of magnitude smaller than that of the LSST). We compare the properties (morphologies, the presence of interactions, rest-frame colours and environment) of our AGN to galaxies in a control sample, which are drawn from the non-variable population and matched in redshift and stellar mass to their AGN counterparts. The fraction of AGN with early-type morphology ($\sim$55 per cent) and the fraction that is interacting ($\sim$23 per cent) are similar to what is observed in the controls, suggesting that these AGN are not primarily triggered by interactions. Similarly, the AGN and controls do not show strong differences in their rest-frame $(u-z)$ colours or local environment, suggesting that neither the recent star formation histories nor the surroundings of the AGN are strongly atypical of the general galaxy population. This study provides a glimpse into forthcoming AGN science using the LSST. With vastly improved statistics, LSST will offer unprecedented insights into AGN demographics, host-galaxy evolution and the processes that fuel supermassive black holes, potentially reshaping our understanding of their place in the Universe.

Circumgalactic gas around massive galaxies generally has a volume-filling component -- an atmosphere -- with a temperature determined by the potential-well depth of the galaxy's halo. If the atmosphere is near hydrostatic equilibrium and is stable to convection, then it can remain nearly homogeneous, as long as it is not too dense. But if its density is great enough, it becomes prone to producing a rain of cold clouds that fall toward the galaxy's center and accrete onto its central black hole. Here we explain how relatively weak magnetic fields enhance a galactic atmosphere's tendency to produce cold clouds and how the cold gas becomes organized into vertically elongated, highly magnetized filaments descending at sub-Keplerian speeds. It is intended to complement recent numerical simulations of the process and to serve as a guide to interpreting both simulations and observations of the filamentary gas in hot galactic atmospheres.

Powerful, radio-loud active galactic nuclei (AGN) are associated with one of the most important forms of AGN feedback, and understanding how they are triggered is key to properly incorporating them into models of galaxy evolution. Here, we present the results of a deep Isaac Newton Telescope/Wide Field Camera imaging survey which, when combined with Gemini/Gemini Multi-Object Spectrograph South images, gives a 98 per cent complete sample of 112 3CR radio galaxies with redshifts $z$ < 0.3, alongside a stellar mass matched control sample. Our results provide strong evidence for significant differences ($\sim$3$σ$) between the triggering mechanisms of the different sub-types of powerful radio AGN. The high-excitation radio galaxies (HERGs) show a high rate of morphological disturbance (62$^{+6}_{-7}$ per cent) -- an excess of $\sim$4$σ$ compared with the control sample -- consistent with them being predominantly triggered in galaxy mergers and interactions. In contrast, the low-excitation radio galaxies (LERGs) show a much lower rate of morphological disturbance (36$^{+7}_{-6}$ per cent), consistent with the control sample, and suggesting a different dominant triggering mechanism, such as the accretion of gas from the hot X-ray haloes of the host galaxies or galaxy clusters. We also demonstrate that, when considering the radio morphology, the FRII HERG sources preferentially reside in disturbed morphologies, a difference of $\sim$3$σ$ to the FRII LERG objects. This suggests that the FRII LERG sources do not solely represent a `switched-off' phase in the HERG lifecycle of the same parent galaxy population as the FRII HERGs.

Barium stars are unusually enriched in barium ([Ba/Fe] >= 1.0 dex) and not predicted by current Galactic chemical evolution models. Previous observations of barium stars have found evidence that they form through mass transfer from a companion asymptotic giant branch (AGB) star or through radiative levitation. The chemical abundance and kinematic information of barium stars may help constrain AGB stellar nucleosynthesis, binary star evolution, and internal evolutionary processes that affect surface abundances. Using ~450,000 stars from the GALactic Archaeology with Hermes (GALAH) survey, we identify nearly 3000 new barium-rich stars and separate them into hot (Teff > 6000 K) and cool (Teff < 6000 K) populations. Cross-matching with Gaia DR3, we find that 47.7% of our barium stars within 1 kpc have elevated re-normalized unit weight error (RUWE >= 1.4), compared to 16.3% of a comparable sample of the GALAH field, suggesting multiplicity plays an important role in the formation of both populations of barium stars. A subset of hot barium stars exhibit low RUWE (RUWE < 1.2) and [alpha/Fe] < -0.2, supporting radiative levitation as an origin as well. We determine Galactic memberships using both kinematics and chemistry and find that barium stars exist in the thin disk, thick disk, and halo though they are slightly more prevalent at lower metallicities. Overall, we show evidence for barium stars produced by mass transfer and for those produced by radiative levitation, with both formation mechanisms occurring ubiquitously across the Galaxy.

To date no exoplanets have been detected outside the Milky Way, and their extragalactic occurrence rates are poorly constrained. Using available data from TESS we perform the first transit survey of the Sagittarius dwarf galaxy stream using 15,176 main sequence stars identified as likely members. We calculate an upper limit of $<$1.01% for hot Jupiters with radii of 1-2 R$_{Jup}$ and periods of 0.6-10 days after detecting zero planets. We compare our calculated occurrence rate upper limits to the upper limits found in the Milky Way globular clusters M4 and 47 Tuc. Our 1-$σ$ occurrence rate upper limit of $<$0.37% for the Sagittarius dwarf galaxy stream, for planets with radii of 1.5-2 R$_{Jup}$ and periods $<$10 days, is lower than the $<$0.57% upper limit measured in 47 Tuc. Similarly, our 2 sigma upper limit of $<$0.78% for planets with radii of 1.4-2 $_{Jup}$ and periods $<$8 days is below the $<$0.81% upper limit measured in M4. We predict that a future analysis of TESS data with a high detection efficiency for hot Jupiter transit depths would require $η_{extragalactic}$ $\geq$ 11,467 target stars to detect a planet of extragalactic origin. Therefore, we predict that a future investigation of TESS data which includes additional extragalactic stellar streams will be able to either detect the first extragalactic origin planet or provide evidence that older, lower metallicity extragalactic environments may experience a lower hot Jupiter occurrence rate than is observed for the Milky Way.

The dynamical nature and formation mechanism(s) of galactic spiral arms remain long-standing problems in astrophysics. Most theoretical work is based on analytic calculations or idealised simulations, which has yielded several theories of spiral structure. The radial profile of the spiral arm rotation speed - the pattern speed - is a key observable prediction of these theories. However, observations that infer spiral pattern speeds reveal a mixed picture with no clear consensus. Here, we expand on theoretical efforts by examining the pattern speed profiles in the Auriga Superstars set of high-resolution cosmological magnetohydrodynamic simulatons of Milky Way-mass spiral disc galaxies. These simulations combine galaxy formation in a cosmological environment with the high dynamical fidelity afforded by an $\sim 800$ $\rm M_{\odot}$ star particle resolution, giving $\sim 100$ million star particles in the disc. We show that several different spiral arm theories are realised among our simulations, including large-scale kinematic density waves, manifold spirals, dynamic (co-rotating) spirals, and overlapping modes. In particular, we demonstrate that a strong tidal interaction leads to clear kinematic density waves, and that manifold spirals are present in a strongly-barred galaxy. Interestingly, we find that the same galaxy may show qualitative evolution of their spiral pattern speed profiles, indicating that the nature of spiral arms can evolve on potentially sub-Gigayear timescales. Our results demonstrate that in the absence of a strong external encounter or a strong bar, galactic spiral structure is highly transitional and complex with no clear long-lived underlying wave.

Tidal disruption events (TDEs) are unique probes of evolving accretion in supermassive black holes. Recent models of TDE disks show that they undergo brief thermal instabilities with temporal super-Eddington accretion at late times, which has been suggested as a possibility to explain the ubiquitous late radio emergence in TDEs. We model the ultraviolet (UV) and optical signatures of such disk instabilities, expected from the accretion power being reprocessed by the optically-thick outflow following super-Eddington accretion. Our model predicts brief UV-bright transients lasting for days, with luminosities of $10^{42}$-$10^{43}$ erg s$^{-1}$ in near-UV and $10^{41}$-$10^{42}$ erg s$^{-1}$ in optical for a typical TDE by a $10^6~M_\odot$ black hole. These could be detectable by near-future surveys such as ULTRASAT, Vera C. Rubin Observatory and Argus Array, for TDEs of redshifts out to $\approx 0.1$. We further conduct a search for these transients in existing nearby TDEs using data from the Zwicky Transient Facility, placing upper limits on the flare rate for each TDE of $1$-$2$ yr$^{-1}$ dependent on the outflow mass. In the era of future surveys, combined UV/optical and radio monitoring would be an important test to the disk instability phenomena, as well as its explanation for the late-time radio emission in TDEs.

In this paper we present PSR J0024$-$7204ai, a 13.026-ms binary pulsar recently discovered in the globular cluster 47 Tucanae by the MeerKAT radio telescope. This is the slowest spinning pulsar known in this globular cluster, and has a $\sim1.67$-day orbit with an eccentricity of $e\approx0.18$. Although it was not yet possible to derive an unambiguous phase-connected timing solution, by combining detections obtained from MeerKAT and archival Parkes data we were able to measure the rate of advance of periastron to high significance, $\dotω$ = 0.1601 $\pm 0.0046$ deg yr$^{-1}$. This value implies a total system mass of $2.41 \pm 0.11\, \mathrm{M}_\odot$ (68.3\% C. L.), which, when combined with the binary mass function, gives a maximum pulsar mass of $\sim 1.7 \, \mathrm{M}_\odot$ and a minimum companion mass of $\sim 0.7\, \mathrm{M}_\odot$. Apart from being the slowest pulsar in 47~Tucanae, its orbit is by far the most eccentric and its companion is the most massive among all known binary pulsars in this globular cluster. One possibility is that system is an old MSP - Carbon-Oxygen White Dwarf binary, whose orbit was perturbed by stellar dynamical interactions in the cluster core. Further follow-up observations of this system will be essential for a more detailed characterisation of this system and its evolution.

In galaxy clusters, hierarchical assembly predicts the formation of stellar substructures and intracluster light (ICL), a diffuse stellar component tracing the global cluster potential. Because these features are extremely faint, alternative tracers such as globular clusters (GCs) provide a powerful tool to study cluster assembly. We use deep VLT/FORS $V$- and $I$-band imaging to investigate the GC population in the nearby Hydra I galaxy cluster ($\sim 45.7$ Mpc). GC candidates were selected from the $VI$ colour-magnitude diagram and divided into blue and red subpopulations. We find a clear spatial dichotomy: red GCs are concentrated around the massive central galaxies NGC 3311 and NGC 3309, while blue GCs are more extended and offset from the centre, coinciding with a secondary peak of X-ray-emitting gas. In the central regions, GC spatial distributions further depend on stellar population properties: young metal-rich GCs are more extended and may be linked to ram-pressure stripping, whereas old metal-poor GCs are more centrally concentrated, possibly originating from disrupted dwarf galaxies. Comparing the GC number density profiles to the surface brightness profile of NGC 3311, we find that the red GCs closely follow the galaxy light, while the blue population significantly deviates from it and traces the global gravitational potential of the cluster. This is also reflected in the specific frequency of blue GCs, which is approximately $\sim 5\times$ higher in the ICL-dominated outskirts than in the inner regions dominated by red GCs. Finally, we present a novel method to constrain the evolution of the galaxy luminosity function of the cluster using GC specific frequencies and colour distributions, yielding a past faint-end slope of $α=-1.81^{+0.16}_{-0.16}$ compared to $α=-1.41^{+0.08}_{-0.05}$ today, consistent with high-redshift observations and cosmological simulations.

Posterior inference from pulsar observations in the form of light curves is commonly performed using Markov chain Monte Carlo methods, which are accurate but computationally expensive. We introduce a framework that accelerates posterior inference while maintaining accuracy by combining learned latent representations with local simulator-guided optimization. A masked U-Net is first pretrained to reconstruct complete light curves from partial observations and to produce informative latent embeddings. Given a query light curve, we identify similar simulated light curves from the simulation bank by measuring similarity in the learned embedding space produced by pretrained U-Net encoder, yielding an initial empirical approximation to the posterior over parameters. This initialization is then refined using a local optimization procedure using hill-climbing updates, guided by a forward simulator, progressively shifting the empirical posterior toward higher-likelihood parameter regions. Experiments on the observed light curve of PSR J0030+0451, captured by NASA's Neutron Star Interior Composition Explorer (NICER), show that our method closely matches posterior estimates obtained using traditional MCMC methods while achieving 120 times reduction in inference time (from 24 hours to 12 minutes), demonstrating the effectiveness of learned representations and simulator-guided optimization for accelerated posterior inference.

The James Webb Space Telescope (JWST) has enabled the discovery of hundreds of supermassive black holes (SMBHs) at redshifts $z\gtrsim 4-7$. A non-negligible fraction of these SMBHs are hosted in galaxies with BH-to-galaxy mass ratios ($M_{\rm BH}/M_\star$) being excessively larger than that for local SMBHs by $\sim 1-2$ dex. The origin of these ``overmassive'' BHs remains elusive, demanding either a heavy seed formation scenario or rapid growth of seed BHs. Their deviation from local scaling relations challenges our understanding of how SMBHs and their host galaxies coevolve across cosmic time. In this paper, we apply phenomenological modelings for BHs and galaxies to dark matter halo merger histories from N-body simulations to investigate the subsequent evolution of JWST-discovered ``overmassive'' SMBHs. We find that early evolution of ``overmassive'' SMBHs is dominated by stunted accretion leading to gradual decreases in $M_{\rm BH}/M_\star$ ratios. In contrast, less massive SMBHs experience super-Eddington accretion during their early evolution, resulting in a slow increase of mass ratios toward $M_{\rm BH}/M_\star \sim 0.01$. Convergence occurs at $M_{\rm BH}\sim 10^8~M_\odot$ with $M_{\rm BH}/M_\star \sim 0.01$. At lower redshift, nearly all SMBHs evolve onto local relations, as expected given that our models adopt empirical relations derived from low-redshift observations. This suggests that the global feedback mechanisms regulating the coevolution of $M_{\rm BH}/M_\star$ ratios are implicitly encoded in local relations in terms of star-formation rate distribution, black hole accretion rate distribution and their active (quiescent) fractions.

PS1-12sk is a type Ibn supernova (SN) found at the host environment showing no obvious ongoing star formation, which challenges the massive star explosion scenario. We use the ejecta-circumstellar medium (CSM) interaction (CSI) and the CSI plus $^{56}$Ni models in the context of double white dwarf (WD) merger to fit the bolometric light curve (LC) of PS1-12sk, since the He emission lines at the photospheric phases indicated the interaction between the SN ejecta and He-rich CSM. We find that the CSI model failed to explain the LC, but the CSI plus $^{56}$Ni model can account for the bolometric LC. The derived masses of the two WDs and $^{56}$Ni are $\sim 0.70 M_\odot$, $\sim 0.40 M_\odot$, and $\sim 0.09\,M_\odot$, respectively. The facts that the ejecta mass ($\sim 0.984 M_\odot$) is well below the Chandrasekhar limit ($\sim 1.4 M_\odot$) and that the $^{56}$Ni mass is comparable to the $^{56}$Ni yields of the explosions of some sub-Chandrasekhar explosion models support the scenario that PS1-12sk might be from a sub-Chandrasekhar explosion induced by the merger of two low-mass WDs. The derived innermost radius ($\sim 13.81 \times 10^{12}$ cm) and the mass of the CSM ($\sim 0.116 M_\odot$) disfavor the possibility that the CSM was formed in the merger phase. We suggest that the flybys before the merger can account for the position and mass of the CSM.

By identifying quasi-periodic oscillations (QPOs) reported in FRB 20240114A (from the Five-hundred-meter Aperture Spherical Telescope) with neutron star crustal torsional oscillations, together with experimental constraints on the incompressibility $K_0$ of symmetric nuclear matter at saturation density, we constrain the mass and radius of an extragalactic neutron star at redshift $z\approx0.13$. Identifying the low-order QPO frequencies as fundamental oscillations, and frequencies of $567.7\,\mathrm{Hz}$ or $655.5\,\mathrm{Hz}$ (rest frame) as first overtone candidates, implies neutron star mass ranges of $1.00$--$1.55\,M_\odot$ or $1.17$--$1.76\,M_\odot$, respectively. The radius is also constrained, with a self-consistent value around $13$~km, further supported by the calculation of the NS structure within the low-mass/low-central density regime. Simultaneously, we also constrain another nuclear saturation parameter, namely the density dependence of the nuclear symmetry energy at saturation density (i.e., the slope parameter), $L$, and determine it to be $L=59.5-96.8$ MeV with $\sim 10\%$ systematic uncertainty, which is broadly consistent with previous constraints on $L$ obtained from experiments and astronomical observations. Thus, a mapping of FRB QPOs to crustal torsional modes seems reasonable. This can be confirmed with upcoming FRB surveys over a broad range of redshifts and more elaborate data analyses.