arXiv Daily Digest - 2026-02-19

Latent space models are widely used for analyzing high-dimensional discrete data matrices, such as patient-feature matrices in electronic health records (EHRs), by capturing complex dependence structures through low-dimensional embeddings. However, estimation becomes challenging in the imbalanced regime, where one matrix dimension is much larger than the other. In EHR applications, cohort sizes are often limited by disease prevalence or data availability, whereas the feature space remains extremely large due to the breadth of medical coding system. Motivated by the increasing availability of external semantic embeddings, such as pre-trained embeddings of clinical concepts in EHRs, we propose a knowledge-embedded latent projection model that leverages semantic side information to regularize representation learning. Specifically, we model column embeddings as smooth functions of semantic embeddings via a mapping in a reproducing kernel Hilbert space. We develop a computationally efficient two-step estimation procedure that combines semantically guided subspace construction via kernel principal component analysis with scalable projected gradient descent. We establish estimation error bounds that characterize the trade-off between statistical error and approximation error induced by the kernel projection. Furthermore, we provide local convergence guarantees for our non-convex optimization procedure. Extensive simulation studies and a real-world EHR application demonstrate the effectiveness of the proposed method.

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

Fast weight architectures offer a promising alternative to attention-based transformers for long-context modeling by maintaining constant memory overhead regardless of context length. However, their potential is limited by the next-token prediction (NTP) training paradigm. NTP optimizes single-token predictions and ignores semantic coherence across multiple tokens following a prefix. Consequently, fast weight models, which dynamically update their parameters to store contextual information, learn suboptimal representations that fail to capture long-range dependencies. We introduce REFINE (Reinforced Fast weIghts with Next sEquence prediction), a reinforcement learning framework that trains fast weight models under the next-sequence prediction (NSP) objective. REFINE selects informative token positions based on prediction entropy, generates multi-token rollouts, assigns self-supervised sequence-level rewards, and optimizes the model with group relative policy optimization (GRPO). REFINE is applicable throughout the training lifecycle of pre-trained language models: mid-training, post-training, and test-time training. Our experiments on LaCT-760M and DeltaNet-1.3B demonstrate that REFINE consistently outperforms supervised fine-tuning with NTP across needle-in-a-haystack retrieval, long-context question answering, and diverse tasks in LongBench. REFINE provides an effective and versatile framework for improving long-context modeling in fast weight architectures.

Large language models (LLMs) perform strongly on biological benchmarks, raising concerns that they may help novice actors acquire dual-use laboratory skills. Yet, whether this translates to improved human performance in the physical laboratory remains unclear. To address this, we conducted a pre-registered, investigator-blinded, randomized controlled trial (June-August 2025; n = 153) evaluating whether LLMs improve novice performance in tasks that collectively model a viral reverse genetics workflow. We observed no significant difference in the primary endpoint of workflow completion (5.2% LLM vs. 6.6% Internet; P = 0.759), nor in the success rate of individual tasks. However, the LLM arm had numerically higher success rates in four of the five tasks, most notably for the cell culture task (68.8% LLM vs. 55.3% Internet; P = 0.059). Post-hoc Bayesian modeling of pooled data estimates an approximate 1.4-fold increase (95% CrI 0.74-2.62) in success for a "typical" reverse genetics task under LLM assistance. Ordinal regression modelling suggests that participants in the LLM arm were more likely to progress through intermediate steps across all tasks (posterior probability of a positive effect: 81%-96%). Overall, mid-2025 LLMs did not substantially increase novice completion of complex laboratory procedures but were associated with a modest performance benefit. These results reveal a gap between in silico benchmarks and real-world utility, underscoring the need for physical-world validation of AI biosecurity assessments as model capabilities and user proficiency evolve.

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

Interpretability research on large language models (LLMs) has yielded important insights into model behaviour, yet recurring pitfalls persist: findings that do not generalise, and causal interpretations that outrun the evidence. Our position is that causal inference specifies what constitutes a valid mapping from model activations to invariant high-level structures, the data or assumptions needed to achieve it, and the inferences it can support. Specifically, Pearl's causal hierarchy clarifies what an interpretability study can justify. Observations establish associations between model behaviour and internal components. Interventions (e.g., ablations or activation patching) support claims how these edits affect a behavioural metric (\eg, average change in token probabilities) over a set of prompts. However, counterfactual claims -- i.e., asking what the model output would have been for the same prompt under an unobserved intervention -- remain largely unverifiable without controlled supervision. We show how causal representation learning (CRL) operationalises this hierarchy, specifying which variables are recoverable from activations and under what assumptions. Together, these motivate a diagnostic framework that helps practitioners select methods and evaluations matching claims to evidence such that findings generalise.

Machine unlearning aims to remove specific data points from a trained model, often striving to emulate "perfect retraining", i.e., producing the model that would have been obtained had the deleted data never been included. We demonstrate that this approach, and security definitions that enable it, carry significant privacy risks for the remaining (undeleted) data points. We present a reconstruction attack showing that for certain tasks, which can be computed securely without deletions, a mechanism adhering to perfect retraining allows an adversary controlling merely $ω(1)$ data points to reconstruct almost the entire dataset merely by issuing deletion requests. We survey existing definitions for machine unlearning, showing they are either susceptible to such attacks or too restrictive to support basic functionalities like exact summation. To address this problem, we propose a new security definition that specifically safeguards undeleted data against leakage caused by the deletion of other points. We show that our definition permits several essential functionalities, such as bulletin boards, summations, and statistical learning.

Single-cell RNA sequencing (scRNA-seq) data exhibit strong and reproducible statistical structure. This has motivated the development of large-scale foundation models, such as TranscriptFormer, that use transformer-based architectures to learn a generative model for gene expression by embedding genes into a latent vector space. These embeddings have been used to obtain state-of-the-art (SOTA) performance on downstream tasks such as cell-type classification, disease-state prediction, and cross-species learning. Here, we ask whether similar performance can be achieved without utilizing computationally intensive deep learning-based representations. Using simple, interpretable pipelines that rely on careful normalization and linear methods, we obtain SOTA or near SOTA performance across multiple benchmarks commonly used to evaluate single-cell foundation models, including outperforming foundation models on out-of-distribution tasks involving novel cell types and organisms absent from the training data. Our findings highlight the need for rigorous benchmarking and suggest that the biology of cell identity can be captured by simple linear representations of single cell gene expression data.

The digital services economy consists of online platforms that facilitate interactions between service providers and consumers. This ecosystem is characterized by short-term, often one-off, transactions between parties that have no prior familiarity. To establish trust among users, platforms employ rating systems which allow users to report on the quality of their previous interactions. However, while arguably crucial for these platforms to function, rating systems can perpetuate negative biases against marginalised groups. This paper investigates how to design platforms around biased reputation systems, reducing discrimination while maintaining incentives for all service providers to offer high quality service for users. We introduce an evolutionary game theoretical model to study how digital platforms can perpetuate or counteract rating-based discrimination. We focus on the platforms' decisions to promote service providers who have high reputations or who belong to a specific protected group. Our results demonstrate a fundamental trade-off between user experience and fairness: promoting highly-rated providers benefits users, but lowers the demand for marginalised providers against which the ratings are biased. Our results also provide evidence that intervening by tuning the demographics of the search results is a highly effective way of reducing unfairness while minimally impacting users. Furthermore, we show that even when precise measurements on the level of rating bias affecting marginalised service providers is unavailable, there is still potential to improve upon a recommender system which ignores protected characteristics. Altogether, our model highlights the benefits of proactive anti-discrimination design in systems where ratings are used to promote cooperative behaviour.

Multiple hypothesis testing with false discovery rate (FDR) control is a fundamental problem in statistical inference, with broad applications in genomics, drug screening, and outlier detection. In many such settings, researchers may have access not only to real experimental observations but also to auxiliary or synthetic data -- from past, related experiments or generated by generative models -- that can provide additional evidence about the hypotheses of interest. We introduce SynthBH, a synthetic-powered multiple testing procedure that safely leverages such synthetic data. We prove that SynthBH guarantees finite-sample, distribution-free FDR control under a mild PRDS-type positive dependence condition, without requiring the pooled-data p-values to be valid under the null. The proposed method adapts to the (unknown) quality of the synthetic data: it enhances the sample efficiency and may boost the power when synthetic data are of high quality, while controlling the FDR at a user-specified level regardless of their quality. We demonstrate the empirical performance of SynthBH on tabular outlier detection benchmarks and on genomic analyses of drug-cancer sensitivity associations, and further study its properties through controlled experiments on simulated data.

Compositional generalization, the ability to reason about novel combinations of familiar concepts, is fundamental to human cognition and a critical challenge for machine learning. Object-centric (OC) representations, which encode a scene as a set of objects, are often argued to support such generalization, but systematic evidence in visually rich settings is limited. We introduce a Visual Question Answering benchmark across three controlled visual worlds (CLEVRTex, Super-CLEVR, and MOVi-C) to measure how well vision encoders, with and without object-centric biases, generalize to unseen combinations of object properties. To ensure a fair and comprehensive comparison, we carefully account for training data diversity, sample size, representation size, downstream model capacity, and compute. We use DINOv2 and SigLIP2, two widely used vision encoders, as the foundation models and their OC counterparts. Our key findings reveal that (1) OC approaches are superior in harder compositional generalization settings; (2) original dense representations surpass OC only on easier settings and typically require substantially more downstream compute; and (3) OC models are more sample efficient, achieving stronger generalization with fewer images, whereas dense encoders catch up or surpass them only with sufficient data and diversity. Overall, object-centric representations offer stronger compositional generalization when any one of dataset size, training data diversity, or downstream compute is constrained.

In this work, we study the hardness of approximation of the fair $k$-center problem. Here the data points are partitioned into groups and the task is to choose a prescribed number of data points from each group, called centers, while minimizing the maximum distance from any point to its closest center. Although a polynomial-time $3$-approximation is known for this problem in general metrics, it has remained open whether this approximation guarantee is tight or could be further improved, especially since the unconstrained $k$-center problem admits a polynomial-time factor-$2$ approximation. We resolve this open question by proving that, for every $ε>0$, achieving a $(3-ε)$-approximation is NP-hard, assuming $\text{P} \neq \text{NP}$. Our inapproximability results hold even when only two disjoint groups are present and at least one center must be chosen from each group. Further, it extends to the canonical one-per-group setting with $k$-groups (for arbitrary $k$), where exactly one center must be selected from each group. Consequently, the factor-$3$ barrier for fair $k$-center in general metric spaces is inherent, and existing $3$-approximation algorithms are optimal up to lower-order terms even in these restricted regimes. This result stands in sharp contrast to the $k$-supplier formulation, where both the unconstrained and fair variants admit factor-$3$ approximation in polynomial time.

Current audio language models are predominantly text-first, either extending pre-trained text LLM backbones or relying on semantic-only audio tokens, limiting general audio modeling. This paper presents a systematic empirical study of native audio foundation models that apply next-token prediction to audio at scale, jointly modeling semantic content, acoustic details, and text to support both general audio generation and cross-modal capabilities. We provide comprehensive empirical insights for building such models: (1) We systematically investigate design choices -- data sources, text mixture ratios, and token composition -- establishing a validated training recipe. (2) We conduct the first scaling law study for discrete audio models via IsoFLOP analysis on 64 models spanning $3{\times}10^{18}$ to $3{\times}10^{20}$ FLOPs, finding that optimal data grows 1.6$\times$ faster than optimal model size. (3) We apply these lessons to train SODA (Scaling Open Discrete Audio), a suite of models from 135M to 4B parameters on 500B tokens, comparing against our scaling predictions and existing models. SODA serves as a flexible backbone for diverse audio/text tasks -- we demonstrate this by fine-tuning for voice-preserving speech-to-speech translation, using the same unified architecture.

Matched molecular pairs (MMPs) capture the local chemical edits that medicinal chemists routinely use to design analogs, but existing ML approaches either operate at the whole-molecule level with limited edit controllability or learn MMP-style edits from restricted settings and small models. We propose a variable-to-variable formulation of analog generation and train a foundation model on large-scale MMP transformations (MMPTs) to generate diverse variables conditioned on an input variable. To enable practical control, we develop prompting mechanisms that let the users specify preferred transformation patterns during generation. We further introduce MMPT-RAG, a retrieval-augmented framework that uses external reference analogs as contextual guidance to steer generation and generalize from project-specific series. Experiments on general chemical corpora and patent-specific datasets demonstrate improved diversity, novelty, and controllability, and show that our method recovers realistic analog structures in practical discovery scenarios.

In order for close proximity satellites to safely perform their missions, the relative states of all satellites and pieces of debris must be well understood. This presents a problem for ground based tracking and orbit determination since it may not be practical to achieve the required accuracy. Using space-based sensors allows for more accurate relative state estimates, especially if multiple satellites are allowed to communicate. Of interest to this work is the case where several communicating satellites each need to maintain a local catalog of communicating and non-communicating objects using angles-only limited field of view (FOV) measurements. However, this introduces the problem of efficiently scheduling and coordinating observations among the agents. This paper presents a decentralized task allocation algorithm to address this problem and quantifies its performance in terms of fuel usage and overall catalog uncertainty via numerical simulation. It was found that the new method significantly outperforms the uncertainty-fuel Pareto frontier formed by current approaches.

Clustering-based Approximate Nearest Neighbor Search (ANNS) organizes a set of points into partitions, and searches only a few of them to find the nearest neighbors of a query. Despite its popularity, there are virtually no analytical tools to determine the suitability of clustering-based ANNS for a given dataset -- what we call "searchability." To address that gap, we present two measures for flat clusterings of high-dimensional points in Euclidean space. First is Clustering-Neighborhood Stability Measure (clustering-NSM), an internal measure of clustering quality -- a function of a clustering of a dataset -- that we show to be predictive of ANNS accuracy. The second, Point-Neighborhood Stability Measure (point-NSM), is a measure of clusterability -- a function of the dataset itself -- that is predictive of clustering-NSM. The two together allow us to determine whether a dataset is searchable by clustering-based ANNS given only the data points. Importantly, both are functions of nearest neighbor relationships between points, not distances, making them applicable to various distance functions including inner product.

Automated unit test generation for C remains a formidable challenge due to the semantic gap between high-level program intent and the rigid syntactic constraints of pointer arithmetic and manual memory management. While Large Language Models (LLMs) exhibit strong generative capabilities, direct intent-to-code synthesis frequently suffers from the leap-to-code failure mode, where models prematurely emit code without grounding in program structure, constraints, and semantics. This will result in non-compilable tests, hallucinated function signatures, low branch coverage, and semantically irrelevant assertions that cannot properly capture bugs. We introduce SPARC, a neuro-symbolic, scenario-based framework that bridges this gap through four stages: (1) Control Flow Graph (CFG) analysis, (2) an Operation Map that grounds LLM reasoning in validated utility helpers, (3) Path-targeted test synthesis, and (4) an iterative, self-correction validation loop using compiler and runtime feedback. We evaluate SPARC on 59 real-world and algorithmic subjects, where it outperforms the vanilla prompt generation baseline by 31.36% in line coverage, 26.01% in branch coverage, and 20.78% in mutation score, matching or exceeding the symbolic execution tool KLEE on complex subjects. SPARC retains 94.3% of tests through iterative repair and produces code with significantly higher developer-rated readability and maintainability. By aligning LLM reasoning with program structure, SPARC provides a scalable path for industrial-grade testing of legacy C codebases.

AI agents are increasingly deployed to execute important tasks. While rising accuracy scores on standard benchmarks suggest rapid progress, many agents still continue to fail in practice. This discrepancy highlights a fundamental limitation of current evaluations: compressing agent behavior into a single success metric obscures critical operational flaws. Notably, it ignores whether agents behave consistently across runs, withstand perturbations, fail predictably, or have bounded error severity. Grounded in safety-critical engineering, we provide a holistic performance profile by proposing twelve concrete metrics that decompose agent reliability along four key dimensions: consistency, robustness, predictability, and safety. Evaluating 14 agentic models across two complementary benchmarks, we find that recent capability gains have only yielded small improvements in reliability. By exposing these persistent limitations, our metrics complement traditional evaluations while offering tools for reasoning about how agents perform, degrade, and fail.

As autonomous agents powered by LLM are increasingly deployed in society, understanding their collective behaviour in social dilemmas becomes critical. We introduce an evaluation framework where LLMs generate strategies encoded as algorithms, enabling inspection prior to deployment and scaling to populations of hundreds of agents -- substantially larger than in previous work. We find that more recent models tend to produce worse societal outcomes compared to older models when agents prioritise individual gain over collective benefits. Using cultural evolution to model user selection of agents, our simulations reveal a significant risk of convergence to poor societal equilibria, particularly when the relative benefit of cooperation diminishes and population sizes increase. We release our code as an evaluation suite for developers to assess the emergent collective behaviour of their models.

The widespread deployment of large language models (LLMs) across linguistic communities necessitates reliable multilingual safety alignment. However, recent efforts to extend alignment to other languages often require substantial resources, either through large-scale, high-quality supervision in the target language or through pairwise alignment with high-resource languages, which limits scalability. In this work, we propose a resource-efficient method for improving multilingual safety alignment. We introduce a plug-and-play Multi-Lingual Consistency (MLC) loss that can be integrated into existing monolingual alignment pipelines. By improving collinearity between multilingual representation vectors, our method encourages directional consistency at the multilingual semantic level in a single update. This allows simultaneous alignment across multiple languages using only multilingual prompt variants without requiring additional response-level supervision in low-resource languages. We validate the proposed method across different model architectures and alignment paradigms, and demonstrate its effectiveness in enhancing multilingual safety with limited impact on general model utility. Further evaluation across languages and tasks indicates improved cross-lingual generalization, suggesting the proposed approach as a practical solution for multilingual consistency alignment under limited supervision.

The solar dynamo relies on the regeneration of the poloidal magnetic field through processes strongly modulated by nonlinear feedbacks such as tilt quenching (TQ) and latitude quenching (LQ). These mechanisms play a decisive role in regulating the buildup of the Sun's polar field and, in turn, the amplitude of future solar cycles. In this work, we employ Physics-Informed Neural Networks (PINN) to solve the surface flux transport (SFT) equation, embedding physical constraints directly into the neural network framework. By systematically varying transport parameters, we isolate the relative contributions of TQ and LQ to polar dipole buildup. We use the residual dipole moment as a diagnostic for cycle-to-cycle amplification and show that TQ suppression strengthens with increasing diffusivity, while LQ dominates in advection-dominated regimes. The ratio $ΔD_{\mathrm{LQ}}/ΔD_{\mathrm{TQ}}$ exhibits a smooth inverse-square dependence on the dynamo effectivity range, refining previous empirical fits with improved accuracy and reduced scatter. The results further reveal that the need for a decay term is not essential for PINN set-up due to the training process. Compared with the traditional 1D SFT model, the PINN framework achieves significantly lower error metrics and more robust recovery of nonlinear trends. Our results suggest that the nonlinear interplay between LQ and TQ can naturally produce alternations between weak and strong cycles, providing a physical explanation for the observed even-odd cycle modulation. These findings demonstrate the potential of PINN as an accurate, efficient, and physically consistent tool for solar cycle prediction.

Agent Skill framework, now widely and officially supported by major players such as GitHub Copilot, LangChain, and OpenAI, performs especially well with proprietary models by improving context engineering, reducing hallucinations, and boosting task accuracy. Based on these observations, an investigation is conducted to determine whether the Agent Skill paradigm provides similar benefits to small language models (SLMs). This question matters in industrial scenarios where continuous reliance on public APIs is infeasible due to data-security and budget constraints requirements, and where SLMs often show limited generalization in highly customized scenarios. This work introduces a formal mathematical definition of the Agent Skill process, followed by a systematic evaluation of language models of varying sizes across multiple use cases. The evaluation encompasses two open-source tasks and a real-world insurance claims data set. The results show that tiny models struggle with reliable skill selection, while moderately sized SLMs (approximately 12B - 30B) parameters) benefit substantially from the Agent Skill approach. Moreover, code-specialized variants at around 80B parameters achieve performance comparable to closed-source baselines while improving GPU efficiency. Collectively, these findings provide a comprehensive and nuanced characterization of the capabilities and constraints of the framework, while providing actionable insights for the effective deployment of Agent Skills in SLM-centered environments.

Polymer literature contains a large and growing body of experimental knowledge, yet much of it is buried in unstructured text and inconsistent terminology, making systematic retrieval and reasoning difficult. Existing tools typically extract narrow, study-specific facts in isolation, failing to preserve the cross-study context required to answer broader scientific questions. Retrieval-augmented generation (RAG) offers a promising way to overcome this limitation by combining large language models (LLMs) with external retrieval, but its effectiveness depends strongly on how domain knowledge is represented. In this work, we develop two retrieval pipelines: a dense semantic vector-based approach (VectorRAG) and a graph-based approach (GraphRAG). Using over 1,000 polyhydroxyalkanoate (PHA) papers, we construct context-preserving paragraph embeddings and a canonicalized structured knowledge graph supporting entity disambiguation and multi-hop reasoning. We evaluate these pipelines through standard retrieval metrics, comparisons with general state-of-the-art systems such as GPT and Gemini, and qualitative validation by a domain chemist. The results show that GraphRAG achieves higher precision and interpretability, while VectorRAG provides broader recall, highlighting complementary trade-offs. Expert validation further confirms that the tailored pipelines, particularly GraphRAG, produce well-grounded, citation-reliable responses with strong domain relevance. By grounding every statement in evidence, these systems enable researchers to navigate the literature, compare findings across studies, and uncover patterns that are difficult to extract manually. More broadly, this work establishes a practical framework for building materials science assistants using curated corpora and retrieval design, reducing reliance on proprietary models while enabling trustworthy literature analysis at scale.

The RNA inverse folding problem aims to identify nucleotide sequences that preferentially adopt a given target secondary structure. While various heuristic and machine learning-based approaches have been proposed, many require a large number of sequence evaluations, which limits their applicability when experimental validation is costly. We propose a method to solve the problem using a factorization machine with quadratic-optimization annealing (FMQA). FMQA is a discrete black-box optimization method reported to obtain high-quality solutions with a limited number of evaluations. Applying FMQA to the problem requires converting nucleotides into binary variables. However, the influence of integer-to-nucleotide assignments and binary-integer encoding on the performance of FMQA has not been thoroughly investigated, even though such choices determine the structure of the surrogate model and the search landscape, and thus can directly affect solution quality. Therefore, this study aims both to establish a novel FMQA framework for RNA inverse folding and to analyze the effects of these assignments and encoding methods. We evaluated all 24 possible assignments of the four nucleotides to the ordered integers (0-3), in combination with four binary-integer encoding methods. Our results demonstrated that one-hot and domain-wall encodings outperform binary and unary encodings in terms of the normalized ensemble defect value. In domain-wall encoding, nucleotides assigned to the boundary integers (0 and 3) appeared with higher frequency. In the RNA inverse folding problem, assigning guanine and cytosine to these boundary integers promoted their enrichment in stem regions, which led to more thermodynamically stable secondary structures than those obtained with one-hot encoding.

Neural Collapse (NC) refers to the emergence of highly symmetric geometric structures in the representations of deep neural networks during the terminal phase of training. Despite its prevalence, the theoretical understanding of NC remains limited. Existing analyses largely ignore the role of the optimizer, thereby suggesting that NC is universal across optimization methods. In this work, we challenge this assumption and demonstrate that the choice of optimizer plays a critical role in the emergence of NC. The phenomenon is typically quantified through NC metrics, which, however, are difficult to track and analyze theoretically. To overcome this limitation, we introduce a novel diagnostic metric, NC0, whose convergence to zero is a necessary condition for NC. Using NC0, we provide theoretical evidence that NC cannot emerge under decoupled weight decay in adaptive optimizers, as implemented in AdamW. Concretely, we prove that SGD, SignGD with coupled weight decay (a special case of Adam), and SignGD with decoupled weight decay (a special case of AdamW) exhibit qualitatively different NC0 dynamics. Also, we show the accelerating effect of momentum on NC (beyond convergence of train loss) when trained with SGD, being the first result concerning momentum in the context of NC. Finally, we conduct extensive empirical experiments consisting of 3,900 training runs across various datasets, architectures, optimizers, and hyperparameters, confirming our theoretical results. This work provides the first theoretical explanation for optimizer-dependent emergence of NC and highlights the overlooked role of weight-decay coupling in shaping the implicit biases of optimizers.

The rapid proliferation of Large Language Models (LLMs) has revolutionized Natural Language Processing (NLP) but has simultaneously created a "resource divide." State-of-the-art legal intelligence systems typically rely on massive parameter counts (7B+) and cloud-based inference, rendering them inaccessible to practitioners in resource-constrained environments and posing significant data sovereignty risks. This paper introduces Quecto-V1, a domain-specific Small Language Model (SLM) engineered to democratize access to Indian legal intelligence. Built upon a custom configuration of the GPT-2 architecture (124 million parameters), Quecto-V1 was trained from scratch exclusively on a corpus of Indian statutes, including the Indian Penal Code (IPC), the Code of Criminal Procedure (CrPC), and the Constitution of India. Unlike generalist models, which prioritize broad world knowledge, our approach maximizes "lexical density" within the legal domain. Furthermore, we address the deployment bottleneck by applying post-training 8-bit quantization (GGUF format), compressing the model to a memory footprint of under 150 MB. Our empirical analysis demonstrates that Quecto-V1 achieves high fidelity in retrieving statutory definitions and penal provisions, outperforming general-purpose SLMs in domain-specific exact match tasks while running entirely offline on consumer-grade CPUs. We further present an ablation study showing that 8-bit quantization yields a 74% reduction in model size with less than 3.5% degradation in retrieval accuracy compared to full-precision baselines. These findings suggest that for specialized, high-stakes domains like law, domain-specific training coupled with aggressive quantization offers a viable, privacy-preserving alternative to monolithic cloud models.

Evaluating the social intelligence of Large Language Models (LLMs) increasingly requires moving beyond static text generation toward dynamic, adversarial interaction. We introduce the Adversarial Resource Extraction Game (AREG), a benchmark that operationalizes persuasion and resistance as a multi-turn, zero-sum negotiation over financial resources. Using a round-robin tournament across frontier models, AREG enables joint evaluation of offensive (persuasion) and defensive (resistance) capabilities within a single interactional framework. Our analysis provides evidence that these capabilities are weakly correlated ($ρ= 0.33$) and empirically dissociated: strong persuasive performance does not reliably predict strong resistance, and vice versa. Across all evaluated models, resistance scores exceed persuasion scores, indicating a systematic defensive advantage in adversarial dialogue settings. Further linguistic analysis suggests that interaction structure plays a central role in these outcomes. Incremental commitment-seeking strategies are associated with higher extraction success, while verification-seeking responses are more prevalent in successful defenses than explicit refusal. Together, these findings indicate that social influence in LLMs is not a monolithic capability and that evaluation frameworks focusing on persuasion alone may overlook asymmetric behavioral vulnerabilities.

The rare-event sampling problem has long been the central limiting factor in molecular dynamics (MD), especially in biomolecular simulation. Recently, diffusion models such as BioEmu have emerged as powerful equilibrium samplers that generate independent samples from complex molecular distributions, eliminating the cost of sampling rare transition events. However, a sampling problem remains when computing observables that rely on states which are rare in equilibrium, for example folding free energies. Here, we introduce enhanced diffusion sampling, enabling efficient exploration of rare-event regions while preserving unbiased thermodynamic estimators. The key idea is to perform quantitatively accurate steering protocols to generate biased ensembles and subsequently recover equilibrium statistics via exact reweighting. We instantiate our framework in three algorithms: UmbrellaDiff (umbrella sampling with diffusion models), $Δ$G-Diff (free-energy differences via tilted ensembles), and MetaDiff (a batchwise analogue for metadynamics). Across toy systems, protein folding landscapes and folding free energies, our methods achieve fast, accurate, and scalable estimation of equilibrium properties within GPU-minutes to hours per system -- closing the rare-event sampling gap that remained after the advent of diffusion-model equilibrium samplers.

The average reward is a fundamental performance metric in reinforcement learning (RL) focusing on the long-run performance of an agent. Differential temporal difference (TD) learning algorithms are a major advance for average reward RL as they provide an efficient online method to learn the value functions associated with the average reward in both on-policy and off-policy settings. However, existing convergence guarantees require a local clock in learning rates tied to state visit counts, which practitioners do not use and does not extend beyond tabular settings. We address this limitation by proving the almost sure convergence of on-policy $n$-step differential TD for any $n$ using standard diminishing learning rates without a local clock. We then derive three sufficient conditions under which off-policy $n$-step differential TD also converges without a local clock. These results strengthen the theoretical foundations of differential TD and bring its convergence analysis closer to practical implementations.

Recent success in natural language processing has motivated growing interest in large-scale foundation models for neuroimaging data. Such models often require discretization of continuous neural time series data, a process referred to as 'tokenization'. However, the impact of different tokenization strategies for neural data is currently poorly understood. In this work, we present a systematic evaluation of sample-level tokenization strategies for transformer-based large neuroimaging models (LNMs) applied to magnetoencephalography (MEG) data. We compare learnable and non-learnable tokenizers by examining their signal reconstruction fidelity and their impact on subsequent foundation modeling performance (token prediction, biological plausibility of generated data, preservation of subject-specific information, and performance on downstream tasks). For the learnable tokenizer, we introduce a novel approach based on an autoencoder. Experiments were conducted on three publicly available MEG datasets spanning different acquisition sites, scanners, and experimental paradigms. Our results show that both learnable and non-learnable discretization schemes achieve high reconstruction accuracy and broadly comparable performance across most evaluation criteria, suggesting that simple fixed sample-level tokenization strategies can be used in the development of neural foundation models. The code is available at https://github.com/OHBA-analysis/Cho2026_Tokenizer.

Abstracting from a low level to a more explanatory high level of description, and ideally while preserving causal structure, is fundamental to scientific practice, to causal inference problems, and to robust, efficient and interpretable AI. We present a general account of abstractions between low and high level models as natural transformations, focusing on the case of causal models. This provides a new formalisation of causal abstraction, unifying several notions in the literature, including constructive causal abstraction, Q-$τ$ consistency, abstractions based on interchange interventions, and `distributed' causal abstractions. Our approach is formalised in terms of category theory, and uses the general notion of a compositional model with a given set of queries and semantics in a monoidal, cd- or Markov category; causal models and their queries such as interventions being special cases. We identify two basic notions of abstraction: downward abstractions mapping queries from high to low level; and upward abstractions, mapping concrete queries such as Do-interventions from low to high. Although usually presented as the latter, we show how common causal abstractions may, more fundamentally, be understood in terms of the former. Our approach also leads us to consider a new stronger notion of `component-level' abstraction, applying to the individual components of a model. In particular, this yields a novel, strengthened form of constructive causal abstraction at the mechanism-level, for which we prove characterisation results. Finally, we show that abstraction can be generalised to further compositional models, including those with a quantum semantics implemented by quantum circuits, and we take first steps in exploring abstractions between quantum compositional circuit models and high-level classical causal models as a means to explainable quantum AI.

Large language models (LLMs) are increasingly applied as automatic evaluators for natural language generation assessment often using pairwise comparative judgements. Existing approaches typically rely on single judges or aggregate multiple judges assuming equal reliability. In practice, LLM judges vary substantially in performance across tasks and aspects, and their judgment probabilities may be biased and inconsistent. Furthermore, human-labelled supervision for judge calibration may be unavailable. We first empirically demonstrate that inconsistencies in LLM comparison probabilities exist and show that it limits the effectiveness of direct probability-based ranking. To address this, we study the LLM-as-a-jury setting and propose BT-sigma, a judge-aware extension of the Bradley-Terry model that introduces a discriminator parameter for each judge to jointly infer item rankings and judge reliability from pairwise comparisons alone. Experiments on benchmark NLG evaluation datasets show that BT-sigma consistently outperforms averaging-based aggregation methods, and that the learned discriminator strongly correlates with independent measures of the cycle consistency of LLM judgments. Further analysis reveals that BT-sigma can be interpreted as an unsupervised calibration mechanism that improves aggregation by modelling judge reliability.

Current state-of-the-art multi-vector models are obtained through a small Knowledge Distillation (KD) training step on top of strong single-vector models, leveraging the large-scale pre-training of these models. In this paper, we study the pre-training of multi-vector models and show that large-scale multi-vector pre-training yields much stronger multi-vector models. Notably, a fully ColBERT-pre-trained model, ColBERT-Zero, trained only on public data, outperforms GTE-ModernColBERT as well as its base model, GTE-ModernBERT, which leverages closed and much stronger data, setting new state-of-the-art for model this size. We also find that, although performing only a small KD step is not enough to achieve results close to full pre-training, adding a supervised step beforehand allows to achieve much closer performance while skipping the most costly unsupervised phase. Finally, we find that aligning the fine-tuning and pre-training setups is crucial when repurposing existing models. To enable exploration of our results, we release various checkpoints as well as code used to train them.

Transformer models achieve state-of-the-art performance across domains and tasks, yet their deeply layered representations make their predictions difficult to interpret. Existing explainability methods rely on final-layer attributions, capture either local token-level attributions or global attention patterns without unification, and lack context-awareness of inter-token dependencies and structural components. They also fail to capture how relevance evolves across layers and how structural components shape decision-making. To address these limitations, we proposed the \textbf{Context-Aware Layer-wise Integrated Gradients (CA-LIG) Framework}, a unified hierarchical attribution framework that computes layer-wise Integrated Gradients within each Transformer block and fuses these token-level attributions with class-specific attention gradients. This integration yields signed, context-sensitive attribution maps that capture supportive and opposing evidence while tracing the hierarchical flow of relevance through the Transformer layers. We evaluate the CA-LIG Framework across diverse tasks, domains, and transformer model families, including sentiment analysis and long and multi-class document classification with BERT, hate speech detection in a low-resource language setting with XLM-R and AfroLM, and image classification with Masked Autoencoder vision Transformer model. Across all tasks and architectures, CA-LIG provides more faithful attributions, shows stronger sensitivity to contextual dependencies, and produces clearer, more semantically coherent visualizations than established explainability methods. These results indicate that CA-LIG provides a more comprehensive, context-aware, and reliable explanation of Transformer decision-making, advancing both the practical interpretability and conceptual understanding of deep neural models.

Municipal meeting minutes are formal records documenting the discussions and decisions of local government, yet their content is often lengthy, dense, and difficult for citizens to navigate. Automatic summarization can help address this challenge by producing concise summaries for each discussion subject. Despite its potential, research on summarizing discussion subjects in municipal meeting minutes remains largely unexplored, especially in low-resource languages, where the inherent complexity of these documents adds further challenges. A major bottleneck is the scarcity of datasets containing high-quality, manually crafted summaries, which limits the development and evaluation of effective summarization models for this domain. In this paper, we present CitiLink-Summ, a new corpus of European Portuguese municipal meeting minutes, comprising 100 documents and 2,322 manually hand-written summaries, each corresponding to a distinct discussion subject. Leveraging this dataset, we establish baseline results for automatic summarization in this domain, employing state-of-the-art generative models (e.g., BART, PRIMERA) as well as large language models (LLMs), evaluated with both lexical and semantic metrics such as ROUGE, BLEU, METEOR, and BERTScore. CitiLink-Summ provides the first benchmark for municipal-domain summarization in European Portuguese, offering a valuable resource for advancing NLP research on complex administrative texts.

The growing demand for large language models (LLMs) requires serving systems to handle many concurrent requests with diverse service level objectives (SLOs). This exacerbates head-of-line (HoL) blocking during the compute-intensive prefill phase, where long-running requests monopolize resources and delay higher-priority ones, leading to widespread time-to-first-token (TTFT) SLO violations. While chunked prefill enables interruptibility, it introduces an inherent trade-off between responsiveness and throughput: reducing chunk size improves response latency but degrades computational efficiency, whereas increasing chunk size maximizes throughput but exacerbates blocking. This necessitates an adaptive preemption mechanism. However, dynamically balancing execution granularity against scheduling overheads remains a key challenge. In this paper, we propose FlowPrefill, a TTFT-goodput-optimized serving system that resolves this conflict by decoupling preemption granularity from scheduling frequency. To achieve adaptive prefill scheduling, FlowPrefill introduces two key innovations: 1) Operator-Level Preemption, which leverages operator boundaries to enable fine-grained execution interruption without the efficiency loss associated with fixed small chunking; and 2) Event-Driven Scheduling, which triggers scheduling decisions only upon request arrival or completion events, thereby supporting efficient preemption responsiveness while minimizing control-plane overhead. Evaluation on real-world production traces shows that FlowPrefill improves maximum goodput by up to 5.6$\times$ compared to state-of-the-art systems while satisfying heterogeneous SLOs.

Machine learning models are increasingly trained or fine-tuned on synthetic data. Recursively training on such data has been observed to significantly degrade performance in a wide range of tasks, often characterized by a progressive drift away from the target distribution. In this work, we theoretically analyze this phenomenon in the setting of score-based diffusion models. For a realistic pipeline where each training round uses a combination of synthetic data and fresh samples from the target distribution, we obtain upper and lower bounds on the accumulated divergence between the generated and target distributions. This allows us to characterize different regimes of drift, depending on the score estimation error and the proportion of fresh data used in each generation. We also provide empirical results on synthetic data and images to illustrate the theory.

Cops and Robbers is a pursuit evasion game played on a graph, first introduced independently by Quilliot \cite{quilliot1978jeux} and Nowakowski and Winkler \cite{NOWAKOWSKI1983235} over four decades ago. A main interest in recent the literature is identifying the cop number of graph families. The cop number of a graph, $c(G)$, is defined as the minimum number of cops required to guarantee capture of the robber. Determining the cop number is computationally difficult and exact algorithms for this are typically restricted to small graph families. This paper investigates whether classical machine learning methods and graph neural networks can accurately predict a graph's cop number from its structural properties and identify which properties most strongly influence this prediction. Of the classical machine learning models, tree-based models achieve high accuracy in prediction despite class imbalance, whereas graph neural networks achieve comparable results without explicit feature engineering. The interpretability analysis shows that the most predictive features are related to node connectivity, clustering, clique structure, and width parameters, which aligns with known theoretical results. Our findings suggest that machine learning approaches can be used in complement with existing cop number algorithms by offering scalable approximations where computation is infeasible.

Modern AI models are not static. They go through multiple updates in their lifecycles. Thus, exploiting the model dynamics to create stronger Membership Inference (MI) attacks and tighter privacy audits are timely questions. Though the literature empirically shows that using a sequence of model updates can increase the power of MI attacks, rigorous analysis of the `optimal' MI attacks is limited to static models with infinite samples. Hence, we develop an `optimal' MI attack, SeMI*, that uses the sequence of model updates to identify the presence of a target inserted at a certain update step. For the empirical mean computation, we derive the optimal power of SeMI*, while accessing a finite number of samples with or without privacy. Our results retrieve the existing asymptotic analysis. We observe that having access to the model sequence avoids the dilution of MI signals unlike the existing attacks on the final model, where the MI signal vanishes as training data accumulates. Furthermore, an adversary can use SeMI* to tune both the insertion time and the canary to yield tighter privacy audits. Finally, we conduct experiments across data distributions and models trained or fine-tuned with DP-SGD demonstrating that practical variants of SeMI* lead to tighter privacy audits than the baselines.

Street-view image attribute classification is a vital downstream task of image classification, enabling applications such as autonomous driving, urban analytics, and high-definition map construction. It remains computationally demanding whether training from scratch, initialising from pre-trained weights, or fine-tuning large models. Although pre-trained vision-language models such as CLIP offer rich image representations, existing adaptation or fine-tuning methods often rely on their global image embeddings, limiting their ability to capture fine-grained, localised attributes essential in complex, cluttered street scenes. To address this, we propose CLIP-MHAdapter, a variant of the current lightweight CLIP adaptation paradigm that appends a bottleneck MLP equipped with multi-head self-attention operating on patch tokens to model inter-patch dependencies. With approximately 1.4 million trainable parameters, CLIP-MHAdapter achieves superior or competitive accuracy across eight attribute classification tasks on the Global StreetScapes dataset, attaining new state-of-the-art results while maintaining low computational cost. The code is available at https://github.com/SpaceTimeLab/CLIP-MHAdapter.

Operational rigor determines whether human-agent collaboration succeeds or fails. Scientific data pipelines need the equivalent of DevOps -- SciOps -- yet common approaches fragment provenance across disconnected systems without transactional guarantees. DataJoint 2.0 addresses this gap through the relational workflow model: tables represent workflow steps, rows represent artifacts, foreign keys prescribe execution order. The schema specifies not only what data exists but how it is derived -- a single formal system where data structure, computational dependencies, and integrity constraints are all queryable, enforceable, and machine-readable. Four technical innovations extend this foundation: object-augmented schemas integrating relational metadata with scalable object storage, semantic matching using attribute lineage to prevent erroneous joins, an extensible type system for domain-specific formats, and distributed job coordination designed for composability with external orchestration. By unifying data structure, data, and computational transformations, DataJoint creates a substrate for SciOps where agents can participate in scientific workflows without risking data corruption.

Reliable global streamflow forecasting is essential for flood preparedness and water resource management, yet data-driven models often suffer from a performance gap when transitioning from historical reanalysis to operational forecast products. This paper introduces AIFL (Artificial Intelligence for Floods), a deterministic LSTM-based model designed for global daily streamflow forecasting. Trained on 18,588 basins curated from the CARAVAN dataset, AIFL utilises a novel two-stage training strategy to bridge the reanalysis-to-forecast domain shift. The model is first pre-trained on 40 years of ERA5-Land reanalysis (1980-2019) to capture robust hydrological processes, then fine-tuned on operational Integrated Forecasting System (IFS) control forecasts (2016-2019) to adapt to the specific error structures and biases of operational numerical weather prediction. To our knowledge, this is the first global model trained end-to-end within the CARAVAN ecosystem. On an independent temporal test set (2021-2024), AIFL achieves high predictive skill with a median modified Kling-Gupta Efficiency (KGE') of 0.66 and a median Nash-Sutcliffe Efficiency (NSE) of 0.53. Benchmarking results show that AIFL is highly competitive with current state-of-the-art global systems, achieving comparable accuracy while maintaining a transparent and reproducible forcing pipeline. The model demonstrates exceptional reliability in extreme-event detection, providing a streamlined and operationally robust baseline for the global hydrological community.

Can a machine write good poetry? Any positive answer raises fundamental questions about the nature and value of art. We report a seven-month poetry workshop in which a large language model was shaped into a digital poet through iterative in-context expert feedback, without retraining. Across sessions, the model developed a distinctive style and a coherent corpus, supported by quantitative and qualitative analyses, and it produced a pen name and author image. In a blinded authorship test with 50 humanities students and graduates (three AI poems and three poems by well-known poets each), judgments were at chance: human poems were labeled human 54% of the time and AI poems 52%, with 95% confidence intervals including 50%. After the workshop, a commercial publisher released a poetry collection authored by the model. These results show that workshop-style prompting can support long-horizon creative shaping and renew debates on creativity and authorship.

Urban demand forecasting plays a critical role in optimizing routing, dispatching, and congestion management within Intelligent Transportation Systems. By leveraging data fusion and analytics techniques, traffic demand forecasting serves as a key intermediate measure for identifying emerging spatial and temporal demand patterns. In this paper, we tackle this challenge by proposing two gradient boosting model variations, one for classiffication and one for regression, both capable of generating demand forecasts at various temporal horizons, from 5 minutes up to one hour. Our overall approach effectively integrates temporal and contextual features, enabling accurate predictions that are essential for improving the efficiency of shared (micro-) mobility services. To evaluate its effectiveness, we utilize open shared mobility data derived from e-scooter and e-bike networks in five metropolitan areas. These real-world datasets allow us to compare our approach with state-of-the-art methods as well as a Generative AI-based model, demonstrating its effectiveness in capturing the complexities of modern urban mobility. Ultimately, our methodology offers novel insights on urban micro-mobility management, helping to tackle the challenges arising from rapid urbanization and thus, contributing to more sustainable, efficient, and livable cities.

Large-scale sharing of dialogue-based data is instrumental for advancing the science of teaching and learning, yet rigorous de-identification remains a major barrier. In mathematics tutoring transcripts, numeric expressions frequently resemble structured identifiers (e.g., dates or IDs), leading generic Personally Identifiable Information (PII) detection systems to over-redact core instructional content and reduce dataset utility. This work asks how PII can be detected in math tutoring transcripts while preserving their educational utility. To address this challenge, we investigate the "numeric ambiguity" problem and introduce MathEd-PII, the first benchmark dataset for PII detection in math tutoring dialogues, created through a human-in-the-loop LLM workflow that audits upstream redactions and generates privacy-preserving surrogates. The dataset contains 1,000 tutoring sessions (115,620 messages; 769,628 tokens) with validated PII annotations. Using a density-based segmentation method, we show that false PII redactions are disproportionately concentrated in math-dense regions, confirming numeric ambiguity as a key failure mode. We then compare four detection strategies: a Presidio baseline and LLM-based approaches with basic, math-aware, and segment-aware prompting. Math-aware prompting substantially improves performance over the baseline (F1: 0.821 vs. 0.379) while reducing numeric false positives, demonstrating that de-identification must incorporate domain context to preserve analytic utility. This work provides both a new benchmark and evidence that utility-preserving de-identification for tutoring data requires domain-aware modeling.

Inference-time algorithms are an emerging paradigm in which pre-trained models are used as subroutines to solve downstream tasks. Such algorithms have been proposed for tasks ranging from inverse problems and guided image generation to reasoning. However, the methods currently deployed in practice are heuristics with a variety of failure modes -- and we have very little understanding of when these heuristics can be efficiently improved. In this paper, we consider the task of sampling from a reward-tilted diffusion model -- that is, sampling from $p^{\star}(x) \propto p(x) \exp(r(x))$ -- given a reward function $r$ and pre-trained diffusion oracle for $p$. We provide a fine-grained analysis of the computational tractability of this task for quadratic rewards $r(x) = x^\top A x + b^\top x$. We show that linear-reward tilts are always efficiently sampleable -- a simple result that seems to have gone unnoticed in the literature. We use this as a building block, along with a conceptually new ingredient -- the Hubbard-Stratonovich transform -- to provide an efficient algorithm for sampling from low-rank positive-definite quadratic tilts, i.e. $r(x) = x^\top A x$ where $A$ is positive-definite and of rank $O(1)$. For negative-definite tilts, i.e. $r(x) = - x^\top A x$ where $A$ is positive-definite, we prove that the problem is intractable even if $A$ is of rank 1 (albeit with exponentially-large entries).

Sparse recovery is among the most well-studied problems in learning theory and high-dimensional statistics. In this work, we investigate the statistical and computational landscapes of sparse recovery with $\ell_\infty$ error guarantees. This variant of the problem is motivated by \emph{variable selection} tasks, where the goal is to estimate the support of a $k$-sparse signal in $\mathbb{R}^d$. Our main contribution is a provable separation between the \emph{oblivious} (``for each'') and \emph{adaptive} (``for all'') models of $\ell_\infty$ sparse recovery. We show that under an oblivious model, the optimal $\ell_\infty$ error is attainable in near-linear time with $\approx k\log d$ samples, whereas in an adaptive model, $\gtrsim k^2$ samples are necessary for any algorithm to achieve this bound. This establishes a surprising contrast with the standard $\ell_2$ setting, where $\approx k \log d$ samples suffice even for adaptive sparse recovery. We conclude with a preliminary examination of a \emph{partially-adaptive} model, where we show nontrivial variable selection guarantees are possible with $\approx k\log d$ measurements.

We introduce MetaDOAR, a lightweight meta-controller that augments the Double Oracle / PSRO paradigm with a learned, partition-aware filtering layer and Q-value caching to enable scalable multi-agent reinforcement learning on very large cyber-network environments. MetaDOAR learns a compact state projection from per node structural embeddings to rapidly score and select a small subset of devices (a top-k partition) on which a conventional low-level actor performs focused beam search utilizing a critic agent. Selected candidate actions are evaluated with batched critic forwards and stored in an LRU cache keyed by a quantized state projection and local action identifiers, dramatically reducing redundant critic computation while preserving decision quality via conservative k-hop cache invalidation. Empirically, MetaDOAR attains higher player payoffs than SOTA baselines on large network topologies, without significant scaling issues in terms of memory usage or training time. This contribution provide a practical, theoretically motivated path to efficient hierarchical policy learning for large-scale networked decision problems.

Variational Quantum Circuits (VQCs) have emerged as a promising paradigm for quantum machine learning in the NISQ era. While parameter sharing in VQCs can reduce the parameter space dimensionality and potentially mitigate the barren plateau phenomenon, it introduces a complex trade-off that has been largely overlooked. This paper investigates how parameter sharing, despite creating better global optima with fewer parameters, fundamentally alters the optimization landscape through deceptive gradients -- regions where gradient information exists but systematically misleads optimizers away from global optima. Through systematic experimental analysis, we demonstrate that increasing degrees of parameter sharing generate more complex solution landscapes with heightened gradient magnitudes and measurably higher deceptiveness ratios. Our findings reveal that traditional gradient-based optimizers (Adam, SGD) show progressively degraded convergence as parameter sharing increases, with performance heavily dependent on hyperparameter selection. We introduce a novel gradient deceptiveness detection algorithm and a quantitative framework for measuring optimization difficulty in quantum circuits, establishing that while parameter sharing can improve circuit expressivity by orders of magnitude, this comes at the cost of significantly increased landscape deceptiveness. These insights provide important considerations for quantum circuit design in practical applications, highlighting the fundamental mismatch between classical optimization strategies and quantum parameter landscapes shaped by parameter sharing.

We analyze independent policy-gradient (PG) learning in $N$-player linear-quadratic (LQ) stochastic differential games. Each player employs a distributed policy that depends only on its own state and updates the policy independently using the gradient of its own objective. We establish global linear convergence of these methods to an equilibrium by showing that the LQ game admits an $α$-potential structure, with $α$ determined by the degree of pairwise interaction asymmetry. For pairwise-symmetric interactions, we construct an affine distributed equilibrium by minimizing the potential function and show that independent PG methods converge globally to this equilibrium, with complexity scaling linearly in the population size and logarithmically in the desired accuracy. For asymmetric interactions, we prove that independent projected PG algorithms converge linearly to an approximate equilibrium, with suboptimality proportional to the degree of asymmetry. Numerical experiments confirm the theoretical results across both symmetric and asymmetric interaction networks.

We introduce MerLean, a fully automated agentic framework for autoformalization in quantum computation. MerLean extracts mathematical statements from \LaTeX{} source files, formalizes them into verified Lean~4 code built on Mathlib, and translates the result back into human-readable \LaTeX{} for semantic review. We evaluate MerLean on three theoretical quantum computing papers producing 2,050 Lean declarations from 114 statements in total. MerLean achieves end-to-end formalization on all three papers, reducing the verification burden to only the newly introduced definitions and axioms. Our results demonstrate that agentic autoformalization can scale to frontier research, offering both a practical tool for machine-verified peer review and a scalable engine for mining high-quality synthetic data to train future reasoning models. Our approach can also be generalized to any other rigorous research in mathematics and theoretical physics.

The inverse design of RNA three-dimensional (3D) structures is crucial for engineering functional RNAs in synthetic biology and therapeutics. While recent deep learning approaches have advanced this field, they are typically optimized and evaluated using native sequence recovery, which is a limited surrogate for structural fidelity, since different sequences can fold into similar 3D structures and high recovery does not necessarily indicate correct folding. To address this limitation, we propose RIDER, an RNA Inverse DEsign framework with Reinforcement learning that directly optimizes for 3D structural similarity. First, we develop and pre-train a GNN-based generative diffusion model conditioned on the target 3D structure, achieving a 9% improvement in native sequence recovery over state-of-the-art methods. Then, we fine-tune the model with an improved policy gradient algorithm using four task-specific reward functions based on 3D self-consistency metrics. Experimental results show that RIDER improves structural similarity by over 100% across all metrics and discovers designs that are distinct from native sequences.

Video recognition models are typically trained on fixed taxonomies which are often too coarse, collapsing distinctions in object, manner or outcome under a single label. As tasks and definitions evolve, such models cannot accommodate emerging distinctions and collecting new annotations and retraining to accommodate such changes is costly. To address these challenges, we introduce category splitting, a new task where an existing classifier is edited to refine a coarse category into finer subcategories, while preserving accuracy elsewhere. We propose a zero-shot editing method that leverages the latent compositional structure of video classifiers to expose fine-grained distinctions without additional data. We further show that low-shot fine-tuning, while simple, is highly effective and benefits from our zero-shot initialization. Experiments on our new video benchmarks for category splitting demonstrate that our method substantially outperforms vision-language baselines, improving accuracy on the newly split categories without sacrificing performance on the rest. Project page: https://kaitingliu.github.io/Category-Splitting/.

Safe reinforcement learning (Safe RL) aims to ensure policy performance while satisfying safety constraints. However, most existing Safe RL methods assume benign environments, making them vulnerable to adversarial perturbations commonly encountered in real-world settings. In addition, existing gradient-based adversarial attacks typically require access to the policy's gradient information, which is often impractical in real-world scenarios. To address these challenges, we propose an adversarial attack framework to reveal vulnerabilities of Safe RL policies. Using expert demonstrations and black-box environment interaction, our framework learns a constraint model and a surrogate (learner) policy, enabling gradient-based attack optimization without requiring the victim policy's internal gradients or the ground-truth safety constraints. We further provide theoretical analysis establishing feasibility and deriving perturbation bounds. Experiments on multiple Safe RL benchmarks demonstrate the effectiveness of our approach under limited privileged access.

We study online conformal prediction for non-stationary data streams subject to unknown distribution drift. While most prior work studied this problem under adversarial settings and/or assessed performance in terms of gaps of time-averaged marginal coverage, we instead evaluate performance through training-conditional cumulative regret. We specifically focus on independently generated data with two types of distribution shift: abrupt change points and smooth drift. When non-conformity score functions are pretrained on an independent dataset, we propose a split-conformal style algorithm that leverages drift detection to adaptively update calibration sets, which provably achieves minimax-optimal regret. When non-conformity scores are instead trained online, we develop a full-conformal style algorithm that again incorporates drift detection to handle non-stationarity; this approach relies on stability - rather than permutation symmetry - of the model-fitting algorithm, which is often better suited to online learning under evolving environments. We establish non-asymptotic regret guarantees for our online full conformal algorithm, which match the minimax lower bound under appropriate restrictions on the prediction sets. Numerical experiments corroborate our theoretical findings.

We study transfer learning for a linear regression task using several least-squares pretrained models that can be overparameterized. We formulate the target learning task as optimization that minimizes squared errors on the target dataset with penalty on the distance of the learned model from the pretrained models. We analytically formulate the test error of the learned target model and provide the corresponding empirical evaluations. Our results elucidate when using more pretrained models can improve transfer learning. Specifically, if the pretrained models are overparameterized, using sufficiently many of them is important for beneficial transfer learning. However, the learning may be compromised by overparameterization bias of pretrained models, i.e., the minimum $\ell_2$-norm solution's restriction to a small subspace spanned by the training examples in the high-dimensional parameter space. We propose a simple debiasing via multiplicative correction factor that can reduce the overparameterization bias and leverage more pretrained models to learn a target predictor.

Kolmogorov-Arnold Networks (KANs) have recently emerged as a compelling alternative to multilayer perceptrons, offering enhanced interpretability via functional decomposition. However, existing KAN architectures, including spline-, wavelet-, radial-basis variants, etc., suffer from high computational cost and slow convergence, limiting scalability and practical applicability. Here, we introduce Feature-Enriched Kolmogorov-Arnold Networks (FEKAN), a simple yet effective extension that preserves all the advantages of KAN while improving computational efficiency and predictive accuracy through feature enrichment, without increasing the number of trainable parameters. By incorporating these additional features, FEKAN accelerates convergence, increases representation capacity, and substantially mitigates the computational overhead characteristic of state-of-the-art KAN architectures. We investigate FEKAN across a comprehensive set of benchmarks, including function-approximation tasks, physics-informed formulations for diverse partial differential equations (PDEs), and neural operator settings that map between input and output function spaces. For function approximation, we systematically compare FEKAN against a broad family of KAN variants, FastKAN, WavKAN, ReLUKAN, HRKAN, ChebyshevKAN, RBFKAN, and the original SplineKAN. Across all tasks, FEKAN demonstrates substantially faster convergence and consistently higher approximation accuracy than the underlying baseline architectures. We also establish the theoretical foundations for FEKAN, showing its superior representation capacity compared to KAN, which contributes to improved accuracy and efficiency.

This paper presents a capacity-constrained incentive-based demand response approach for residential smart grids. It aims to maintain electricity grid capacity limits and prevent congestion by financially incentivising end users to reduce or shift their energy consumption. The proposed framework adopts a hierarchical architecture in which a service provider adjusts hourly incentive rates based on wholesale electricity prices and aggregated residential load. The financial interests of both the service provider and end users are explicitly considered. A deep reinforcement learning approach is employed to learn optimal real-time incentive rates under explicit capacity constraints. Heterogeneous user preferences are modelled through appliance-level home energy management systems and dissatisfaction costs. Using real-world residential electricity consumption and price data from three households, simulation results show that the proposed approach effectively reduces peak demand and smooths the aggregated load profile. This leads to an approximately 22.82% reduction in the peak-to-average ratio compared to the no-demand-response case.

We extend directed quantum circuit synthesis (DQCS) with reinforcement learning from purely discrete gate selection to parameterized quantum state preparation with continuous single-qubit rotations \(R_x\), \(R_y\), and \(R_z\). We compare two training regimes: a one-stage agent that jointly selects the gate type, the affected qubit(s), and the rotation angle; and a two-stage variant that first proposes a discrete circuit and subsequently optimizes the rotation angles with Adam using parameter-shift gradients. Using Gymnasium and PennyLane, we evaluate Proximal Policy Optimization (PPO) and Advantage Actor--Critic (A2C) on systems comprising two to ten qubits and on targets of increasing complexity with \(λ\) ranging from one to five. Whereas A2C does not learn effective policies in this setting, PPO succeeds under stable hyperparameters (one-stage: learning rate approximately \(5\times10^{-4}\) with a self-fidelity-error threshold of 0.01; two-stage: learning rate approximately \(10^{-4}\)). Both approaches reliably reconstruct computational basis states (between 83\% and 99\% success) and Bell states (between 61\% and 77\% success). However, scalability saturates for \(λ\) of approximately three to four and does not extend to ten-qubit targets even at \(λ=2\). The two-stage method offers only marginal accuracy gains while requiring around three times the runtime. For practicality under a fixed compute budget, we therefore recommend the one-stage PPO policy, provide explicit synthesized circuits, and contrast with a classical variational baseline to outline avenues for improved scalability.

Jailbreak prompts are a practical and evolving threat to large language models (LLMs), particularly in agentic systems that execute tools over untrusted content. Many attacks exploit long-context hiding, semantic camouflage, and lightweight obfuscations that can evade single-pass guardrails. We present RLM-JB, an end-to-end jailbreak detection framework built on Recursive Language Models (RLMs), in which a root model orchestrates a bounded analysis program that transforms the input, queries worker models over covered segments, and aggregates evidence into an auditable decision. RLM-JB treats detection as a procedure rather than a one-shot classification: it normalizes and de-obfuscates suspicious inputs, chunks text to reduce context dilution and guarantee coverage, performs parallel chunk screening, and composes cross-chunk signals to recover split-payload attacks. On AutoDAN-style adversarial inputs, RLM-JB achieves high detection effectiveness across three LLM backends (ASR/Recall 92.5-98.0%) while maintaining very high precision (98.99-100%) and low false positive rates (0.0-2.0%), highlighting a practical sensitivity-specificity trade-off as the screening backend changes.

This paper introduces ParlaCAP, a large-scale dataset for analyzing parliamentary agenda setting across Europe, and proposes a cost-effective method for building domain-specific policy topic classifiers. Applying the Comparative Agendas Project (CAP) schema to the multilingual ParlaMint corpus of over 8 million speeches from 28 parliaments of European countries and autonomous regions, we follow a teacher-student framework in which a high-performing large language model (LLM) annotates in-domain training data and a multilingual encoder model is fine-tuned on these annotations for scalable data annotation. We show that this approach produces a classifier tailored to the target domain. Agreement between the LLM and human annotators is comparable to inter-annotator agreement among humans, and the resulting model outperforms existing CAP classifiers trained on manually-annotated but out-of-domain data. In addition to the CAP annotations, the ParlaCAP dataset offers rich speaker and party metadata, as well as sentiment predictions coming from the ParlaSent multilingual transformer model, enabling comparative research on political attention and representation across countries. We illustrate the analytical potential of the dataset with three use cases, examining the distribution of parliamentary attention across policy topics, sentiment patterns in parliamentary speech, and gender differences in policy attention.

Prompting schemes such as Chain of Thought, Tree of Thoughts, and Graph of Thoughts can significantly enhance the reasoning capabilities of large language models. However, most existing schemes require users to define static, problem-specific reasoning structures that lack adaptability to dynamic or unseen problem types. Additionally, these schemes are often under-optimized in terms of hyperparameters, prompts, runtime, and prompting cost. To address these limitations, we introduce Framework of Thoughts (FoT)--a general-purpose foundation framework for building and optimizing dynamic reasoning schemes. FoT comes with built-in features for hyperparameter tuning, prompt optimization, parallel execution, and intelligent caching, unlocking the latent performance potential of reasoning schemes. We demonstrate FoT's capabilities by implementing three popular schemes--Tree of Thoughts, Graph of Thoughts, and ProbTree--within FoT. We empirically show that FoT enables significantly faster execution, reduces costs, and achieves better task scores through optimization. We release our codebase to facilitate the development of future dynamic and efficient reasoning schemes.

One of the central challenges in the computational analysis of liquid chromatography-tandem mass spectrometry (LC-MS/MS) data is to identify the compounds underlying the output spectra. In recent years, this problem is increasingly tackled using deep learning methods. A common strategy involves predicting a molecular fingerprint vector from an input mass spectrum, which is then used to search for matches in a chemical compound database. While various loss functions are employed in training these predictive models, their impact on model performance remains poorly understood. In this study, we investigate commonly used loss functions, deriving novel regret bounds that characterize when Bayes-optimal decisions for these objectives must diverge. Our results reveal a fundamental trade-off between the two objectives of (1) fingerprint similarity and (2) molecular retrieval. Optimizing for more accurate fingerprint predictions typically worsens retrieval results, and vice versa. Our theoretical analysis shows this trade-off depends on the similarity structure of candidate sets, providing guidance for loss function and fingerprint selection.

Hazard and survival functions are natural, interpretable targets in time-to-event prediction, but their inherent non-additivity fundamentally limits standard additive explanation methods. We introduce Survival Functional Decomposition (SurvFD), a principled approach for analyzing feature interactions in machine learning survival models. By decomposing higher-order effects into time-dependent and time-independent components, SurvFD offers a previously unrecognized perspective on survival explanations, explicitly characterizing when and why additive explanations fail. Building on this theoretical decomposition, we propose SurvSHAP-IQ, which extends Shapley interactions to time-indexed functions, providing a practical estimator for higher-order, time-dependent interactions. Together, SurvFD and SurvSHAP-IQ establish an interaction- and time-aware interpretability approach for survival modeling, with broad applicability across time-to-event prediction tasks.

Generalized additive models (GAMs) offer interpretability through independent univariate feature effects but underfit when interactions are present in data. GA$^2$Ms add selected pairwise interactions which improves accuracy, but sacrifices interpretability and limits model auditing. We propose \emph{Conditionally Additive Local Models} (CALMs), a new model class, that balances the interpretability of GAMs with the accuracy of GA$^2$Ms. CALMs allow multiple univariate shape functions per feature, each active in different regions of the input space. These regions are defined independently for each feature as simple logical conditions (thresholds) on the features it interacts with. As a result, effects remain locally additive while varying across subregions to capture interactions. We further propose a principled distillation-based training pipeline that identifies homogeneous regions with limited interactions and fits interpretable shape functions via region-aware backfitting. Experiments on diverse classification and regression tasks show that CALMs consistently outperform GAMs and achieve accuracy comparable with GA$^2$Ms. Overall, CALMs offer a compelling trade-off between predictive accuracy and interpretability.

Soft prompt tuning leverages continuous embeddings to capture task-specific information in large pre-trained language models (LLMs), achieving competitive performance in few-shot settings. However, soft prompts rely on high-dimensional, implicit representations and lack explicit semantics and traceable training behaviors, which limits their interpretability. To address this limitation, we propose a soft prompt tuning optimization method based on topological morphological evolution. Specifically, we employ persistent homology from topological data analysis (TDA) to quantify the structural representations of soft prompts in continuous parameter space and their training process evolution. Quantitative analysis shows that topologically stable and compact soft prompts achieve better downstream performance. Based on this empirical observation, we construct a loss function for optimizing soft prompt tuning, termed Topological Soft Prompt Loss (TSLoss). TSLoss guides the model to learn structurally stable adaptations by quantifying inter-parameter connectivity and redundancy. Extensive experiments show that training with TSLoss accelerates convergence and improves tuning performance, providing an interpretable method to understand and optimize soft prompt tuning from structural and topological perspectives.

The Asset Administration Shell (AAS) is an emerging technology for the implementation of digital twins in the field of manufacturing. Software is becoming increasingly important, not only in general but specifically in relation to manufacturing, especially with regard to digital manufacturing and a shift towards the usage of artificial intelligence. This increases the need not only to model software, but also to integrate services directly into the AAS. The existing literature contains individual solutions to implement such software-heavy AAS. However, there is no systematic analysis of software architectures that integrate software services directly into the AAS. This paper aims to fill this research gap and differentiate architectures based on software quality criteria as well as typical manufacturing use cases. This work may be considered as an interpretation guideline for software-heavy AAS, both in academia and for practitioners.

Analytical diffusion models offer a mathematically transparent path to generative modeling by formulating the denoising score as an empirical-Bayes posterior mean. However, this interpretability comes at a prohibitive cost: the standard formulation necessitates a full-dataset scan at every timestep, scaling linearly with dataset size. In this work, we present the first systematic study addressing this scalability bottleneck. We challenge the prevailing assumption that the entire training data is necessary, uncovering the phenomenon of Posterior Progressive Concentration: the effective golden support of the denoising score is not static but shrinks asymptotically from the global manifold to a local neighborhood as the signal-to-noise ratio increases. Capitalizing on this, we propose Dynamic Time-Aware Golden Subset Diffusion (GoldDiff), a training-free framework that decouples inference complexity from dataset size. Instead of static retrieval, GoldDiff uses a coarse-to-fine mechanism to dynamically pinpoint the ''Golden Subset'' for inference. Theoretically, we derive rigorous bounds guaranteeing that our sparse approximation converges to the exact score. Empirically, GoldDiff achieves a $\bf 71 \times$ speedup on AFHQ while matching or achieving even better performance than full-scan baselines. Most notably, we demonstrate the first successful scaling of analytical diffusion to ImageNet-1K, unlocking a scalable, training-free paradigm for large-scale generative modeling.

Looping, reusing a block of layers across depth, and depth growing, training shallow-to-deep models by duplicating middle layers, have both been linked to stronger reasoning, but their relationship remains unclear. We provide a mechanistic unification: looped and depth-grown models exhibit convergent depth-wise signatures, including increased reliance on late layers and recurring patterns aligned with the looped or grown block. These shared signatures support the view that their gains stem from a common form of iterative computation. Building on this connection, we show that the two techniques are adaptable and composable: applying inference-time looping to the middle blocks of a depth-grown model improves accuracy on some reasoning primitives by up to $2\times$, despite the model never being trained to loop. Both approaches also adapt better than the baseline when given more in-context examples or additional supervised fine-tuning data. Additionally, depth-grown models achieve the largest reasoning gains when using higher-quality, math-heavy cooldown mixtures, which can be further boosted by adapting a middle block to loop. Overall, our results position depth growth and looping as complementary, practical methods for inducing and scaling iterative computation to improve reasoning.

Large language models (LLMs) often struggle to learn from corrective feedback within a conversational context. They are rarely proactive in soliciting this feedback, even when faced with ambiguity, which can make their dialogues feel static, one-sided, and lacking the adaptive qualities of human conversation. To address these limitations, we draw inspiration from social meta-learning (SML) in humans - the process of learning how to learn from others. We formulate SML as a finetuning methodology, training LLMs to solicit and learn from language feedback in simulated pedagogical dialogues, where static tasks are converted into interactive social learning problems. SML effectively teaches models to use conversation to solve problems they are unable to solve in a single turn. This capability generalises across domains; SML on math problems produces models that better use feedback to solve coding problems and vice versa. Furthermore, despite being trained only on fully-specified problems, these models are better able to solve underspecified tasks where critical information is revealed over multiple turns. When faced with this ambiguity, SML-trained models make fewer premature answer attempts and are more likely to ask for the information they need. This work presents a scalable approach to developing AI systems that effectively learn from language feedback.

Existing Multi-Agent Systems (MAS) typically rely on static, homogeneous model configurations, limiting their ability to exploit the distinct strengths of differently post-trained models. To address this, we introduce Team-of-Thoughts, a novel MAS architecture that leverages the complementary capabilities of heterogeneous agents via an orchestrator-tool paradigm. Our framework introduces two key mechanisms to optimize performance: (1) an orchestrator calibration scheme that identifies models with superior coordination capabilities, and (2) a self-assessment protocol where tool agents profile their own domain expertise to account for variations in post-training skills. During inference, the orchestrator dynamically activates the most suitable tool agents based on these proficiency profiles. Experiments on five reasoning and code generation benchmarks show that Team-of-Thoughts delivers consistently superior task performance. Notably, on AIME24 and LiveCodeBench, our approach achieves accuracies of 96.67% and 72.53%, respectively, substantially outperforming homogeneous role-play baselines, which score 80% and 65.93%.

Causal discovery seeks to uncover causal relations from data, typically represented as causal graphs, and is essential for predicting the effects of interventions. While expert knowledge is required to construct principled causal graphs, many statistical methods have been proposed to leverage observational data with varying formal guarantees. Causal Assumption-based Argumentation (ABA) is a framework that uses symbolic reasoning to ensure correspondence between input constraints and output graphs, while offering a principled way to combine data and expertise. We explore the use of large language models (LLMs) as imperfect experts for Causal ABA, eliciting semantic structural priors from variable names and descriptions and integrating them with conditional-independence evidence. Experiments on standard benchmarks and semantically grounded synthetic graphs demonstrate state-of-the-art performance, and we additionally introduce an evaluation protocol to mitigate memorisation bias when assessing LLMs for causal discovery.

Federated Learning (FL) enables collaborative model training without exposing clients' private data, and has been widely adopted in privacy-sensitive scenarios. However, FL faces two critical security threats: curious servers that may launch inference attacks to reconstruct clients' private data, and compromised clients that can launch poisoning attacks to disrupt model aggregation. Existing solutions mitigate these attacks by combining mainstream privacy-preserving techniques with defensive aggregation strategies. However, they either incur high computation and communication overhead or perform poorly under non-independent and identically distributed (Non-IID) data settings. To tackle these challenges, we propose SRFed, an efficient Byzantine-robust and privacy-preserving FL framework for Non-IID scenarios. First, we design a decentralized efficient functional encryption (DEFE) scheme to support efficient model encryption and non-interactive decryption. DEFE also eliminates third-party reliance and defends against server-side inference attacks. Second, we develop a privacy-preserving defensive model aggregation mechanism based on DEFE. This mechanism filters poisonous models under Non-IID data by layer-wise projection and clustering-based analysis. Theoretical analysis and extensive experiments show that SRFed outperforms state-of-the-art baselines in privacy protection, Byzantine robustness, and efficiency.

Estimating consumer preferences is central to many problems in economics and marketing. This paper develops a flexible framework for learning individual preferences from partial ranking information by interpreting observed rankings as collections of pairwise comparisons with logistic choice probabilities. We model latent utility as the sum of interpretable product attributes, item fixed effects, and a low-rank user-item factor structure, enabling both interpretability and information sharing across consumers and items. We further correct for selection in which comparisons are observed: a comparison is recorded only if both items enter the consumer's consideration set, inducing exposure bias toward frequently encountered items. We model pair observability as the product of item-level observability propensities and estimate these propensities with a logistic model for the marginal probability that an item is observable. Preference parameters are then estimated by maximizing an inverse-probability-weighted (IPW), ridge-regularized log-likelihood that reweights observed comparisons toward a target comparison population. To scale computation, we propose a stochastic gradient descent (SGD) algorithm based on inverse-probability resampling, which draws comparisons in proportion to their IPW weights. In an application to transaction data from an online wine retailer, the method improves out-of-sample recommendation performance relative to a popularity-based benchmark, with particularly strong gains in predicting purchases of previously unconsumed products.

C-RASP is a simple programming language that was recently shown to capture concepts expressible by transformers. In this paper, we develop new algorithmic techniques for automatically verifying C-RASPs. To this end, we establish a connection to the verification of synchronous dataflow programs in Lustre, which enables us to exploit state-of-the-art model checkers utilizing highly optimized SMT-solvers. Our second contribution addresses learning a C-RASP program in the first place. To this end, we provide a new algorithm for learning a C-RASP from examples using local search. We demonstrate efficacy of our implementation for benchmarks of C-RASPs in the literature, in particular in connection to the following applications: (1) transformer program optimization, and (2) constrained learning of transformer programs (based on a partial specification).

Machine-translated data is widely used in multilingual NLP, particularly when native text is scarce. However, translated text differs systematically from native text. This phenomenon is known as translationese, and it reflects both traces of the source language and characteristic properties of translation itself. In this paper, we study how training on machine-translated data affects small English language models, focusing on how translationese from different source languages shapes linguistic acceptability judgments and language modelling for different domains. We train models on English text translated from 24 typologically and resource-diverse source languages, enabling a systematic analysis of how source language and corpus properties influence what models learn. Our results show that the source language has a clear impact on model behavior: general perplexity is more driven by the lexical diversity of the translated corpus, while grammatical performance is strongly correlated to typological similarity to English, given enough data.

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

The rapid advancement of large language models (LLMs) necessitates evaluation frameworks that reflect real-world academic rigor and multilingual complexity. This paper introduces IndicEval, a scalable benchmarking platform designed to assess LLM performance using authentic high-stakes examination questions from UPSC, JEE, and NEET across STEM and humanities domains in both English and Hindi. Unlike synthetic benchmarks, IndicEval grounds evaluation in real examination standards, enabling realistic measurement of reasoning, domain knowledge, and bilingual adaptability. The framework automates assessment using Zero-Shot, Few-Shot, and Chain-of-Thought (CoT) prompting strategies and supports modular integration of new models and languages. Experiments conducted on Gemini 2.0 Flash, GPT-4, Claude, and LLaMA 3-70B reveal three major findings. First, CoT prompting consistently improves reasoning accuracy, with substantial gains across subjects and languages. Second, significant cross-model performance disparities persist, particularly in high-complexity examinations. Third, multilingual degradation remains a critical challenge, with marked accuracy drops in Hindi compared to English, especially under Zero-Shot conditions. These results highlight persistent gaps in bilingual reasoning and domain transfer. Overall, IndicEval provides a practice-oriented, extensible foundation for rigorous, equitable evaluation of LLMs in multilingual educational settings and offers actionable insights for improving reasoning robustness and language adaptability.

Low-Rank Adaptation (LoRA) fine-tunes large models by learning low-rank updates on top of frozen weights, dramatically reducing trainable parameters and memory. In this work, we address the gap between training with full steps with low-rank projections (SVDLoRA) and LoRA fine-tuning. We propose LoRSum, a memory-efficient subroutine that closes this gap for gradient descent by casting LoRA optimization as a proximal sub-problem and solving it efficiently with alternating least squares updates, which we prove to be an implicit block power method. We recover several recently proposed preconditioning methods for LoRA as special cases, and show that LoRSum can also be used for updating a low-rank momentum. In order to address full steps with preconditioned gradient descent, we propose a scaled variant of LoRSum that uses structured metrics such as K-FAC and Shampoo, and we show that storing the diagonal of these metrics still allows them to perform well while remaining memory-efficient. Experiments on a synthetic task, CIFAR-100, and language-model fine-tuning on GLUE, SQuAD v2, and WikiText-103, show that our method can match or improve LoRA baselines given modest compute overhead, while avoiding full-matrix SVD projections and retaining LoRA-style parameter efficiency.

Generative model evaluation commonly relies on high-dimensional embedding spaces to compute distances between samples. We show that dataset representations in these spaces are affected by the hubness phenomenon, which distorts nearest neighbor relationships and biases distance-based metrics. Building on the classical Iterative Contextual Dissimilarity Measure (ICDM), we introduce Generative ICDM (GICDM), a method to correct neighborhood estimation for both real and generated data. We introduce a multi-scale extension to improve empirical behavior. Extensive experiments on synthetic and real benchmarks demonstrate that GICDM resolves hubness-induced failures, restores reliable metric behavior, and improves alignment with human judgment.

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

As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a growing need for hardware-aware neural architectures that enable efficient, low-latency, and energy-conscious local processing. We present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on a SoC FPGA and evaluated on two open-source datasets. For classification task, our baseline floating-point model achieves 92.7% accuracy on SHD dataset - only 2.4% below the state of the art - while requiring over 10x and 67x fewer parameters. On SSC, our models achieve 66.9-71.0% accuracy. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by up to 19.3% while reducing resource usage and latency. For SSC, we report the first hardware-accelerated evaluation. We further demonstrate the first end-to-end FPGA implementation of event-audio keyword spotting, combining graph convolutional layers with recurrent sequence modelling. The system achieves up to 95% word-end detection accuracy, with only 10.53 microsecond latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS.

Conventional large language model (LLM) fairness alignment largely focuses on mitigating bias along single sensitive attributes, overlooking fairness as an inherently multidimensional and context-specific value. This approach risks creating systems that achieve narrow fairness metrics while exacerbating disparities along untargeted attributes, a phenomenon known as bias spillover. While extensively studied in machine learning, bias spillover remains critically underexplored in LLM alignment. In this work, we investigate how targeted gender alignment affects fairness across nine sensitive attributes in three state-of-the-art LLMs (Mistral 7B, Llama 3.1 8B, Qwen 2.5 7B). Using Direct Preference Optimization and the BBQ benchmark, we evaluate fairness under ambiguous and disambiguous contexts. Our findings reveal noticeable bias spillover: while aggregate results show improvements, context-aware analysis exposes significant degradations in ambiguous contexts, particularly for physical appearance ($p< 0.001$ across all models), sexual orientation, and disability status. We demonstrate that improving fairness along one attribute can inadvertently worsen disparities in others under uncertainty, highlighting the necessity of context-aware, multi-attribute fairness evaluation frameworks.

Releasing data once and for all under noninteractive Local Differential Privacy (LDP) enables complete data reusability, but the resulting noise may create bias in subsequent analyses. In this work, we leverage the Weierstrass transform to characterize this bias in binary classification. We prove that inverting this transform leads to a bias-correction method to compute unbiased estimates of nonlinear functions on examples released under LDP. We then build a novel stochastic gradient descent algorithm called Inverse Weierstrass Private SGD (IWP-SGD). It converges to the true population risk minimizer at a rate of $\mathcal{O}(1/n)$, with $n$ the number of examples. We empirically validate IWP-SGD on binary classification tasks using synthetic and real-world datasets.

Automated feature engineering (AFE) enables AI systems to autonomously construct high-utility representations from raw tabular data. However, existing AFE methods rely on statistical heuristics, yielding brittle features that fail under distribution shift. We introduce CAFE, a framework that reformulates AFE as a causally-guided sequential decision process, bridging causal discovery with reinforcement learning-driven feature construction. Phase I learns a sparse directed acyclic graph over features and the target to obtain soft causal priors, grouping features as direct, indirect, or other based on their causal influence with respect to the target. Phase II uses a cascading multi-agent deep Q-learning architecture to select causal groups and transformation operators, with hierarchical reward shaping and causal group-level exploration strategies that favor causally plausible transformations while controlling feature complexity. Across 15 public benchmarks (classification with macro-F1; regression with inverse relative absolute error), CAFE achieves up to 7% improvement over strong AFE baselines, reduces episodes-to-convergence, and delivers competitive time-to-target. Under controlled covariate shifts, CAFE reduces performance drop by ~4x relative to a non-causal multi-agent baseline, and produces more compact feature sets with more stable post-hoc attributions. These findings underscore that causal structure, used as a soft inductive prior rather than a rigid constraint, can substantially improve the robustness and efficiency of automated feature engineering.

Large language models are increasingly integrated into academic writing workflows; however, the most widely used \LaTeX\ editors remain AI-peripheral -- offering compilation and collaboration, but no native intelligence. This separation forces researchers to leave their editing environment for AI assistance, fragmenting document context and interrupting writing flow. We present Bibby AI (trybibby.com), a native, AI-first \LaTeX\ editor that unifies the complete research writing lifecycle within a single interface. Bibby embeds an AI writing assistant, smart citation search, AI table and equation generation, an AI paper reviewer, abstract generator, literature review drafting, a deep research assistant, and real-time \LaTeX\ error detection and auto-fix -- all natively, without plugins or copy-paste workflows. We introduce LaTeXBench-500, a benchmark of 500 real-world compilation errors across six categories. Bibby achieves 91.4\% detection accuracy and 83.7\% one-click fix accuracy, outperforming Overleaf's native diagnostics (61.2\%) and OpenAI Prism (78.3 / 64.1\%) by large margins. Bibby demonstrates that a privacy-preserving, research-first AI editor can meaningfully accelerate every stage of academic manuscript preparation. We found that Bibby AI is a far superior alternative to overleaf latex and better than OpenAI Prism functionalities and AI.

Designing Optical Character Recognition (OCR) systems for India requires balancing linguistic diversity, document heterogeneity, and deployment constraints. In this paper, we study two training strategies for building multilingual OCR systems with Vision-Language Models through the Chitrapathak series. We first follow a popular multimodal approach, pairing a generic vision encoder with a strong multilingual language model and training the system end-to-end for OCR. Alternatively, we explore fine-tuning an existing OCR model, despite not being trained for the target languages. Through extensive evaluation on multilingual Indic OCR benchmarks and deployment-oriented metrics, we find that the second strategy consistently achieves better accuracy-latency trade-offs. Chitrapathak-2 achieves 3-6x speedup over its predecessor with being state-of-the-art (SOTA) in Telugu (6.69 char ANLS) and second best in the rest. In addition, we present Parichay, an independent OCR model series designed specifically for 9 Indian government documents to extract structured key fields, achieving 89.8% Exact Match score with a faster inference. Together, these systems achieve SOTA performance and provide practical guidance for building production-scale OCR pipelines in the Indian context.

Agentic systems, AI architectures that autonomously execute multi-step workflows to achieve complex goals, are often built using repeated large language model (LLM) calls for closed-set decision tasks such as routing, shortlisting, gating, and verification. While convenient, this design makes deployments slow and expensive due to cumulative latency and token usage. We propose TabAgent, a framework for replacing generative decision components in closed-set selection tasks with a compact textual-tabular classifier trained on execution traces. TabAgent (i) extracts structured schema, state, and dependency features from trajectories (TabSchema), (ii) augments coverage with schema-aligned synthetic supervision (TabSynth), and (iii) scores candidates with a lightweight classifier (TabHead). On the long-horizon AppWorld benchmark, TabAgent maintains task-level success while eliminating shortlist-time LLM calls, reducing latency by approximately 95% and inference cost by 85-91%. Beyond tool shortlisting, TabAgent generalizes to other agentic decision heads, establishing a paradigm for learned discriminative replacements of generative bottlenecks in production agent architectures.

Multiagent AI systems require consistent communication, but we lack methods to verify that agents share the same understanding of the terms used. Natural language is interpretable but vulnerable to semantic drift, while learned protocols are efficient but opaque. We propose a certification protocol based on the stimulus-meaning model, where agents are tested on shared observable events and terms are certified if empirical disagreement falls below a statistical threshold. In this protocol, agents restricting their reasoning to certified terms ("core-guarded reasoning") achieve provably bounded disagreement. We also outline mechanisms for detecting drift (recertification) and recovering shared vocabulary (renegotiation). In simulations with varying degrees of semantic divergence, core-guarding reduces disagreement by 72-96%. In a validation with fine-tuned language models, disagreement is reduced by 51%. Our framework provides a first step towards verifiable agent-to-agent communication.

Generating diagnostic text from histopathology whole slide images (WSIs) is challenging due to the gigapixel scale of the input and the requirement for precise, domain specific language. We propose a hierarchical vision language framework that combines a frozen pathology foundation model with a Transformer decoder for report generation. To make WSI processing tractable, we perform multi resolution pyramidal patch selection (downsampling factors 2^3 to 2^6) and remove background and artifacts using Laplacian variance and HSV based criteria. Patch features are extracted with the UNI Vision Transformer and projected to a 6 layer Transformer decoder that generates diagnostic text via cross attention. To better represent biomedical terminology, we tokenize the output using BioGPT. Finally, we add a retrieval based verification step that compares generated reports with a reference corpus using Sentence BERT embeddings; if a high similarity match is found, the generated report is replaced with the retrieved ground truth reference to improve reliability.

Evaluating machine unlearning methods remains technically challenging, with recent benchmarks requiring complex setups and significant engineering overhead. We introduce a unified and extensible benchmarking suite that simplifies the evaluation of unlearning algorithms using the KLoM (KL divergence of Margins) metric. Our framework provides precomputed model ensembles, oracle outputs, and streamlined infrastructure for running evaluations out of the box. By standardizing setup and metrics, it enables reproducible, scalable, and fair comparison across unlearning methods. We aim for this benchmark to serve as a practical foundation for accelerating research and promoting best practices in machine unlearning. Our code and data are publicly available.

Replay attacks remain a critical vulnerability for automatic speaker verification systems, particularly in real-time voice assistant applications. In this work, we propose acoustic maps as a novel spatial feature representation for replay speech detection from multi-channel recordings. Derived from classical beamforming over discrete azimuth and elevation grids, acoustic maps encode directional energy distributions that reflect physical differences between human speech radiation and loudspeaker-based replay. A lightweight convolutional neural network is designed to operate on this representation, achieving competitive performance on the ReMASC dataset with approximately 6k trainable parameters. Experimental results show that acoustic maps provide a compact and physically interpretable feature space for replay attack detection across different devices and acoustic environments.

We propose an agentic data augmentation method for Aspect-Based Sentiment Analysis (ABSA) that uses iterative generation and verification to produce high quality synthetic training examples. To isolate the effect of agentic structure, we also develop a closely matched prompting-based baseline using the same model and instructions. Both methods are evaluated across three ABSA subtasks (Aspect Term Extraction (ATE), Aspect Sentiment Classification (ATSC), and Aspect Sentiment Pair Extraction (ASPE)), four SemEval datasets, and two encoder-decoder models: T5-Base and Tk-Instruct. Our results show that the agentic augmentation outperforms raw prompting in label preservation of the augmented data, especially when the tasks require aspect term generation. In addition, when combined with real data, agentic augmentation provides higher gains, consistently outperforming prompting-based generation. These benefits are most pronounced for T5-Base, while the more heavily pretrained Tk-Instruct exhibits smaller improvements. As a result, augmented data helps T5-Base achieve comparable performance with its counterpart.

Generative models are increasingly used in recommender systems, both for modeling user behavior as event sequences and for integrating large language models into recommendation pipelines. A key challenge in this setting is the extremely large cardinality of item spaces, which makes training generative models difficult and introduces a vocabulary gap between natural language and item identifiers. Semantic identifiers (semantic IDs), which represent items as sequences of low-cardinality tokens, have recently emerged as an effective solution to this problem. However, existing approaches generate semantic identifiers of fixed length, assigning the same description length to all items. This is inefficient, misaligned with natural language, and ignores the highly skewed frequency structure of real-world catalogs, where popular items and rare long-tail items exhibit fundamentally different information requirements. In parallel, the emergent communication literature studies how agents develop discrete communication protocols, often producing variable-length messages in which frequent concepts receive shorter descriptions. Despite the conceptual similarity, these ideas have not been systematically adopted in recommender systems. In this work, we bridge recommender systems and emergent communication by introducing variable-length semantic identifiers for recommendation. We propose a discrete variational autoencoder with Gumbel-Softmax reparameterization that learns item representations of adaptive length under a principled probabilistic framework, avoiding the instability of REINFORCE-based training and the fixed-length constraints of prior semantic ID methods.

Artificial intelligence can rapidly propose candidate phases and structures from X-ray diffraction (XRD), but these hypotheses often fail in downstream refinement because peak intensities cannot be stably assigned under severe overlap and diffraction consistency is enforced only weakly. Here we introduce WPEM, a physics-constrained whole-pattern decomposition and refinement workflow that turns Bragg's law into an explicit constraint within a batch expectation--maximization framework. WPEM models the full profile as a probabilistic mixture density and iteratively infers component-resolved intensities while keeping peak centres Bragg-consistent, producing a continuous, physically admissible intensity representation that remains stable in heavily overlapped regions and in the presence of mixed radiation or multiple phases. We benchmark WPEM on standard reference patterns (\ce{PbSO4} and \ce{Tb2BaCoO5}), where it yields lower $R_{\mathrm{p}}$/$R_{\mathrm{wp}}$ than widely used packages (FullProf and TOPAS) under matched refinement conditions. We further demonstrate generality across realistic experimental scenarios, including phase-resolved decomposition of a multiphase Ti--15Nb thin film, quantitative recovery of \ce{NaCl}--\ce{Li2CO3} mixture compositions, separation of crystalline peaks from amorphous halos in semicrystalline polymers, high-throughput operando lattice tracking in layered cathodes, automated refinement of a compositionally disordered Ru--Mn oxide solid solution (CCDC 2530452), and quantitative phase-resolved deciphering of an ancient Egyptian make-up sample from synchrotron powder XRD. By providing Bragg-consistent, uncertainty-aware intensity partitioning as a refinement-ready interface, WPEM closes the gap between AI-generated hypotheses and diffraction-admissible structure refinement on challenging XRD data.

This letter proposes a novel Multihop Dual Modular Clock Algorithm (M-DMCA) for efficient node discovery in cognitive radio-based emergency response networks. M-DMCA supports dual-channel selection per timeslot and incorporates a three-way handshake mechanism to significantly reduce rendezvous time. Performance evaluation under a worst-case scenario with 20 nodes, asymmetric channel sets of size 20, channel similarity index (m) as 2, and high primary radio activity shows that M-DMCA achieves a 24% reduction in rendezvous time compared to the multihop Extended Modular Clock Algorithm (EMCA), outperforming existing rendezvous protocols.

We study reward-free and reward-agnostic exploration in episodic finite-horizon Markov decision processes (MDPs), where an agent explores an unknown environment without observing external rewards. Reward-free exploration aims to enable $ε$-optimal policies for any reward revealed after exploration, while reward-agnostic exploration targets $ε$-optimality for rewards drawn from a small finite class. In the reward-agnostic setting, Li, Yan, Chen, and Fan achieve minimax sample complexity, but only for restrictively small accuracy parameter $ε$. We propose a new algorithm that significantly relaxes the requirement on $ε$. Our approach is novel and of technical interest by itself. Our algorithm employs an online learning procedure with carefully designed rewards to construct an exploration policy, which is used to gather data sufficient for accurate dynamics estimation and subsequent computation of an $ε$-optimal policy once the reward is revealed. Finally, we establish a tight lower bound for reward-free exploration, closing the gap between known upper and lower bounds.

Extreme Edge Computing (XEC) distributes streaming workloads across consumer-owned devices, exploiting their proximity to users and ubiquitous availability. Many such workloads are AI-driven, requiring continuous neural network inference for tasks like object detection and video analytics. Distributed Inference (DI), which partitions model execution across multiple edge devices, enables these streaming services to meet strict throughput and latency requirements. Yet consumer devices exhibit volatile computational availability due to competing applications and unpredictable usage patterns. This volatility poses a fundamental challenge: how can we quantify the probability that a device, or ensemble of devices, will maintain the processing rate required by a streaming service? This paper presents an analytical framework for computational reliability in XEC, defined as the probability that instantaneous capacity meets demand at a specified Quality of Service (QoS) threshold. We derive closed-form reliability expressions under two information regimes: Minimal Information (MI), requiring only declared operational bounds, and historical data, which refines estimates via Maximum Likelihood Estimation from past observations. The framework extends to multi-device deployments, providing reliability expressions for series, parallel, and partitioned workload configurations. We derive optimal workload allocation rules and analytical bounds for device selection, equipping orchestrators with tractable tools to evaluate deployment feasibility and configure distributed streaming systems. We validate the framework using real-time object detection with YOLO11m model as a representative DI streaming workload; experiments on emulated XED environments demonstrate close agreement between analytical predictions, Monte Carlo sampling, and empirical measurements across diverse capacity and demand configurations.

Accurate estimation of the relative concentrations of chromophores in a spectroscopic photoacoustic (sPA) image can reveal immense structural, functional, and molecular information about physiological processes. However, due to nonlinearities and ill-posedness inherent to sPA imaging, concentration estimation is intractable. The Spectroscopic Photoacoustic Optical Inversion Autoencoder (SPOI-AE) aims to address the sPA optical inversion and spectral unmixing problems without assuming linearity. Herein, SPOI-AE was trained and tested on \textit{in vivo} mouse lymph node sPA images with unknown ground truth chromophore concentrations. SPOI-AE better reconstructs input sPA pixels than conventional algorithms while providing biologically coherent estimates for optical parameters, chromophore concentrations, and the percent oxygen saturation of tissue. SPOI-AE's unmixing accuracy was validated using a simulated mouse lymph node phantom ground truth.

Semantics has enabled 3D scene understanding and affordance-driven object interaction. However, robots operating in real-world environments face a critical limitation: they cannot anticipate how objects move. Long-horizon mobile manipulation requires closing the gap between semantics, geometry, and kinematics. In this work, we present MoMa-SG, a novel framework for building semantic-kinematic 3D scene graphs of articulated scenes containing a myriad of interactable objects. Given RGB-D sequences containing multiple object articulations, we temporally segment object interactions and infer object motion using occlusion-robust point tracking. We then lift point trajectories into 3D and estimate articulation models using a novel unified twist estimation formulation that robustly estimates revolute and prismatic joint parameters in a single optimization pass. Next, we associate objects with estimated articulations and detect contained objects by reasoning over parent-child relations at identified opening states. We also introduce the novel Arti4D-Semantic dataset, which uniquely combines hierarchical object semantics including parent-child relation labels with object axis annotations across 62 in-the-wild RGB-D sequences containing 600 object interactions and three distinct observation paradigms. We extensively evaluate the performance of MoMa-SG on two datasets and ablate key design choices of our approach. In addition, real-world experiments on both a quadruped and a mobile manipulator demonstrate that our semantic-kinematic scene graphs enable robust manipulation of articulated objects in everyday home environments. We provide code and data at: https://momasg.cs.uni-freiburg.de.

In the age of digital epidemiology, epidemiologists are faced by an increasing amount of data of growing complexity and dimensionality. Machine learning is a set of powerful tools that can help to analyze such enormous amounts of data. This chapter lays the methodological foundations for successfully applying machine learning in epidemiology. It covers the principles of supervised and unsupervised learning and discusses the most important machine learning methods. Strategies for model evaluation and hyperparameter optimization are developed and interpretable machine learning is introduced. All these theoretical parts are accompanied by code examples in R, where an example dataset on heart disease is used throughout the chapter.

Meshless methods are used to solve partial differential equations by approximating differential operators at a node as a weighted sum of values at its neighbours. One of the algorithms for generating nodes suitable for meshless numerical analysis is an n-dimensional Poisson disc sampling based method. It can handle complex geometries and supports variable node density, a crucial feature for adaptive analysis. We modify this method for parallel execution using coupled spatial indexing and work distribution hypertrees. The latter is prebuilt according to the node density function, ensuring that each leaf represents a balanced work unit. Threads advance separate fronts and claim work hypertree leaves as needed while avoiding leaves neighbouring those claimed by other threads. Node placement constraints and the partially prebuilt spatial hypertree are combined to eliminate the need to lock the tree while it is being modified. Thread collision handling is managed by the work hypertree at the leaf level, drastically reducing the number of required mutex acquisitions for point insertion collision checks. We explore the behaviour of the proposed algorithm and compare the performance with existing attempts at parallelisation and consider the requirements for adapting the developed algorithm to distributed systems.

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

Since Text-to-Speech systems typically don't produce waveforms directly, recent spoof detection studies use resynthesized waveforms from vocoders and neural audio codecs to simulate an attacker. Unlike vocoders, which are specifically designed for speech synthesis, neural audio codecs were originally developed for compressing audio for storage and transmission. However, their ability to discretize speech also sparked interest in language-modeling-based speech synthesis. Owing to this dual functionality, codec resynthesized data may be labeled as either bonafide or spoof. So far, very little research has addressed this issue. In this study, we present a challenging extension of the ASVspoof 5 dataset constructed for this purpose. We examine how different labeling choices affect detection performance and provide insights into labeling strategies.

In this work, we apply and compare two state-of-the-art eXplainability Artificial Intelligence (XAI) methods, the Integrated Gradients (IG) and the SHapley Additive exPlanations (SHAP), that explain the fault diagnosis decisions of a highly accurate Long Short-Time Memory (LSTM) classifier. The classifier is trained to detect faults in a benchmark non-linear chemical process, the Tennessee Eastman Process (TEP). It is highlighted how XAI methods can help identify the subsystem of the process where the fault occurred. Using our knowledge of the process, we note that in most cases the same features are indicated as the most important for the decision, while insome cases the SHAP method seems to be more informative and closer to the root cause of the fault. Finally, since the used XAI methods are model-agnostic, the proposed approach is not limited to the specific process and can also be used in similar problems.

We study the implicit bias of momentum-based optimizers on homogeneous models. We first extend existing results on the implicit bias of steepest descent in homogeneous models to normalized steepest descent with an optional learning rate schedule. We then show that for smooth homogeneous models, momentum steepest descent algorithms like Muon (spectral norm), MomentumGD ($\ell_2$ norm), and Signum ($\ell_\infty$ norm) are approximate steepest descent trajectories under a decaying learning rate schedule, proving that these algorithms too have a bias towards KKT points of the corresponding margin maximization problem. We extend the analysis to Adam (without the stability constant), which maximizes the $\ell_\infty$ margin, and to Muon-Signum and Muon-Adam, which maximize a hybrid norm. Our experiments corroborate the theory and show that the identity of the margin maximized depends on the choice of optimizer. Overall, our results extend earlier lines of work on steepest descent in homogeneous models and momentum-based optimizers in linear models.

Zero-knowledge proof generation imposes stringent timing and reliability constraints on blockchain systems. For ZK-rollups, delayed proofs cause finality lag and economic loss; for Ethereum's emerging L1 zkEVM, proofs must complete within the 12-second slot window to enable stateless validation. The Ethereum Foundation's Ethproofs initiative coordinates multiple independent zkVMs across proving clusters to achieve real-time block proving, yet no principled orchestration framework addresses the joint challenges of (i) strict head-of-chain ordering, (ii) sub-slot latency bounds, (iii) fault-tolerant task reassignment, and (iv) prover-agnostic workflow composition. We present push0, a cloud-native proof orchestration system that decouples prover binaries from scheduling infrastructure. push0 employs an event-driven dispatcher--collector architecture over persistent priority queues, enforcing block-sequential proving while exploiting intra-block parallelism. We formalize requirements drawn from production ZK-rollup operations and the Ethereum real-time proving specification, then demonstrate via production Kubernetes cluster experiments that push0 achieves 5 ms median orchestration overhead with 99--100% scaling efficiency at 32 dispatchers for realistic workloads--overhead negligible (less than 0.1%) relative to typical proof computation times of 7+ seconds. Controlled Docker experiments validate these results, showing comparable performance (3--10 ms P50) when network variance is eliminated. Production deployment on the Zircuit zkrollup (14+ million mainnet blocks since March 2025) provides ecological validity for these controlled experiments. Our design enables seamless integration of heterogeneous zkVMs, supports automatic task recovery via message persistence, and provides the scheduling primitives necessary for both centralized rollup operators and decentralized multi-prover networks.

We propose the Subtractive Modulative Network (SMN), a novel, parameter-efficient Implicit Neural Representation (INR) architecture inspired by classical subtractive synthesis. The SMN is designed as a principled signal processing pipeline, featuring a learnable periodic activation layer (Oscillator) that generates a multi-frequency basis, and a series of modulative mask modules (Filters) that actively generate high-order harmonics. We provide both theoretical analysis and empirical validation for our design. Our SMN achieves a PSNR of $40+$ dB on two image datasets, comparing favorably against state-of-the-art methods in terms of both reconstruction accuracy and parameter efficiency. Furthermore, consistent advantage is observed on the challenging 3D NeRF novel view synthesis task. Supplementary materials are available at https://inrainbws.github.io/smn/.

This work presents HAWX, a hardware-aware scalable exploration framework that employs multi-level sensitivity scoring at different DNN abstraction levels (operator, filter, layer, and model) to guide selective integration of heterogeneous AxC blocks. Supported by predictive models for accuracy, power, and area, HAWX accelerates the evaluation of candidate configurations, achieving over 23* speedup in a layer-level search with two candidate approximate blocks and more than (3*106)* speedup at the filter-level search only for LeNet-5, while maintaining accuracy comparable to exhaustive search. Experiments across state-of-the-art DNN benchmarks such as VGG-11, ResNet-18, and EfficientNetLite demonstrate that the efficiency benefits of HAWX scale exponentially with network size. The HAWX hardware-aware search algorithm supports both spatial and temporal accelerator architectures, leveraging either off-the-shelf approximate components or customized designs.

Spatial audio understanding aims to enable machines to interpret complex auditory scenes, particularly when sound sources move over time. In this work, we study Spatial Audio Question Answering (Spatial AQA) with a focus on movement reasoning, where a model must infer object motion, position, and directional changes directly from stereo audio. First, we introduce a movement-centric spatial audio augmentation framework that synthesizes diverse motion patterns from isolated mono audio events, enabling controlled and scalable training data generation. Second, we propose an end-to-end multimodal finetuning approach with a thinking mode, which allows audio-language models to produce explicit intermediate reasoning steps before predicting an answer. Third, we investigate the impact of query-conditioned source separation as a preprocessing stage and compare three inference regimes: no masking, an audio grounding model (AGM), and ground-truth masks. Our results show that reasoning amplifies the benefits of source separation, with thinking mode showing significant improvement of +5.1% when a single event is present in the question. These findings highlight the interplay between movement modeling, reasoning, and separation quality, offering new insights for advancing spatial audio understanding.

Skeletal muscle-based biohybrid actuators have proved to be a promising component in soft robotics, offering efficient movement. However, their intrinsic biological variability and nonlinearity pose significant challenges for controllability and predictability. To address these issues, this study investigates the application of supervised learning, a form of machine learning, to model and predict the behavior of biohybrid machines (BHMs), focusing on a muscle ring anchored on flexible polymer pillars. First, static prediction models (i.e., random forest and neural network regressors) are trained to estimate the maximum exerted force achieved from input variables such as muscle sample, electrical stimulation parameters, and baseline exerted force. Second, a dynamic modeling framework, based on Long Short-Term Memory networks, is developed to serve as a digital twin, replicating the time series of exerted forces observed in response to electrical stimulation. Both modeling approaches demonstrate high predictive accuracy. The best performance of the static models is characterized by R2 of 0.9425, whereas the dynamic model achieves R2 of 0.9956. The static models can enable optimization of muscle actuator performance for targeted applications and required force outcomes, while the dynamic model provides a foundation for developing robustly adaptive control strategies in future biohybrid robotic systems.

With the introduction of cyber-physical genome sequencing and editing technologies, such as CRISPR, researchers can more easily access tools to investigate and create remedies for a variety of topics in genetics and health science (e.g. agriculture and medicine). As the field advances and grows, new concerns present themselves in the ability to predict the off-target behavior. In this work, we explore the underlying biological and chemical model from a data driven perspective. Additionally, we present a machine learning based solution named \textit{Guide-Guard} to predict the behavior of the system given a gRNA in the CRISPR gene-editing process with 84\% accuracy. This solution is able to be trained on multiple different genes at the same time while retaining accuracy.

In the fast-evolving field of artificial intelligence, where models are increasingly growing in complexity and size, the availability of labeled data for training deep learning models has become a significant challenge. Addressing complex problems like object detection demands considerable time and resources for data labeling to achieve meaningful results. For companies developing such applications, this entails extensive investment in highly skilled personnel or costly outsourcing. This research work aims to demonstrate that enhancing feature extractors can substantially alleviate this challenge, enabling models to learn more effective representations with less labeled data. Utilizing a self-supervised learning strategy, we present a model trained on unlabeled data that outperforms state-of-the-art feature extractors pre-trained on ImageNet and particularly designed for object detection tasks. Moreover, the results demonstrate that our approach encourages the model to focus on the most relevant aspects of an object, thus achieving better feature representations and, therefore, reinforcing its reliability and robustness.

Multi-objective optimisation using BRIGHT has proven insightful and effective in prostate cancer brachytherapy treatment planning. BRachytherapy via artificially Intelligent GOMEA-Heuristic based Treatment planning (BRIGHT) generates multiple treatment plans, each with a different trade-off between tumour coverage and organs-at-risk sparing. BRIGHT was recently extended to cervical cancer brachytherapy. In this study, we present a novel, custom-developed graphical user interface (GUI) that enables plan navigation, pairwise comparisons, dose distribution visualisation, and possibility for adjustments - essential for efficient clinical use of BRIGHT. End-user validation of BRIGHT with the dedicated GUI was conducted for cervical cancer brachytherapy by emulating clinical practice in ten previously treated patients. A multidisciplinary brachytherapy team used BRIGHT to create new treatment plans. GUI usability was assessed using the System Usability Scale (SUS). BRIGHT plan quality was compared to clinical practice via blinded one-on-one comparisons. The GUI offered helpful features for plan navigation and evaluation, giving users quick insight into whether planning aims are achievable and what treatment options are available. The overall SUS score was 83.3, indicating an 'excellent' system. BRIGHT outperformed clinical practice in five out of ten patients regarding the coverage-sparing trade-off and performed equally well in the remaining five. The BRIGHT plan was preferred over the clinical plan in eight out of ten patients, four of which showed clinically relevant differences. The clinical plan was preferred in two patients, neither with clinically relevant differences. In conclusion, BRIGHT, with its dedicated GUI, is a clinically viable and user-friendly tool for treatment planning in cervical cancer brachytherapy.

Accurate and computationally efficient 3D medical image segmentation remains a critical challenge in clinical workflows. Transformer-based architectures often demonstrate superior global contextual modeling but at the expense of excessive parameter counts and memory demands, restricting their clinical deployment. We propose RefineFormer3D, a lightweight hierarchical transformer architecture that balances segmentation accuracy and computational efficiency for volumetric medical imaging. The architecture integrates three key components: (i) GhostConv3D-based patch embedding for efficient feature extraction with minimal redundancy, (ii) MixFFN3D module with low-rank projections and depthwise convolutions for parameter-efficient feature extraction, and (iii) a cross-attention fusion decoder enabling adaptive multi-scale skip connection integration. RefineFormer3D contains only 2.94M parameters, substantially fewer than contemporary transformer-based methods. Extensive experiments on ACDC and BraTS benchmarks demonstrate that RefineFormer3D achieves 93.44\% and 85.9\% average Dice scores respectively, outperforming or matching state-of-the-art methods while requiring significantly fewer parameters. Furthermore, the model achieves fast inference (8.35 ms per volume on GPU) with low memory requirements, supporting deployment in resource-constrained clinical environments. These results establish RefineFormer3D as an effective and scalable solution for practical 3D medical image segmentation.

Weight-space models learn directly from the parameters of neural networks, enabling tasks such as predicting their accuracy on new datasets. Naive methods -- like applying MLPs to flattened parameters -- perform poorly, making the design of better weight-space architectures a central challenge. While prior work leveraged permutation symmetries in standard networks to guide such designs, no analogous analysis or tailored architecture yet exists for Kolmogorov-Arnold Networks (KANs). In this work, we show that KANs share the same permutation symmetries as MLPs, and propose the KAN-graph, a graph representation of their computation. Building on this, we develop WS-KAN, the first weight-space architecture that learns on KANs, which naturally accounts for their symmetry. We analyze WS-KAN's expressive power, showing it can replicate an input KAN's forward pass - a standard approach for assessing expressiveness in weight-space architectures. We construct a comprehensive ``zoo'' of trained KANs spanning diverse tasks, which we use as benchmarks to empirically evaluate WS-KAN. Across all tasks, WS-KAN consistently outperforms structure-agnostic baselines, often by a substantial margin. Our code is available at https://github.com/BarSGuy/KAN-Graph-Metanetwork.

Recommender systems shape individual choices through feedback loops in which user behavior and algorithmic recommendations coevolve over time. The systemic effects of these loops remain poorly understood, in part due to unrealistic assumptions in existing simulation studies. We propose a feedback-loop model that captures implicit feedback, periodic retraining, probabilistic adoption of recommendations, and heterogeneous recommender systems. We apply the framework on online retail and music streaming data and analyze systemic effects of the feedback loop. We find that increasing recommender adoption may lead to a progressive diversification of individual consumption, while collective demand is redistributed in model- and domain-dependent ways, often amplifying popularity concentration. Temporal analyses further reveal that apparent increases in individual diversity observed in static evaluations are illusory: when adoption is fixed and time unfolds, individual diversity consistently decreases across all models. Our results highlight the need to move beyond static evaluations and explicitly account for feedback-loop dynamics when designing recommender systems.

Existing evaluations of agents with memory typically assess memorization and action in isolation. One class of benchmarks evaluates memorization by testing recall of past conversations or text but fails to capture how memory is used to guide future decisions. Another class focuses on agents acting in single-session tasks without the need for long-term memory. However, in realistic settings, memorization and action are tightly coupled: agents acquire memory while interacting with the environment, and subsequently rely on that memory to solve future tasks. To capture this setting, we introduce MemoryArena, a unified evaluation gym for benchmarking agent memory in multi-session Memory-Agent-Environment loops. The benchmark consists of human-crafted agentic tasks with explicitly interdependent subtasks, where agents must learn from earlier actions and feedback by distilling experiences into memory, and subsequently use that memory to guide later actions to solve the overall task. MemoryArena supports evaluation across web navigation, preference-constrained planning, progressive information search, and sequential formal reasoning, and reveals that agents with near-saturated performance on existing long-context memory benchmarks like LoCoMo perform poorly in our agentic setting, exposing a gap in current evaluations for agents with memory.

Fault injection attacks on embedded neural network models have been shown as a potent threat. Numerous works studied resilience of models from various points of view. As of now, there is no comprehensive study that would evaluate the influence of number representations used for model parameters against electromagnetic fault injection (EMFI) attacks. In this paper, we investigate how four different number representations influence the success of an EMFI attack on embedded neural network models. We chose two common floating-point representations (32-bit, and 16-bit), and two integer representations (8-bit, and 4-bit). We deployed four common image classifiers, ResNet-18, ResNet-34, ResNet-50, and VGG-11, on an embedded memory chip, and utilized a low-cost EMFI platform to trigger faults. Our results show that while floating-point representations exhibit almost a complete degradation in accuracy (Top-1 and Top-5) after a single fault injection, integer representations offer better resistance overall. Especially, when considering the the 8-bit representation on a relatively large network (VGG-11), the Top-1 accuracies stay at around 70% and the Top-5 at around 90%.

This study investigates generative artificial intelligence (GenAI) usage of university students who study alongside their professional career. Previous literature has paid little attention to part-time students and the intersectional use of GenAI between education and business. This study examines with a grounded theory approach the characteristics of GenAI usage of part-time students. Eleven students from a distance learning university were interviewed. Three causal and four intervening conditions, as well as strategies were identified, to influence the use of GenAI. The study highlights both the potential and challenges of GenAI usage in education and business. While GenAI can significantly enhance productivity and learning outcomes, concerns about ethical implications, reliability, and the risk of academic misconduct persist. The developed grounded model offers a comprehensive understanding of GenAI usage among students, providing valuable insights for educators, policymakers, and developers of GenAI tools seeking to bridge the gap between education and business.

Probing is widely adopted in computer vision to faithfully evaluate self-supervised learning (SSL) embeddings, as fine-tuning may misrepresent their inherent quality. In contrast, audio SSL models still rely on fine-tuning because simple probing fails to unlock their full potential and alters their rankings when competing for SOTA on AudioSet. Hence, a robust and efficient probing mechanism is required to guide the trajectory of audio SSL towards reliable and reproducible methods. We introduce Convex Gated Probing (CGP), a prototype-based method that drastically closes the gap between fine-tuning and probing in audio. CGP efficiently utilizes all frozen layers via a gating mechanism and exposes the location of latent task-relevant information. Guided by CGP, we rework the entire SSL pipeline of current SOTA audio models that use legacy implementations of prior SSL methods. By refining data preprocessing, model architecture, and pre-training recipe, we introduce Better Audio Transformer (BAT), and establish new SOTA on audio benchmarks.

The prevalence of malicious packages in open-source repositories, such as PyPI, poses a critical threat to the software supply chain. While Large Language Models (LLMs) have emerged as a promising tool for automated security tasks, their effectiveness in detecting malicious packages and indicators remains underexplored. This paper presents a systematic evaluation of 13 LLMs for detecting malicious software packages. Using a curated dataset of 4,070 packages (3,700 benign and 370 malicious), we evaluate model performance across two tasks: binary classification (package detection) and multi-label classification (identification of specific malicious indicators). We further investigate the impact of prompting strategies, temperature settings, and model specifications on detection accuracy. We find a significant "granularity gap" in LLMs' capabilities. While GPT-4.1 achieves near-perfect performance in binary detection (F1 $\approx$ 0.99), performance degrades by approximately 41\% when the task shifts to identifying specific malicious indicators. We observe that general models are best for filtering out the majority of threats, while specialized coder models are better at detecting attacks that follow a strict, predictable code structure. Our correlation analysis indicates that parameter size and context width have negligible explanatory power regarding detection accuracy. We conclude that while LLMs are powerful detectors at the package level, they lack the semantic depth required for precise identification at the granular indicator level.

Achieving cooperation among self-interested agents remains a fundamental challenge in multi-agent reinforcement learning. Recent work showed that mutual cooperation can be induced between "learning-aware" agents that account for and shape the learning dynamics of their co-players. However, existing approaches typically rely on hardcoded, often inconsistent, assumptions about co-player learning rules or enforce a strict separation between "naive learners" updating on fast timescales and "meta-learners" observing these updates. Here, we demonstrate that the in-context learning capabilities of sequence models allow for co-player learning awareness without requiring hardcoded assumptions or explicit timescale separation. We show that training sequence model agents against a diverse distribution of co-players naturally induces in-context best-response strategies, effectively functioning as learning algorithms on the fast intra-episode timescale. We find that the cooperative mechanism identified in prior work-where vulnerability to extortion drives mutual shaping-emerges naturally in this setting: in-context adaptation renders agents vulnerable to extortion, and the resulting mutual pressure to shape the opponent's in-context learning dynamics resolves into the learning of cooperative behavior. Our results suggest that standard decentralized reinforcement learning on sequence models combined with co-player diversity provides a scalable path to learning cooperative behaviors.

Large Language Models (LLMs) are beginning to reshape how media professionals verify information, yet automated support for detecting check-worthy claims a key step in the fact-checking process remains limited. We introduce the Multi-Check-Worthy (MultiCW) dataset, a balanced multilingual benchmark for check-worthy claim detection spanning 16 languages, 7 topical domains, and 2 writing styles. It consists of 123,722 samples, evenly distributed between noisy (informal) and structured (formal) texts, with balanced representation of check-worthy and non-check-worthy classes across all languages. To probe robustness, we also introduce an equally balanced out-of-distribution evaluation set of 27,761 samples in 4 additional languages. To provide baselines, we benchmark 3 common fine-tuned multilingual transformers against a diverse set of 15 commercial and open LLMs under zero-shot settings. Our findings show that fine-tuned models consistently outperform zero-shot LLMs on claim classification and show strong out-of-distribution generalization across languages, domains, and styles. MultiCW provides a rigorous multilingual resource for advancing automated fact-checking and enables systematic comparisons between fine-tuned models and cutting-edge LLMs on the check-worthy claim detection task.

Just as the $λ$-calculus uses three primitives (abstraction, application, variable) as the foundation of functional programming, Overlay-Calculus uses three primitives (record, definition, inheritance) as the foundation of declarative programming. It trivially embeds the $λ$-calculus, although the entire semantics builds on only naive set theory; as a consequence, all constructs including inheritance are inherently commutative, idempotent, and associative; the linearization problem of multiple inheritance simply does not arise. This induces a fully abstract semantics of the lazy $λ$-calculus with respect to Böhm tree equivalence. Overlay-Calculus is distilled from the Overlay language, a practical implementation in which we observed further emergent phenomena: the Expression Problem dissolves, programs are CPS-agnostic, records natively encode random-access memory, and self-reference resolves to multiple targets. These properties suggest applications to configuration languages, dependency injection, object-oriented programming, composable effect systems, modular software architectures, file-system-as-compiler, general-purpose programming, and no-code development.

Arabic dialects have long been under-represented in Natural Language Processing (NLP) research due to their non-standardization and high variability, which pose challenges for computational modeling. Recent advances in the field, such as Large Language Models (LLMs), offer promising avenues to address this gap by enabling Arabic to be modeled as a pluricentric language rather than a monolithic system. This paper presents Aladdin-FTI, our submission to the AMIYA shared task. The proposed system is designed to both generate and translate dialectal Arabic (DA). Specifically, the model supports text generation in Moroccan, Egyptian, Palestinian, Syrian, and Saudi dialects, as well as bidirectional translation between these dialects, Modern Standard Arabic (MSA), and English. The code and trained model are publicly available.

Scaling language models to long contexts is often bottlenecked by the size of the key-value (KV) cache. In deployed settings, long contexts are typically managed through compaction in token space via summarization. However, summarization can be highly lossy, substantially harming downstream performance. Recent work on Cartridges has shown that it is possible to train highly compact KV caches in latent space that closely match full-context performance, but at the cost of slow and expensive end-to-end optimization. This work describes an approach for fast context compaction in latent space through Attention Matching, which constructs compact keys and values to reproduce attention outputs and preserve attention mass at a per-KV-head level. We show that this formulation naturally decomposes into simple subproblems, some of which admit efficient closed-form solutions. Within this framework, we develop a family of methods that significantly push the Pareto frontier of compaction time versus quality, achieving up to 50x compaction in seconds on some datasets with little quality loss.

We present the first high-probability regret bound for classical online Q-learning in infinite-horizon discounted Markov decision processes, without relying on optimism or bonus terms. We first analyze Boltzmann Q-learning with decaying temperature and show that its regret depends critically on the suboptimality gap of the MDP: for sufficiently large gaps, the regret is sublinear, while for small gaps it deteriorates and can approach linear growth. To address this limitation, we study a Smoothed $ε_n$-Greedy exploration scheme that combines $ε_n$-greedy and Boltzmann exploration, for which we prove a gap-robust regret bound of near-$\tilde{O}(N^{9/10})$. To analyze these algorithms, we develop a high-probability concentration bound for contractive Markovian stochastic approximation with iterate- and time-dependent transition dynamics. This bound may be of independent interest as the contraction factor in our bound is governed by the mixing time and is allowed to converge to one asymptotically.

We analyze speech embeddings from structured clinical interviews of psychotic patients and healthy controls by treating language production as a high-dimensional dynamical process. Lyapunov exponent (LE) spectra are computed from word-level and answer-level embeddings generated by two distinct large language models, allowing us to assess the stability of the conclusions with respect to different embedding presentations. Word-level embeddings exhibit uniformly contracting dynamics with no positive LE, while answer-level embeddings, in spite of the overall contraction, display a number of positive LEs and higher-dimensional attractors. The resulting LE spectra robustly separate psychotic from healthy speech, while differentiation within the psychotic group is not statistically significant overall, despite a tendency of the most severe cases to occupy distinct dynamical regimes. These findings indicate that nonlinear dynamical invariants of speech embeddings provide a physics-inspired probe of disordered cognition whose conclusions remain stable across embedding models.

Encoding classical data into quantum states is a central bottleneck in quantum machine learning: many widely used encodings are circuit-inefficient, requiring deep circuits and substantial quantum resources, which limits scalability on quantum hardware. In this work, we propose TNQE, a circuit-efficient quantum data encoding framework built on structured unitary tensor network (TN) representations. TNQE first represents each classical input via a TN decomposition and then compiles the resulting tensor cores into an encoding circuit through two complementary core-to-circuit strategies. To make this compilation trainable while respecting the unitary nature of quantum operations, we introduce a unitary-aware constraint that parameterizes TN cores as learnable block unitaries, enabling them to be directly optimized and directly encoded as quantum operators. The proposed TNQE framework enables explicit control over circuit depth and qubit resources, allowing the construction of shallow, resource-efficient circuits. Across a range of benchmarks, TNQE achieves encoding circuits as shallow as $0.04\times$ the depth of amplitude encoding, while naturally scaling to high-resolution images ($256 \times 256$) and demonstrating practical feasibility on real quantum hardware.

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

In this work, we develop, for the first time, a supervised classification framework with class-dependent rewards (CDR) to predict $\geq$MM flares within 24 hr. We construct multiple datasets, covering knowledge-informed features and line-of sight (LOS) magnetograms. We also apply three deep learning models (CNN, CNN-BiLSTM, and Transformer) and three CDR counterparts (CDR-CNN, CDR-CNN-BiLSTM, and CDR-Transformer). First, we analyze the importance of LOS magnetic field parameters with the Transformer, then compare its performance using LOS-only, vector-only, and combined magnetic field parameters. Second, we compare flare prediction performance based on CDR models versus deep learning counterparts. Third, we perform sensitivity analysis on reward engineering for CDR models. Fourth, we use the SHAP method for model interpretability. Finally, we conduct performance comparison between our models and NASA/CCMC. The main findings are: (1)Among LOS feature combinations, R_VALUE and AREA_ACR consistently yield the best results. (2)Transformer achieves better performance with combined LOS and vector magnetic field data than with either alone. (3)Models using knowledge-informed features outperform those using magnetograms. (4)While CNN and CNN-BiLSTM outperform their CDR counterparts on magnetograms, CDR-Transformer is slightly superior to its deep learning counterpart when using knowledge-informed features. Among all models, CDR-Transformer achieves the best performance. (5)The predictive performance of the CDR models is not overly sensitive to the reward choices.(6)Through SHAP analysis, the CDR model tends to regard TOTUSJH as more important, while the Transformer tends to prioritize R_VALUE more.(7)Under identical prediction time and active region (AR) number, the CDR-Transformer shows superior predictive capabilities compared to NASA/CCMC.

Speech emotion recognition (SER) has traditionally relied on categorical or dimensional labels. However, this technique is limited in representing both the diversity and interpretability of emotions. To overcome this limitation, we focus on color attributes, such as hue, saturation, and value, to represent emotions as continuous and interpretable scores. We annotated an emotional speech corpus with color attributes via crowdsourcing and analyzed them. Moreover, we built regression models for color attributes in SER using machine learning and deep learning, and explored the multitask learning of color attribute regression and emotion classification. As a result, we demonstrated the relationship between color attributes and emotions in speech, and successfully developed color attribute regression models for SER. We also showed that multitask learning improved the performance of each task.

Interactive large language model (LLM) agents operating via multi-turn dialogue and multi-step tool calling are increasingly used in production. Benchmarks for these agents must both reliably compare models and yield on-policy training data. Prior agentic benchmarks (e.g., tau-bench, tau2-bench, AppWorld) rely on fully deterministic backends, which are costly to build and iterate. We propose Proxy State-Based Evaluation, an LLM-driven simulation framework that preserves final state-based evaluation without a deterministic database. Specifically, a scenario specifies the user goal, user/system facts, expected final state, and expected agent behavior, and an LLM state tracker infers a structured proxy state from the full interaction trace. LLM judges then verify goal completion and detect tool/user hallucinations against scenario constraints. Empirically, our benchmark produces stable, model-differentiating rankings across families and inference-time reasoning efforts, and its on-/off-policy rollouts provide supervision that transfers to unseen scenarios. Careful scenario specification yields near-zero simulator hallucination rates as supported by ablation studies. The framework also supports sensitivity analyses over user personas. Human-LLM judge agreement exceeds 90%, indicating reliable automated evaluation. Overall, proxy state-based evaluation offers a practical, scalable alternative to deterministic agentic benchmarks for industrial LLM agents.

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

We revisit the classical problem of universal prediction of stochastic sequences with a finite time horizon $T$ known to the learner. The question we investigate is whether it is possible to derive vanishing regret bounds that hold with high probability, complementing existing bounds from the literature that hold in expectation. We propose such high-probability bounds which have a very similar form as the prior expectation bounds. For the case of universal prediction of a stochastic process over a countable alphabet, our bound states a convergence rate of $\mathcal{O}(T^{-1/2} δ^{-1/2})$ with probability as least $1-δ$ compared to prior known in-expectation bounds of the order $\mathcal{O}(T^{-1/2})$. We also propose an impossibility result which proves that it is not possible to improve the exponent of $δ$ in a bound of the same form without making additional assumptions.

Circuit cutting decomposes a large quantum circuit into a collection of smaller subcircuits. The outputs of these subcircuits are then classically reconstructed to recover the original expectation values. While prior work characterises cutting overhead largely in terms of subcircuit counts and sampling complexity, its end-to-end impact on iterative, estimator-driven training pipelines remains insufficiently measured from a systems perspective. In this paper, we propose a cut-aware estimator execution pipeline that treats circuit cutting as a staged distributed workload and instruments each estimator query into partitioning, subexperiment generation, parallel execution, and classical reconstruction phases. Using logged runtime traces and learning outcomes on two binary classification workloads (Iris and MNIST), we quantify cutting overheads, scaling limits, and sensitivity to injected stragglers, and we evaluate whether accuracy and robustness are preserved under matched training budgets. Our measurements show that cutting introduces substantial end-to-end overheads that grow with the number of cuts, and that reconstruction constitutes a dominant fraction of per-query time, bounding achievable speed-up under increased parallelism. Despite these systems costs, test accuracy and robustness are preserved in the measured regimes, with configuration-dependent improvements observed in some cut settings. These results indicate that practical scaling of circuit cutting for learning workloads hinges on reducing and overlapping reconstruction and on scheduling policies that account for barrier-dominated critical paths.

Learning latent actions from action-free video has emerged as a powerful paradigm for scaling up controllable world model learning. Latent actions provide a natural interface for users to iteratively generate and manipulate videos. However, most existing approaches rely on monolithic inverse and forward dynamics models that learn a single latent action to control the entire scene, and therefore struggle in complex environments where multiple entities act simultaneously. This paper introduces Factored Latent Action Model (FLAM), a factored dynamics framework that decomposes the scene into independent factors, each inferring its own latent action and predicting its own next-step factor value. This factorized structure enables more accurate modeling of complex multi-entity dynamics and improves video generation quality in action-free video settings compared to monolithic models. Based on experiments on both simulation and real-world multi-entity datasets, we find that FLAM outperforms prior work in prediction accuracy and representation quality, and facilitates downstream policy learning, demonstrating the benefits of factorized latent action models.

Time series data are prone to noise in various domains, and training samples may contain low-predictability patterns that deviate from the normal data distribution, leading to training instability or convergence to poor local minima. Therefore, mitigating the adverse effects of low-predictability samples is crucial for time series analysis tasks such as time series forecasting (TSF) and time series classification (TSC). While many deep learning models have achieved promising performance, few consider how to identify and penalize low-predictability samples to improve model performance from the training perspective. To fill this gap, we propose a general Amortized Predictability-aware Training Framework (APTF) for both TSF and TSC. APTF introduces two key designs that enable the model to focus on high-predictability samples while still learning appropriately from low-predictability ones: (i) a Hierarchical Predictability-aware Loss (HPL) that dynamically identifies low-predictability samples and progressively expands their loss penalty as training evolves, and (ii) an amortization model that mitigates predictability estimation errors caused by model bias, further enhancing HPL's effectiveness. The code is available at https://github.com/Meteor-Stars/APTF.

We investigate space-time trade-offs for population protocols in sparse interaction graphs. In complete interaction graphs, optimal space-time trade-offs are known for the leader election and exact majority problems. However, it has remained open if other graph families exhibit similar space-time complexity trade-offs, as existing lower bound techniques do not extend beyond highly dense graphs. In this work, we show that -- unlike in complete graphs -- population protocols on bounded-degree trees do not exhibit significant asymptotic space-time trade-offs for leader election and exact majority. For these problems, we give constant-space protocols that have near-optimal worst-case expected stabilisation time. These new protocols achieve a linear speed-up compared to the state-of-the-art. Our results are based on two novel protocols, which we believe are of independent interest. First, we give a new fast self-stabilising 2-hop colouring protocol for general interaction graphs, whose stabilisation time we bound using a stochastic drift argument. Second, we give a self-stabilising tree orientation algorithm that builds a rooted tree in optimal time on any tree. As a consequence, we can use simple constant-state protocols designed for directed trees to solve leader election and exact majority fast. For example, we show that ``directed'' annihilation dynamics solve exact majority in $O(n^2 \log n)$ steps on directed trees.

Modeling multiscale patterns is crucial for long-term time series forecasting (TSF). However, redundancy and noise in time series, together with semantic gaps between non-adjacent scales, make the efficient alignment and integration of multi-scale temporal dependencies challenging. To address this, we propose SEMixer, a lightweight multiscale model designed for long-term TSF. SEMixer features two key components: a Random Attention Mechanism (RAM) and a Multiscale Progressive Mixing Chain (MPMC). RAM captures diverse time-patch interactions during training and aggregates them via dropout ensemble at inference, enhancing patch-level semantics and enabling MLP-Mixer to better model multi-scale dependencies. MPMC further stacks RAM and MLP-Mixer in a memory-efficient manner, achieving more effective temporal mixing. It addresses semantic gaps across scales and facilitates better multiscale modeling and forecasting performance. We not only validate the effectiveness of SEMixer on 10 public datasets, but also on the \textit{2025 CCF AlOps Challenge} based on 21GB real wireless network data, where SEMixer achieves third place. The code is available at the link https://github.com/Meteor-Stars/SEMixer.

Bayesian quadrature is a probabilistic, model-based approach to numerical integration, the estimation of intractable integrals, or expectations. Although Bayesian quadrature was popularised already in the 1980s, no systematic and comprehensive treatment has been published. The purpose of this survey is to fill this gap. We review the mathematical foundations of Bayesian quadrature from different points of view; present a systematic taxonomy for classifying different Bayesian quadrature methods along the three axes of modelling, inference, and sampling; collect general theoretical guarantees; and provide a controlled numerical study that explores and illustrates the effect of different choices along the axes of the taxonomy. We also provide a realistic assessment of practical challenges and limitations to application of Bayesian quadrature methods and include an up-to-date and nearly exhaustive bibliography that covers not only machine learning and statistics literature but all areas of mathematics and engineering in which Bayesian quadrature or equivalent methods have seen use.

Surface extraction from implicit neural representations modelling a single class surface is a well-known task. However, there exist no surface extraction methods from an implicit representation of multiple classes that guarantee topological correctness and no holes. In this work, we lay the groundwork by introducing a 2D boundary extraction algorithm for the multi-class case focusing on topological consistency and water-tightness, which also allows for setting minimum detail restraint on the approximation. Finally, we evaluate our algorithm using geological modelling data, showcasing its adaptiveness and ability to honour complex topology.

Deep learning has improved automated electrocardiogram (ECG) classification, but limited insight into prediction reliability hinders its use in safety-critical settings. This paper proposes UCTECG-Net, an uncertainty-aware hybrid architecture that combines one-dimensional convolutions and Transformer encoders to process raw ECG signals and their spectrograms jointly. Evaluated on the MIT-BIH Arrhythmia and PTB Diagnostic datasets, UCTECG-Net outperforms LSTM, CNN1D, and Transformer baselines in terms of accuracy, precision, recall and F1 score, achieving up to 98.58% accuracy on MIT-BIH and 99.14% on PTB. To assess predictive reliability, we integrate three uncertainty quantification methods (Monte Carlo Dropout, Deep Ensembles, and Ensemble Monte Carlo Dropout) into all models and analyze their behavior using an uncertainty-aware confusion matrix and derived metrics. The results show that UCTECG-Net, particularly with Ensemble or EMCD, provides more reliable and better-aligned uncertainty estimates than competing architectures, offering a stronger basis for risk-aware ECG decision support.

This paper introduces a novel approach to sea ice modeling using Graph Neural Networks (GNNs), utilizing the natural graph structure of sea ice, where nodes represent individual ice pieces, and edges model the physical interactions, including collisions. This concept is developed within a one-dimensional framework as a foundational step. Traditional numerical methods, while effective, are computationally intensive and less scalable. By utilizing GNNs, the proposed model, termed the Collision-captured Network (CN), integrates data assimilation (DA) techniques to effectively learn and predict sea ice dynamics under various conditions. The approach was validated using synthetic data, both with and without observed data points, and it was found that the model accelerates the simulation of trajectories without compromising accuracy. This advancement offers a more efficient tool for forecasting in marginal ice zones (MIZ) and highlights the potential of combining machine learning with data assimilation for more effective and efficient modeling.

Neural operators offer an effective framework for learning solutions of partial differential equations for many physical systems in a resolution-invariant and data-driven manner. Existing neural operators, however, often suffer from instability in multi-layer iteration and long-horizon rollout, which stems from the unconstrained Euclidean latent space updates that violate the geometric and conservation laws. To address this challenge, we propose to constrain manifolds with low-rank Lie algebra parameterization that performs group action updates on the latent representation. Our method, termed Manifold Constraining based on Lie group (MCL), acts as an efficient \emph{plug-and-play} module that enforces geometric inductive bias to existing neural operators. Extensive experiments on various partial differential equations, such as 1-D Burgers and 2-D Navier-Stokes, over a wide range of parameters and steps demonstrate that our method effectively lowers the relative prediction error by 30-50\% at the cost of 2.26\% of parameter increase. The results show that our approach provides a scalable solution for improving long-term prediction fidelity by addressing the principled geometric constraints absent in the neural operator updates.

Neurochaos Learning (NL) has shown promise in recent times over traditional deep learning due to its two key features: ability to learn from small sized training samples, and low compute requirements. In prior work, NL has been implemented and extensively tested on separable and time series data, and demonstrated its superior performance on both classification and regression tasks. In this paper, we investigate the next step in NL, viz., applying NL to linked data, in particular, data that is represented in the form of knowledge graphs. We integrate linked data into NL by implementing node aggregation on knowledge graphs, and then feeding the aggregated node features to the simplest NL architecture: ChaosNet. We demonstrate the results of our implementation on homophilic graph datasets as well as heterophilic graph datasets of verying heterophily. We show better efficacy of our approach on homophilic graphs than on heterophilic graphs. While doing so, we also present our analysis of the results, as well as suggestions for future work.

Large language models (LLMs) are trained on web-scale corpora that exhibit steep power-law distributions, in which the distribution of knowledge is highly long-tailed, with most appearing infrequently. While scaling has improved average-case performance, persistent failures on low-frequency, domain-specific, cultural, and temporal knowledge remain poorly characterized. This paper develops a structured taxonomy and analysis of long-Tail Knowledge in large language models, synthesizing prior work across technical and sociotechnical perspectives. We introduce a structured analytical framework that synthesizes prior work across four complementary axes: how long-Tail Knowledge is defined, the mechanisms by which it is lost or distorted during training and inference, the technical interventions proposed to mitigate these failures, and the implications of these failures for fairness, accountability, transparency, and user trust. We further examine how existing evaluation practices obscure tail behavior and complicate accountability for rare but consequential failures. The paper concludes by identifying open challenges related to privacy, sustainability, and governance that constrain long-Tail Knowledge representation. Taken together, this paper provides a unifying conceptual framework for understanding how long-Tail Knowledge is defined, lost, evaluated, and manifested in deployed language model systems.

In this thesis, I refine our understanding as to what conclusions we can reach from coreference-based evaluations by expanding existing evaluation practices and considering the extent to which evaluation results are either converging or conflicting. First, I analyze standard coreference evaluations and show that their design often leads to non-generalizable conclusions due to issues of measurement validity - including contestedness (multiple, competing definitions of coreference) and convergent validity (evaluation results that rank models differently across benchmarks). Second, I propose and implement a novel evaluation focused on testing systems' ability to infer the relative plausibility of events, a key aspect of resolving coreference. Through this extended evaluation, I find that contemporary language models demonstrate strong performance on standard benchmarks - improving over earlier baseline systems within certain domains and types of coreference - but remain sensitive to the evaluation conditions: they often fail to generalize in ways one would expect a human to be capable of when evaluation contexts are slightly modified. Taken together, these findings clarify both the strengths, such as improved accuracy over baselines on widely used evaluations, and the limitations of the current NLP paradigm, including weaknesses in measurement validity, and suggest directions for future work in developing better evaluation methods and more genuinely generalizable systems.

Adaptation methods have been a workhorse for unlocking the transformative power of pre-trained diffusion models in diverse applications. Existing approaches often abstract adaptation objectives as a reward function and steer diffusion models to generate high-reward samples. However, these approaches can incur high computational overhead due to additional training, or rely on stringent assumptions on the reward such as differentiability. Moreover, despite their empirical success, theoretical justification and guarantees are seldom established. In this paper, we propose DOIT (Doob-Oriented Inference-time Transformation), a training-free and computationally efficient adaptation method that applies to generic, non-differentiable rewards. The key framework underlying our method is a measure transport formulation that seeks to transport the pre-trained generative distribution to a high-reward target distribution. We leverage Doob's $h$-transform to realize this transport, which induces a dynamic correction to the diffusion sampling process and enables efficient simulation-based computation without modifying the pre-trained model. Theoretically, we establish a high probability convergence guarantee to the target high-reward distribution via characterizing the approximation error in the dynamic Doob's correction. Empirically, on D4RL offline RL benchmarks, our method consistently outperforms state-of-the-art baselines while preserving sampling efficiency.

Multimodal systems are vulnerable to partial or complete loss of input channels at deployment, which undermines reliability in real-world settings. This paper presents ModalImmune, a training framework that enforces modality immunity by intentionally and controllably collapsing selected modality information during training so the model learns joint representations that are robust to destructive modality influence. The framework combines a spectrum-adaptive collapse regularizer, an information-gain guided controller for targeted interventions, curvature-aware gradient masking to stabilize destructive updates, and a certified Neumann-truncated hyper-gradient procedure for automatic meta-parameter adaptation. Empirical evaluation on standard multimodal benchmarks demonstrates that ModalImmune improves resilience to modality removal and corruption while retaining convergence stability and reconstruction capacity.

Coordinating large populations of interacting agents is a central challenge in multi-agent reinforcement learning (MARL), where the size of the joint state-action space scales exponentially with the number of agents. Mean-field methods alleviate this burden by aggregating agent interactions, but these approaches assume homogeneous interactions. Recent graphon-based frameworks capture heterogeneity, but are computationally expensive as the number of agents grows. Therefore, we introduce $\texttt{GMFS}$, a $\textbf{G}$raphon $\textbf{M}$ean-$\textbf{F}$ield $\textbf{S}$ubsampling framework for scalable cooperative MARL with heterogeneous agent interactions. By subsampling $κ$ agents according to interaction strength, we approximate the graphon-weighted mean-field and learn a policy with sample complexity $\mathrm{poly}(κ)$ and optimality gap $O(1/\sqrtκ)$. We verify our theory with numerical simulations in robotic coordination, showing that $\texttt{GMFS}$ achieves near-optimal performance.

Permanent citizens' assemblies are ongoing deliberative bodies composed of randomly selected citizens, organized into panels that rotate over time. Unlike one-off panels, which represent the population in a single snapshot, permanent assemblies enable shifting participation across multiple rounds. This structure offers a powerful framework for ensuring that different groups of individuals are represented over time across successive panels. In particular, it allows smaller groups of individuals that may not warrant representation in every individual panel to be represented across a sequence of them. We formalize this temporal sortition framework by requiring proportional representation both within each individual panel and across the sequence of panels. Building on the work of Ebadian and Micha (2025), we consider a setting in which the population lies in a metric space, and the goal is to achieve both proportional representation, ensuring that every group of citizens receives adequate representation, and individual fairness, ensuring that each individual has an equal probability of being selected. We extend the notion of representation to a temporal setting by requiring that every initial segment of the panel sequence, viewed as a cumulative whole, proportionally reflects the structure of the population. We present algorithms that provide varying guarantees of proportional representation, both within individual panels and across any sequence of panels, while also maintaining individual fairness over time.

Several complex physical systems are governed by multi-scale partial differential equations (PDEs) that exhibit both smooth low-frequency components and localized high-frequency structures. Existing physics-informed neural network (PINN) methods typically train with fixed coordinate system inputs, where geometric misalignment with these structures induces gradient stiffness and ill-conditioning that hinder convergence. To address this issue, we introduce a mapping paradigm that reshapes the input coordinates through differentiable geometric compactification mappings and couples the geometric structure of PDEs with the spectral properties of residual operators. Based on this paradigm, we propose Geometric Compactification (GC)-PINN, a framework that introduces three mapping strategies for periodic boundaries, far-field scale expansion, and localized singular structures in the input domain without modifying the underlying PINN architecture. Extensive empirical evaluation demonstrates that this approach yields more uniform residual distributions and higher solution accuracy on representative 1D and 2D PDEs, while improving training stability and convergence speed.

Driven by our mission of "uplifting the world with memory," this paper explores the design concept of "memory" that is essential for achieving artificial superintelligence (ASI). Rather than proposing novel methods, we focus on several alternative approaches whose potential benefits are widely imaginable, yet have remained largely unexplored. The currently dominant paradigm, which can be termed "extract then store," involves extracting information judged to be useful from experiences and saving only the extracted content. However, this approach inherently risks the loss of information, as some valuable knowledge particularly for different tasks may be discarded in the extraction process. In contrast, we emphasize the "store then on-demand extract" approach, which seeks to retain raw experiences and flexibly apply them to various tasks as needed, thus avoiding such information loss. In addition, we highlight two further approaches: discovering deeper insights from large collections of probabilistic experiences, and improving experience collection efficiency by sharing stored experiences. While these approaches seem intuitively effective, our simple experiments demonstrate that this is indeed the case. Finally, we discuss major challenges that have limited investigation into these promising directions and propose research topics to address them.

Despite the continuous research and evolution of language models, they sometimes underperform previous versions. Existing approaches to overcome these challenges are resource-intensive, highlighting the need for alternatives that enable immediate action. We assume that each language model has a local module inside that is suitable for a specific function. First, this work identifies a set of modules showing consistent and local activation changes under an inference workload through activation-based analysis. Subsequently, we transplant an internal module that is properly activated for a specific task into the target model, leading to immediate and measurable functional changes without additional training or fine-tuning. To experimentally demonstrate the effectiveness of the transplant technique, we quantify the relationship between transplant strength and performance improvement under different conditions for two language models. In the cross-generation setting, we find that transplanting activation-selected modules can substantially improve the underperforming model, reaching up to twice the target baseline and achieving gap-based recovery above 100%. Moreover, in transplant experiments between a base model and its instruction-tuned counterpart, transplantation improves the underperforming model toward the stronger baseline, yielding up to about 2.33 times the target baseline with gap-based recovery reaching up to 100% in the best case. These results show that meaningful capacity transfer can be realized through the implantation of highly localized modules implied by language models. Overall, this work provides empirical evidence for task-localized modularity in language models and presents a new research area: model transplantation.

LLM-for-time series (TS) methods typically treat time shallowly, injecting positional or prompt-based cues once at the input of a largely frozen decoder, which limits temporal reasoning as this information degrades through the layers. We introduce Temporal-Prior Conditioning (TPC), which elevates time to a first-class modality that conditions the model at multiple depths. TPC attaches a small set of learnable time series tokens to the patch stream; at selected layers these tokens cross-attend to temporal embeddings derived from compact, human-readable temporal descriptors encoded by the same frozen LLM, then feed temporal context back via self-attention. This disentangles time series signal and temporal information while maintaining a low parameter budget. We show that by training only the cross-attention modules and explicitly disentangling time series signal and temporal information, TPC consistently outperforms both full fine-tuning and shallow conditioning strategies, achieving state-of-the-art performance in long-term forecasting across diverse datasets. Code available at: https://github.com/fil-mp/Deep_tpc

Robots executing iterative tasks in complex, uncertain environments require control strategies that balance robustness, safety, and high performance. This paper introduces a safe information-theoretic learning model predictive control (SIT-LMPC) algorithm for iterative tasks. Specifically, we design an iterative control framework based on an information-theoretic model predictive control algorithm to address a constrained infinite-horizon optimal control problem for discrete-time nonlinear stochastic systems. An adaptive penalty method is developed to ensure safety while balancing optimality. Trajectories from previous iterations are utilized to learn a value function using normalizing flows, which enables richer uncertainty modeling compared to Gaussian priors. SIT-LMPC is designed for highly parallel execution on graphics processing units, allowing efficient real-time optimization. Benchmark simulations and hardware experiments demonstrate that SIT-LMPC iteratively improves system performance while robustly satisfying system constraints.

Operational disruptions in retail payments can induce behavioral responses that outlast technical recovery and may amplify liquidity stress. We propose a multi-agent model linking card payment outages to trust dynamics, channel avoidance, and threshold-gated withdrawals. Customers and merchants interact through repeated payment attempts, while customers additionally influence one another on a Watts-Strogatz small-world network. Customers update bounded memory variables capturing accumulated negative experience (scar) and perceived systemic risk (rumor), with merchants contributing persistent broadcast signals that may lag operational recovery. We prove that, under mild conditions on memory persistence and threshold gating, aggregate withdrawal pressure can peak strictly after the outage nadir, including during the recovery phase. Simulations reproduce behavioral hysteresis and confirm delayed peaks of outflows. We further study payment substitution via instant transfer: substitution consistently reduces peak avoidance, yet its effect on cumulative outflows is non-monotonic under realistic merchant broadcast persistence. Robustness experiments across random seeds show stable qualitative behavior. The model highlights why "status green" is not equivalent to risk resolution and motivates incident response strategies that address perception, merchant messaging, and post-recovery communication in addition to technical remediation.

We study the \emph{Submodular Welfare Problem} (SWP), where items are partitioned among agents with monotone submodular utilities to maximize the total welfare under \emph{bandit feedback}. Classical SWP assumes full value-oracle access, achieving $(1-1/e)$ approximations via continuous-greedy algorithms. We extend this to a \emph{multi-agent combinatorial bandit} framework (\textsc{MA-CMAB}), where actions are partitions under full-bandit feedback with non-communicating agents. Unlike prior single-agent or separable multi-agent CMAB models, our setting couples agents through shared allocation constraints. We propose an explore-then-commit strategy with randomized assignments, achieving $\tilde{\mathcal{O}}(T^{2/3})$ regret against a $(1-1/e)$ benchmark, the first such guarantee for partition-based submodular welfare problem under bandit feedback.

Energy theft poses a significant threat to the stability and efficiency of smart grids, leading to substantial economic losses and operational challenges. Traditional centralized machine learning approaches for theft detection require aggregating user data, raising serious concerns about privacy and data security. These issues are further exacerbated in smart meter environments, where devices are often resource-constrained and lack the capacity to run heavy models. In this work, we propose a privacy-preserving federated learning framework for energy theft detection that addresses both privacy and computational constraints. Our approach leverages a lightweight multilayer perceptron (MLP) model, suitable for deployment on low-power smart meters, and integrates basic differential privacy (DP) by injecting Gaussian noise into local model updates before aggregation. This ensures formal privacy guarantees without compromising learning performance. We evaluate our framework on a real-world smart meter dataset under both IID and non-IID data distributions. Experimental results demonstrate that our method achieves competitive accuracy, precision, recall, and AUC scores while maintaining privacy and efficiency. This makes the proposed solution practical and scalable for secure energy theft detection in next-generation smart grid infrastructures.

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

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

Mobile edge computing (MEC) based wireless metaverse services offer an untethered, immersive experience to users, where the superior quality of experience (QoE) needs to be achieved under stringent latency constraints and visual quality demands. To achieve this, MEC-based intelligent resource allocation for virtual reality users needs to be supported by coordination across MEC servers to harness distributed data. Federated learning (FL) is a promising solution, and can be combined with reinforcement learning (RL) to develop generalized policies across MEC-servers. However, conventional FL incurs transmitting the full model parameters across the MEC-servers and the cloud, and suffer performance degradation due to naive global aggregation, especially in heterogeneous multi-radio access technology environments. To address these challenges, this paper proposes Federated Split Decision Transformer (FSDT), an offline RL framework where the transformer model is partitioned between MEC servers and the cloud. Agent-specific components (e.g., MEC-based embedding and prediction layers) enable local adaptability, while shared global layers in the cloud facilitate cooperative training across MEC servers. Experimental results demonstrate that FSDT enhances QoE for up to 10% in heterogeneous environments compared to baselines, while offloadingnearly 98% of the transformer model parameters to the cloud, thereby reducing the computational burden on MEC servers.

Modern AI agents are powerful but often fail to align with the idiosyncratic, evolving preferences of individual users. Prior approaches typically rely on static datasets, either training implicit preference models on interaction history or encoding user profiles in external memory. However, these approaches struggle with new users and with preferences that change over time. We introduce Personalized Agents from Human Feedback (PAHF), a framework for continual personalization in which agents learn online from live interaction using explicit per-user memory. PAHF operationalizes a three-step loop: (1) seeking pre-action clarification to resolve ambiguity, (2) grounding actions in preferences retrieved from memory, and (3) integrating post-action feedback to update memory when preferences drift. To evaluate this capability, we develop a four-phase protocol and two benchmarks in embodied manipulation and online shopping. These benchmarks quantify an agent's ability to learn initial preferences from scratch and subsequently adapt to persona shifts. Our theoretical analysis and empirical results show that integrating explicit memory with dual feedback channels is critical: PAHF learns substantially faster and consistently outperforms both no-memory and single-channel baselines, reducing initial personalization error and enabling rapid adaptation to preference shifts.

Non-autoregressive (NAR) generation reduces decoding latency by predicting many tokens in parallel, but iterative refinement often suffers from error accumulation and distribution shift under self-generated drafts. Masked diffusion language models (MDLMs) and their remasking samplers (e.g., ReMDM) can be viewed as modern NAR iterative refinement, where generation repeatedly revises a partially observed draft. In this work we show that \emph{training alone} can substantially improve the step-efficiency of MDLM/ReMDM sampling. We propose \textsc{DSL} (Discrete Stochastic Localization), which trains a single SNR-invariant denoiser across a continuum of corruption levels, bridging intermediate draft noise and mask-style endpoint corruption within one Diffusion Transformer. On OpenWebText, \textsc{DSL} fine-tuning yields large MAUVE gains at low step budgets, surpassing the MDLM+ReMDM baseline with \(\sim\)4$\times$ fewer denoiser evaluations, and matches autoregressive quality at high budgets. Analyses show improved self-correction and uncertainty calibration, making remasking markedly more compute-efficient.

Physics-informed neural networks and neural operators often suffer from severe optimization difficulties caused by ill-conditioned gradients, multi-scale spectral behavior, and stiffness induced by physical constraints. Recently, the Muon optimizer has shown promise by performing orthogonalized updates in the singular-vector basis of the gradient, thereby improving geometric conditioning. However, its unit-singular-value updates may lead to overly aggressive steps and lack explicit stability guarantees when applied to physics-informed learning. In this work, we propose SpecMuon, a spectral-aware optimizer that integrates Muon's orthogonalized geometry with a mode-wise relaxed scalar auxiliary variable (RSAV) mechanism. By decomposing matrix-valued gradients into singular modes and applying RSAV updates individually along dominant spectral directions, SpecMuon adaptively regulates step sizes according to the global loss energy while preserving Muon's scale-balancing properties. This formulation interprets optimization as a multi-mode gradient flow and enables principled control of stiff spectral components. We establish rigorous theoretical properties of SpecMuon, including a modified energy dissipation law, positivity and boundedness of auxiliary variables, and global convergence with a linear rate under the Polyak-Lojasiewicz condition. Numerical experiments on physics-informed neural networks, DeepONets, and fractional PINN-DeepONets demonstrate that SpecMuon achieves faster convergence and improved stability compared with Adam, AdamW, and the original Muon optimizer on benchmark problems such as the one-dimensional Burgers equation and fractional partial differential equations.

Training LLMs as interactive agents for multi-turn decision-making remains challenging, particularly in long-horizon tasks with sparse and delayed rewards, where agents must execute extended sequences of actions before receiving meaningful feedback. Most existing reinforcement learning (RL) approaches model LLM agents as flat policies operating at a single time scale, selecting one action at each turn. In sparse-reward settings, such flat policies must propagate credit across the entire trajectory without explicit temporal abstraction, which often leads to unstable optimization and inefficient credit assignment. We propose HiPER, a novel Hierarchical Plan-Execute RL framework that explicitly separates high-level planning from low-level execution. HiPER factorizes the policy into a high-level planner that proposes subgoals and a low-level executor that carries them out over multiple action steps. To align optimization with this structure, we introduce a key technique called hierarchical advantage estimation (HAE), which carefully assigns credit at both the planning and execution levels. By aggregating returns over the execution of each subgoal and coordinating updates across the two levels, HAE provides an unbiased gradient estimator and provably reduces variance compared to flat generalized advantage estimation. Empirically, HiPER achieves state-of-the-art performance on challenging interactive benchmarks, reaching 97.4\% success on ALFWorld and 83.3\% on WebShop with Qwen2.5-7B-Instruct (+6.6\% and +8.3\% over the best prior method), with especially large gains on long-horizon tasks requiring multiple dependent subtasks. These results highlight the importance of explicit hierarchical decomposition for scalable RL training of multi-turn LLM agents.

We argue that uncertainty is a key and understudied limitation of LLMs' performance in creative writing, which is often characterized as trite and cliché-ridden. Literary theory identifies uncertainty as a necessary condition for creative expression, while current alignment strategies steer models away from uncertain outputs to ensure factuality and reduce hallucination. We formalize this tension by quantifying the "uncertainty gap" between human-authored stories and model-generated continuations. Through a controlled information-theoretic analysis of 28 LLMs on high-quality storytelling datasets, we demonstrate that human writing consistently exhibits significantly higher uncertainty than model outputs. We find that instruction-tuned and reasoning models exacerbate this trend compared to their base counterparts; furthermore, the gap is more pronounced in creative writing than in functional domains, and strongly correlates to writing quality. Achieving human-level creativity requires new uncertainty-aware alignment paradigms that can distinguish between destructive hallucinations and the constructive ambiguity required for literary richness.

Emotional expression underpins natural communication and effective human-computer interaction. We present Emotion Collider (EC-Net), a hyperbolic hypergraph framework for multimodal emotion and sentiment modeling. EC-Net represents modality hierarchies using Poincare-ball embeddings and performs fusion through a hypergraph mechanism that passes messages bidirectionally between nodes and hyperedges. To sharpen class separation, contrastive learning is formulated in hyperbolic space with decoupled radial and angular objectives. High-order semantic relations across time steps and modalities are preserved via adaptive hyperedge construction. Empirical results on standard multimodal emotion benchmarks show that EC-Net produces robust, semantically coherent representations and consistently improves accuracy, particularly when modalities are partially available or contaminated by noise. These findings indicate that explicit hierarchical geometry combined with hypergraph fusion is effective for resilient multimodal affect understanding.

Real-world deployments routinely face distribution shifts, group imbalances, and adversarial perturbations, under which the traditional Empirical Risk Minimization (ERM) framework can degrade severely. Distributionally Robust Optimization (DRO) addresses this issue by optimizing the worst-case expected loss over an uncertainty set of distributions, offering a principled approach to robustness. Meanwhile, as training data in DRO always involves sensitive information, safeguarding it against leakage under Differential Privacy (DP) is essential. In contrast to classical DP-ERM, DP-DRO has received much less attention due to its minimax optimization structure with uncertainty constraint. To bridge the gap, we provide a comprehensive study of DP-(finite-sum)-DRO with $ψ$-divergence and non-convex loss. First, we study DRO with general $ψ$-divergence by reformulating it as a minimization problem, and develop a novel $(\varepsilon, δ)$-DP optimization method, called DP Double-Spider, tailored to this structure. Under mild assumptions, we show that it achieves a utility bound of $\mathcal{O}(\frac{1}{\sqrt{n}}+ (\frac{\sqrt{d \log (1/δ)}}{n \varepsilon})^{2/3})$ in terms of the gradient norm, where $n$ denotes the data size and $d$ denotes the model dimension. We further improve the utility rate for specific divergences. In particular, for DP-DRO with KL-divergence, by transforming the problem into a compositional finite-sum optimization problem, we develop a DP Recursive-Spider method and show that it achieves a utility bound of $\mathcal{O}((\frac{\sqrt{d \log(1/δ)}}{n\varepsilon})^{2/3} )$, matching the best-known result for non-convex DP-ERM. Experimentally, we demonstrate that our proposed methods outperform existing approaches for DP minimax optimization.

Chain-of-thought (CoT) reasoning sometimes fails to faithfully reflect the true computation of a large language model (LLM), hampering its utility in explaining how LLMs arrive at their answers. Moreover, optimizing for faithfulness and interpretability in reasoning often degrades task performance. To address this tradeoff and improve CoT faithfulness, we propose Reasoning Execution by Multiple Listeners (REMUL), a multi-party reinforcement learning approach. REMUL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful. A speaker model generates a reasoning trace, which is truncated and passed to a pool of listener models who "execute" the trace, continuing the trace to an answer. Speakers are rewarded for producing reasoning that is clear to listeners, with additional correctness regularization via masked supervised finetuning to counter the tradeoff between faithfulness and performance. On multiple reasoning benchmarks (BIG-Bench Extra Hard, MuSR, ZebraLogicBench, and FOLIO), REMUL consistently and substantially improves three measures of faithfulness -- hint attribution, early answering area over the curve (AOC), and mistake injection AOC -- while also improving accuracy. Our analysis finds that these gains are robust across training domains, translate to legibility gains, and are associated with shorter and more direct CoTs.

This paper is concerned with the distributed composite optimization problem over networks, where agents aim to minimize a sum of local smooth components and a common nonsmooth term. Leveraging the probabilistic local updates mechanism, we propose a communication-efficient Adapt-Then-Combine (ATC) framework, FlexATC, unifying numerous ATC-based distributed algorithms. Under stepsizes independent of the network topology and the number of local updates, we establish sublinear and linear convergence rates for FlexATC in convex and strongly convex settings, respectively. Remarkably, in the strong convex setting, the linear rate is decoupled from the objective functions and network topology, and FlexATC permits communication to be skipped in most iterations without any deterioration of the linear rate. In addition, the proposed unified theory demonstrates for the first time that local updates provably lead to communication acceleration for ATC-based distributed algorithms. Numerical experiments further validate the efficacy of the proposed framework and corroborate the theoretical results.

Cross-subject generalization in EEG-based brain-computer interfaces (BCIs) remains challenging due to individual variability in neural signals. We investigate whether spectral representations offer more stable features for cross-subject transfer than temporal waveforms. Through correlation analyses across three EEG paradigms (SSVEP, P300, and Motor Imagery), we find that spectral features exhibit consistently higher cross-subject similarity than temporal signals. Motivated by this observation, we introduce ASPEN, a hybrid architecture that combines spectral and temporal feature streams via multiplicative fusion, requiring cross-modal agreement for features to propagate. Experiments across six benchmark datasets reveal that ASPEN is able to dynamically achieve the optimal spectral-temporal balance depending on the paradigm. ASPEN achieves the best unseen-subject accuracy on three of six datasets and competitive performance on others, demonstrating that multiplicative multimodal fusion enables effective cross-subject generalization.

There are a number of existing studies analysing the convergence behaviour of graph neural networks on large random graphs. Unfortunately, the majority of these studies do not model correlations between node features, which would naturally exist in a variety of real-life networks. Consequently, the derived limitations of GNNs, resulting from such convergence behaviour, is not truly reflective of the expressive power of GNNs when applied to realistic graphs. In this paper, we will introduce a novel method to generate random graphs that have correlated node features. The node features will be sampled in such a manner to ensure correlation between neighbouring nodes. As motivation for our choice of sampling scheme, we will appeal to properties exhibited by real-life graphs, particularly properties that are captured by the Barabási-Albert model. A theoretical analysis will strongly indicate that convergence can be avoided in some cases, which we will empirically validate on large random graphs generated using our novel method. The observed divergent behaviour provides evidence that GNNs may be more expressive than initial studies would suggest, especially on realistic graphs.

As multimodal systems increasingly process sensitive personal data, the ability to selectively revoke specific data modalities has become a critical requirement for privacy compliance and user autonomy. We present Missing-by-Design (MBD), a unified framework for revocable multimodal sentiment analysis that combines structured representation learning with a certifiable parameter-modification pipeline. Revocability is critical in privacy-sensitive applications where users or regulators may request removal of modality-specific information. MBD learns property-aware embeddings and employs generator-based reconstruction to recover missing channels while preserving task-relevant signals. For deletion requests, the framework applies saliency-driven candidate selection and a calibrated Gaussian update to produce a machine-verifiable Modality Deletion Certificate. Experiments on benchmark datasets show that MBD achieves strong predictive performance under incomplete inputs and delivers a practical privacy-utility trade-off, positioning surgical unlearning as an efficient alternative to full retraining.

Probabilistic bits (p-bits) offer an energy-efficient hardware abstraction for stochastic optimization; however, existing p-bit-based simulated annealing accelerators suffer from poor scalability and limited support for fully connected graphs due to fan-out and memory overhead. This paper presents an energy-efficient FPGA architecture for stochastic simulated quantum annealing (SSQA) that addresses these challenges. The proposed design combines a spin-serial and replica-parallel update schedule with a dual-BRAM delay-line architecture, enabling scalable support for fully connected Ising models while eliminating fan-out growth in logic resources. By exploiting SSQA, the architecture achieves fast convergence using only final replica states, significantly reducing memory requirements compared to conventional p-bit-based annealers. Implemented on a Xilinx ZC706 FPGA, the proposed system solves an 800-node MAX-CUT benchmark and achieves up to 50% reduction in energy consumption and over 90\% reduction in logic resources compared with prior FPGA-based p-bit annealing architectures. These results demonstrate the practicality of quantum-inspired, p-bit-based annealing hardware for large-scale combinatorial optimization under strict energy and resource constraints.

We study density estimation in Kullback-Leibler divergence: given an i.i.d. sample from an unknown density $p$, the goal is to construct an estimator $\widehat p$ such that $\mathrm{KL}(p,\widehat p)$ is small with high probability. We consider two settings involving a finite dictionary of $M$ densities: (i) model aggregation, where $p$ belongs to the dictionary, and (ii) convex aggregation (mixture density estimation), where $p$ is a mixture of densities from the dictionary. Crucially, we make no assumption on the base densities: their ratios may be unbounded and their supports may differ. For both problems, we identify the best possible high-probability guarantees in terms of the dictionary size, sample size, and confidence level. These optimal rates are higher than those achievable when density ratios are bounded by absolute constants; for mixture density estimation, they match existing lower bounds in the special case of discrete distributions. Our analysis of the mixture case hinges on two new covering results. First, we provide a sharp, distribution-free upper bound on the local Hellinger entropy of the class of mixtures of $M$ distributions. Second, we prove an optimal ratio covering theorem for convex sets: for every convex compact set $K\subset \mathbb{R}_+^d$, there exists a subset $A\subset K$ with at most $2^{8d}$ elements such that each element of $K$ is coordinate-wise dominated by an element of $A$ up to a universal constant factor. This geometric result is of independent interest; notably, it yields new cardinality estimates for $\varepsilon$-approximate Pareto sets in multi-objective optimization when the attainable set of objective vectors is convex.

This study aimed to comprehend how user domain knowledge and artificial intelligence (AI) literacy impact the effective use of human-AI interactive building energy management system (BEMS). While prior studies have investigated the potential of integrating large language models (LLMs) into BEMS or building energy modeling, very few studies have examined how user interact with such systems. We conducted a systematic role-playing experiment, where 85 human subjects interacted with an advanced generative pre-trained transformer (OpenAI GPT-4o). Participants were tasked with identifying the top five behavioral changes that could reduce home energy use with the GPT model that functioned as an LLM-integrated BEMS. Then, the collected prompt-response data and participant conclusions were analyzed using an analytical framework that hierarchically assessed and scored human-AI interactions and their home energy analysis approaches. Also, participants were classified into four groups based on their self-evaluated domain knowledge of building energy use and AI literacy, and Kruskal-Wallis H tests with post-hoc pairwise comparisons were conducted across 20 quantifiable metrics. Key takeaways include: most participants employed concise prompts (median: 16.2 words) and relied heavily on GPT's analytical capabilities; and notably, only 1 of 20 metrics, appliance identification rate, showed statistically significant group differences (p=0.037), driven by AI literacy rather than domain knowledge, suggesting an equalizing effect of LLMs across expertise levels. This study provides foundational insights into human-AI collaboration dynamics and promising development directions in the context of LLM-integrated BEMS and contributes to realizing human-centric LLM-integrated energy systems.

The rapid proliferation of AI-generated content on the Web presents a structural risk to information retrieval, as search engines and Retrieval-Augmented Generation (RAG) systems increasingly consume evidence produced by the Large Language Models (LLMs). We characterize this ecosystem-level failure mode as Retrieval Collapse, a two-stage process where (1) AI-generated content dominates search results, eroding source diversity, and (2) low-quality or adversarial content infiltrates the retrieval pipeline. We analyzed this dynamic through controlled experiments involving both high-quality SEO-style content and adversarially crafted content. In the SEO scenario, a 67\% pool contamination led to over 80\% exposure contamination, creating a homogenized yet deceptively healthy state where answer accuracy remains stable despite the reliance on synthetic sources. Conversely, under adversarial contamination, baselines like BM25 exposed $\sim$19\% of harmful content, whereas LLM-based rankers demonstrated stronger suppression capabilities. These findings highlight the risk of retrieval pipelines quietly shifting toward synthetic evidence and the need for retrieval-aware strategies to prevent a self-reinforcing cycle of quality decline in Web-grounded systems.

Text-to-video diffusion models deliver impressive results but remain slow because of the sequential denoising of 3D latents. Existing approaches to speed up inference either require expensive model retraining or use heuristic-based step skipping, which struggles to maintain video quality as the number of denoising steps decreases. Our work, CHAI, aims to use cross-inference caching to reduce latency while maintaining video quality. We introduce Cache Attention as an effective method for attending to shared objects/scenes across cross-inference latents. This selective attention mechanism enables effective reuse of cached latents across semantically related prompts, yielding high cache hit rates. We show that it is possible to generate high-quality videos using Cache Attention with as few as 8 denoising steps. When integrated into the overall system, CHAI is 1.65x - 3.35x faster than baseline OpenSora 1.2 while maintaining video quality.

Large language models (LLMs) are increasingly used as agents to solve complex tasks such as question answering (QA), scientific debate, and software development. A standard evaluation procedure aggregates multiple responses from LLM agents into a single final answer, often via majority voting, and compares it against reference answers. However, this process can obscure the quality and distributional characteristics of the original responses. In this paper, we propose a novel evaluation framework based on the empirical cumulative distribution function (ECDF) of cosine similarities between generated responses and reference answers. This enables a more nuanced assessment of response quality beyond exact match metrics. To analyze the response distributions across different agent configurations, we further introduce a clustering method for ECDFs using their distances and the $k$-medoids algorithm. Our experiments on a QA dataset demonstrate that ECDFs can distinguish between agent settings with similar final accuracies but different quality distributions. The clustering analysis also reveals interpretable group structures in the responses, offering insights into the impact of temperature, persona, and question topics.

Learning with shared representation is widely recognized as an effective way to separate commonalities from heterogeneity across various heterogeneous sources. Most existing work includes all related data sources via simultaneously training a common feature extractor and source-specific heads. It is well understood that data sources with low relevance or poor quality may hinder representation learning. In this paper, we further dive into the question of which data sources should be learned jointly by focusing on the traditionally deemed ``good'' collection of sources, in which individual sources have similar relevance and qualities with respect to the true underlying common structure. Towards tractability, we focus on the linear setting where sources share a low-dimensional subspace. We find that source screening can play a central role in statistically optimal subspace estimation. We show that, for a broad class of problem instances, training on a carefully selected subset of sources suffices to achieve minimax optimality, even when a substantial portion of data is discarded. We formalize the notion of an informative subpopulation, develop algorithms and practical heuristics for identifying such subsets, and validate their effectiveness through both theoretical analysis and empirical evaluations on synthetic and real-world datasets.

Approximate nearest neighbor (ANN) search is widely used in the retrieval stage of large-scale recommendation systems. In this stage, candidate items are indexed using their learned embedding vectors, and ANN search is executed for each user (or item) query to retrieve a set of relevant items. However, ANN-based retrieval has two key limitations. First, item embeddings and their indices are typically learned in separate stages: indexing is often performed offline after embeddings are trained, which can yield suboptimal retrieval quality-especially for newly created items. Second, although ANN offers sublinear query time, it must still be run for every request, incurring substantial computation cost at industry scale. In this paper, we propose MultiFaceted Learnable Index (MFLI), a scalable, real-time retrieval paradigm that learns multifaceted item embeddings and indices within a unified framework and eliminates ANN search at serving time. Specifically, we construct a multifaceted hierarchical codebook via residual quantization of item embeddings and co-train the codebook with the embeddings. We further introduce an efficient multifaceted indexing structure and mechanisms that support real-time updates. At serving time, the learned hierarchical indices are used directly to identify relevant items, avoiding ANN search altogether. Extensive experiments on real-world data with billions of users show that MFLI improves recall on engagement tasks by up to 11.8\%, cold-content delivery by up to 57.29\%, and semantic relevance by 13.5\% compared with prior state-of-the-art methods. We also deploy MFLI in the system and report online experimental results demonstrating improved engagement, less popularity bias, and higher serving efficiency.

This paper introduces and demonstrates a computational pipeline for the statistical analysis of shape graph datasets, namely geometric networks embedded in 2D or 3D spaces. Unlike traditional abstract graphs, our purpose is not only to retrieve and distinguish variations in the connectivity structure of the data but also geometric differences of the network branches. Our proposed approach relies on the extraction of a specifically curated and explicit set of topological, geometric and directional features, designed to satisfy key invariance properties. We leverage the resulting feature representation for tasks such as group comparison, clustering and classification on cohorts of shape graphs. The effectiveness of this representation is evaluated on several real-world datasets including urban road/street networks, neuronal traces and astrocyte imaging. These results are benchmarked against several alternative methods, both feature-based and not.

Large Language Models cannot reliably acquire new knowledge post-deployment -- even when relevant text resources exist, models fail to transform them into actionable knowledge without retraining. Retrieval-Augmented Generation attempts to bridge this gap by surfacing relevant documents at inference time, yet similarity-based retrieval often fails to identify context that actually improves task performance. We introduce Evolutionary Context Search (ECS), an evolutionary method that searches context combinations using accuracy on a small development set, requiring only inference calls without weight updates. ECS moves beyond semantic similarity to discover non-obvious context pairings that significantly boost performance. Our empirical results show that ECS improves BackendBench by 27\% and $τ$-bench airline by 7\%. The evolved contexts are model-agnostic, as those evolved with Gemini-3-Flash transfer effectively to Claude Sonnet and DeepSeek. This suggests that ECS opens a path toward automated context discovery for skill acquisition -- an efficient alternative to manual prompt engineering or costly fine-tuning.

Online media platforms often need to measure how frequently users are exposed to specific content attributes in order to evaluate trade-offs in A/B experiments. A direct approach is to sample content, label it using a high-quality rubric (e.g., an expert-reviewed LLM prompt), and estimate impression-weighted prevalence. However, repeatedly running such labeling for every experiment arm and segment is too costly and slow to serve as a default measurement at scale. We present a scalable \emph{surrogate-based prevalence measurement} framework that decouples expensive labeling from per-experiment evaluation. The framework calibrates a surrogate signal to reference labels offline and then uses only impression logs to estimate prevalence for arbitrary experiment arms and segments. We instantiate this framework using \emph{score bucketing} as the surrogate: we discretize a model score into buckets, estimate bucket-level prevalences from an offline labeled sample, and combine these calibrated bucket level prevalences with the bucket distribution of impressions in each arm to obtain fast, log-based estimates. Across multiple large-scale A/B tests, we validate that the surrogate estimates closely match the reference estimates for both arm-level prevalence and treatment--control deltas. This enables scalable, low-latency prevalence measurement in experimentation without requiring per-experiment labeling jobs.

Computed Tomography (CT) is one of the most widely used and diagnostically information-dense imaging modalities, covering critical organs such as the heart, lungs, liver, and colon. Clinical interpretation relies on both slice-driven local features (e.g., sub-centimeter nodules, lesion boundaries) and volume-driven spatial representations (e.g., tumor infiltration, inter-organ anatomical relations). However, existing Large Vision-Language Models (LVLMs) remain fragmented in CT slice versus volumetric understanding: slice-driven LVLMs show strong generalization but lack cross-slice spatial consistency, while volume-driven LVLMs explicitly capture volumetric semantics but suffer from coarse granularity and poor compatibility with slice inputs. The absence of a unified modeling paradigm constitutes a major bottleneck for the clinical translation of medical LVLMs. We present OmniCT, a powerful unified slice-volume LVLM for CT scenarios, which makes three contributions: (i) Spatial Consistency Enhancement (SCE): volumetric slice composition combined with tri-axial positional embedding that introduces volumetric consistency, and an MoE hybrid projection enables efficient slice-volume adaptation; (ii) Organ-level Semantic Enhancement (OSE): segmentation and ROI localization explicitly align anatomical regions, emphasizing lesion- and organ-level semantics; (iii) MedEval-CT: the largest slice-volume CT dataset and hybrid benchmark integrates comprehensive metrics for unified evaluation. OmniCT consistently outperforms existing methods with a substantial margin across diverse clinical tasks and satisfies both micro-level detail sensitivity and macro-level spatial reasoning. More importantly, it establishes a new paradigm for cross-modal medical imaging understanding.

Cross-border insider threats pose a critical challenge to government financial schemes, particularly when dealing with distributed, privacy-sensitive data across multiple jurisdictions. Existing approaches face fundamental limitations: they cannot effectively share intelligence across borders due to privacy constraints, lack reasoning capabilities to understand complex multi-step attack patterns, and fail to capture intricate graph-structured relationships in financial networks. We introduce FedGraph-AGI, a novel federated learning framework integrating Artificial General Intelligence (AGI) reasoning with graph neural networks for privacy-preserving cross-border insider threat detection. Our approach combines: (1) federated graph neural networks preserving data sovereignty; (2) Mixture-of-Experts (MoE) aggregation for heterogeneous jurisdictions; and (3) AGI-powered reasoning via Large Action Models (LAM) performing causal inference over graph data. Through experiments on a 50,000-transaction dataset across 10 jurisdictions, FedGraph-AGI achieves 92.3% accuracy, significantly outperforming federated baselines (86.1%) and centralized approaches (84.7%). Our ablation studies reveal AGI reasoning contributes 6.8% improvement, while MoE adds 4.4%. The system maintains epsilon = 1.0 differential privacy while achieving near-optimal performance and scales efficiently to 50+ clients. This represents the first integration of AGI reasoning with federated graph learning for insider threat detection, opening new directions for privacy-preserving cross-border intelligence sharing.

Code translation, the automatic conversion of programs between languages, is a growing use case for Large Language Models (LLMs). However, direct one-shot translation often fails to preserve program intent, leading to errors in control flow, type handling, and I/O behavior. We propose an algorithm-based pipeline that introduces a language-neutral intermediate specification to capture these details before code generation. This study empirically evaluates the extent to which structured planning can improve translation accuracy and reliability relative to direct translation. We conduct an automated paired experiment - direct and algorithm-based to translate between Python and Java using five widely used LLMs on the Avatar and CodeNet datasets. For each combination (model, dataset, approach, and direction), we compile and execute the translated program and run the tests provided. We record compilation results, runtime behavior, timeouts (e.g., infinite loop), and test outcomes. We compute accuracy from these tests, counting a translation as correct only if it compiles, runs without exceptions or timeouts, and passes all tests. We then map every failed compile-time and runtime case to a unified, language-aware taxonomy and compare subtype frequencies between the direct and algorithm-based approaches. Overall, the Algorithm-based approach increases micro-average accuracy from 67.7% to 78.5% (10.8% increase). It eliminates lexical and token errors by 100%, reduces incomplete constructs by 72.7%, and structural and declaration issues by 61.1%. It also substantially lowers runtime dependency and entry-point failures by 78.4%. These results demonstrate that algorithm-based pipelines enable more reliable, intent-preserving code translation, providing a foundation for robust multilingual programming assistants.

Large Language Models (LLMs) are increasingly deployed in applications that interact with the physical world, such as navigation, robotics, or mapping, making robust geospatial reasoning a critical capability. Despite that, LLMs' ability to reason about GPS coordinates and real-world geography remains underexplored. We introduce GPSBench, a dataset of 57,800 samples across 17 tasks for evaluating geospatial reasoning in LLMs, spanning geometric coordinate operations (e.g., distance and bearing computation) and reasoning that integrates coordinates with world knowledge. Focusing on intrinsic model capabilities rather than tool use, we evaluate 14 state-of-the-art LLMs and find that GPS reasoning remains challenging, with substantial variation across tasks: models are generally more reliable at real-world geographic reasoning than at geometric computations. Geographic knowledge degrades hierarchically, with strong country-level performance but weak city-level localization, while robustness to coordinate noise suggests genuine coordinate understanding rather than memorization. We further show that GPS-coordinate augmentation can improve in downstream geospatial tasks, and that finetuning induces trade-offs between gains in geometric computation and degradation in world knowledge. Our dataset and reproducible code are available at https://github.com/joey234/gpsbench

Reliable and cost-effective maintenance is essential for railway safety, particularly at the wheel-rail interface, which is prone to wear and failure. Predictive maintenance frameworks increasingly leverage sensor-generated time-series data, yet traditional methods require manual feature engineering, and deep learning models often degrade in online settings with evolving operational patterns. This work presents a semantic-aware, label-efficient continual learning framework for railway fault diagnostics. Accelerometer signals are encoded via a Variational AutoEncoder into latent representations capturing the normal operational structure in a fully unsupervised manner. Importantly, semantic metadata, including axle counts, wheel indexes, and strain-based deformations, is extracted via AI-driven peak detection on fiber Bragg grating sensors (resistant to electromagnetic interference) and fused with the VAE embeddings, enhancing anomaly detection under unknown operational conditions. A lightweight gradient boosting supervised classifier stabilizes anomaly scoring with minimal labels, while a replay-based continual learning strategy enables adaptation to evolving domains without catastrophic forgetting. Experiments show the model detects minor imperfections due to flats and polygonization, while adapting to evolving operational conditions, such as changes in train type, speed, load, and track profiles, captured using a single accelerometer and strain gauge in wayside monitoring.

With the rapid advancement of large language models (LLMs), efficiently serving LLM inference under limited GPU resources has become a critical challenge. Recently, an increasing number of studies have explored applying serverless computing paradigms to LLM serving in order to maximize resource utilization. However, LLM inference workloads are highly diverse, and modern GPU clusters are inherently heterogeneous, making it necessary to dynamically adjust deployment configurations online to better adapt to the elastic and dynamic nature of serverless environments. At the same time, enabling such online reconfiguration is particularly challenging due to the stateful nature of LLM inference and the massive size of model parameters. In this paper, we propose a dynamic pipeline reconfiguration approach that enables online adjustment of pipeline configurations while minimizing service downtime and performance degradation. Our method allows the system to select the optimal pipeline configuration in response to changing workloads. Experimental results on heterogeneous GPU platforms, including NVIDIA A100 and L40s, demonstrate that our migration mechanism incurs less than 50 ms of service downtime, while introducing under 10% overhead on both time-to-first-token (TTFT) and time-per-output-token (TPOT).

The Internet of Battlefield Things (IoBT) relies on heterogeneous, bandwidth-constrained, and intermittently connected tactical networks that face rapidly evolving cyber threats. In this setting, intrusion detection cannot depend on continuous central collection of raw traffic due to disrupted links, latency, operational security limits, and non-IID traffic across zones. We present Zone-Adaptive Intrusion Detection (ZAID), a collaborative detection and model-improvement framework for unseen attack types, where "zero-day" refers to previously unobserved attack families and behaviours (not vulnerability disclosure timing). ZAID combines a universal convolutional model for generalisable traffic representations, an autoencoder-based reconstruction signal as an auxiliary anomaly score, and lightweight adapter modules for parameter-efficient zone adaptation. To support cross-zone generalisation under constrained connectivity, ZAID uses federated aggregation and pseudo-labelling to leverage locally observed, weakly labelled behaviours. We evaluate ZAID on ToN_IoT using a zero-day protocol that excludes MITM, DDoS, and DoS from supervised training and introduces them during zone-level deployment and adaptation. ZAID achieves up to 83.16% accuracy on unseen attack traffic and transfers to UNSW-NB15 under the same procedure, with a best accuracy of 71.64%. These results indicate that parameter-efficient, zone-personalised collaboration can improve the detection of previously unseen attacks in contested IoBT environments.

Post-training endows pretrained LLMs with a variety of desirable skills, including instruction-following, reasoning, and others. However, these post-trained LLMs only encode knowledge up to a cut-off date, necessitating continual adaptation. Unfortunately, existing solutions cannot simultaneously learn new knowledge from an adaptation document corpora and mitigate the forgetting of earlier learned capabilities. To address this, we introduce Distillation via Split Contexts (DiSC), a simple context-distillation based approach for continual knowledge adaptation. \methodname~derives student and teacher distributions by conditioning on distinct segments of the training example and minimizes the KL divergence between the shared tokens. This allows us to efficiently apply context-distillation without requiring explicit generation steps during training. We run experiments on four post-trained models and two adaptation domains. Compared to prior finetuning and distillation methods for continual adaptation, DiSC consistently reports the best trade-off between learning new knowledge and mitigating forgetting of previously learned skills like instruction-following, reasoning, and factual knowledge.

Any-order autoregressive models (AO-ARMs) offer a promising path toward efficient masked diffusion by enabling native key-value caching, but competitive performance has so far required two-stream attention, typically motivated as a means of decoupling token content from position. In this work, we argue that two-stream attention may be serving a more subtle role. We identify a structural-semantic tradeoff in any-order generation: the hidden representation at each step must simultaneously attend to semantically informative tokens for prediction and structurally recent tokens for summarization, objectives that compete for attention capacity in a single stream but can specialize across two streams. To isolate this tradeoff from position-content separation, we propose Decoupled RoPE, a modification to rotary position embeddings that provides target position information without revealing target content. Decoupled RoPE performs competitively at short sequence lengths--where semantic and structural proximity coincide--but degrades as sequence length increases and the two orderings diverge. These results suggest that the success of two-stream attention stems not merely from separating position from content, but from circumventing the deeper structural-semantic tradeoff inherent to any-order generation.

Background: Symbolic models, particularly decision trees, are widely used in software engineering for explainable analytics in defect prediction, configuration tuning, and software quality assessment. Most of these models rely on correlational split criteria, such as variance reduction or information gain, which identify statistical associations but cannot imply causation between X and Y. Recent empirical studies in software engineering show that both correlational models and causal discovery algorithms suffer from pronounced instability. This instability arises from two complementary issues: 1-Correlation-based methods conflate association with causation. 2-Causal discovery algorithms rely on heuristic approximations to cope with the NP-hard nature of structure learning, causing their inferred graphs to vary widely under minor input perturbations. Together, these issues undermine trust, reproducibility, and the reliability of explanations in real-world SE tasks. Objective: This study investigates whether incorporating causality-aware split criteria into symbolic models can improve their stability and robustness, and whether such gains come at the cost of predictive or optimization performance. We additionally examine how the stability of human expert judgments compares to that of automated models. Method: Using 120+ multi-objective optimization tasks from the MOOT repository of multi-objective optimization tasks, we evaluate stability through a preregistered bootstrap-ensemble protocol that measures variance with win-score assignments. We compare the stability of human causal assessments with correlation-based decision trees (EZR). We would also compare the causality-aware trees, which leverage conditional-entropy split criteria and confounder filtering. Stability and performance differences are analyzed using statistical methods (variance, Gini Impurity, KS test, Cliff's delta)

The response of the climate system to increased greenhouse gases and other radiative perturbations is governed by a combination of fast and slow feedbacks. Slow feedbacks are typically activated in response to changes in ocean temperatures on decadal timescales and manifest as changes in climatic state with no recent historical analogue. However, fast feedbacks are activated in response to rapid atmospheric physical processes on weekly timescales, and they are already operative in the present-day climate. This distinction implies that the physics of fast radiative feedbacks is present in the historical meteorological reanalyses used to train many recent successful machine-learning-based (ML) emulators of weather and climate. In addition, these feedbacks are functional under the historical boundary conditions pertaining to the top-of-atmosphere radiative balance and sea-surface temperatures. Together, these factors imply that we can use historically trained ML weather emulators to study the response of radiative-convective equilibrium (RCE), and hence the global hydrological cycle, to perturbations in carbon dioxide and other well-mixed greenhouse gases. Without retraining on prospective Earth system conditions, we use ML weather emulators to quantify the fast precipitation response to reduced and elevated carbon dioxed concentrations with no recent historical precedent. We show that the responses from historically trained emulators agree with those produced by full-physics Earth System Models (ESMs). In conclusion, we discuss the prospects for and advantages from using ESMs and ML emulators to study fast processes in global climate.

Discrete image tokenization is a key bottleneck for scalable visual generation: a tokenizer must remain compact for efficient latent-space priors while preserving semantic structure and using discrete capacity effectively. Existing quantizers face a trade-off: vector-quantized tokenizers learn flexible geometries but often suffer from biased straight-through optimization, codebook under-utilization, and representation collapse at large vocabularies. Structured scalar or implicit tokenizers ensure stable, near-complete utilization by design, yet rely on fixed discretization geometries that may allocate capacity inefficiently under heterogeneous latent statistics. We introduce Learnable Geometric Quantization (LGQ), a discrete image tokenizer that learns discretization geometry end-to-end. LGQ replaces hard nearest-neighbor lookup with temperature-controlled soft assignments, enabling fully differentiable training while recovering hard assignments at inference. The assignments correspond to posterior responsibilities of an isotropic Gaussian mixture and minimize a variational free-energy objective, provably converging to nearest-neighbor quantization in the low-temperature limit. LGQ combines a token-level peakedness regularizer with a global usage regularizer to encourage confident yet balanced code utilization without imposing rigid grids. Under a controlled VQGAN-style backbone on ImageNet across multiple vocabulary sizes, LGQ achieves stable optimization and balanced utilization. At 16K codebook size, LGQ improves rFID by 11.88% over FSQ while using 49.96% fewer active codes, and improves rFID by 6.06% over SimVQ with 49.45% lower effective representation rate, achieving comparable fidelity with substantially fewer active entries. Our GitHub repository is available at: https://github.com/KurbanIntelligenceLab/LGQ

Research on mental state reasoning in language models (LMs) has the potential to inform theories of human social cognition--such as the theory that mental state reasoning emerges in part from language exposure--and our understanding of LMs themselves. Yet much published work on LMs relies on a relatively small sample of closed-source LMs, limiting our ability to rigorously test psychological theories and evaluate LM capacities. Here, we replicate and extend published work on the false belief task by assessing LM mental state reasoning behavior across 41 open-weight models (from distinct model families). We find sensitivity to implied knowledge states in 34% of the LMs tested; however, consistent with prior work, none fully ``explain away'' the effect in humans. Larger LMs show increased sensitivity and also exhibit higher psychometric predictive power. Finally, we use LM behavior to generate and test a novel hypothesis about human cognition: both humans and LMs show a bias towards attributing false beliefs when knowledge states are cued using a non-factive verb (``John thinks...'') than when cued indirectly (``John looks in the...''). Unlike the primary effect of knowledge states, where human sensitivity exceeds that of LMs, the magnitude of the human knowledge cue effect falls squarely within the distribution of LM effect sizes-suggesting that distributional statistics of language can in principle account for the latter but not the former in humans. These results demonstrate the value of using larger samples of open-weight LMs to test theories of human cognition and evaluate LM capacities.

Where should we intervene in a language model (LM) to control behaviors that are diffused across many tokens of a long-form response? We introduce Generative Causal Mediation (GCM), a procedure for selecting model components, e.g., attention heads, to steer a binary concept (e.g., talk in verse vs. talk in prose) from contrastive long-form responses. In GCM, we first construct a dataset of contrasting inputs and responses. Then, we quantify how individual model components mediate the contrastive concept and select the strongest mediators for steering. We evaluate GCM on three tasks--refusal, sycophancy, and style transfer--across three language models. GCM successfully localizes concepts expressed in long-form responses and consistently outperforms correlational probe-based baselines when steering with a sparse set of attention heads. Together, these results demonstrate that GCM provides an effective approach for localizing and controlling the long-form responses of LMs.

Analog processing-using-memory (PUM; a.k.a. in-memory computing) makes use of electrical interactions inside memory arrays to perform bulk matrix-vector multiplication (MVM) operations. However, many popular matrix-based kernels need to execute non-MVM operations, which analog PUM cannot directly perform. To retain its energy efficiency, analog PUM architectures augment memory arrays with CMOS-based domain-specific fixed-function hardware to provide complete kernel functionality, but the difficulty of integrating such specialized CMOS logic with memory arrays has largely limited analog PUM to being an accelerator for machine learning inference, or for closely related kernels. An opportunity exists to harness analog PUM for general-purpose computation: recent works have shown that memory arrays can also perform Boolean PUM operations, albeit with very different supporting hardware and electrical signals than analog PUM. We propose DARTH-PUM, a general-purpose hybrid PUM architecture that tackles key hardware and software challenges to integrating analog PUM and digital PUM. We propose optimized peripheral circuitry, coordinating hardware to manage and interface between both types of PUM, an easy-to-use programming interface, and low-cost support for flexible data widths. These design elements allow us to build a practical PUM architecture that can execute kernels fully in memory, and can scale easily to cater to domains ranging from embedded applications to large-scale data-driven computing. We show how three popular applications (AES encryption, convolutional neural networks, large-language models) can map to and benefit from DARTH-PUM, with speedups of 59.4x, 14.8x, and 40.8x over an analog+CPU baseline.

Developing autonomous driving systems for complex traffic environments requires balancing multiple objectives, such as avoiding collisions, obeying traffic rules, and making efficient progress. In many situations, these objectives cannot be satisfied simultaneously, and explicit priority relations naturally arise. Also, driving rules require context, so it is important to formally model the environment scenarios within which such rules apply. Existing benchmarks for evaluating autonomous vehicles lack such combinations of multi-objective prioritized rules and formal environment models. In this work, we introduce ScenicRules, a benchmark for evaluating autonomous driving systems in stochastic environments under prioritized multi-objective specifications. We first formalize a diverse set of objectives to serve as quantitative evaluation metrics. Next, we design a Hierarchical Rulebook framework that encodes multiple objectives and their priority relations in an interpretable and adaptable manner. We then construct a compact yet representative collection of scenarios spanning diverse driving contexts and near-accident situations, formally modeled in the Scenic language. Experimental results show that our formalized objectives and Hierarchical Rulebooks align well with human driving judgments and that our benchmark effectively exposes agent failures with respect to the prioritized objectives. Our benchmark can be accessed at https://github.com/BerkeleyLearnVerify/ScenicRules/.

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

Rapidly increasing context lengths have led to the assumption that large language models (LLMs) can directly reason over entire codebases. Concurrently, recent advances in LLMs have enabled strong performance on software engineering benchmarks, particularly when paired with agentic workflows. In this work, we systematically evaluate whether current LLMs can reliably perform long-context code debugging and patch generation. Using SWE-bench Verified as a controlled experimental setting, we first evaluate state-of-the-art models within an agentic harness (mini-SWE-agent), where performance improves substantially: GPT-5-nano achieves up to a 31\% resolve rate on 100 samples, and open-source models such as Deepseek-R1-0528 obtain competitive results. However, token-level analysis shows that successful agentic trajectories typically remain under 20k tokens, and that longer accumulated contexts correlate with lower success rates, indicating that agentic success primarily arises from task decomposition into short-context steps rather than effective long-context reasoning. To directly test long-context capability, we construct a data pipeline where we artificially inflate the context length of the input by placing the relevant files into the context (ensuring perfect retrieval recall); we then study single-shot patch generation under genuinely long contexts (64k-128k tokens). Despite this setup, performance degrades sharply: Qwen3-Coder-30B-A3B achieves only a 7\% resolve rate at 64k context, while GPT-5-nano solves none of the tasks. Qualitative analysis reveals systematic failure modes, including hallucinated diffs, incorrect file targets, and malformed patch headers. Overall, our findings highlight a significant gap between nominal context length and usable context capacity in current LLMs, and suggest that existing agentic coding benchmarks do not meaningfully evaluate long-context reasoning.

Adapting one's thought process based on corrective feedback is an essential ability in human learning, particularly in collaborative settings. In contrast, the current large language model training paradigm relies heavily on modeling vast, static corpora. While effective for knowledge acquisition, it overlooks the interactive feedback loops essential for models to adapt dynamically to their context. In this work, we propose a framework that treats this interactive in-context learning ability not as an emergent property, but as a distinct, trainable skill. We introduce a scalable method that transforms single-turn verifiable tasks into multi-turn didactic interactions driven by information asymmetry. We first show that current flagship models struggle to integrate corrective feedback on hard reasoning tasks. We then demonstrate that models trained with our approach dramatically improve the ability to interactively learn from language feedback. More specifically, the multi-turn performance of a smaller model nearly reaches that of a model an order of magnitude larger. We also observe robust out-of-distribution generalization: interactive training on math problems transfers to diverse domains like coding, puzzles and maze navigation. Our qualitative analysis suggests that this improvement is due to an enhanced in-context plasticity. Finally, we show that this paradigm offers a unified path to self-improvement. By training the model to predict the teacher's critiques, effectively modeling the feedback environment, we convert this external signal into an internal capability, allowing the model to self-correct even without a teacher.

Generative Artificial Intelligence (AI), such as large language models (LLMs), has become a transformative force across science, industry, and society. As these systems grow in popularity, web data becomes increasingly interwoven with this AI-generated material and it is increasingly difficult to separate them from naturally generated content. As generative models are updated regularly, later models will inevitably be trained on mixtures of human-generated data and AI-generated data from earlier versions, creating a recursive training process with data contamination. Existing theoretical work has examined only highly simplified settings, where both the real data and the generative model are discrete or Gaussian, where it has been shown that such recursive training leads to model collapse. However, real data distributions are far more complex, and modern generative models are far more flexible than Gaussian and linear mechanisms. To fill this gap, we study recursive training in a general framework with minimal assumptions on the real data distribution and allow the underlying generative model to be a general universal approximator. In this framework, we show that contaminated recursive training still converges, with a convergence rate equal to the minimum of the baseline model's convergence rate and the fraction of real data used in each iteration. To the best of our knowledge, this is the first (positive) theoretical result on recursive training without distributional assumptions on the data. We further extend the analysis to settings where sampling bias is present in data collection and support all theoretical results with empirical studies.

This paper introduces a novel, open-source MARL simulation framework for studying implicit cooperation in LEMs, modeled as a decentralized partially observable Markov decision process and implemented as a Gymnasium environment for MARL. Our framework features a modular market platform with plug-and-play clearing mechanisms, physically constrained agent models (including battery storage), a realistic grid network, and a comprehensive analytics suite to evaluate emergent coordination. The main contribution is a novel method to foster implicit cooperation, where agents' observations and rewards are enhanced with system-level key performance indicators to enable them to independently learn strategies that benefit the entire system and aim for collectively beneficial outcomes without explicit communication. Through representative case studies (available in a dedicated GitHub repository in https://github.com/salazarna/marlem, we show the framework's ability to analyze how different market configurations (such as varying storage deployment) impact system performance. This illustrates its potential to facilitate emergent coordination, improve market efficiency, and strengthen grid stability. The proposed simulation framework is a flexible, extensible, and reproducible tool for researchers and practitioners to design, test, and validate strategies for future intelligent, decentralized energy systems.

This paper proposes implicit cooperation, a framework enabling decentralized agents to approximate optimal coordination in local energy markets without explicit peer-to-peer communication. We formulate the problem as a decentralized partially observable Markov decision problem that is solved through a multi-agent reinforcement learning task in which agents use stigmergic signals (key performance indicators at the system level) to infer and react to global states. Through a 3x3 factorial design on an IEEE 34-node topology, we evaluated three training paradigms (CTCE, CTDE, DTDE) and three algorithms (PPO, APPO, SAC). Results identify APPO-DTDE as the optimal configuration, achieving a coordination score of 91.7% relative to the theoretical centralized benchmark (CTCE). However, a critical trade-off emerges between efficiency and stability: while the centralized benchmark maximizes allocative efficiency with a peer-to-peer trade ratio of 0.6, the fully decentralized approach (DTDE) demonstrates superior physical stability. Specifically, DTDE reduces the variance of grid balance by 31% compared to hybrid architectures, establishing a highly predictable, import-biased load profile that simplifies grid regulation. Furthermore, topological analysis reveals emergent spatial clustering, where decentralized agents self-organize into stable trading communities to minimize congestion penalties. While SAC excelled in hybrid settings, it failed in decentralized environments due to entropy-driven instability. This research proves that stigmergic signaling provides sufficient context for complex grid coordination, offering a robust, privacy-preserving alternative to expensive centralized communication infrastructure.

Estimating population quantities such as mean outcomes from user feedback is fundamental to platform evaluation and social science, yet feedback is often missing not at random (MNAR): users with stronger opinions are more likely to respond, so standard estimators are biased and the estimand is not identified without additional assumptions. Existing approaches typically rely on strong parametric assumptions or bespoke auxiliary variables that may be unavailable in practice. In this paper, we develop a partial identification framework in which sharp bounds on the estimand are obtained by solving a pair of linear programs whose constraints encode the observed data structure. This formulation naturally incorporates outcome predictions from pretrained models, including large language models (LLMs), as additional linear constraints that tighten the feasible set. We call these predictions weak shadow variables: they satisfy a conditional independence assumption with respect to missingness but need not meet the completeness conditions required by classical shadow-variable methods. When predictions are sufficiently informative, the bounds collapse to a point, recovering standard identification as a special case. In finite samples, to provide valid coverage of the identified set, we propose a set-expansion estimator that achieves slower-than-$\sqrt{n}$ convergence rate in the set-identified regime and the standard $\sqrt{n}$ rate under point identification. In simulations and semi-synthetic experiments on customer-service dialogues, we find that LLM predictions are often ill-conditioned for classical shadow-variable methods yet remain highly effective in our framework. They shrink identification intervals by 75--83\% while maintaining valid coverage under realistic MNAR mechanisms.

Railway crossings present complex safety challenges where driver behavior varies by location, time, and conditions. Traditional approaches analyze crossings individually, limiting the ability to identify shared behavioral patterns across locations. We propose a multi-view tensor decomposition framework that captures behavioral similarities across three temporal phases: Approach (warning activation to gate lowering), Waiting (gates down to train passage), and Clearance (train passage to gate raising). We analyze railway crossing videos from multiple locations using TimeSformer embeddings to represent each phase. By constructing phase-specific similarity matrices and applying non-negative symmetric CP decomposition, we discover latent behavioral components with distinct temporal signatures. Our tensor analysis reveals that crossing location appears to be a stronger determinant of behavior patterns than time of day, and that approach-phase behavior provides particularly discriminative signatures. Visualization of the learned component space confirms location-based clustering, with certain crossings forming distinct behavioral clusters. This automated framework enables scalable pattern discovery across multiple crossings, providing a foundation for grouping locations by behavioral similarity to inform targeted safety interventions.

The prefill stage in long-context LLM inference remains a computational bottleneck. Recent token-ranking heuristics accelerate inference by selectively processing a subset of semantically relevant tokens. However, existing methods suffer from unstable token importance estimation, often varying between layers. Evaluating token-ranking quality independently from heuristic-specific architectures is challenging. To address this, we introduce an Answer-Informed Oracle, which defines ground-truth token importance by measuring attention from generated answers back to the prompt. This oracle reveals that existing heuristics exhibit high variance across layers: rankings can degrade sharply at specific layers, a failure mode invisible to end-to-end benchmarks. The diagnosis suggests a simple fix: aggregate scores across layers rather than relying on any single one. We implement this as Cross-Layer Attention Aggregation (CLAA), which closes the gap to the oracle upper bound and reduces Time-to-First-Token (TTFT) by up to 39\% compared to the Full KV Cache baseline.

Mental health disorders affect over 1 billion people worldwide, yet access to care remains limited by workforce shortages and cost constraints. While AI systems show therapeutic promise, current alignment approaches optimize objectives independently, failing to balance patient preferences with clinical safety. We survey 335 individuals with lived mental health experience to collect preference rankings across therapeutic dimensions, then develop a multi-objective alignment framework using direct preference optimization. We train reward models for six criteria -- empathy, safety, active listening, self-motivated change, trust/rapport, and patient autonomy -- and systematically compare multi-objective approaches against single-objective optimization, supervised fine-tuning, and parameter merging. Multi-objective DPO (MODPO) achieves superior balance (77.6% empathy, 62.6% safety) compared to single-objective optimization (93.6% empathy, 47.8% safety), and therapeutic criteria outperform general communication principles by 17.2%. Blinded clinician evaluation confirms MODPO is consistently preferred, with LLM-evaluator agreement comparable to inter-clinician reliability.

Speculative decoding accelerates Large Language Model (LLM) inference by verifying multiple drafted tokens in parallel. However, for Mixture-of-Experts (MoE) models, this parallelism introduces a severe bottleneck: large draft trees activate many unique experts, significantly increasing memory pressure and diminishing speedups from speculative decoding relative to autoregressive decoding. Prior methods reduce speculation depth when MoE verification becomes expensive. We propose MoE-Spec, a training-free verification-time expert budgeting method that decouples speculation depth from memory cost by enforcing a fixed expert capacity limit at each layer, loading only the experts that contribute most to verification and dropping the long tail of rarely used experts that drive bandwidth overhead. Experiments across multiple model scales and datasets show that this method yields 10--30\% higher throughput than state-of-the-art speculative decoding baselines (EAGLE-3) at comparable quality, with flexibility to trade accuracy for further latency reductions through tighter budgets.

Background: Large language models have demonstrated strong performance on general medical examinations, but subspecialty clinical reasoning remains challenging due to rapidly evolving guidelines and nuanced evidence hierarchies. Methods: We evaluated January Mirror, an evidence-grounded clinical reasoning system, against frontier LLMs (GPT-5, GPT-5.2, Gemini-3-Pro) on a 120-question endocrinology board-style examination. Mirror integrates a curated endocrinology and cardiometabolic evidence corpus with a structured reasoning architecture to generate evidence-linked outputs. Mirror operated under a closed-evidence constraint without external retrieval. Comparator LLMs had real-time web access to guidelines and primary literature. Results: Mirror achieved 87.5% accuracy (105/120; 95% CI: 80.4-92.3%), exceeding a human reference of 62.3% and frontier LLMs including GPT-5.2 (74.6%), GPT-5 (74.0%), and Gemini-3-Pro (69.8%). On the 30 most difficult questions (human accuracy less than 50%), Mirror achieved 76.7% accuracy. Top-2 accuracy was 92.5% for Mirror versus 85.25% for GPT-5.2. Conclusions: Mirror provided evidence traceability: 74.2% of outputs cited at least one guideline-tier source, with 100% citation accuracy on manual verification. Curated evidence with explicit provenance can outperform unconstrained web retrieval for subspecialty clinical reasoning and supports auditability for clinical deployment.

We present a workflow and associated toolkit to automate the creation of graphical user interfaces (GUI) for executables run from command line interfaces (CLI). The workflow consists of three phases, namely (Step 1) the plugin design, (Step 2) the formal (platform independent) specification of the GUI, and (Step 3) the plugin code generation for the targeted platforms. Our architecture is aligned with the Model--View--Presenter (MVP) pattern: steps one and two build the Model and View descriptions, while step three implements the Presenter layer that binds inputs, invokes the CLI, and updates outputs. Once Step one has been (manually) completed, steps two and three are fully automated. The decoupled MVP design and platform-specific generator modules enable reuse of logic, portability across ecosystems, and significant reductions in engineering effort for complex interactive applications. We primarily use our workflow to generate GUI in structural bioinformatics for CLI executables from the Structural Bioinformatics Library (SBL), targeting three platforms, namely VMD, Pymol and Web servers. The workflow can be used as a guideline, while its implementation available in the package Plugin_manager from the SBL, see https://sbl.inria.fr/doc/Plugin_manager-user-manual.html.

As machine learning (ML) continues its rapid expansion, the environmental cost of model training and inference has become a critical societal concern. Existing benchmarks overwhelmingly focus on standard performance metrics such as accuracy, BLEU, or mAP, while largely ignoring energy consumption and carbon emissions. This single-objective evaluation paradigm is increasingly misaligned with the practical requirements of large-scale deployment, particularly in energy-constrained environments such as mobile devices, developing regions, and climate-aware enterprises. In this paper, we propose AI-CARE, an evaluation tool for reporting energy consumption, and carbon emissions of ML models. In addition, we introduce the carbon-performance tradeoff curve, an interpretable tool that visualizes the Pareto frontier between performance and carbon cost. We demonstrate, through theoretical analysis and empirical validation on representative ML workloads, that carbon-aware benchmarking changes the relative ranking of models and encourages architectures that are simultaneously accurate and environmentally responsible. Our proposal aims to shift the research community toward transparent, multi-objective evaluation and align ML progress with global sustainability goals. The tool and documentation are available at https://github.com/USD-AI-ResearchLab/ai-care.

The rapid rise of large language models (LLMs) is reshaping the landscape of automatic assessment in education. While these systems demonstrate substantial advantages in adaptability to diverse question types and flexibility in output formats, they also introduce new challenges related to output uncertainty, stemming from the inherently probabilistic nature of LLMs. Output uncertainty is an inescapable challenge in automatic assessment, as assessment results often play a critical role in informing subsequent pedagogical actions, such as providing feedback to students or guiding instructional decisions. Unreliable or poorly calibrated uncertainty estimates can lead to unstable downstream interventions, potentially disrupting students' learning processes and resulting in unintended negative consequences. To systematically understand this challenge and inform future research, we benchmark a broad range of uncertainty quantification methods in the context of LLM-based automatic assessment. Although the effectiveness of these methods has been demonstrated in many tasks across other domains, their applicability and reliability in educational settings, particularly for automatic grading, remain underexplored. Through comprehensive analyses of uncertainty behaviors across multiple assessment datasets, LLM families, and generation control settings, we characterize the uncertainty patterns exhibited by LLMs in grading scenarios. Based on these findings, we evaluate the strengths and limitations of different uncertainty metrics and analyze the influence of key factors, including model families, assessment tasks, and decoding strategies, on uncertainty estimates. Our study provides actionable insights into the characteristics of uncertainty in LLM-based automatic assessment and lays the groundwork for developing more reliable and effective uncertainty-aware grading systems in the future.

Large Language Models (LLMs) have advanced Automated Heuristic Design (AHD) in combinatorial optimization (CO) in the past few years. However, existing discovery pipelines often require extensive manual trial-and-error or reliance on domain expertise to adapt to new or complex problems. This stems from tightly coupled internal mechanisms that limit systematic improvement of the LLM-driven design process. To address this challenge, we propose a structured framework for LLM-driven AHD that explicitly decomposes the heuristic discovery process into modular stages: a forward pass for evaluation, a backward pass for analytical feedback, and an update step for program refinement. This separation provides a clear abstraction for iterative refinement and enables principled improvements of individual components. We validate our framework across four diverse real-world CO domains, where it consistently outperforms baselines, achieving up to $0.17$ improvement in QYI on unseen test sets. Finally, we show that several popular AHD methods are restricted instantiations of our framework. By integrating them in our structured pipeline, we can upgrade the components modularly and significantly improve their performance.

Autonomous agentic workflows that iteratively refine their own behavior hold considerable promise, yet their failure modes remain poorly characterized. We investigate optimization instability, a phenomenon in which continued autonomous improvement paradoxically degrades classifier performance, using Pythia, an open-source framework for automated prompt optimization. Evaluating three clinical symptoms with varying prevalence (shortness of breath at 23%, chest pain at 12%, and Long COVID brain fog at 3%), we observed that validation sensitivity oscillated between 1.0 and 0.0 across iterations, with severity inversely proportional to class prevalence. At 3% prevalence, the system achieved 95% accuracy while detecting zero positive cases, a failure mode obscured by standard evaluation metrics. We evaluated two interventions: a guiding agent that actively redirected optimization, amplifying overfitting rather than correcting it, and a selector agent that retrospectively identified the best-performing iteration successfully prevented catastrophic failure. With selector agent oversight, the system outperformed expert-curated lexicons on brain fog detection by 331% (F1) and chest pain by 7%, despite requiring only a single natural language term as input. These findings characterize a critical failure mode of autonomous AI systems and demonstrate that retrospective selection outperforms active intervention for stabilization in low-prevalence classification tasks.

Despite universal GenAI adoption, students cannot distinguish task performance from actual learning and lack skills to leverage AI for learning, leading to worse exam performance when AI use remains unreflective. Yet few interventions teaching students to prompt AI as a tutor rather than solution provider have been validated at scale through randomized controlled trials (RCTs). To bridge this gap, we conducted a semester-long RCT (N=979) with four ICAP framework-based instructional conditions varying in engagement intensity with a pre-test, immediate and delayed post-test and surveys. Mixed methods analysis results showed: (1) All conditions significantly improved prompting skills, with gains increasing progressively from Condition 1 to Condition 4, validating ICAP's cognitive engagement hierarchy; (2) for students with similar pre-test scores, higher learning gain in immediate post-test predict higher final exam score, though no direct between-group differences emerged; (3) Our interventions are suitable and scalable solutions for diverse educational contexts, resources and learners. Together, this study makes empirical and theoretical contributions: (1) theoretically, we provided one of the first large-scale RCTs examining how cognitive engagement shapes learning in prompting literacy and clarifying the relationship between learning-oriented prompting skills and broader academic performance; (2) empirically, we offered timely design guidance for transforming GenAI classroom policies into scalable, actionable prompting literacy instruction to advance learning in the era of Generative AI.

In this study, we propose an implementation methodology of real-time few-shot learning on tiny FPGA SoCs such as the PYNQ-Z1 board with arbitrary fixed-point bit-widths. Tensil-based conventional design environments limited hardware implementations to fixed-point bit-widths of 16 or 32 bits. To address this, we adopt the FINN framework, enabling implementations with arbitrary bit-widths. Several customizations and minor adjustments are made, including: 1.Optimization of Transpose nodes to resolve data format mismatches, 2.Addition of handling for converting the final reduce mean operation to Global Average Pooling (GAP). These adjustments allow us to reduce the bit-width while maintaining the same accuracy as the conventional realization, and achieve approximately twice the throughput in evaluations using CIFAR-10 dataset.

Multiword expressions (MWEs) have been widely studied in cross-lingual annotation frameworks such as PARSEME. However, Korean MWEs remain underrepresented in these efforts. In particular, Korean multiword adpositions lack systematic analysis, annotated resources, and integration into existing multilingual frameworks. In this paper, we study a class of Korean functional multiword expressions: postpositional verb-based constructions (PVCs). Using data from Korean Wikipedia, we survey and analyze several PVC expressions and contrast them with non-MWEs and light verb constructions (LVCs) with similar structure. Building on this analysis, we propose annotation guidelines designed to support future work in Korean multiword adpositions and facilitate alignment with cross-lingual frameworks.

Molecular crystal structure prediction represents a grand challenge in computational chemistry due to large sizes of constituent molecules and complex intra- and intermolecular interactions. While generative modeling has revolutionized structure discovery for molecules, inorganic solids, and metal-organic frameworks, extending such approaches to fully periodic molecular crystals is still elusive. Here, we present MolCrystalFlow, a flow-based generative model for molecular crystal structure prediction. The framework disentangles intramolecular complexity from intermolecular packing by embedding molecules as rigid bodies and jointly learning the lattice matrix, molecular orientations, and centroid positions. Centroids and orientations are represented on their native Riemannian manifolds, allowing geodesic flow construction and graph neural network operations that respects geometric symmetries. We benchmark our model against state-of-the-art generative models for large-size periodic crystals and rule-based structure generation methods on two open-source molecular crystal datasets. We demonstrate an integration of MolCrystalFlow model with universal machine learning potential to accelerate molecular crystal structure prediction, paving the way for data-driven generative discovery of molecular crystals.

Vision-language foundation models have emerged as powerful general-purpose representation learners with strong potential for multimodal understanding, but their deterministic embeddings often fail to provide the reliability required for high-stakes biomedical applications. This work introduces MedProbCLIP, a probabilistic vision-language learning framework for chest X-ray and radiology report representation learning and bidirectional retrieval. MedProbCLIP models image and text representations as Gaussian embeddings through a probabilistic contrastive objective that explicitly captures uncertainty and many-to-many correspondences between radiographs and clinical narratives. A variational information bottleneck mitigates overconfident predictions, while MedProbCLIP employs multi-view radiograph encoding and multi-section report encoding during training to provide fine-grained supervision for clinically aligned correspondence, yet requires only a single radiograph and a single report at inference. Evaluated on the MIMIC-CXR dataset, MedProbCLIP outperforms deterministic and probabilistic baselines, including CLIP, CXR-CLIP, and PCME++, in both retrieval and zero-shot classification. Beyond accuracy, MedProbCLIP demonstrates superior calibration, risk-coverage behavior, selective retrieval reliability, and robustness to clinically relevant corruptions, underscoring the value of probabilistic vision-language modeling for improving the trustworthiness and safety of radiology image-text retrieval systems.

Graph neural networks (GNNs) are a powerful framework for learning representations from graph-structured data, but their direct implementation on near-term quantum hardware remains challenging due to circuit depth, multi-qubit interactions, and qubit scalability constraints. In this work, we introduce a fully quantum graph convolutional architecture designed explicitly for unsupervised learning in the noisy intermediate-scale quantum (NISQ) regime. Our approach combines a variational quantum feature extraction layer with an edge-local and qubit-efficient quantum message-passing mechanism inspired by the Quantum Alternating Operator Ansatz (QAOA) framework. Unlike prior models that rely on global operations or multi-controlled unitaries, our model decomposes message passing into pairwise interactions along graph edges using only hardware-native single- and two-qubit gates. This design reduces the qubit requirement from $O(Nn)$ to $O(n)$ for a graph with $N$ nodes and $n$-qubit feature registers, enabling implementation on current quantum devices regardless of graph size. We train the model using the Deep Graph Infomax objective to perform unsupervised node representation learning. Experiments on the Cora citation network and a large-scale genomic SNP dataset demonstrate that our model remains competitive with prior quantum and hybrid approaches.

Conformal prediction provides distribution-free coverage guaranties for regression; yet existing methods assume Euclidean output spaces and produce prediction regions that are poorly calibrated when responses lie on Riemannian manifolds. We propose \emph{adaptive geodesic conformal prediction}, a framework that replaces Euclidean residuals with geodesic nonconformity scores and normalizes them by a cross-validated difficulty estimator to handle heteroscedastic noise. The resulting prediction regions, geodesic caps on the sphere, have position-independent area and adapt their size to local prediction difficulty, yielding substantially more uniform conditional coverage than non-adaptive alternatives. In a synthetic sphere experiment with strong heteroscedasticity and a real-world geomagnetic field forecasting task derived from IGRF-14 satellite data, the adaptive method markedly reduces conditional coverage variability and raises worst-case coverage much closer to the nominal level, while coordinate-based baselines waste a large fraction of coverage area due to chart distortion.

Neural solvers have achieved impressive progress in addressing simple routing problems, particularly excelling in computational efficiency. However, their advantages under complex constraints remain nascent, for which current constraint-handling schemes via feasibility masking or implicit feasibility awareness can be inefficient or inapplicable for hard constraints. In this paper, we present Construct-and-Refine (CaR), the first general and efficient constraint-handling framework for neural routing solvers based on explicit learning-based feasibility refinement. Unlike prior construction-search hybrids that target reducing optimality gaps through heavy improvements yet still struggle with hard constraints, CaR achieves efficient constraint handling by designing a joint training framework that guides the construction module to generate diverse and high-quality solutions well-suited for a lightweight improvement process, e.g., 10 steps versus 5k steps in prior work. Moreover, CaR presents the first use of construction-improvement-shared representation, enabling potential knowledge sharing across paradigms by unifying the encoder, especially in more complex constrained scenarios. We evaluate CaR on typical hard routing constraints to showcase its broader applicability. Results demonstrate that CaR achieves superior feasibility, solution quality, and efficiency compared to both classical and neural state-of-the-art solvers.

Checkpoint/Restart (C/R) saves the running state of the programs periodically, which consumes considerable system resources. We observe that not every piece of data is involved in the computation in typical HPC applications; such unused data should be excluded from checkpointing for better storage/compute efficiency. To find out, we propose a systematic approach that leverages automatic differentiation (AD) to scrutinize every element within variables (e.g., arrays) for checkpointing allowing us to identify critical/uncritical elements and eliminate uncritical elements from checkpointing. Specifically, we inspect every single element within a variable for checkpointing with an AD tool to determine whether the element has an impact on the application output or not. We empirically validate our approach with eight benchmarks from the NAS Parallel Benchmark (NPB) suite. We successfully visualize critical/uncritical elements/regions within a variable with respect to its impact (yes or no) on the application output. We find patterns/distributions of critical/uncritical elements/regions quite interesting and follow the physical formulation/logic of the algorithm.The evaluation on NPB benchmarks shows that our approach saves storage for checkpointing by up to 20%.

We introduce the Massive Audio Embedding Benchmark (MAEB), a large-scale benchmark covering 30 tasks across speech, music, environmental sounds, and cross-modal audio-text reasoning in 100+ languages. We evaluate 50+ models and find that no single model dominates across all tasks: contrastive audio-text models excel at environmental sound classification (e.g., ESC50) but score near random on multilingual speech tasks (e.g., SIB-FLEURS), while speech-pretrained models show the opposite pattern. Clustering remains challenging for all models, with even the best-performing model achieving only modest results. We observe that models excelling on acoustic understanding often perform poorly on linguistic tasks, and vice versa. We also show that the performance of audio encoders on MAEB correlates highly with their performance when used in audio large language models. MAEB is derived from MAEB+, a collection of 98 tasks. MAEB is designed to maintain task diversity while reducing evaluation cost, and it integrates into the MTEB ecosystem for unified evaluation across text, image, and audio modalities. We release MAEB and all 98 tasks along with code and a leaderboard at https://github.com/embeddings-benchmark/mteb.

We introduce ODYN, a novel all-shifted primal-dual non-interior-point quadratic programming (QP) solver designed to efficiently handle challenging dense and sparse QPs. ODYN combines all-shifted nonlinear complementarity problem (NCP) functions with proximal method of multipliers to robustly address ill-conditioned and degenerate problems, without requiring linear independence of the constraints. It exhibits strong warm-start performance and is well suited to both general-purpose optimization, and robotics and AI applications, including model-based control, estimation, and kernel-based learning methods. We provide an open-source implementation and benchmark ODYN on the Maros-Mészáros test set, demonstrating state-of-the-art convergence performance in small-to-high-scale problems. The results highlight ODYN's superior warm-starting capabilities, which are critical in sequential and real-time settings common in robotics and AI. These advantages are further demonstrated by deploying ODYN as the backend of an SQP-based predictive control framework (OdynSQP), as the implicitly differentiable optimization layer for deep learning (ODYNLayer), and the optimizer of a contact-dynamics simulation (ODYNSim).

Purpose of Review Imaging derived fractional flow reserve (FFR) is rapidly evolving beyond conventional computational fluid dynamics (CFD) based pipelines toward machine learning (ML), deep learning (DL), and physics informed approaches that enable fast, wire free, and scalable functional assessment of coronary stenosis. This review synthesizes recent advances in CT and angiography based FFR, with particular emphasis on emerging physics informed neural networks and neural operators (PINNs and PINOs) and key considerations for their clinical translation. Recent Findings ML/DL approaches have markedly improved automation and computational speed, enabling prediction of pressure and FFR from anatomical descriptors or angiographic contrast dynamics. However, their real-world performance and generalizability can remain variable and sensitive to domain shift, due to multi-center heterogeneity, interpretability challenges, and differences in acquisition protocols and image quality. Physics informed learning introduces conservation structure and boundary condition consistency into model training, improving generalizability and reducing dependence on dense supervision while maintaining rapid inference. Recent evaluation trends increasingly highlight deployment oriented metrics, including calibration, uncertainty quantification, and quality control gatekeeping, as essential for safe clinical use. Summary The field is converging toward imaging derived FFR methods that are faster, more automated, and more reliable. While ML/DL offers substantial efficiency gains, physics informed frameworks such as PINNs and PINOs may provide a more robust balance between speed and physical consistency. Prospective multi center validation and standardized evaluation will be critical to support broad and safe clinical adoption.

Capability emergence during neural network training remains mechanistically opaque. We track five geometric measures across five model scales (405K-85M parameters), 120+ emergence events in eight algorithmic tasks, and three Pythia language models (160M-2.8B). We find: (1) training begins with a universal representation collapse to task-specific floors that are scale-invariant across a 210X parameter range (e.g., modular arithmetic collapses to RANKME ~ 2.0 regardless of model size); (2) collapse propagates top-down through layers (32/32 task X model consistency), contradicting bottom-up feature-building intuition; (3) a geometric hierarchy in which representation geometry leads emergence (75-100% precursor rate for hard tasks), while the local learning coefficient is synchronous (0/24 precursor) and Hessian measures lag. We also delineate prediction limits: geometric measures encode coarse task difficulty but not fine-grained timing (within-class concordance 27%; when task ordering reverses across scales, prediction fails at 26%). On Pythia, global geometric patterns replicate but per-task precursor signals do not -- the precursor relationship requires task-training alignment that naturalistic pre-training does not provide. Our contribution is the geometric anatomy of emergence and its boundary conditions, not a prediction tool.

The objective of Vertical Federated Learning (VFL) is to collectively train a model using features available on different devices while sharing the same users. This paper focuses on the saddle point reformulation of the VFL problem via the classical Lagrangian function. We first demonstrate how this formulation can be solved using deterministic methods. More importantly, we explore various stochastic modifications to adapt to practical scenarios, such as employing compression techniques for efficient information transmission, enabling partial participation for asynchronous communication, and utilizing coordinate selection for faster local computation. We show that the saddle point reformulation plays a key role and opens up possibilities to use mentioned extension that seem to be impossible in the standard minimization formulation. Convergence estimates are provided for each algorithm, demonstrating their effectiveness in addressing the VFL problem. Additionally, alternative reformulations are investigated, and numerical experiments are conducted to validate performance and effectiveness of the proposed approach.

After all these years and all these other shared memory programming frameworks, OpenMP is still the most popular one. However, its greater levels of non-deterministic execution makes debugging and testing more challenging. The ability to record and deterministically replay the program execution is key to address this challenge. However, scalably replaying OpenMP programs is still an unresolved problem. In this paper, we propose two novel techniques that use Distributed Clock (DC) and Distributed Epoch (DE) recording schemes to eliminate excessive thread synchronization for OpenMP record and replay. Our evaluation on representative HPC applications with ReOMP, which we used to realize DC and DE recording, shows that our approach is 2-5x more efficient than traditional approaches that synchronize on every shared-memory access. Furthermore, we demonstrate that our approach can be easily combined with MPI-level replay tools to replay non-trivial MPI+OpenMP applications. We achieve this by integrating \toolname into ReMPI, an existing scalable MPI record-and-replay tool, with only a small MPI-scale-independent runtime overhead.

Emerging analog computing substrates, such as oscillator-based Ising machines, offer rapid convergence times for combinatorial optimization but often suffer from limited scalability due to physical implementation constraints. To tackle real-world problems involving thousands of variables, problem decomposition is required; however, performing this step on standard CPUs introduces significant latency, preventing the high-speed solver from operating at full capacity. This work presents a heterogeneous system that offloads the decomposition workload to an FPGA, tightly integrated with a custom 28nm Ising solver. By migrating the decomposition logic to reconfigurable hardware and utilizing parallel processing elements, the system minimizes the communication latency typically associated with host-device interactions. Our evaluation demonstrates that this co-design approach effectively bridges the speed gap between digital preprocessing and analog solving, achieving nearly 2$\times$ speedup and an energy efficiency improvement of over two orders of magnitude compared to optimized software baselines running on modern CPUs.

Flow and diffusion models are typically pre-trained on limited available data (e.g., molecular samples), covering only a fraction of the valid design space (e.g., the full molecular space). As a consequence, they tend to generate samples from only a narrow portion of the feasible domain. This is a fundamental limitation for scientific discovery applications, where one typically aims to sample valid designs beyond the available data distribution. To this end, we address the challenge of leveraging access to a verifier (e.g., an atomic bonds checker), to adapt a pre-trained flow model so that its induced density expands beyond regions of high data availability, while preserving samples validity. We introduce formal notions of strong and weak verifiers and propose algorithmic frameworks for global and local flow expansion via probability-space optimization. Then, we present Flow Expander (FE), a scalable mirror descent scheme that provably tackles both problems by verifier-constrained entropy maximization over the flow process noised state space. Next, we provide a thorough theoretical analysis of the proposed method, and state convergence guarantees under both idealized and general assumptions. Ultimately, we empirically evaluate our method on both illustrative, yet visually interpretable settings, and on a molecular design task showcasing the ability of FE to expand a pre-trained flow model increasing conformer diversity while preserving validity.

Large language models (LLMs) can translate natural language into optimization code, but silent failures pose a critical risk: code that executes and returns solver-feasible solutions may encode semantically incorrect formulations, creating a feasibility-correctness gap of up to 90 percentage points on compositional problems. We introduce ReLoop, addressing silent failures from two complementary directions. Structured generation decomposes code production into a four-stage reasoning chain (understand, formalize, synthesize, verify) that mirrors expert modeling practice, with explicit variable-type reasoning and self-verification to prevent formulation errors at their source. Behavioral verification detects errors that survive generation by testing whether the formulation responds correctly to solver-based parameter perturbation, without requiring ground truth -- an external semantic signal that bypasses the self-consistency problem inherent in LLM-based code review. The two mechanisms are complementary: structured generation dominates on complex compositional problems, while behavioral verification becomes the largest single contributor on problems with localized formulation defects. Together with execution recovery via IIS-enhanced diagnostics, ReLoop raises correctness from 22.6% to 31.1% and execution from 72.1% to 100.0% on the strongest model, with consistent gains across five models spanning three paradigms (foundation, SFT, RL) and three benchmarks. We additionally release RetailOpt-190, 190 compositional retail optimization scenarios targeting the multi-constraint interactions where LLMs most frequently fail.

We study a type of Multi-Armed Bandit (MAB) problems in which arms with a Gaussian reward feedback are clustered. Such an arm setting finds applications in many real-world problems, for example, mmWave communications and portfolio management with risky assets, as a result of the universality of the Gaussian distribution. Based on the Thompson Sampling algorithm with Gaussian prior (TSG) algorithm for the selection of the optimal arm, we propose our Thompson Sampling with Clustered arms under Gaussian prior (TSCG) specific to the 2-level hierarchical structure. We prove that by utilizing the 2-level structure, we can achieve a lower regret bound than we do with ordinary TSG. In addition, when the reward is Unimodal, we can reach an even lower bound on the regret by our Unimodal Thompson Sampling algorithm with Clustered Arms under Gaussian prior (UTSCG). Each of our proposed algorithms are accompanied by theoretical evaluation of the upper regret bound, and our numerical experiments confirm the advantage of our proposed algorithms.

Inspired by non-equilibrium thermodynamics, diffusion models have achieved state-of-the-art performance in generative modeling. However, their iterative sampling nature results in high inference latency. While recent distillation techniques accelerate sampling, they discard intermediate trajectory steps. This sparse supervision leads to a loss of structural information and introduces significant discretization errors. To mitigate this, we propose B-DENSE, a novel framework that leverages multi-branch trajectory alignment. We modify the student architecture to output $K$-fold expanded channels, where each subset corresponds to a specific branch representing a discrete intermediate step in the teacher's trajectory. By training these branches to simultaneously map to the entire sequence of the teacher's target timesteps, we enforce dense intermediate trajectory alignment. Consequently, the student model learns to navigate the solution space from the earliest stages of training, demonstrating superior image generation quality compared to baseline distillation frameworks.

Dataset documentation is widely recognized as essential for the responsible development of automated systems. Despite growing efforts to support documentation through different kinds of artifacts, little is known about the motivations shaping documentation tool design or the factors hindering their adoption. We present a systematic review supported by mixed-methods analysis of 59 dataset documentation publications to examine the motivations behind building documentation tools, how authors conceptualize documentation practices, and how these tools connect to existing systems, regulations, and cultural norms. Our analysis shows four persistent patterns in dataset documentation conceptualization that potentially impede adoption and standardization: unclear operationalizations of documentation's value, decontextualized designs, unaddressed labor demands, and a tendency to treat integration as future work. Building on these findings, we propose a shift in Responsible AI tool design toward institutional rather than individual solutions, and outline actions the HCI community can take to enable sustainable documentation practices.

The rapid expansion of renewable energy, particularly wind and solar power, has made reliable forecasting critical for power system operations. While recent deep learning models have achieved strong average accuracy, the increasing frequency and intensity of climate-driven extreme weather events pose severe threats to grid stability and operational security. Consequently, developing robust forecasting models that can withstand volatile conditions has become a paramount challenge. In this paper, we present R$^2$Energy, a large-scale benchmark for NWP-assisted renewable energy forecasting. It comprises over 10.7 million high-fidelity hourly records from 902 wind and solar stations across four provinces in China, providing the diverse meteorological conditions necessary to capture the wide-ranging variability of renewable generation. We further establish a standardized, leakage-free forecasting paradigm that grants all models identical access to future Numerical Weather Prediction (NWP) signals, enabling fair and reproducible comparison across state-of-the-art representative forecasting architectures. Beyond aggregate accuracy, we incorporate regime-wise evaluation with expert-aligned extreme weather annotations, uncovering a critical ``robustness gap'' typically obscured by average metrics. This gap reveals a stark robustness-complexity trade-off: under extreme conditions, a model's reliability is driven by its meteorological integration strategy rather than its architectural complexity. R$^2$Energy provides a principled foundation for evaluating and developing forecasting models for safety-critical power system applications.

High-speed optical-resolution photoacoustic microscopy (OR-PAM) with bidirectional raster scanning doubles imaging speed but introduces coupled domain shift and geometric misalignment between forward and backward scan lines. Existing registration methods, constrained by brightness constancy assumptions, achieve limited alignment quality, while recent generative approaches address domain shift through complex architectures that lack temporal awareness across frames. We propose GPEReg-Net, a scene-appearance disentanglement framework that separates domain-invariant scene features from domain-specific appearance codes via Adaptive Instance Normalization (AdaIN), enabling direct image-to-image registration without explicit deformation field estimation. To exploit temporal structure in sequential acquisitions, we introduce a Global Position Encoding (GPE) module that combines learnable position embeddings with sinusoidal encoding and cross-frame attention, allowing the network to leverage context from neighboring frames for improved temporal coherence. On the OR-PAM-Reg-4K benchmark (432 test samples), GPEReg-Net achieves NCC of 0.953, SSIM of 0.932, and PSNR of 34.49dB, surpassing the state-of-the-art by 3.8% in SSIM and 1.99dB in PSNR while maintaining competitive NCC. Code is available at https://github.com/JiahaoQin/GPEReg-Net.

Document understanding in real-world applications often requires processing heterogeneous, multi-page document packets containing multiple documents stitched together. Despite recent advances in visual document understanding, the fundamental task of document packet splitting, which involves separating a document packet into individual units, remains largely unaddressed. We present the first comprehensive benchmark dataset, DocSplit, along with novel evaluation metrics for assessing the document packet splitting capabilities of large language models. DocSplit comprises five datasets of varying complexity, covering diverse document types, layouts, and multimodal settings. We formalize the DocSplit task, which requires models to identify document boundaries, classify document types, and maintain correct page ordering within a document packet. The benchmark addresses real-world challenges, including out-of-order pages, interleaved documents, and documents lacking clear demarcations. We conduct extensive experiments evaluating multimodal LLMs on our datasets, revealing significant performance gaps in current models' ability to handle complex document splitting tasks. The DocSplit benchmark datasets and proposed novel evaluation metrics provide a systematic framework for advancing document understanding capabilities essential for legal, financial, healthcare, and other document-intensive domains. We release the datasets to facilitate future research in document packet processing.

Complex change is often described as "evolutionary" in economics, policy, and technology, yet most system dynamics models remain constrained to fixed state spaces and equilibrium-seeking behavior. This paper argues that evolutionary dynamics should be treated as a core system-thinking problem rather than as a biological metaphor. We introduce Stability-Driven Assembly (SDA) as a minimal, non-equilibrium framework in which stochastic interactions combined with differential persistence generate endogenous selection without genes, replication, or predefined fitness functions. In SDA, longer-lived patterns accumulate in the population, biasing future interactions and creating feedback between population composition and system dynamics. This feedback yields fitness-proportional sampling as an emergent property, realizing a natural genetic algorithm driven solely by stability. Using SDA, we demonstrate why equilibrium-constrained models, even when simulated numerically, cannot exhibit open-ended evolution: evolutionary systems require population-dependent, non-stationary dynamics in which structure and dynamics co-evolve. We conclude by discussing implications for system dynamics, economics, and policy modeling, and outline how agent-based and AI-enabled approaches may support evolutionary models capable of sustained novelty and structural emergence.

A P300 ERP-based Brain-Computer Interface (BCI) speller is an assistive communication tool. It searches for the P300 event-related potential (ERP) elicited by target stimuli, distinguishing it from the neural responses to non-target stimuli embedded in electroencephalogram (EEG) signals. Conventional methods require a lengthy calibration procedure to construct the binary classifier, which reduced overall efficiency. Thus, we proposed a unified framework with minimum calibration effort such that, given a small amount of labeled calibration data, we employed an adaptive semi-supervised EM-GMM algorithm to update the binary classifier. We evaluated our method based on character-level prediction accuracy, information transfer rate (ITR), and BCI utility. We applied calibration on training data and reported results on testing data. Our results indicate that, out of 15 participants, 9 participants exceed the minimum character-level accuracy of 0.7 using either on our adaptive method or the benchmark, and 7 out of these 9 participants showed that our adaptive method performed better than the benchmark. The proposed semi-supervised learning framework provides a practical and efficient alternative to improve the overall spelling efficiency in the real-time BCI speller system, particularly in contexts with limited labeled data.

Reliable spacecraft attitude control depends on accurate prediction of attitude dynamics, particularly when model-based strategies such as Model Predictive Control (MPC) are employed, where performance is limited by the quality of the internal system model. For spacecraft with complex dynamics, obtaining accurate physics-based models can be difficult, time-consuming, or computationally heavy. Learning-based system identification presents a compelling alternative; however, models trained exclusively on data frequently exhibit fragile stability properties and limited extrapolation capability. This work explores Physics-Informed Neural Networks (PINNs) for modeling spacecraft attitude dynamics and contrasts it with a conventional data-driven approach. A comprehensive dataset is generated using high-fidelity numerical simulations, and two learning methodologies are investigated: a purely data-driven pipeline and a physics-regularized approach that incorporates prior knowledge into the optimization process. The results indicate that embedding physical constraints during training leads to substantial improvements in predictive reliability, achieving a 68.17% decrease in mean relative error relative. When deployed within an MPC architecture, the physics-informed models yield superior closed-loop tracking performance and improved robustness to uncertainty. Furthermore, a hybrid control formulation that merges the learned nonlinear dynamics with a nominal linear model enables consistent steady-state convergence and significantly faster response, reducing settling times by 61.52%-76.42% under measurement noise and reaction wheel friction.

We develop a general framework to discover scientific algorithms and apply it to three problems in computational cosmology. Our code, MadEvolve, is similar to Google's AlphaEvolve, but places a stronger emphasis on free parameters and their optimization. Our code starts with a baseline human algorithm implementation, and then optimizes its performance metrics by making iterative changes to its code. As a further convenient feature, MadEvolve automatically generates a report that compares the input algorithm with the evolved algorithm, describes the algorithmic innovations and lists the free parameters and their function. Our code supports both auto-differentiable, gradient-based parameter optimization and gradient-free optimization methods. We apply MadEvolve to the reconstruction of cosmological initial conditions, 21cm foreground contamination reconstruction and effective baryonic physics in N-body simulations. In all cases, we find substantial improvements over the base algorithm. We make MadEvolve and our three tasks publicly available at madevolve.org.

We present a simple experiment that exposes a fundamental limitation in vision-language models (VLMs): the inability to accurately localize filled cells in binary grids when those cells lack textual identity. We generate fifteen 15x15 grids with varying density (10.7%-41.8% filled cells) and render each as two image types -- text symbols (. and #) and filled squares without gridlines -- then ask three frontier VLMs (Claude Opus, ChatGPT 5.2, and Gemini 3 Thinking) to transcribe them. In the text-symbol condition, Claude and ChatGPT achieve approximately 91% cell accuracy and 84% F1, while Gemini achieves 84% accuracy and 63% F1. In the filled-squares condition, all three models collapse to 60-73% accuracy and 29-39% F1. Critically, all conditions pass through the same visual encoder -- the text symbols are images, not tokenized text. The text-vs-squares F1 gap ranges from 34 to 54 points across models, demonstrating that VLMs behave as if they possess a high-fidelity text-recognition pathway for spatial reasoning that dramatically outperforms their native visual pathway. Each model exhibits a distinct failure mode in the squares condition -- systematic under-counting (Claude), massive over-counting (ChatGPT), and template hallucination (Gemini) -- but all share the same underlying deficit: severely degraded spatial localization for non-textual visual elements.

Model Context Protocols (MCPs) provide a unified platform for agent systems to discover, select, and orchestrate tools across heterogeneous execution environments. As MCP-based systems scale to incorporate larger tool catalogs and multiple concurrently connected MCP servers, traditional tool-by-tool invocation increases coordination overhead, fragments state management, and limits support for wide-context operations. To address these scalability challenges, recent MCP designs have incorporated code execution as a first-class capability, an approach called Code Execution MCP (CE-MCP). This enables agents to consolidate complex workflows, such as SQL querying, file analysis, and multi-step data transformations, into a single program that executes within an isolated runtime environment. In this work, we formalize the architectural distinction between context-coupled (traditional) and context-decoupled (CE-MCP) models, analyzing their fundamental scalability trade-offs. Using the MCP-Bench framework across 10 representative servers, we empirically evaluate task behavior, tool utilization patterns, execution latency, and protocol efficiency as the scale of connected MCP servers and available tools increases, demonstrating that while CE-MCP significantly reduces token usage and execution latency, it introduces a vastly expanded attack surface. We address this security gap by applying the MAESTRO framework, identifying sixteen attack classes across five execution phases-including specific code execution threats such as exception-mediated code injection and unsafe capability synthesis. We validate these vulnerabilities through adversarial scenarios across multiple LLMs and propose a layered defense architecture comprising containerized sandboxing and semantic gating. Our findings provide a rigorous roadmap for balancing scalability and security in production-ready executable agent workflows.

Generative large vision-language models (LVLMs) have recently achieved impressive performance gains, and their user base is growing rapidly. However, the security of LVLMs, in particular in a long-context multi-turn setting, is largely underexplored. In this paper, we consider the realistic scenario in which an attacker uploads a manipulated image to the web/social media. A benign user downloads this image and uses it as input to the LVLM. Our novel stealthy Visual Memory Injection (VMI) attack is designed such that on normal prompts the LVLM exhibits nominal behavior, but once the user gives a triggering prompt, the LVLM outputs a specific prescribed target message to manipulate the user, e.g. for adversarial marketing or political persuasion. Compared to previous work that focused on single-turn attacks, VMI is effective even after a long multi-turn conversation with the user. We demonstrate our attack on several recent open-weight LVLMs. This article thereby shows that large-scale manipulation of users is feasible with perturbed images in multi-turn conversation settings, calling for better robustness of LVLMs against these attacks. We release the source code at https://github.com/chs20/visual-memory-injection

Object detection has compelling applications over a range of domains, including human-computer interfaces, security and video surveillance, navigation and road traffic monitoring, transportation systems, industrial automation healthcare, the world of Augmented Reality (AR) and Virtual Reality (VR), environment monitoring and activity identification. Applications of real time object detection in all these areas provide dynamic analysis of the visual information that helps in immediate decision making. Furthermore, advanced deep learning algorithms leverage the progress in the field of object detection providing more accurate and efficient solutions. There are some outstanding deep learning algorithms for object detection which includes, Faster R CNN(Region-based Convolutional Neural Network),Mask R-CNN, Cascade R-CNN, YOLO (You Only Look Once), SSD (Single Shot Multibox Detector), RetinaNet etc. This article goes into great detail on how deep learning algorithms are used to enhance real time object recognition. It provides information on the different object detection models available, open benchmark datasets, and studies on the use of object detection models in a range of applications. Additionally, controlled studies are provided to compare various strategies and produce some illuminating findings. Last but not least, a number of encouraging challenges and approaches are offered as suggestions for further investigation in both relevant deep learning approaches and object recognition.

Stochastic-gradient MCMC methods enable scalable Bayesian posterior sampling but often suffer from sensitivity to minibatch size and gradient noise. To address this, we propose Stochastic Gradient Lattice Random Walk (SGLRW), an extension of the Lattice Random Walk discretization. Unlike conventional Stochastic Gradient Langevin Dynamics (SGLD), SGLRW introduces stochastic noise only through the off-diagonal elements of the update covariance; this yields greater robustness to minibatch size while retaining asymptotic correctness. Furthermore, as comparison we analyze a natural analogue of SGLD utilizing gradient clipping. Experimental validation on Bayesian regression and classification demonstrates that SGLRW remains stable in regimes where SGLD fails, including in the presence of heavy-tailed gradient noise, and matches or improves predictive performance.

Recent advances in photonic inverse design have demonstrated the ability to automatically synthesize compact, high-performance photonic components that surpass conventional, hand-designed structures, offering a promising path toward scalable and functionality-rich photonic hardware. However, the practical deployment of inverse-designed PICs is bottlenecked by manufacturability: their irregular, subwavelength geometries are highly sensitive to fabrication variations, leading to large performance degradation, low yield, and a persistent gap between simulated optimality and fabricated performance. Unlike electronics, photonics lacks a systematic, flexible mask optimization flow. Fabrication deviations in photonic components cause large optical response drift and compounding error in cascaded circuits, while calibrating fabrication models remains costly and expertise-heavy, often requiring repeated fabrication cycles that are inaccessible to most designers. To bridge this gap, we introduce PRISM, a photonics-informed inverse lithography workflow that makes photonic mask optimization data-efficient, reliable, and optics-informed. PRISM (i) synthesizes compact, informative calibration patterns to minimize required fabrication data, (ii) trains a physics-grounded differentiable fabrication model, enabling gradient-based optimization, and (iii) performs photonics-informed inverse mask optimization that prioritizes performance-critical features beyond geometry matching. Across multiple inverse-designed components with both electron-beam lithography and deep ultra-violet photolithography processes, PRISM significantly boosts post-fabrication performance and yield while reducing calibration area and turnaround time, enabling and democratizing manufacturable and high-yield inverse-designed photonic hardware at scale.

Density Functional Theory (DFT) is widely used for first-principles simulations in chemistry and materials science, but its computational cost remains a key limitation for large systems. Motivated by recent advances in ML-based exchange-correlation (XC) functionals, this paper introduces a differentiable framework that integrates machine learning models into density functional theory (DFT) for solids and other periodic systems. The framework defines a clean API for neural network models that can act as drop in replacements for conventional exchange-correlation (XC) functionals and enables gradients to flow through the full self-consistent DFT workflow. The framework is implemented in Python using a PyTorch backend, making it fully differentiable and easy to use with standard deep learning tools. We integrate the implementation with the DeepChem library to promote the reuse of established models and to lower the barrier for experimentation. In initial benchmarks against established electronic structure packages (GPAW and PySCF), our models achieve relative errors on the order of 5-10%.

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.

State-of-the-art Vision-Language-Action (VLA) models excel at semantic generalization but struggle to generalize to unseen physical motions in novel environments. We introduce DreamZero, a World Action Model (WAM) built upon a pretrained video diffusion backbone. Unlike VLAs, WAMs learn physical dynamics by predicting future world states and actions, using video as a dense representation of how the world evolves. By jointly modeling video and action, DreamZero learns diverse skills effectively from heterogeneous robot data without relying on repetitive demonstrations. This results in over 2x improvement in generalization to new tasks and environments compared to state-of-the-art VLAs in real robot experiments. Crucially, through model and system optimizations, we enable a 14B autoregressive video diffusion model to perform real-time closed-loop control at 7Hz. Finally, we demonstrate two forms of cross-embodiment transfer: video-only demonstrations from other robots or humans yield a relative improvement of over 42% on unseen task performance with just 10-20 minutes of data. More surprisingly, DreamZero enables few-shot embodiment adaptation, transferring to a new embodiment with only 30 minutes of play data while retaining zero-shot generalization.

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 suffer from late-stage performance collapse, leading to degraded reasoning quality and unstable training. Our analysis shows that the magnitude of token-wise policy gradients in RL is negatively correlated with token probability and local policy entropy. We find that training instability can be caused 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. To mitigate this instability, we design S2T (silencing spurious tokens) mechanism to efficiently identify spurious tokens through characteristic signals with low probability, low entropy, and positive advantage, and then to suppress their gradient perturbations during optimization. Incorporating this mechanism into a group-based objective, we propose Spurious-Token-Aware Policy Optimization (STAPO), which promotes stable and effective large-scale model refinement. 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\% ($ρ_{\mathrm{T}}$=1.0, top-p=1.0) and 3.69\% ($ρ_{\mathrm{T}}$=0.7, top-p=0.9) over GRPO, 20-Entropy and JustRL.

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 JavaScript. 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.

The growing complexity of healthcare systems requires advanced computational models for real-time monitoring, secure data exchange, and intelligent decision-making. Digital Twins (DTs) provide virtual representations of physical healthcare entities, enabling continuous patient monitoring and personalized care. However, classical DT frameworks face limitations in scalability, computational efficiency, and security. Recent studies have introduced Quantum Digital Twins (QDTs) to enhance performance through quantum computing, addressing challenges such as quantum-resistant security and efficient task offloading in healthcare environments. Despite these advances, most existing QDT models remain constrained by fundamental challenges related to quantum hardware limitations, hybrid classical-quantum system integration, cloud-based quantum access, scalability, and clinical trust. This paper provides a comprehensive review of QDTs for healthcare, with a particular focus on identifying and analyzing the key challenges that currently hinder their real-world adoption. Furthermore, it outlines critical research directions and enabling strategies aimed at advancing the development of secure, reliable, and clinically viable quantum digital twin systems for next-generation healthcare applications.

This paper addresses graph learning in Gaussian Graphical Models (GGMs). In this context, data matrices often come with auxiliary metadata (e.g., textual descriptions associated with each node) that is usually ignored in traditional graph estimation processes. To fill this gap, we propose a graph learning approach based on Laplacian-constrained GGMs that jointly leverages the node signals and such metadata. The resulting formulation yields an optimization problem, for which we develop an efficient majorization-minimization (MM) algorithm with closed-form updates at each iteration. Experimental results on a real-world financial dataset demonstrate that the proposed method significantly improves graph clustering performance compared to state-of-the-art approaches that use either signals or metadata alone, thus illustrating the interest of fusing both sources of information.

Can the privacy vulnerability of individual data points be assessed without retraining models or explicitly simulating attacks? We answer affirmatively by showing that exposure to membership inference attack (MIA) is fundamentally governed by a data point's influence on the learned model. We formalize this in the linear setting by establishing a theoretical correspondence between individual MIA risk and the leverage score, identifying it as a principled metric for vulnerability. This characterization explains how data-dependent sensitivity translates into exposure, without the computational burden of training shadow models. Building on this, we propose a computationally efficient generalization of the leverage score for deep learning. Empirical evaluations confirm a strong correlation between the proposed score and MIA success, validating this metric as a practical surrogate for individual privacy risk assessment.

While leveraging LLMs to automatically generate SystemVerilog assertions (SVAs) from natural language specifications holds great potential, existing techniques face a key challenge: LLMs often lack sufficient understanding of IC design, leading to poor assertion quality in a single pass. Therefore, verifying whether the generated assertions effectively cover the functional specifications and designing feedback mechanisms based on this coverage remain significant hurdles. To address these limitations, this paper introduces CoverAssert, a novel iterative framework for optimizing SVA generation with LLMs. The core contribution is a lightweight mechanism for matching generated assertions with specific functional descriptions in the specifications. CoverAssert achieves this by clustering the joint representations of semantic features of LLM-generated assertions and structural features extracted from abstract syntax trees (ASTs) about signals related to assertions, and then mapping them back to the specifications to analyze functional coverage quality. Leveraging this capability, CoverAssert constructs a feedback loop based on functional coverage to guide LLMs in prioritizing uncovered functional points, thereby iteratively improving assertion quality. Experimental evaluations on four open-source designs demonstrate that integrating CoverAssert with state-of-the-art generators, AssertLLM and Spec2Assertion, achieves average improvements of 9.57 % in branch coverage, 9.64 % in statement coverage, and 15.69 % in toggle coverage.

Benchmarking spatial reasoning in multimodal large language models (MLLMs) has attracted growing interest in computer vision due to its importance for embodied AI and other agentic systems that require precise interaction with the physical world. However, spatial reasoning on Earth imagery has lagged behind, as it uniquely involves grounding objects in georeferenced images and quantitatively reasoning about distances, directions, and topological relations using both visual cues and vector geometry coordinates (e.g., 2D bounding boxes, polylines, and polygons). Existing benchmarks for Earth imagery primarily focus on 2D spatial grounding, image captioning, and coarse spatial relations (e.g., simple directional or proximity cues). They lack support for quantitative direction and distance reasoning, systematic topological relations, and complex object geometries beyond bounding boxes. To fill this gap, we propose \textbf{EarthSpatialBench}, a comprehensive benchmark for evaluating spatial reasoning in MLLMs on Earth imagery. The benchmark contains over 325K question-answer pairs spanning: (1) qualitative and quantitative reasoning about spatial distance and direction; (2) systematic topological relations; (3) single-object queries, object-pair queries, and compositional aggregate group queries; and (4) object references expressed via textual descriptions, visual overlays, and explicit geometry coordinates, including 2D bounding boxes, polylines, and polygons. We conducted extensive experiments on both open-source and proprietary models to identify limitations in the spatial reasoning of MLLMs.

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: WEB, containing 377 web-sourced usage examples from Urban Dictionary, and 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 WEB versus 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.

In high-performance computing (HPC) environments, particularly in synchrotron radiation facilities, vast amounts of X-ray images are generated. Processing large-scale X-ray Computed Tomography (X-CT) datasets presents significant computational and storage challenges due to their high dimensionality and data volume. Traditional approaches often require extensive storage capacity and high transmission bandwidth, limiting real-time processing capabilities and workflow efficiency. To address these constraints, we introduce a region-of-interest (ROI)-driven extraction framework (ROIX-Comp) that intelligently compresses X-CT data by identifying and retaining only essential features. Our work reduces data volume while preserving critical information for downstream processing tasks. At pre-processing stage, we utilize error-bounded quantization to reduce the amount of data to be processed and therefore improve computational efficiencies. At the compression stage, our methodology combines object extraction with multiple state-of-the-art lossless and lossy compressors, resulting in significantly improved compression ratios. We evaluated this framework against seven X-CT datasets and observed a relative compression ratio improvement of 12.34x compared to the standard compression.

Knowledge-based Visual Question Answering (KB-VQA) requires models to answer questions by integrating visual information with external knowledge. However, retrieved knowledge is often noisy, partially irrelevant, or misaligned with the visual content, while internal model knowledge is difficult to control and interpret. Naive aggregation of these sources limits reasoning effectiveness and reduces answer accuracy. To address this, we propose MaS-VQA, a selection-driven framework that tightly couples explicit knowledge filtering with implicit knowledge reasoning. MaS-VQA first retrieves candidate passages and applies a Mask-and-Select mechanism to jointly prune irrelevant image regions and weakly relevant knowledge fragments, producing compact, high-signal multimodal knowledge . This filtered knowledge then guides the activation of internal knowledge in a constrained semantic space, enabling complementary co-modeling of explicit and implicit knowledge for robust answer prediction. Experiments on Encyclopedic-VQA and InfoSeek demonstrate consistent performance gains across multiple MLLM backbones, and ablations verify that the selection mechanism effectively reduces noise and enhances knowledge utilization.

Large Language Models (LLMs) are increasingly used by undergraduate students as on-demand tutors, yet their reliability on circuit- and diagram-based digital logic problems remains unclear. We present a human- AI study evaluating three widely used LLMs (GPT, Gemini, and Claude) on 10 undergraduate-level digital logic questions spanning non-standard counters, JK-based state transitions, timing diagrams, frequency division, and finite-state machines. Twenty-four students performed pairwise model comparisons, providing per-question judgments on (i) preferred model, (ii) perceived correctness, (iii) consistency, (iv) verbosity, and (v) confidence, along with global ratings of overall model quality, satisfaction across multiple dimensions (e.g., accuracy and clarity), and perceived mental effort required to verify answers. To benchmark technical validity, we applied an independent judge-based evaluation against official solutions for all ten questions, using strict correctness criteria. Results reveal a consistent gap between perceived helpfulness and formal correctness: for the most sequentially demanding problems (Q1- Q7), none of the evaluated LLMs matched the official answers, despite producing confident, well-structured explanations that students often rated favorably. Error analysis indicates that models frequently default to canonical textbook templates (e.g., standard ripple counters) and struggle to translate circuit structure into exact state evolution and timing behavior. These findings suggest that, without verification scaffolds, LLMs may be unreliable for core digital logic topics and can inadvertently reinforce misconceptions in undergraduate instruction.

We show that deliberately breaking detailed balance in generative diffusion processes can accelerate the reverse process without changing the stationary distribution. Considering the Ornstein--Uhlenbeck process, we decompose the dynamics into a symmetric component and a non-reversible anti-symmetric component that generates rotational probability currents. We then construct an exponentially optimal non-reversible perturbation that improves the long-time relaxation rate while preserving the stationary target. We analyze how such non-reversible control reshapes the macroscopic dynamical regimes of the phase transitions recently identified in generative diffusion models. We derive a general criterion for the speciation time and show that suitable non-reversible perturbations can accelerate speciation. In contrast, the collapse transition is governed by a trace-controlled phase-space contraction mechanism that is fixed by the symmetric component, and the corresponding collapse time remains unchanged under anti-symmetric perturbations. Numerical experiments on Gaussian mixture models support these findings.

Foundation models (FMs) are rapidly reshaping medical imaging, shifting the field from narrowly trained, task-specific networks toward large, general-purpose models that can be adapted across modalities, anatomies, and clinical tasks. In this review, we synthesize the emerging landscape of medical imaging FMs along three major axes: principles of FM design, applications of FMs, and forward-looking challenges and opportunities. Taken together, this review provides a technically grounded, clinically aware, and future-facing roadmap for developing FMs that are not only powerful and versatile but also trustworthy and ready for responsible translation into clinical practice.

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.

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.

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.

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.

The energy and latency of an accelerator running a deep neural network (DNN) depend on how the computation and data movement are scheduled in the accelerator (i.e., mapping). Optimizing mappings is essential to evaluating and designing accelerators. However, the space of mappings is large, and prior works can not guarantee finding optimal mappings because they use heuristics or metaheuristics to narrow down the space. These limitations preclude proper hardware evaluation, since designers can not tell whether performance differences are due to changes in hardware or suboptimal mapping. To address this challenge, we propose the Turbo-Charged Mapper (TCM), a fast mapper that is guaranteed to find optimal mappings. The key to our approach is that we define a new concept in mapping, called dataplacement, which, like the prior concept of dataflow, allows for clear analysis and comparison of mappings. Through it, we identify multiple opportunities to prune redundant and suboptimal mappings, reducing search space by up to 32 orders of magnitude. Leveraging these insights, TCM can perform full mapspace searches, making it the first mapper that can find optimal mappings in feasible runtime. Compared to prior mappers, we show that TCM can find optimal mappings quickly (less than a minute), while prior works can not find optimal mappings (energy-delay-product $21\%$ higher than optimal) even when given $1000\times$ the runtime ($>10$ hours).

The latency and energy of tensor algebra accelerators depend on how data movement and operations are scheduled (i.e., mapped) onto accelerators, so determining the potential of an accelerator architecture requires both a performance model and a mapper to search for the optimal mapping. A key optimization that the mapper must explore is fusion, meaning holding data on-chip between computation steps, which has been shown to reduce energy and latency by reducing DRAM accesses. However, prior mappers cannot find optimal mappings with fusion (i.e., fused mappings) in a feasible runtime because the number of fused mappings to search increases exponentially with the number of workload computation steps. In this paper, we introduce the Fast and Fusiest Mapper (FFM), the first mapper to quickly find optimal mappings in a comprehensive fused mapspace for tensor algebra workloads. FFM shrinks the search space by pruning subsets of mappings (i.e., partial mappings) that are shown to never be a part of optimal mappings, quickly eliminating all suboptimal mappings with those partial mappings as subsets. Then FFM joins partial mappings to construct optimal fused mappings. We evaluate FFM and show that, although the mapspace size grows exponentially with the number of computation steps, FFM's runtime scales approximately linearly. FFM is orders of magnitude faster ($>1000\times$) than prior state-of-the-art approaches at finding optimal mappings for Transformers.

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.

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. Videos are available at https://bigpicturepolicies.github.io/

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.

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

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.

Neural network (NN) accelerators with multi-chip-module (MCM) architectures enable integration of massive computation capability; however, they face challenges of computing resource underutilization and off-chip communication overheads. Traditional parallelization schemes for NN inference on MCM architectures, such as intra-layer parallelism and inter-layer pipelining, show incompetency in breaking through both challenges, limiting the scalability of MCM architectures. We observed that existing works typically deploy layers separately rather than considering them jointly. This underexploited dimension leads to compromises between system computation and communication, thus hindering optimal utilization, especially as hardware/software scale. To address this limitation, we propose Scope, a merged pipeline framework incorporating this overlooked multi-layer dimension, thereby achieving improved throughput and scalability by relaxing tradeoffs between computation, communication and memory costs. This new dimension, however, adds to the complexity of design space exploration (DSE). To tackle this, we develop a series of search algorithms that achieves exponential-to-linear complexity reduction, while identifying solutions that rank in the top 0.05% of performance. Experiments show that Scope achieves up to 1.73x throughput improvement while maintaining similar energy consumption for ResNet-152 inference compared to state-of-the-art approaches.

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 the global average of semantic contents stabilizes 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 a stable structure and consensus 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.

We present a tightly integrated and unified near-memory GPU architecture that delivers 6 to 16 times speedup and 6 to 13 times energy savings across Convolutional Neural Networks, Graph Convolutional Networks, Linear Programming, Large Language Models, and Ising workloads compared to MIAOW GPU. The design includes a custom sparsity-aware near-memory circuit providing about 1.5 times energy savings, and a lightweight softmax circuit providing about 1.6 times energy savings. The architecture supports reconfigurable compute up to INT16 with dynamic resolution updates and scales efficiently across problem sizes. ABI-enabled MI300 and Blackwell systems achieve about 4.5 times speedup over baseline MI300 and Blackwell.

Paralinguistic and non-linguistic aspects of speech strongly influence listener impressions. While most research focuses on absolute impression scoring, this study investigates relative voice impression estimation (RIE), a framework for predicting the perceptual difference between two utterances from the same speaker. The estimation target is a low-dimensional vector derived from subjective evaluations, quantifying the perceptual shift of the second utterance relative to the first along an antonymic axis (e.g., ``Dark--Bright''). To isolate expressive and prosodic variation, we used recordings of a professional speaker reading a text in various styles. We compare three modeling approaches: classical acoustic features commonly used for speech emotion recognition, self-supervised speech representations, and multimodal large language models (MLLMs). Our results demonstrate that models using self-supervised representations outperform methods with classical acoustic features, particularly in capturing complex and dynamic impressions (e.g., ``Cold--Warm'') where classical features fail. In contrast, current MLLMs prove unreliable for this fine-grained pairwise task. This study provides the first systematic investigation of RIE and demonstrates the strength of self-supervised speech models in capturing subtle perceptual variations.

Rapidly evolving AI exhibits increasingly strong autonomy and goal-directed capabilities, accompanied by derivative systemic risks that are more unpredictable, difficult to control, and potentially irreversible. However, current AI safety evaluation systems suffer from critical limitations such as restricted risk dimensions and failed frontier risk detection. The lagging safety benchmarks and alignment technologies can hardly address the complex challenges posed by cutting-edge AI models. To bridge this gap, we propose the "ForesightSafety Bench" AI Safety Evaluation Framework, beginning with 7 major Fundamental Safety pillars and progressively extends to advanced Embodied AI Safety, AI4Science Safety, Social and Environmental AI risks, Catastrophic and Existential Risks, as well as 8 critical industrial safety domains, forming a total of 94 refined risk dimensions. To date, the benchmark has accumulated tens of thousands of structured risk data points and assessment results, establishing a widely encompassing, hierarchically clear, and dynamically evolving AI safety evaluation framework. Based on this benchmark, we conduct systematic evaluation and in-depth analysis of over twenty mainstream advanced large models, identifying key risk patterns and their capability boundaries. The safety capability evaluation results reveals the widespread safety vulnerabilities of frontier AI across multiple pillars, particularly focusing on Risky Agentic Autonomy, AI4Science Safety, Embodied AI Safety, Social AI Safety and Catastrophic and Existential Risks. Our benchmark is released at https://github.com/Beijing-AISI/ForesightSafety-Bench. The project website is available at https://foresightsafety-bench.beijing-aisi.ac.cn/.

Combinatorial optimization problems are central to science and engineering and specialized hardware from quantum annealers to classical Ising machines are being actively developed to address them. These systems typically sample from a fixed energy landscape defined by the problem Hamiltonian encoding the discrete optimization problem. The recently introduced Probabilistic Approximate Optimization Algorithm (PAOA) takes a different approach: it treats the optimization landscape itself as variational, iteratively learning circuit parameters from samples. Here, we demonstrate PAOA on a 64$\times$64 perimeter-gated single-photon avalanche diode (pgSPAD) array fabricated in 0.35 $μ$m CMOS, the first realization of the algorithm using intrinsically stochastic nanodevices. Each p-bit exhibits a device-specific, asymmetric (Gompertz-type) activation function due to dark-count variability. Rather than calibrating devices to enforce a uniform symmetric (logistic/tanh) activation, PAOA learns around device variations, absorbing residual activation and other mismatches into the variational parameters. On canonical 26-spin Sherrington-Kirkpatrick instances, PAOA achieves high approximation ratios with $2p$ parameters ($p$ up to 17 layers), and pgSPAD-based inference closely tracks CPU simulations. These results show that variational learning can accommodate the non-idealities inherent to nanoscale devices, suggesting a practical path toward larger-scale, CMOS-compatible probabilistic computers.

We introduce LaySPA, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design. LaySPA addresses two key challenges: LLMs' limited spatial reasoning and the lack of opacity in design decision making. Instead of operating at the pixel level, we reformulate layout design as a policy learning problem over a structured textual spatial environment that explicitly encodes canvas geometry, element attributes, and inter-element relationships. LaySPA produces dual-level outputs comprising interpretable reasoning traces and structured layout specifications, enabling transparent and controllable design decision making. Layout design policy is optimized via a multi-objective spatial critique that decomposes layout quality into geometric validity, relational coherence, and aesthetic consistency, and is trained using relative group optimization to stabilize learning in open-ended design spaces. Experiments demonstrate that LaySPA improves structural validity and visual quality, outperforming larger proprietary LLMs and achieving performance comparable to specialized SOTA layout generators while requiring fewer annotated samples and reduced latency.

In this work, we report implementation and performance evaluation of memristor-driven fundamental logic gates, including NOT, AND, NAND, OR, NOR, and XOR, and novel and optimized design of the sequential logic circuits, such as D flip-flop, T-flip-flop, JK-flip-flop, and SR-flip-flop. The design, implementation, and optimization of these logic circuits were performed in SPECTRE in Cadence Virtuoso and integrated with 90 nm CMOS technology node. Additionally, we discuss an optimized design of memristor-driven logic gates and sequential logic circuits, and draw a comparative analysis with the other reported state-of-the-art work on sequential circuits. Moreover, the utilized memristor framework was experimentally pre-validated with the experimental data of Y2O3-based memristive devices, which shows significantly low values of variability during switching in both device-to-device (D2D) and cycle-to-cycle (C2C) operation. The performance metrics were calculated in terms of area, power, and delay of these sequential circuits and were found to be reduced by more than ~24%, 60%, and 58%, respectively, as compared to the other state-of-the-art work on sequential circuits. Therefore, the implemented memristor-based design significantly improves the performance of various logic designs, which makes it more area and power-efficient and shows the potential of memristor in designing various low-power, low-cost, ultrafast, and compact circuits.

Electromagnetic (EM) world modeling is emerging as a foundational capability for environment-aware and embodiment-enabled wireless systems. However, most existing mmWave sensing solutions are designed for snapshot-based parameter estimation and rely on hardware-intensive architectures, making scalable and persistent world modeling difficult to achieve. This article rethinks mmWave sensing from a system-level perspective and introduces a generative-space framework, in which sensing is realized through controlled traversal of a low-dimensional excitation space spanning frequency, waveform, and physical embodiment. This perspective decouples spatial observability from rigid antenna arrays and transmit-time multiplexing, enabling flexible and scalable sensing-by-design radios. To illustrate the practicality of this framework, we present a representative realization called Multi-RF Chain Frequency-as-Aperture Clip-on Aperture Fabric (MRC-FaA-CAF), where multiple FMCW sources coordinate frequency-selective modules distributed along guided-wave backbones. This architecture enables interference-free excitation, preserves beat-frequency separability, and maintains low calibration overhead. Case studies show that generative-space-driven sensing can achieve update rates comparable to phased arrays while avoiding dense RF replication and the latency penalties of TDM-MIMO systems. Overall, this work positions generative-space-driven sensing as a practical architectural foundation for mmWave systems that move beyond snapshot sensing toward persistent EM world modeling.

Binary neural networks (BNNs) are increasingly deployed in edge computing applications due to their low hardware complexity and high energy efficiency. However, verifying the robustness of BNNs against input perturbations, including adversarial attacks, remains computationally challenging because the underlying decision problem is inherently combinatorial. In this paper, we propose an Ising- and quantum-inspired framework for BNN robustness verification. We show that, for a broad class of BNN architectures, robustness verification can be formulated as a Quadratic Constrained Boolean Optimization (QCBO) problem and subsequently transformed into a Quadratic Unconstrained Boolean Optimization (QUBO) instance amenable to Ising and quantum-inspired solvers. We demonstrate the feasibility of this formulation on binarized MNIST by solving the resulting QUBOs with a free energy machine (FEM) solver and simulated annealing. We also show the deployment of this framework on quantum annealing and digital annealing platforms. Our results highlight the potential of quantum-inspired computing and Ising computing as a pathway toward trustworthy AI systems.

Recently, the superconducting diode effect (SDE), characterized by unequal critical currents in opposite directions, has been observed experimentally and predicted theoretically in models of bulk superconductors and Josephson junctions (JJs). In this work, we construct a Josephson junction using a recently discovered unconventional coplanar magnet, the $p$-wave magnet (PM), with proximity-induced superconductivity, and demonstrate the emergence of a Josephson diode effect (JDE). The barrier region is formed by another unconventional collinear magnet, namely an altermagnet (AM). We illustrate that apart from time-reversal and inversion symmetries, the mirror operation $M_{yz}$ emerges as the key symmetry constraint. Also, unlike earlier models that realize the JDE using unconventional magnets, this setup does not require Rashba spin-orbit coupling (SOC) or different superconductors across the junction. Moreover, we demonstrate that the realization of the JDE in this framework requires only minimal conditions while maintaining high performance. The effect remains robust across a broad parameter regime, and thus making the system particularly promising for applications in quantum circuits and computing technologies.

We present a systematic construction of non-Hermitian tight-binding lattices whose Bloch spectra are $n$th roots of those of Hermitian parent two-dimensional (2D) lattices, namely graphene and the Lieb lattice. The $n$-roots of these models are constructed from connecting loop modules of unidirectional couplings whose geometrical arrangements match that of the corresponding parent system. Their energy spectrum is shown to consist of $n$ rotated and equivalent branches in the complex energy plane, each matching the real spectrum of the parent model when raised to the $n$th power, together with extra zero-energy flat bands (FBs) accounted for by the generalized index theorem. We show how the low-energy Dirac cones of the parent models translate, for an appropriate choice of phase configuration for the couplings of the $n$-root lattices, as what we call an "exceptional horn" appearing at each branch, with the central Dirac point (DP) converted into zero-energy exceptional points (EPs) of order $n$ or higher at high-symmetry momenta. These exceptional horns reflect the behavior of low-lying excitations that scale with momentum as $E\sim\vert \mathbf{q}\vert^{\frac{1}{n}}$, with $n\geq 3$, as opposed to the linear massless modes that characterize a Dirac cone. Moreover, we derive analytic expressions for the associated Landau levels (LLs), whose energies scale with magnetic flux as $E\simφ^{\frac{1}{2n}}$. For the case of the $n$-root Lieb lattice, the zeroth LL is shown to be exceptional. These results are analytically derived for both $n$-root models and numerically demonstrated for certain values of $n$. Finally, we propose a realistic photonic implementation based on coupled ring resonators with a split configuration of optical gain and loss.

p-spin glasses, characterized by frustrated many-body interactions beyond the conventional pairwise case (p>2), are prototypical disordered systems whose ground-state search is NP-hard and computationally prohibitive for large instances. Solving this problem is not only fundamental for understanding high-order disorder, structural glasses, and topological phases, but also central to a wide spectrum of hard combinatorial optimization tasks. Despite decades of progress, there still lacks an efficient and scalable solver for generic large-scale p-spin models. Here we introduce PLANCK, a physics-inspired deep reinforcement learning framework built on hypergraph neural networks. PLANCK directly optimizes arbitrary high-order interactions, and systematically exploits gauge symmetry throughout both training and inference. Trained exclusively on small synthetic instances, PLANCK exhibits strong zero-shot generalization to systems orders of magnitude larger, and consistently outperforms state-of-the-art thermal annealing methods across all tested structural topologies and coupling distributions. Moreover, without any modification, PLANCK achieves near-optimal solutions for a broad class of NP-hard combinatorial problems, including random k-XORSAT, hypergraph max-cut, and conventional max-cut. The presented framework provides a physics-inspired algorithmic paradigm that bridges statistical mechanics and reinforcement learning. The symmetry-aware design not only advances the tractable frontiers of high-order disordered systems, but also opens a promising avenue for machine-learning-based solvers to tackle previously intractable combinatorial optimization challenges.

We study the mean-field dynamics of a system of $N$ interacting bosons starting from an initially condensated state. For a broad class of mean-field Hamiltonians, including models with arbitrary bounded interactions and models with unbounded interaction potentials, we prove that the probability of having $n$ particles outside the condensate decays exponentially in $n$ for any finite evolution time. Our results strengthen previously known bounds that provide only polynomial control on the probability of having $n$ excitations.

Fluid molecular ferroelectrics are a new class of organic materials where ferroelectricity is found in conjunction with 3D fluidity whilst still retaining spontaneous polarization values comparable to their traditional solid state counterparts. One of the major challenges for soft condensed matter physics is predicting whether a fluid molecular material will form ferroelectric phase with nematic or smectic order. Through the synthesis of forty five systematically varied molecules, and by analogy to solid molecular ferroelectrics, is it shown that subtle hydrogen fluorine substitution allows for tuneable syn-parallel pairing motifs resulting in either specific pairings leading too geometrically constrained lamellar order or diversified pairings stabilising nematic ordering. Large-scale, fully atomistic molecular dynamics simulations reveal that smectic ferroelectricity emerges from discrete lateral pairing modes, whereas nematic phases arise from a multiplicity of equivalent polar configurations. Together, these findings establish experimentally validated design principles for fluid molecular ferroelectrics and provide a predictive framework for engineering functional polar fluids.

Particle-stabilised emulsions are a cornerstone of soft matter science due to their broad application and fundamental relevance. Computer simulations provide key insights into the formation and behaviour of these emulsions, yet current methods are limited by the spatiotemporal scales accessible for study. The principal issue is that particles are resolved individually. In this work, an alternative strategy is introduced based on phase-field theory, for which we establish the framework. By evolving continuous fields, large-scale dynamics can be simulated in a computationally efficient manner. Our approach is then applied to model the complex formation of a bicontinuous interfacially jammed emulsion gel (bijel) via solvent-transfer induced phase separation (STrIPS). By resolving the coupled dynamics of liquid phase separation and nanoparticle adsorption, the model allows for the characterisation of the influence of nanoparticles on the morphology. Higher concentrations of nanoparticles are found to reduce the average domain size of STrIPS bijels, in line with previous experimental evidence. The presented phase-field model thus represents a promising approach for the morphological investigation of complex particle-stabilised emulsions.

The quantum symmetric simple exclusion process (QSSEP) is a recent extension of the symmetric simple exclusion process, designed to model quantum coherent fluctuating effects in noisy diffusive systems. It models stochastic nearest-neighbor fermionic hopping on a lattice, possibly driven out-of-equilibrium by boundary processes. We present a direct formulation in the continuum, and establish how this formulation captures the scaling limit of the discrete version. In the continuum, QSSEP emerges as a non-commutative process, driven by free increments, conditioned on the algebra of functions on the ambiant space to encode spatial correlations. We actually develop a more general framework dealing with conditioned orbits with free increments which may find applications beyond the present context. We view this construction as a preliminary step toward formulating a quantum extension of the macroscopic fluctuation theory.

Exciton-phonon interactions govern the optical response of semiconductors, yet disentangling multiple coupling channels in lead halide perovskites remains challenging. We investigate CsPbBr3 microcrystals using photoluminescence, Raman and reflectance spectroscopy at low temperature, revealing the simultaneous presence of high-energy and Rashba excitons, each accompanied by distinct phonon replica series. High-energy exciton replicas are uniquely spaced by approximately 9 meV, whereas Rashba exciton replicas exhibit a characteristic approximately 6 meV spacing, indicating the specificity of the exciton-phonon coupling. Unsupervised machine learning applied to a large low-temperature photoluminescence dataset reveals these replica features are prevalent. With increasing temperature, replica features broaden and merge, evolving into a dominant longitudinal optical phonon coupling regime at room temperature. This work establishes direct spectroscopic evidence for concurrent, exciton-specific phonon coupling within a single material, offering new pathways to engineer light-matter interactions for optoelectronic and phonon-photon-based quantum device applications.

Systems of instantaneously annihilating or coalescing Brownian motions on the line are considered. The extreme points of the set of entrance laws for this process are shown to be Pfaffian point processes at all times and their kernels are identified.

We consider quantum-to-classical mapping for an arbitrary system of interacting spins at finite temperatures. We prove that, in the large-$S$ limit, the asymptotic form of the partition function coincides with that of a classical model for spins of length $S_C=\sqrt{S(S+1)}$. Quantum corrections to the leading term form a series in powers of $1/[S(S+1)]$. This representation provides a rigorous basis for classical modeling of realistic magnetic Hamiltonians. As an application, the classical Monte Carlo simulations are performed to compute transition temperatures for several topical materials with known interaction parameters, including MnF$_2$, MnTe, Rb$_2$MnF$_4$, MnPSe$_3$, FePS$_3$, FePSe$_3$, CoPS$_3$, CrSBr, and CrI$_3$. The resulting transition temperatures show good agreement with experimental data.

We examine the homological structure of the Barabasi-Albert model, focusing on the time evolution of $Δ$-dimensional simplices and topological holes as functions of time $t$ and the attachment parameter $m$ (the number of edges added by each incoming node). Numerical simulations reveal a non-trivial topological transition (TT) in the $(Δ, m)$ plane, marking a change from a topologically trivial regime to non-trivial topology. This transition signals the emergence of topological complexity in the model, where higher-order structures develop self-similarly across scales. Beyond this transition, the network exhibits self-similar topological growth, evidenced by a power-law decay in the increments of $Δ$-simplices with $m$-dependent exponents. An analogous transition occurs in the Betti numbers, which display self-similar behavior near the TT and an arctangent dependence farther from it. Based on simulation data, we propose explicit scaling relations describing the behavior of both $Δ$-simplices and Betti numbers near the TT. Overall, the analysis reveals a rich, gapful topological transition structure, where topological quantities exhibit discrete jumps at the transition point.

Confinement epitaxy beneath graphene stabilizes exotic material phases by restricting vertical growth and altering lateral diffusion, conditions unattainable on bare substrates. However, achieving long-range interfacial order while maintaining high-quality graphene remains a significant challenge. Here, we demonstrate the synthesis of large-area quasi-free-standing monolayer graphene (QFMLG) via the intercalation of a two-dimensional (2D) Sn. While the triangular Sn(1x1) interface exhibits a robust metallic band structure, the decoupled QFMLG maintains charge neutrality, confirmed by photoemission spectroscopy. Using high-resolution Raman spectroscopy and microscopy, we distinguish between direct intercalation and diffusion-driven expansion, identifying the latter as the critical pathway to superior QFMLG crystalline quality. Temperature-dependent analysis reveals dynamical structural coupling between the decoupled QFMLG and the Sn interface, providing a novel degree of freedom for strain engineering. Beyond uncovering the diffusion-driven mechanism, this work establishes metal intercalation as an effective strategy for tailoring durable graphene-metal heterostructures with tunable properties for next-generation quantum materials platforms.

Diversity-Generating Retroelements (DGRs) create rapid, targeted variation within specific genomic regions in phages and bacteria. They operate through stochastic retro-transcription of a template region (TR) into a variable region (VR), which typically encodes ligand-binding proteins. Despite their prevalence, the evolutionary conditions that maintain such hypermutating systems remain unclear. Here we introduce a two-timescale framework separating fast VR diversification from slow TR evolution, allowing the dynamics of DGR-controlled loci to be analytically understood from the TR design point of view. We quantity the fitness gain provided by the diversification mechanism of DGR in the presence of environmental switching with respect to standard mutagenesis. Our framework accounts for observed patterns of DGR activity in human-gut \textit{Bacteroides} and clarifies when constitutive DGR activation is evolutionarily favored.

Topology in quantum systems is typically considered in infinite crystals in one, two, or higher integer dimensions. Here, we show that one can continuously transform a system between a topological phase associated with one dimension and a topological phase associated with two dimensions without closing the energy gap. In this process, the dimension of the system itself changes. Concretely, we investigate a modified version of the Qi-Wu-Zhang model and develop a procedure to smoothly shrink the width of the system in one direction. By tracking gaps which remain open throughout the modulation, we establish a smooth transition from a two-dimensional to a one-dimensional topological insulator. In between the system exhibits both one- and two-dimensional topology, and the way the system accomplishes the transition is by making the one-dimensional topology more robust as the width decreases, while the two-dimensional topology becomes less robust. Finally, we show how the gaps arise from hybridization of edge states due to the finite width.

We numerically study the dynamics of a three-dimensional contractile fluid droplet in the bulk and under confinement. We show that varying activity leads to a variety of shapes and motile regimes whose motion is driven by an interplay between spontaneous flows and elasticity. In the bulk the droplet self-propels unidirectionally, acquiring either an almost spherical shape at intermediate activity or a peanut-like geometry for larger values. Under confinement, the droplet exhibits a previously unreported oscillating dynamics characterized by periodic hits against opposite walls of a microchannel while moving forward. These results could be of interest for the study of artificial microswimmers and their biological analogs, such as living cells.

The calculation of thermal conductivity in insulating solids at temperatures below the Debye temperature is problematic, due to the breakdown of classical and semi-classical approaches. In this work, we present a fully quantum methodology to compute thermal conductivity based on Path Integral Monte Carlo (PIMC) simulations combined with the Green-Kubo linear response theory. The method is applied to crystalline argon modeled by a Lennard-Jones potential, a paradigmatic system where quantum effects strongly affect both thermodynamic and transport properties. From PIMC simulations, we obtain the temperature-dependent phonon frequencies, lifetimes, and specific heat. From the imaginary time correlations of the energy current, we extract the thermal transport coefficients based on a physically motivated prior. We show that the experimentally observed increase of the thermal conductivity at low temperatures cannot be explained within a standard Peierls-Boltzmann framework or quasi-harmonic approximation using phonon lifetimes alone. Instead, a distinct transport lifetime emerges from the analysis of heat-current correlations. Our results demonstrate that quantum Monte Carlo methods provide a robust, non-perturbative framework to investigate heat transport in insulating solids, beyond the limits of classical molecular dynamics and quasi-harmonic approximations.

Understanding why and how crystalline solids melt remains a central problem in condensed-matter physics. Dislocation loops are fundamental topological excitations that control the thermodynamic stability of crystals, yet their role in setting universal aspects of melting has remained unclear. Here we show, within dislocation-mediated melting theory, that the free-energy condition for loop proliferation leads to a universal ratio between the energy of a minimal dislocation loop and the thermal energy at melting. For minimal dislocation loops that begin to proliferate at the onset of melting, this ratio takes the purely geometric value $\mathcal{E}_* = E_{\rm loop}/(k_B T_m) \approx 25.1$, independent of elastic moduli and chemistry-dependent details. This result provides a microscopic explanation for recent empirical findings by Lunkenheimer \emph{et al.}, who identified a closely related universal energy scale $\approx 24.6$ from viscosity data. The same framework also rationalizes the empirical $2/3$ rule relating the glass-transition and melting temperatures.

The application of rare-earth complexes in electrically driven light sources poses a series of challenges that require specific optimization of the molecular photophysical properties. Here, we present a report on films of three different Eu3+ complexes characterized in terms of emission spectra and fluorescence decay. We compare molecular complexes in powder form and sublimed films, in films on glass and on a metal surface, and in films of thicknesses down to less than 3 nm (< 3 ML), approaching electrical coupling. Our photoluminescence experiments supported by scanning tunneling microscopy of sub-monolayers indicate that Eu3+(trensal) complexes are less affected by sublimation and more stable on the metal surface than typical beta diketonate complexes, making them promising candidates for electroluminescence devices.

We uncover a "junction law" for genuine multipartite entanglement, suggesting that in gapped local systems multipartite entanglement is controlled and effectively localized near junctions where subsystem boundaries meet. Using the Rényi-2 genuine multi-entropy $\mathrm{GM}^{(\mathtt{q})}_2$ as a diagnostic of genuine $\mathtt{q}$-partite entanglement, we establish this behavior in $(2+1)$-dimensional gapped free-fermion lattices with correlation length $ξ$. For partitions with a single junction, $\mathrm{GM}^{(\mathtt{q})}_2$ exhibits a universal scaling crossover in $L/ξ$, growing for $L\llξ$ and saturating to a $ξ$-dependent constant for $L\ggξ$, up to $\mathcal{O}(e^{-L/ξ})$ corrections. In sharp contrast, for partitions without a junction, $\mathrm{GM}^{(\mathtt{q})}_2$ is exponentially suppressed in $L/ξ$ and drops below numerical resolution once $L\ggξ$. We observe the same pattern for $\mathtt{q}=3$ (tripartite) and $\mathtt{q}=4$ (quadripartite) cases, and further corroborate this localization by translating the junction at fixed system size. We also provide a geometric explanation of the junction law in holography. Altogether, these results show that in this gapped free-fermion setting genuine multipartite entanglement is localized within a correlation-length neighborhood of junctions.

Many anisotropic layered materials, despite their strong in-plane birefringence, exhibit substantial visible absorption, which severely restricts cavity lengths and hinders the observation of purely birefringence-governed optical phenomena. Here, we realize a birefringence-driven anisotropic optical cavity using $α$-MoO3 flakes, capitalizing on their ultralow optical loss and pronounced in-plane birefringence. Using angle-resolved polarized Raman (ARPR) spectroscopy, we observe a mode-sensitive enhancement of anisotropy, dependent on both flake thickness and Raman shift. A unified model that incorporates the intrinsic Raman tensor, birefringence, and chromatic dispersion accurately reproduces the experimental data, elucidating how cavity resonances at both excitation and scattered wavelengths interact. Within this framework, the intrinsic phonon anisotropy is quantified, providing invaluable insights for accurately predicting ARPR responses and identifying crystallographic orientation. This work provides fundamental insights into birefringence-governed cavities and opens avenues for high-performance birefringent optics and cavity-enhanced anisotropic phenomena.

Photonic devices depend critically on the dielectric materials from which they are made, with higher refractive indices and lower absorption losses enabling new functionalities and higher performance. However, these two material properties are intrinsically linked through the empirical Moss rule, which states that the refractive index of a dielectric decreases as its band gap energy increases. Materials that surpass this rule, termed super-Mossian dielectrics, combine large refractive indices with wide optical transparency and are therefore ideal candidates for advanced photonic applications. This Review surveys the expanding landscape of high-index dielectric and semiconductor materials, with a particular focus on those that surpass the Moss rule. We discuss how electronic band structures with a large joint density of states near the band edge give rise to super-Mossian behavior and how first-principles computational screening can accelerate their discovery. Finally, we establish how the refractive index sets the performance limits of nanoresonators, waveguides, and metasurfaces, highlighting super-Mossian dielectrics as a promising route toward the next performance leap in photonic technologies.

Graphene-based systems have emerged as a rich platform for exploring emergent quantum phenomena-including superconductivity, magnetism, and correlated insulating behavior-arising from flat electronic bands that enhance many-body interactions. Realizing such flat bands has thus far relied primarily on moiré graphene superlattices or rhombohedral stacking graphene systems, both of which face challenges in reproducibility and tunability. Here, we introduce an artificial Kagome superlattice in bilayer graphene, engineered via nanopatterning of the dielectric substrate to create a precisely defined and electrostatically tunable periodic potential. Magnetotransport measurements reveal the emergence of a stack of correlated insulating states at moderate superlattice potentials, characteristic of strong electron-electron interactions within Kagome-induced flat bands. As temperature increases, these correlated gaps collapse, signaling the thermal suppression of interaction-driven states. Continuum-model calculations confirm the formation of multiple flat minibands and reproduce the observed evolution of band reconstruction. Our results establish dielectric-patterned graphene superlattices as a robust and controllable architecture for realizing flat-band-induced correlated phenomena beyond moiré systems.

Machine learning is rapidly finding its way into the field of computational quantum physics. One of the most popular and widely studied approaches in this direction is to use neural networks to model quantum states (NQS) in the Hamiltonian formulation of quantum mechanics. However, an alternative angle of attack to leverage machine learning in physics is through the path integral formulation, which has so far received far more limited attention. In this paper, we explore how reinforcement learning can be used to compute a class of Euclidean path integrals that yield the thermal density matrix of a quantum system, thereby enabling the computation of the free energy or other thermal expectation values. In particular, we propose a two-step approach with the unique feature that after a variational approximation for a quantity is obtained in a first step, it can then be used to efficiently compute the exact result in a second step. We benchmark this method on several simple systems and then apply it to the quantum rotor chain.

Interfacial thermal transport between graphene and water plays a critical role in a wide range of thermal and energy applications. Although chemical functionalization can significantly enhance graphene-water interfacial thermal conductance, it often degrades graphene's intrinsic in-plane phonon transport. In this work, we perform a systematic deep neural network molecular dynamics study comparing edge-functionalized graphene nanoribbons with surface-functionalized graphene in aqueous environments. We demonstrate that functionalizing only 10% of the ribbon edges with hydroxyl groups increases the graphene-water interfacial thermal conductance by more than eightfold, primarily due to strengthened interfacial interactions and improved wettability at the edges. In contrast to basal-plane oxidation, edge functionalization largely preserves in-plane thermal conductivity. Importantly, hydroxyl edge groups exert competing effects on phonon transport: they introduce additional boundary scattering that suppresses heat conduction, while simultaneously passivating dangling bonds at bare edges, thereby reducing phonon localization and edge-induced scattering. This competition leads to a non-monotonic dependence of in-plane thermal conductivity on edge functionalization ratio. These results establish edge functionalization as an effective strategy for enhancing graphene-water interfacial thermal transport without sacrificing intrinsic phonon transport properties.

Active networks composed of biopolymers and motor proteins provide versatile biomimetic systems that have advanced active matter physics and deepened our understanding of cytoskeletal dynamics and self-organization under diverse stimuli. In these systems, activity arises in aqueous solutions where motor proteins cross-link biopolymers and generate active stress driving the emergent network behavior. Here, we establish the active network in the form of a sessile, multi-component droplet on a substrate and investigate how evaporation influences its dynamics. We focus on how mass loss and compositional changes in the droplet reshape the behavior of the active suspension. We show that capillary and Marangoni flows drive the self-organization of microtubules into a distinctive radial arrangement within the droplet. The cross-linking ability of motor proteins gives rise to a striking non-monotonic wetting behavior, where the extensile stresses generated by the motor proteins strongly affect the characteristic timescale of the contact-line retracting and subsequent expansion. Using a combined experimental and theoretical approach, we demonstrate the crucial role of crosslinking in evaporating microtubule networks, and explain how active stresses together with evaporation-induced flows govern the dynamics of reconstituted microtubule systems and their wetting behavior. Evaporating droplets have recently attracted significant attention in the scientific community, and the findings of the setup presented in this study can have broad implications, ranging from self-organization and mechanical pattern formation in biological systems to questions about the origin of life.

Dense granular flows exhibit nonlocal effects due to stress transmission in microplastic events, especially in quasi-static or slowly sheared regions. Hence, traditional local rheological models fail to capture spatial cooperativity effects that are prominent in many granular systems. The nonlocal granular fluidity (NGF) model addresses this limitation by introducing a diffusive-like partial differential equation for a fluidity field, governed by a key material-dependent parameter: the nonlocal amplitude A. However, determining A from experiments or simulations is known to be difficult and typically requires extensive calibration across multiple geometries. In this work, we present a data-driven platform based on Physics-Informed Neural Networks (PINNs) embedded with the NGF model, capable of solving granular flows in a forward or inverse manner. We show that once trained on transient flow fields, these non-local PINNs can readily infer the material parameters, as well as the pressure and stress fields. These data-driven frameworks allow for accurate recovery of small variations in the nonlocal amplitude, A, which lead to sharp bifurcation-like transitions in the flow field. This approach demonstrates the feasibility of data-driven parameter inference in complex nonlocal models and opens up new possibilities for characterizing granular materials from sparse experimental observations.

Complex systems out of equilibrium often experience intermittent oscillations between quiescent and highly dynamic states. The type of intermittency depends on how energy is pumped into the system, and how it is dissipated. While intermittency is usually driven by stochastic noise or external forcing, energy can also be sourced from field-mediated interactions between particles, which are often nonreciprocal and effectively violate Newton's 3rd law. Here we demonstrate how nonreciprocal interactions produce intermittency in clusters of charged micron-sized particles confined in a plasma sheath. Through three-dimensional particle tracking, we observe that vertical oscillations, induced by fluctuations of the plasma environment, can be parametrically coupled to the horizontal modes. Experiments and simulations show that nonreciprocal interactions strongly amplify this parametric coupling, creating a positive feedback loop that drives explosive growth of both the horizontal and vertical modes. This mechanism triggers abrupt melting transitions from an ordered cluster to an ergodic gas-like state, and leads to intermittent switching between states over long time scales. Overall, our work identifies nonreciprocal interactions as a key mechanism through which strongly coupled finite systems transform interaction-mediated activity into dynamical nonequilibrium states.

In this paper, we introduce a new order parameter called the weighted coordination number (WCN) to study the liquid-liquid (LL) phase separation, using the temperature-dependent coarsening of the LL interface as a possible mechanism for the glass transition. The well-established glass-forming Kob-Andersen binary Lennard-Jones system is used for our studies. The gas-liquid binodal line is reconstructed using the WCNs, and the same approach is extended to study the liquid-liquid binodal line. Systems of various densities are instantaneously quenched from high to low temperatures where a liquid-liquid separation is observed. Densities and the composition of each liquid state are used to check the level rule, along with density and pressure profiles, demonstrating local equilibrium of liquid-liquid phase separation. The transition from the liquid-liquid phase separation in the supercooled region to the glass transition region is modeled by adopting a Markov Network Model to estimate the temperature dependent viscosity using liquid-liquid interfacial information from the classification.

We explore the onset of spontaneous strong-to-weak symmetry breaking (SW-SSB) under U(1)-symmetric (i.e., charge-conserving) open-system dynamics. We define this phenomenon for quantum states and classical probability distributions, and explore it in three complementary models, one of which exhibits nontrivial quantum coherence at short times. Our main conclusions are as follows. In one dimension, the strong symmetry is not spontaneously broken at any finite time; however, correlators probing strong-to-weak symmetry breaking develop order on length scales that grow linearly in time, parametrically faster than charge diffusion. We provide numerical evidence for this scaling in multiple distinct probes of SW-SSB, and derive it from a field-theory analysis. Moreover, we relate this scaling to the problem of inferring the charge inside a subregion by measuring its surroundings, and construct explicit decoding protocols that illustrate its origin. In two dimensions, field theory and numerical simulations support a finite-time Berezinskii-Kosterlitz-Thouless-like SW-SSB transition. Within continuum hydrodynamics, by contrast, SW-SSB happens at infinitesimal time in two or more dimensions. The SW-SSB transition time can thus be interpreted as marking the emergence of a continuum hydrodynamic description, or (more precisely) the timescale beyond which non-hydrodynamic information such as discrete particle worldlines can no longer be inferred. We support this picture by analyzing a model in which we exploit SW-SSB to derive a classical stochastic hydrodynamic description from the underlying quantum dynamics.

In one-dimensional systems, spontaneous symmetry breaking (SSB) states are fragile by nature, as the injection of a non-zero energy density above the ground state is expected to restore the symmetry. This instability implies that local perturbations can lead to macroscopic correlation profiles, a breaking of clustering properties and even macroscopic quantum superpositions. In this work, we investigate the effect of interaction on this phenomenology by considering an interacting model with conserved domain wall number, that possesses a ferromagnetic ground state breaking the Z2 symmetry of the Hamiltonian. We first show that a local quench in this system amplifies quantum interferences, producing a macroscopic magnetisation profile that directly reflects the scattering phase of the model. Then, we use two asymmetry measures, namely the Entanglement Asymmetry (EA) and Quantum Fisher Information (QFI), to characterise the quantum coherence associated with the fluctuations of the magnetisation. By focusing on subsystems comparable in size to the light-cone of the perturbation, we confirm the emergence of macroscopic quantum coherence throughout the whole perturbed region. Finally, we discuss the link between EA and QFI and show that the variance/EA inequality for pure state can be generalised to a QFI/EA inequality for mixed states.

Isotropic integrable spin chains such as the Heisenberg model feature superdiffusive spin transport belonging to an as-yet-unidentified dynamical universality class closely related to that of Kardar, Parisi, and Zhang (KPZ). To determine whether these results extend to more generic one-dimensional models, particularly those realizable in quantum simulators, we investigate spin and energy transport in non-integrable, long-range Heisenberg models using state-of-the-art tensor network methods. Despite the lack of integrability and the asymptotic expectation of diffusion, for power-law models (with exponent $2 < α< \infty$) we observe long-lived $z=3/2$ superdiffusive spin transport and two-point correlators consistent with KPZ scaling functions, up to times $t \sim 10^3/J$. We conjecture that this KPZ-like transport is due to the proximity of such power-law-interacting models to the integrable family of Inozemtsev models, which we show to also exhibit KPZ-like spin transport across all interaction ranges. Finally, we consider anisotropic spin models naturally realized in Rydberg atom arrays and ultracold polar molecules, demonstrating that a wide range of long-lived, non-diffusive transport can be observed in experimental settings.

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.

The Lie theory of non-commutative integrability is used to reconstruct some integrable systems of ordinary differential equations in three dimensional Eucledian space. The Darboux-Brioschi-Halphen system is an example of the Lie integrable system associated with the simple Lie algebra sl(2,R). Other examples are related with solvable three dimensional real Lie algebras of Bianchi B class.

The representation of protein backbone geometry through the discrete nonlinear Schrödinger equation provides a theoretical connection between biological structure and integrable systems. Although the global application of this framework is constrained by chiral degeneracies and non-local interactions we propose that helical peptides can be effectively modeled as piecewise integrable systems in which the discrete Hasimoto map remains applicable within specific geometric boundaries. We delineate these boundaries through an analytic characterization of the mapping between biochemical dihedral angles and Frenet frame parameters for a dataset of 50 helical peptide chains. We demonstrate that the transformation is information-preserving globally but ill-conditioned within the helical basin characterized by a median Jacobian condition number of 31 which suggests that the loss of chiral information arises primarily from local coordinate compression rather than topological singularities. We define a local integrability error $E[n]$ derived from the discrete dispersion relation to show that deviations from integrability are driven predominantly by torsion non-uniformity while curvature remains structurally rigid. This metric identifies integrable islands where the analytic dispersion relation predicts backbone coordinates with sub-angstrom accuracy yielding a median root-mean-square deviation of 0.77\,Å and enables a segmentation strategy that isolates structural defects. We further indicate that the inverse design of peptide backbones is feasible within a quantitatively defined integrability zone where the design constraint reduces essentially to the control of torsion uniformity. These findings advance the Hasimoto formalism from a qualitative descriptor toward a precise quantitative framework for analyzing and designing local protein geometry within the limits of piecewise integrability.

The Lorenz equations [1] are a severe Galerkin-truncation of the Oberbeck-Boussinesq (OB) equations describing Rayleigh-Bénard convection (RBC). Here we examine the mathematical connections between the chaotic lobe-switching behavior of a stochastic form of the Lorenz equations, that model the interaction between the thermal boundary layers and the core circulation, and the mean wind reversals in the experiments of Sreenivasan et al. [2]. Long-time numerical simulations of these stochastic equations, not easily accessible with the OB equations, yield a probability distribution for lobe inter-switch timings that exhibits non-Gaussian, multifractal behavior. In the Gaussian frequency range the simulations mirror the laboratory measurements and the classical Hurst exponent and quadratic variation show Brownian second-moment statistics. Further scrutiny reveals a non-linear cumulant generating function, or moment-exponent function, and thus multifractality. A simple generalized two-scale Cantor-cascade analysis reproduces these properties, showing that multiplicative intermittency, characteristic of turbulence, strongly influences the statistics. This demonstrates that this stochastic Lorenz system is a faithful, low-dimensional surrogate for mean-wind reversals in RBC.

We introduce a compact operator-based technique that solves the paraxial wave equation for a broad class of structured light fields. Using the spatial evolution operator to propagate two families of physically apodized inputs, Gaussian apodized Whittaker integrals and Gaussian apodized Helmholtz fields, we derive closed form expressions that retain the Gaussian width and therefore describe finite energy beams. The method unifies and extends the Helmholtz Gauss families and readily generalizes to nonseparable nondiffracting architectures. Experiments on superposed Bessel Gauss beams confirm the predicted transverse rotations, demonstrating that this operator approach is a fast, transparent, and practical alternative to standard diffraction ntegral treatments

Twinning is a primary deformation mechanism in Mg alloys. This study focuses on tension twins during uniaxial compression of Mg-Y alloys, with three key aspects: the orientation specificity of twin grains, the relative evolution of CRSS with increasing Y content, and the local stress and strain evolution at twin sites. Experimental characterization and crystal plasticity modeling were performed. In Mg-7wt.%Y, TT2-{112-1} tension twins were observed in addition to the common TT1-{101-2} twins. Increasing Y suppressed TT1 formation while promoting TT2 activity. A previously unreported group of crystallographic orientations with a higher global Schmid factor for <c+a> slip was identified, which exhibited TT1 twinning with increasing compression strain. To elucidate Y effects on twin activity and local mechanical fields, both TT1 and TT2 tension twin modes were incorporated into PRISMS-Plasticity, an open-source, finite element-based crystal plasticity solver. Four binary Mg-Y alloys were modeled under compression, and statistical analysis was conducted to correlate initial orientations, stress-strain distributions, and twin activities as functions of Y concentration. The plasticity analysis revealed that increasing Y decreases the CRSS ratio of prismatic and pyramidal slip relative to TT1 twinning, while the slip-to-twin CRSS ratio for TT2 increases, thereby serving as a potential indicator of differential twin activity with Y addition in Mg alloys. Additionally, despite their small volume fraction, TT2 twin sites were predicted higher local strain accumulation locally, relative to the representative volume element and TT1 twins, suggesting their potential influence on localized phenomena such as recrystallization or twin nucleation. These findings provide insight into local mechanical behavior in Mg alloys and support alloy design for advanced engineering applications.

Experimentalists often use wind tunnels to study aerodynamic turbulence, but most wind tunnel imaging techniques are limited in their ability to take non-invasive 3D density measurements of turbulence. Wavefront tomography is a technique that uses multiple wavefront measurements from various viewing angles to non-invasively measure the 3D density field of a turbulent medium. Existing methods make strong assumptions, such as a spline basis representation, to address the ill-conditioned nature of this problem. We formulate this problem as a Bayesian, sparse-view tomographic reconstruction problem and develop a model-based iterative reconstruction algorithm for measuring the volumetric 3D density field inside a wind tunnel. We call this method WindDensity-MBIR and apply it using simulated data to difficult reconstruction scenarios with sparse data, small projection field of view, and limited angular extent. WindDensity-MBIR can recover high-order features in these scenarios within 10% to 25% error even when the tip, tilt, and piston are removed from the wavefront measurements.

Active photonic integrated circuits (PICs) in the visible spectrum are essential for on-chip applications, requiring low-loss waveguides with broad transparency and efficient, low-power phase modulation. Here, we demonstrate a compact, ultra-low-power phase modulator based on a silicon nitride (Si$_3$N$_4$) waveguide integrated with thin-film lead zirconate titanate (PZT) that actuates a bridge-type MEMS. The suspended actuator exploits PZT's strong piezoelectric effect to induce mechanically driven phase shifts, enabling efficient modulation in a Mach--Zehnder interferometer. For 3~mm and 5~mm modulators, phase shifts of $1.45π$ and $2.5π$ are achieved at 10~V, corresponding to a scalability metric ($V_π\cdot L$) of 2.25~V$\cdot$cm at 635~nm. This represents an order-of-magnitude improvement in scalability over stress-optic PZT modulators. The devices also exhibit ultralow power consumption ($\sim 12\,\mathrm{nW}$), $\sim 5\,\mathrm{ms}$ rise time, and optical loss $< 0.75\,\mathrm{dB/cm}$. Furthermore, we demonstrate on-chip beam shaping.

The present study is concerned with a one-dimensional problem in explosive welding that pertains to the collision of lead and steel plates. The metal plates and the surrounding air are represented as separate immiscible phases governed by independent equations of state. A multifluid Godunov-type (finite-volume) computational algorithm, based on the mechanical-equilibrium Euler equations and incorporating pressure relaxation, is used to numerically describe the evolution of the waves resulting from the collision. The position of the interface (contact discontinuity) between immiscible phases is tracked by means of the volume-of-fluid (VOF) method. The numerical model allows one to account for the existence of tensile stresses in metal and registers them as regions of negative pressure. The computed arrival time of the unloading wave at the interface between the plates agrees with the experimental data and with simulation results obtained via different methods.

Free-space optical communications with moving targets, such as satellite terminals, demand ultrafast wavefront sensing and correction. This is typically addressed using a Shack-Hartmann sensor, which pairs a high-speed camera with a lenslet array, but such systems add significant cost, weight, and power demands. In this work, we present a hybrid opto-electric neural network (OENN) wavefront sensor that enables ultra-high-speed operation in a compact, low-cost system. Subwavelength diffractive metasurfaces efficiently encode the incoming wavefront into tailored irradiance patterns, which are then decoded by a lightweight multilayer perceptron (MLP). In simulation and experiment, the hybrid approach achieves average Strehl ratio (SR) improvements exceeding 60% and 45%, respectively, for unseen wavefronts compared to purely electronic sensors with few-pixel inputs. Although larger MLPs allow purely electronic sensors to match the hybrid's SR under static conditions, transient atmosphere modeling shows that their added latency leads to rapid SR degradation with increasing Greenwood frequency, while the hybrid system maintains performance. These results highlight the potential of hybrid OENN architectures to unlock scalable, high-bandwidth free-space communication systems and, more broadly, to advance optical technologies where real-time sensing is constrained by electronic latency.

Thermal fluctuations play a central role in fluid dynamics at mesoscopic scales and must be incorporated into numerical schemes in a manner consistent with statistical mechanics. In this work, we develop a fluctuating lattice Boltzmann formulation based on an orthogonal central-moments-based representation. Stochastic forcing is introduced directly in the space of central moments (CMs) and consistently paired with mode-dependent relaxation, yielding a discrete kinetic model that satisfies the fluctuation-dissipation theorem exactly at the lattice level. Owing to the orthogonality of the basis, the equilibrium covariance matrix of the central moments is diagonal, and each non-conserved mode can be interpreted as an independent discrete Ornstein-Uhlenbeck process with variance fixed by equilibrium thermodynamics. The resulting formulation guarantees exact equipartition of kinetic energy at equilibrium, preserves Galilean invariance, and retains the enhanced numerical stability characteristic of CMs-based collision operators. Explicit fluctuating schemes are constructed for the D2Q9 and D3Q27 lattices. The extension to reduced-velocity discretisation is discussed too. A comprehensive set of numerical tests verifies correct thermalisation, isotropy of equilibrium statistics, and the expected scaling of velocity fluctuations with thermal energy, density, and relaxation time. In contrast to fluctuating BGK formulations, the present method remains stable and well posed in the over-relaxation regime, including in the immediate vicinity of the stability limit. These results demonstrate that CMs-based lattice Boltzmann methods provide a natural and robust framework for fluctuating hydrodynamics, in which dissipation, noise, and kinetic mode structure are consistently aligned at the discrete level.

Photonic integration offers the potential to bring complex high-performance optical systems to the form factor of a compact semiconductor chip. However, the range of system functions accessible critically depends on the extent to which free-space and fiber components can be made integrable. The ultralow-expansion cavity-stabilized laser$-$often used in precision metrology, high-resolution sensors, and advanced systems in atomic physics$-$is one component that currently has no direct parallel on chip. Lasers stabilized to photonically-integrated resonators exist, but exhibit considerably higher frequency noise and are accompanied by large levels of frequency drift. We demonstrate here a new architecture for an ultranarrow linewidth integrated laser based on stabilization to a sinusoidal fringe of an interferometer having a long 25-m unbalanced delay line. Our interferometric laser not only advances the state-of-the-art for on-chip lasers, but we in addition introduce an amplitude locking scheme that greatly suppresses the laser's long-term frequency wander. We achieve a record on-chip fractional frequency noise of $5.6 \times 10^{-14}$, corresponding to a linewidth of 12 Hz centered at 1348 nm. To showcase the utility of this laser, we divide the optical carrier to microwave frequencies, demonstrating the ability to outperform state-of-the-art quartz crystal oscillators by 15 dB or more.

In South Korea, age-disaggregated tuberculosis (TB) data at the district level are not publicly available due to privacy constraints, limiting fine-scale analyses of healthcare accessibility. To address this limitation, we present a high-resolution, district-level dataset on tuberculosis (TB) fatalities and hospital accessibility in South Korea, covering the years 2014 to 2022 across 228 districts. The dataset is constructed using a reconstruction method that infers age-disaggregated TB cases and fatalities at the district level by integrating province-level age-specific statistics with district-level spatial and demographic data, enabling analyses that account for both spatial heterogeneity and age structure. Building on an existing hospital allocation framework, we extend the objective function to an age-weighted formulation and apply it to the reconstructed dataset to minimize TB fatalities under different age-weighting schemes. We demonstrate that incorporating age structure can give rise to distinct optimized hospital allocation patterns, even when the total number of minimized fatalities is similar, revealing trade-offs between efficiency and demographic targeting. In addition, the dataset supports temporal analyses of TB burden, hospital availability, and demographic variation over time, and provides a testbed for spatial epidemiology and optimization studies that require high-resolution demographic and healthcare data.

Many numerical studies validate droplet wall impact using only maximum spreading diameter, yet this metric alone cannot ensure correct droplet dynamics. We present a combined dynamic contact angle (DCA) model that merges the geometric accuracy of the generalized Hoffman-Voinov-Tanner law with the kinematic consistency of a Hoffman function-based approach, improving predictions of droplet spreading and receding. We simulate water-glycerol droplet impact on sapphire glass at Weber numbers 20 -- 250 and assess both contact angle formulations. Simulated radial velocity fields are processed in Python using SciPy and compared with Particle Image Velocimetry measurements in the longitudinal section of the spreading droplet. The Hoffman function-based model captures the main droplet kinematic trends and provides more consistent receding dynamics. The generalized Hoffman-Voinov-Tanner law matches the maximum spreading diameter within 7%. However, during receding, it shows a median absolute error in radial velocity up to three times higher than that of the Hoffman function-based solution. Average radial velocity and spreading velocity can differ from experimental trends even when maximum spreading is reproduced. These findings support validation combining geometric and kinematic metrics and motivate the combined model for predicting spreading and receding. Using the maximum spreading factor $β_{max}$ as the ratio of the maximum spreading diameter over the initial droplet diameter and the characteristic capillary number $Ca_{char}$ defined from the mean internal horizontal velocity at 300 micrometer above the substrate, we introduce a $(β_{max},\,Ca_{char})$ diagram to relate spreading characteristics to internal flow dynamics. We hypothesize that, given sufficient data, the contact-line geometry may be used to estimate internal kinematics.

The recent reconstruction of the Drosophila brain provides a neural network of unprecedented size and level of details. In this work, we study the geometrical properties of this system by applying network embedding techniques to the graph of synaptic connections. Since previous analysis have revealed an inhomogeneous degree distribution, we first employ a hyperbolic embedding approach that maps the neural network onto a point cloud in the two-dimensional hyperbolic space. In general, hyperbolic embedding methods exploit the exponentially growing volume of hyperbolic space with increasing distance from the origin, allowing for an approximately uniform spatial distribution of nodes even in scale-free, small-world networks. By evaluating multiple embedding quality metrics, we find that the network structure is well captured by the resulting two-dimensional hyperbolic embedding, and in fact is more congruent with this representation than with the original neuron coordinates in three-dimensional Euclidean space. In order to examine the network geometry in a broader context, we also apply the well-known Euclidean network embedding approach Node2vec, where the dimension of the embedding space, $d$ can be set arbitrarily. In 3 dimensions, the Euclidean embedding of the network yields lower quality scores compared to the original neuron coordinates. However, as a function of the embedding dimension the scores show an improving tendency, surpassing the level of the 2d hyperbolic embedding roughly at $d=16$, and reaching a maximum around $d=64$. Since network embeddings can serve as valuable inputs for a variety of downstream machine learning tasks, our results offer new perspectives on the structure and representation of this recently revealed and biologically significant neural network.

We report a direct demonstration of quantum-enhanced sensing in the Fourier domain by comparing single- and two-photon interference in a fiber-based interferometer under strictly identical noise conditions. The simultaneous acquisition of both signals provides a common-mode reference that enables a fair and unambiguous benchmark of quantum advantage. Spectral analysis of the interferometric outputs reveals that quantum correlations do not increase the amplitude of the modulation peak, but instead lower the associated noise floor, resulting in the expected 3 dB improvement in signal-to-noise ratio. This enhancement persists in the sub-shot-noise regime, where the classical signal becomes buried in the spectral background while the two-photon contribution remains resolvable. These observations establish Fourier-domain quantum super-sensitivity as an operational and broadly applicable resource for precision interferometric sensing.

Transportable all-optical atomic clocks represent the next-generation devices for precision time keeping, ushering a new era in encompassing a wide range of PNT (Positioning, Navigation and Timing) applications in the civil and strategic sectors. Their performance relies on ultra-stable, narrow-linewidth lasers, frequency stabilized to a compact portable optical cavity. Among various designs, the cubic spacer-based ultra-stable cavity is particularly well-suited for transportable applications due to its low sensitivity to vibrations, owing to its symmetric geometry and robust mounting structure. While longer cavities offer a lower fundamental thermal noise floor, one needs to strike a balance between transportability and size. In this aspect, the 7.5 cm dual-axis cubic cavity offers a lower fundamental thermal noise floor in comparison to smaller counterparts, while still retaining a reasonable SWaP (Size, Weight and Power) for terrestrial and aerial PNT applications. Its dual-axis design also enables multi-wavelength laser stabilization, making it a promising candidate for future transportable clock applications. This work presents a detailed study of the 7.5 cm dual-axis cubic cavity using FEM (Finite Element Method) to evaluate its mechanical and thermal stability. We analyze the impact of various geometric factors, mounting forces, and machining imperfections, while also modelling thermal effects such as conduction, radiation, and mirror heating within a vacuum chamber and thermally shielded environment. Our findings provide design insights for developing robust dual-axis optical reference cavities, advancing the deployment of portable atomic clocks for next-generation applications in PNT, geodesy, VLBI (Very Long Baseline Interferometry) and deep space missions.

Natural birth and death are fundamental mechanisms of population dynamics in ecosystems and have played pivotal roles in shaping population dynamics. Nevertheless, in studies of cyclic competition systems governed by the rock-paper-scissors (RPS) game, these mechanisms have often been ignored in analyses of biodiversity. On the other hand, given the prevalence and profound impact on biodiversity, understanding how higher-order interactions (HOIs) can affect biodiversity is one of the most challenging issues, and thus HOIs have been continuously studied for their effects on biodiversity in systems of cyclic competing populations, with a focus on neutral species. However, in real ecosystems, species can evolve and die naturally or be preyed upon by predators, whereas previous studies have considered only classic reaction rules among three species with a neutral, nonparticipant species. To identify how neutral species can affect the biodiversity of the RPS system when species' natural birth and death are assumed, we consider a model of neutral species in higher-order interactions within the spatial RPS system, assuming birth-and-death processes. Extensive simulations show that when neutral species interfere positively, they dominate the available space, thereby reducing the proportion of other species. Conversely, when the interference is harmful, the density of competing species increases. In addition, unlike traditional RPS dynamics, biodiversity can be effectively maintained even in high-mobility regimes. Our study reaffirms the critical role of neutral species in preserving biodiversity.

Photochromic materials enable dynamic optical modulation through reversible transitions between distinct absorption states, with broad potential for smart windows, adaptive optics, and reconfigurable photonic devices. Micron-scale photochromic hybrid films present a particularly attractive platform for these applications, combining straightforward preparation with substantial optical modulation and scalability for high-volume fabrication. However, rational design of such films remains fundamentally constrained by the absence of well-defined optical constants. Unlike homogeneous thin films, micron-scale hybrid photochromic materials comprise active particles dispersed non-uniformly within polymer matrices. Conventional first-principles electromagnetic simulations face substantial computational costs and discrepancies between simulated and experimental particle distributions. Here, we introduce a data-driven framework that extracts effective optical constants directly from minimal experimental transmittance measurements. Our dual-state effective model approximates the complex inhomogeneous photochromic layer as a compressed homogeneous medium characterized by pseudo-refractive indices and pseudo-extinction coefficients for both pristine and UV-irradiated states. Through systematic optimization against experimental data from tungsten oxide-polyvinylpyrrolidone hybrid films, we determine wavelength-dependent pseudo-optical constants and compression ratios that enable accurate prediction of optical modulation within the tested thickness range. Our methodology establishes a framework for engineering hybrid photochromic systems and demonstrates how data-driven modeling can overcome limitations in characterizing complex nanostructured materials.

Chemiosmosis maintains life in nonequilibrium through ion transport across membranes, yet the origin of this order remains unclear. We develop a minimal model in which ion pump orientation and the intracellular electrochemical potential mutually reinforce each other. This model shows that fluctuations can induce collective pump alignment and the formation of a membrane potential. The alignment undergoes a phase transition from disordered to ordered, analogous to the Ising model. Our results provide a self-organizing mechanism for the emergence of bioelectricity, with implications for the origin of life.

Flat-top (FT) solitons are optical pulses that arise from the balance of dispersion and self-phase modulation in media with the competing cubic-quintic nonlinearity. Previously, FT solitons were studied only in the case of the second-order dispersion ($m=2$). Following the recent observation of pure-quartic solitons (corresponding to $m=4$), we here construct families of FT solitons in the setting with pure-high-even-order dispersion (PHEOD), including $m=4,6,8$, and $10$, and address interactions between them. The PHEOD solitons are completely stable, and, unlike the conventional solitons, they feature oscillatory tails. Interactions between the PHEOD solitons are anomalous, featuring repulsion and attraction between in- and out-of-phase solitons, respectively. These results expand the variety of optical solitons maintained by diverse dispersive nonlinear media.

Structured illumination microscopy (SIM) has attained high spatiotemporal delineation of subcellular architecture, yet offers limited insight into chemical composition. Here, we present Chem-SIM, a structured-illumination fluorescence detected mid-infrared photothermal microscopy, for super-resolution chemical imaging of microorganisms and mammalian cells. A computational pipeline combining Poisson maximum-likelihood demodulation and spectral normalization across wavenumber is implemented to robustly recover the weak IR-induced fluorescence intensity change under low photon budgets and convert the fluorescence intensity modulation to chemical fingerprints. A photothermal gating scheme further rejects water backgrounds in aqueous samples, while the IR pump maintains cellular activity at near-physiological temperature. Chem-SIM preserves full vibrational fingerprints, achieves SIM-grade lateral resolution, and operates in a high-throughput, camera-based format with minimal modifications and low photothermal load. At the single-bacterium level, Chem-SIM distinguishes stationary phase from log phase cells through chemical content mapping. In ovarian cancer cells, Chem-SIM delivers readouts of lipid chemistry under deuterium fatty acid treatment and resolves lipid droplets dynamics in live cells. Together, Chem-SIM provides an accessible route to super-resolved mapping of organelle chemistry, metabolism, and dynamics.

We compare the ability of a simulated annealing program and an evolutionary algorithm to find molecules with large molecular average hyperpolarizabilities. This property is an important component of nonlinear optical materials. Both optimization programs represent molecules as SMILES strings, a method that is widely used by chemists to describe molecular structure using short ASCII strings. Our results suggest that both approaches are comparable and can be used to solve a variety of more realistic problems of interest to chemists and material scientists.

Nonlinear optical (NLO) materials are essential for many photonic, telecommunication, and laser technologies, yet discovering better NLO molecules is computationally challenging due to the vast chemical space and competing objectives. We compare evolutionary algorithms for molecular design, targeting four objectives: maximizing the ratio of first-to-second hyperpolarizability $(β/γ)$, optimizing HOMO-LUMO gap and linear polarizability to target ranges, and minimizing energy per atom. We encode molecules as SMILES strings and evaluate their properties using quantum-chemical calculations. We compare NSGA-II, MAP-Elites, MOME, a single-objective $(μ+λ)$ evolutionary algorithm, and simulated annealing. Quality diversity methods maintain archives across a measure space defined by atom and bond count, enabling the discovery of structurally diverse molecules. Our results demonstrate that NSGA-II consistently earns high scores in every objective, leading to high-quality molecules, but MOME does a better job exploring a wide range of possibilities, resulting in higher global hypervolume and MOQD scores. However, each method has strengths and weaknesses, and produced many promising molecules.

We present a biologically inspired quantum neural network that encodes neuronal populations as fully connected qubits governed by the Lipkin-Meshkov-Glick (LMG) quantum Hamiltonian and stabilized by a synaptic-efficacy feedback implementing activity-dependent homeostatic control. The framework links collective quantum many-body modes and attractor structure to population homeostasis and rhythmogenesis, outlining scalable computational primitives -- stable set points, controllable oscillations, and size-dependent robustness -- that position LMG-based architectures as promising blueprints for bio-inspired quantum brains on future quantum hardware.

It is with great regret that Prof. Wolfgang Helfrich passed away on 28 September 2025 in Berlin. As the founder of the membrane liquid crystal model, Prof. Helfrich made outstanding contributions to membrane physics and liquid crystal display technology. This review article is written in his memory. Biomembranes, primarily composed of lipid bilayers, are not merely passive barriers but dynamic and complex materials whose shapes are governed by the principles of soft matter physics. This review explores the shape problem in biomembranes through the lens of material science and liquid crystal theory. Beginning with classical analogies to crystals and soap bubbles, it details the application of the Helfrich elastic model to explain the biconcave shape of red blood cells. The discussion extends to multi-layer systems, drawing parallels between the focal conic structures of smectic liquid crystals, the geometries of fullerenes and carbon nanotubes, and the reversible transitions in peptide assemblies. Furthermore, it examines icosahedral self-assemblies and shape formation in two-dimensional lipid monolayers at air/water interfaces. At the end of the paper, we find that the shapes such as cylinders, spheres, tori, bicocave discoids and Delaunay surfaces form a group. This result is merely an intrinsic geometric feature of these shapes and is independent of the biomembrane equation. When the pressure on the membrane, surface tension, and bending modules meet certain conditions, the biomembrane will take on these shapes. The review concludes by highlighting the unifying power of continuum elastic theories in describing a vast array of membrane morphologies across biological and synthetic systems.

We present an implementation of edge AI to compress data on an in-memory analog content-addressable memory (ACAM) device. A variational autoencoder is trained on a simulated sample of energy measurements from incident high-energy electrons on a generic three-layer scintillator-based calorimeter. The encoding part is distilled into tabular format by regressing the latent space variables using decision trees, which is then programmed on a memristor-based ACAM. In real-time, the ACAM compresses 48 continuously valued incoming energies measured by the calorimeter sensors into the latent space, achieving a compression factor of 12x, which is transmitted off-detector for decompression. The performance result of the ACAM, obtained using the Structural Simulation Toolkit, the SST open source framework, gives a latency value of 24 ns and a throughput of 330M compressions per second, i.e., 3 ns between successive inputs, and an average energy consumption of 4.1 nJ per compression.

Microwave synthesizers are central to test and measurement systems across applications including wireless communications, radar, spectroscopy, and time and frequency metrology. State-of-the-art microwave sources, however, are fundamentally constrained by trade-offs between frequency tunability and spectral purity. Electro-optic frequency division (eOFD) is an emerging technique for dividing down the purity of optical sources to the microwave domain. Previously reported eOFD-based synthesizers generally have limited tunability due to feedback stabilization requirements. Here we demonstrate a feed-forward eOFD architecture in which the frequency tunability of a microwave source is preserved while optical spectral purity is divided through feed-forward cancellation, without any downstream electronic frequency synthesis. By canceling the phase noise of the microwave source without feedback, this eOFD approach removes loop bandwidth and source noise constraints observed in prior eOFD architectures. We achieve octave-spanning tunability, including the entire X-band, with phase noise below -140 dBc/Hz at kilohertz offsets and a high-frequency noise floor between -155 dBc/Hz and -145 dBc/Hz for carrier frequencies from 8 to 16 GHz. This performance corresponds to single-femtosecond integrated timing jitter, enabling, to our knowledge, the first demonstration of coherent, optically referenced microwave synthesis under wide tuning with this level of spectral purity.

Despite significant advances in lead-free perovskite photovoltaics, achieving a balance among environmental safety and high optoelectronic performance remains challenging. The inorganic double perovskite Cs2AgBiBr6 has emerged as a promising candidate owing to its robust three-dimensional crystal structure and suitable visible-range bandgap. However, best power conversion efficiencies (PCEs) for Cs2AgBiBr6 solar cells reported so far - 6.37% experimentally and 27.78% in numerical studies - remain below the theoretical performance potential, largely due to suboptimal charge transport layers, and interface-related recombination losses. Here, we address this gap using a 3D finite-element method (FEM) implemented in COMSOL Multiphysics, which couples optical simulations with semiconductor drift-diffusion transport. To our knowledge, this work represents the first comprehensive 3D FEM-based study of a double halide perovskite solar cell. Screening of 25 electron transport layer (ETL)-hole transport layer (HTL) combinations identifies CeO2 and P3HT as the optimal ETL and HTL respectively. Device performance is further analyzed through systematic variation of layer thicknesses, doping concentrations and defect densities within the FTO/CeO2/Cs2AgBiBr6/P3HT/Au architecture. Under optimized parameters, the simulated device achieves a PCE of 31.76%, representing the theoretical upper bound predicted by the model. Overall, this work demonstrates 3D physics-based device engineering as a decisive pathway for overcoming efficiency bottlenecks in lead-free double perovskite photovoltaics.

Understanding how bound states in the continuum (BICs) emerge in periodic metasurfaces is essential for the controlled design of high-Q resonances and their systematic manipulation. Here, we investigate the singular value decomposition (SVD) of the effective transition matrix and the scattering matrix of periodic metasurfaces within a parameter range where the metasurface sustains a BIC. Our analysis yields general and practically applicable conditions on the singular values and singular vectors that enable BIC formation. At the BIC eigenfrequency, the inverse of the largest singular value of both matrices vanishes, and the corresponding left (right) singular vector is orthogonal to outgoing (incoming) plane waves that propagate in the directions of open diffraction orders. Our SVD-based approach predicts the spectral position of the BIC and provides detailed information about its properties, including the expansion coefficients in the multipole and plane-wave bases, as well as its behavior under perturbations that transform the BIC into a quasi-BIC. The approach is numerically validated by considering both symmetry-protected and accidental BICs in arrays of scatterers supporting electromagnetic or acoustic multipole resonances. The presented SVD framework offers a broadly applicable foundation for engineering BICs and quasi-BICs in complex metasurfaces, potentially enabling new routes for wave-based devices with tailored radiative properties.

Friction is the essential mediator of terrestrial locomotion, yet in robotic systems it is almost always treated as a passive property fixed by surface materials and conditions. Here, we introduce ultrasonic lubrication as a method to actively control friction in robotic locomotion. By exciting resonant structures at ultrasonic frequencies, contact interfaces can dynamically switch between "grip" and "slip" states, enabling locomotion. We developed two friction control modules, a cylindrical design for lumen-like environments and a flat-plate design for external surfaces, and integrated them into bio-inspired systems modeled after inchworm and wasp ovipositor locomotion. Both systems achieved bidirectional locomotion with nearly perfect locomotion efficiencies that exceeded 90%. Friction characterization experiments further demonstrated substantial friction reduction across various surfaces, including rigid, soft, granular, and biological tissue interfaces, under dry and wet conditions, and on surfaces with different levels of roughness, confirming the broad applicability of ultrasonic lubrication to locomotion tasks. These findings establish ultrasonic lubrication as a viable active friction control mechanism for robotic locomotion, with the potential to reduce design complexity and improve efficiency of robotic locomotion systems.

Optical topologies in the form of Skyrmions have attracted significant interest of late, where their integer Skyrmion number has been shown to be robust to complex media. Here we create the first fractional Skyrmions by structuring light as a vectorial superposition of non-integer orbital angular momentum. We unravel the map structure to reveal a new phenomenon, the abrupt transition jumps in skyrmion number, which serves to reinforce the integer nature of skyrmion topologies. Our experimental demonstration agrees well with simulation, opening a new spectrum of optical topologies to explore, with exciting possibilities in optical communication and sensing.

Multiple scattering of waves in complex media can be harnessed and tailored for diverse phenomena in sound and light. Despite the tremendous progress enabled by technologies such as time-reversal propagation and wavefront shaping, the full control of transmission matrix remains a significant challenge. In this work, we propose a multi-scattering-based approach to design reflectionless complex media with a unitary transmission matrix of arbitrary structures. As such, the perfect transmission of waves through such a medium performs a unitary operation. Based on this principle, we experimentally demonstrated braiding of multiple waveguide modes in an acoustic waveguide via multiple scattering and showed non-Abelian characteristics arising from the concatenation of distinct complex media. Furthermore, we show that the principle can be extended for realizing arbitrary unitary operations beyond braiding. Our scheme uses generalized Wigner-Smith operators to design the optimal acoustic complex media with near-arbitrary targeted functionalities. The scheme is generally applicable beyond acoustics, with broad implications to other wave types. Our results demonstrate unprecedented control over multiple-scattering waves and are relevant to applications that require precise control over propagation, such as multiplexed communications, wave-based logic operations, and computations.

Transformation optics (TO) is a new method to design metamaterials that can manipulate electromagnetic fields. Inspired by the traditional TO techniques which is mostly based on the solid metamaterials with a limited range of tunability, a novel streamline tracing-based transformation optofluidics (STTOF) method is proposed to manipulate the light path by analytically designating the light-carrying streamlines of the flow in a two-dimensional circular bounded domain. A dipole flow model is built to analytically calculate the streamlines of the flow field inside the domain which allocates the optical/fluidic source and sink pairs at arbitrary positions. Liquid core/liquid cladding (L2) configuration is used in the experiment to trace the light via a specific streamline. Experimental results verify that the light paths agree well with the theoretical predictions, and demonstrate that a good range of tunability can be achieved by adjusting the flow rates and the source-sink positions of optical/fluidic source and sink pairs.

Tunable, low-loss, narrowband, and frequency-stabilized filters play a critical role in the realization of transceivers for efficient signal processing in telecommunications and sensing applications in the presence of strong interference and noise. We systematically design, fabricate, and demonstrate multi-cavity, multipole integrated photonic bandpass filters with gigahertz to sub-gigahertz bandwidths. The filters feature ultra-wideband tunable center frequencies exceeding 400 GHz, ultra-low insertion loss, steep roll-off, and compact footprints, making them suitable for radio-frequency, microwave, and millimeter-wave front-end applications. Using a set of design principles for multi-cavity high-order filters together with supporting nanofabrication techniques in a silicon nitride platform, the demonstrated filters achieve record-high figures of merit and advance the state of the art in integrated photonic filtering for radio-frequency, microwave, and millimeter-wave systems. The resonator structures, which are the key building blocks enabling the reported performance, combine a large free spectral range of approximately 70 GHz with a high quality factor of 2.1 times ten to the power of seven. This performance is enabled by combining wide multimode waveguide segments with narrow single-mode regions connected through adiabatic tapers. To the best of our knowledge, the experimentally demonstrated bandwidth of 520 MHz, tuning range of one free spectral range, filter insertion loss of 2 dB, and out-of-band rejection ratios of up to 55 dB represent record performance metrics for integrated photonic filtering of radio-frequency, millimeter-wave, and terahertz signals.

Germanium-tin (GeSn) photodiodes potentiate a viable solution to integrate SWIR and extended SWIR detection technology into CMOS processing line. However, challenges in the growth of thick, high quality GeSn limit the device absorber thickness, making it impossible to ascertain the performance limit of GeSn photodiodes. An in-depth understanding of their device physics and a clear optimization pathway towards commercial-grade devices remain elusive. This work presents a systematic empirical study of GeSn photodiodes with thick absorber (2 to 8% Sn content, up to 2630 nm thick), showing high responsivity up to 0.59 A.W-1 at 1.55 μm and 0.43 A.W-1 at 2 μm wavelengths, low dark current density down to 2 x 10-2 A.cm-2, and high detection cutoff wavelengths up to 2.1 and 2.5 μm at 5% and 8% Sn, respectively. Using specific doping design (P-i-N and N-i-P), an in-depth analysis is presented on the impact of junction position, p-type background carrier concentration, bulk/ surface defects and photocarrier diffusion length - on photodetection performance. Different optimization strategies for GeSn photodiodes, in particular at high Sn content, are proposed.

Anomalous local photocurrent generation via second-order nonlinear and thermoelectric responses is a signature of many topological semimetals. The emergence of these photocurrents is inherently linked to symmetry breaking and anisotropy of their crystal lattices. Studies of type-II Weyl semimetals of group C$_{2v}$ (WTe$_2$, MoTe$_2$, TaIrTe$_4$) have reported anomalous, nonlocal photocurrents localized to crystals edges or far from electrodes, which are highly dependent on the geometry of the material sample. While originally attributed to a nonlinear charge current response, it was recently shown that these currents could instead be attributed to the anisotropic Seebeck coefficients of the materials. Here, we confirm that anomalous photocurrents observed in TaIrTe$_4$ under either visible or far-infrared far-field illumination originate from the large transverse thermoelectric effect. We engineer the mutual orientation of crystal edges and electrodes as well as the thermal environment of TaIrTe$_4$ to control and amplify its spatial photocurrent response. We show that substrate engineering can locally enhance photocurrent. This framework of thermal device engineering can enable broadband photo detection schemes by leveraging spectral and spatial dependence of photocurrents for applications like wavefront sensing, beam positioning, and edge detection.

The accurate calculation of guided electromagnetic modes in optical nanofibres is critical for applications in nanophotonics, from quantum interfaces to vectorial light sensing. Standard textbook methods rely on solving a $4\times4$ matrix eigenvalue problem to find the modal fields. While widely used, this approach has a subtle but significant flaw: the final determination of the field amplitudes requires finding the numerical null space of a theoretically singular matrix, an ill-conditioned problem that introduces large relative errors in the small but physically crucial longitudinal field components. In this work, we introduce a fundamentally more robust and efficient semi-analytical method. By starting from the foundational symmetries of the cylindrical waveguide and employing a judicious normalisation of the field amplitudes, we demonstrate that the problem can be analytically reduced to a much simpler $2\times2$ system. This reformulation yields two decisive advantages: the dispersion relation is obtained numerically from a simple and well-behaved transcendental equation, and more importantly, the modal field amplitudes are subsequently determined \emph{analytically}. Our approach completely bypasses the numerical null space calculation, thereby ensuring the accuracy of the full vectorial field structure. This method provides a powerful and reliable tool for the design and analysis of nanofibre-based devices, particularly for applications in chiral quantum optics and nanophotonics where precise knowledge of field polarisation and specifically of the longitudinal components is paramount.

Lensless digital holographic microscopy (LDHM) relies on interference between an unscattered reference wave and a weakly scattered object wave - an assumption that rapidly fails in turbid samples under multiple scattering. To overcome this limitation, we present near-infrared LDHM (NIR-LDHM), a in-line holographic platform that operates up to the silicon cutoff (~1100 nm) using a conventional board-level CMOS sensor designed for visible (VIS) imaging. Using tissue-mimicking milk scattering layers and calibrated resolution targets, we quantify reconstruction performance versus wavelength, scattering strength, and sample-sensor distance. NIR-LDHM maintains resolvable features through scattering layers up to ~1.4 mm, whereas visible regime fails to resolve features below ~350 um. Importantly, despite a detector quantum efficiency of only 0.19% at 1100 nm, robust reconstructions are obtained under low-photon-budget conditions. We further identify advantageous mechanism by which increasing the sample-sensor distance from ~3 to 12 mm improves lateral resolution by twofold under strong scattering. Finally, we demonstrate wide-field, label-free amplitude-phase imaging of uncleared mouse tissues, resolving internal structure in brain and liver slices up to ~250 um and ~60 um, respectively. By extending lensless complex-field microscopy into strongly scattering regimes with minimal hardware changes, this work has relevance to computational imaging through complex media and biophotonics.

We report on a novel dual-structure scaffold fabricated using two-photon polymerisation (2PP), integrating electrically conductive and non-conductive regions within a single architecture, targeting neural tissue regeneration. Poly(ethylene glycol) diacrylate (PEGDA) was combined with 20 nm gold nanoparticles to create a conductive microstructure, encapsulated within a biocompatible PEGDA lattice designed to support neuronal growth possibility after neural injuries. Electrical resistance measurements confirmed the feasibility of integrating conductive pathways into precision-fabricated microarchitectures.

We present a two-well terahertz (THz) quantum cascade laser designed for high temperature operation based on the InGaAs/AlInAs material system. The lighter effective mass and higher energy barriers increase the gain at high temperatures (T > 150K). When processed in copper-based double metal waveguides the devices show laser action up to a maximum operating temperature of 188K with a maximum current density of 1.4kA/cm$^2$. The low Joule heating due to reduced active region thickness and low electrical bias allows operation at 10% duty cycle up to a temperature of 170K.

Dielectric waveguide (DWG) interconnects frequently utilize multimode waveguides due to their low dispersion in the fundamental mode. However, these links are more vulnerable to cross-modal coupling that significantly impacts their overall performance. This study presents a technique aimed at minimizing the coupling of energy into higher-order modes within weakly coupled rectangular dielectric waveguides that are excited by a linear taper. The approach involves wrapping the waveguide with a material of a higher dielectric constant than both the core and the cladding of the waveguide. The added material significantly improves the modal confinement factor of the fundamental mode to the core, leading to a much smaller coupling to the parasitic higher-order cladding modes. The new waveguide with the additional material cladding is analyzed, and semi-analytical approximate expressions are derived to predict the mode profiles and cutoffs. Design equations are given to choose the thickness of the wrapping material. The waveguide is fabricated and compared against a similar cross-section waveguide without the additional wrapping material. Unlike the unwrapped waveguide, the group delay (GD) results of the proposed (wrapped) waveguide closely match the EM-simulated GD of the fundamental mode, confirming the significant isolation of higher-order modes. The measured GD of the proposed DWG is 50 ps/m, while the expected fundamental mode GD from EM simulations is 35 ps/m. On the contrary, the unwrapped DWG shows a measured GD of 200 ps/m with significant oscillations, while the expected fundamental mode GD from EM simulations is 25 ps/m, demonstrating that wrapping the waveguide significantly improves the GD by reducing the higher-order mode propagation.

Extracellular matrix (ECM) constitutes a key basement structure to human organisms by acting as a complex network of large proteins and carbohydrates that provide structural support to surrounding cells. Remodeling in the extracellular matrix's structural fibers leads to insight into the development of diseases such as cancer, fibrosis and carcinoma. While standard tissues visualization in the ECM involves multiple lengthy histopathological staining protocols, Mueller matrix-based polarimetry provides label-free tissue slices' microstructural information and optical properties. This work aims to identify three types of fiber tissues commonly found in the ECM of gastrointestinal tissue specimens by analyzing their polarization properties. To address decomposition methods' reliance on restrictive hypotheses and inability with an individual polarization-based parameter to determine the nature of a given biological tissue; this study employs Uniform Manifold Approximation and Projection (UMAP) method to offer greater discriminative power and flexibility. Subsequently, polarization-based features will be extracted and compared between fiber regions statistically to discern potential diagnostic differences. By providing colorized images, this work aims to demonstrate the feasibility of distinguishing different fibers with polarization approach, offering insights for future clinical development while complementing existing staining methods for pathological tissue specimens.

Structured light beams with space-variant polarization can be efficiently generated using voltage-tunable nematic liquid-crystal (Q-plate). By appropriately selecting the input state and the retardation of the Q-plate, an optical field acquires a spatially structured polarization distribution that is capable of encoding non-trivial topological information across the beam profile. These features can be directly read out through interaction with plasmonic nano-structures, such as circular and spiral slits. Here we show that, upon illumination, polarization-dependent excitation of surface plasmons converts the hidden topology of the polarization structure into observable intensity distributions, including plasmonic vortices and characteristic interference patterns, while the tunability of the input parameters enables a rich variety of distinct topological forms.

The manipulation of optical properties, including reflection, refraction, polarization, phase, and frequency, has long been central to advancing photonic and optoelectronic technologies. However, existing electro-optical approaches rely on volatile mechanisms that require continuous power consumption. Here, we demonstrate strong, nonvolatile modulation of optical dispersion in monolayer tungsten disulfide (ML WS2) using patterned ferroelectric domains in aluminum scandium nitride (AlScN). By locally poling ferroelectric domains into opposite states, we achieve substantial manipulation of the complex refractive index (Delta n > 0.7, Delta k > 0.4) and excitonic energy shifts (~50 meV) in ML WS2, comparable to previous gate-tuning approaches while eliminating continuous power consumption. We introduce an asymmetric screening model that reveals how ferroelectric polarization induces carrier-density-dependent Coulomb screening, leading to distinct excitonic behaviors between electron- and hole-doped regions. Furthermore, we demonstrate a gate-free lateral p-n homojunction with a rectification ratio of 6e10^3, formed through spatial carrier redistribution. These findings establish ferroelectric/2D heterostructures as a powerful platform for nonvolatile optical dispersion engineering, enabling energy-efficient, reconfigurable photonic and optoelectronic devices.

A key question for most applications involving reconfigurable linear wave systems is how accurately a desired linear operator can be realized by configuring the system's tunable elements. The relevance of this question spans from hybrid-MIMO analog combiners via computational meta-imagers to programmable wave-domain signal processing. Yet, no electromagnetically consistent bounds have been derived for the fidelity with which a desired operator can be realized in a real-world reconfigurable wave system. Here, we derive such bounds based on an electromagnetically consistent multiport-network model (capturing mutual coupling between tunable elements) and accounting for real-world hardware constraints (lossy, 1-bit-programmable elements). Specifically, we formulate the operator-synthesis task as a quadratically constrained fractional-quadratic problem and compute rigorous fidelity upper bounds based on semidefinite relaxation. We apply our technique to three distinct experimental setups. The first two setups are, respectively, a free-space and a rich-scattering $4\times 4$ MIMO channel at 2.45 GHz parameterized by a reconfigurable intelligent surface (RIS) comprising 100 1-bit-programmable elements. The third setup is a $4\times 4$ MIMO channel at 19 GHz from four feeds of a dynamic metasurface antenna (DMA) to four users. We systematically study how the achievable fidelity scales with the number of tunable elements, and we probe the tightness of our bounds by trying to find optimized configurations approaching the bounds with standard discrete-optimization techniques. We observe a strong influence of the coupling strength between tunable elements on our fidelity bound. For the two RIS-based setups, our bound attests to insufficient wave-domain flexibility for the considered operator synthesis.

Spectral and spatial fringes in polarized light are produced by the interference of transmitted and reflected waves at the interface between materials with different indexes of refraction. These instrumental artifacts can affect the accuracy of optical designs conceived for high-sensitivity spectroscopy and polarimetry. We review the fundamental mechanisms that are responsible for these artifacts and the possible design pathways that allow us to mitigate them. In order to do so, we also present an approximate treatment of the problem of the transmission and reflection of light through (possibly absorptive) birefringent layers, relying on a few fundamental results that can be found in the already extensive literature on the subject. Unfortunately, many of these results remain the domain of a niche of investigators working in the field of thin films and optical coatings, and are often overlooked even by experienced designers of spectro-polarimetric instrumentation. The treatment presented in this work is limited to isotropic materials and uniaxial crystals, which are the most common types of optics employed in polarimetric instrumentation, and it fundamentally relies on the approximation of small birefringence for its implementation. An extensive set of modeling examples is provided to highlight the salient characteristics of polarization fringes, as well as to assess how approximate treatments such as this compare to exact but more computational expensive formulations of the problem such as Berreman's calculus.

Light detection and ranging (LiDAR) systems are pivotal for precise distance and velocity measurement, yet widespread deployment requires solutions that balance their performance, robustness, and simplicity. Here, we propose a novel chaotic LiDAR system based on a semiconductor vertical-cavity surface-emitting laser (VCSEL) with delayed orthogonal polarization feedback. By exploiting the intrinsic competition between the transverse electric (TE) and transverse magnetic (TM) modes, the system generates a polarization-multiplexed dynamics: a chaotic TM mode serves as the reference, while a feedback-modulated TE mode probes the target. This all-in-one source eliminates the need for external optical modulators or complex coherent detection. The system's dynamics is finely tunable via a half-wave ($λ$/2) plate in the feedback loop and the laser injection current, enabling real-time optimization of the cross-correlation signal-to-noise ratio. Experimental results demonstrate precise linear ranging with a resolution of approximately 1.2 cm. Furthermore, the system exhibits strong inherent resistance to external optical interference, maintaining accurate ranging even in the presence of a secondary laser source. This compact, tunable, and interference-resilient platform offers a promising pathway toward low-cost, high-performance LiDAR for applications in autonomous navigation, robotics, and industrial metrology.

Moiré patterns have recently become a very active field in nanophotonics. Those structures exhibit novel photonic properties unattainable with traditional photonic crystals. Especially, moiré magic configurations have been shown to allow intriguing slow light modes with zero group velocity. Starting from macroscopic moiré patterns in the everyday life, we will then shift to the subwavelength scale of moiré photonic crystals and detail some of their unusual properties.

The optical frequency comb has made a significant impact in precision spectroscopy and on our ability to probe atomic, molecular and, recently, nuclear transitions to further our understanding of their fundamental properties and how their dynamics and complex interactions affect the observed world. To expand the energy scales and types of systems that can be studied, frequency comb sources from terahertz to vacuum ultraviolet frequencies and beyond have been pursued. Dual-comb spectroscopy, enabled by the development of these frequency comb sources, allows broadband absorption measurements of complicated spectra, exceeding the limitations of direct, single-comb spectroscopy. To date, however, the dual-comb approach has not been able to directly access many important transitions that lie at challenging vacuum ultraviolet wavelengths. Here, we demonstrate dual-comb spectroscopy in the vacuum ultraviolet utilizing intracavity high harmonic generation. This multi-harmonic source is used to measure molecular absorbance spectra at $λ=210$~nm and $λ=149$~nm from room-temperature samples of acetylene and ammonia, respectively. These measurements resolve the Doppler broadened structure of congested molecular spectra with absolute frequency accuracy. Noise contributions to the vacuum ultraviolet dual-comb spectroscopy measurements are characterized, guiding future efforts and technological development in this region.

Compact Free-Electron Lasers (FELs) offering broad, continuous spectral tunability are traditionally constrained by fixed-parameter magnetic structures and the necessity for high-energy electron beams. High-gain Harmonic Lasing (HL) has long been proposed as a solution to overcome these limitations; however, a robust experimental verification of this principle has remained absent. Here, we report the first experimental demonstration of high-gain HL. By employing a frequency-tunable electron beam density modulation to dominate the fundamental instability, we achieved sustained FEL amplification at the 3rd and 5th harmonics of the wiggler. The HL mode generated output power comparable to conventional fundamental operation with enhanced stability and narrower spectral bandwidth. Notably, we demonstrate that HL extends the spectral coverage by a factor of two under fixed facility constraints, achieving pulse energies up to 540 μJ. These results establish high-gain HL as a versatile mechanism for advancing compact, wavelength-flexible FEL facilities.

Developing vortex nanolasers is highly desirable for on-chip multidimensional large-capacity information processing. Topological optical modes hold great promise for achieving coherent emission with diverse functionalities. However, the development of robust and ultracompact topological charge lasing operation remains insufficiently explored. Here, we theoretically propose a translationally deformed topological charge vortex nanocavity with a low mode volume of 0.32 $(λ/n)^3$, and experimentally demonstrate the corresponding lasing emission with a low lasing threshold of around 0.74 $μ$W. The designed topological nanocavity, constructed by translationally deformed photonic crystals, supports an ultracompact optical mode carrying a topological charge characterized by polarization winding. The well-defined topological charge characteristics of the fabricated device are revealed in both near- and far-field polarization-resolved optical profiles. Our work opens a promising avenue for versatile topological photonic integration and gives new potential for exploring intriguing structured light-matter interactions under the topological photonics scenario.

We report a high-energy, bidirectional, dissipative soliton mode-locked fiber laser operating in the 1550 nm normal-dispersion regime. By leveraging intracavity dispersion management and a Lyot filtering mechanism, the laser achieves flat-top optical spectra with 10-dB bandwidths exceeding 20 nm in both directions. Single-pulse energies of 2.7 nJ and 1.5 nJ are achieved for the clockwise and counter-clockwise directions, respectively. Furthermore, the all-fiber configuration exhibits superior noise performance and inherent common-mode noise suppression. To facilitate broadband and high-speed dual-comb spectroscopy, we employ a dynamic repetition rate difference control technique via pump power modulation, enabling zero-crossing dynamic scanning. This approach achieves a spectral measurement bandwidth of approximately 16 nm at an acquisition rate of 500 Hz. Compared to static operation, this represents a nearly two-order-of-magnitude improvement in acquisition speed and achieves a fivefold measurement bandwidth beyond the Nyquist aliasing limit. Experimental results demonstrate that the system maintains robust coherence even under dynamic modulation. By implementing a phase-correction algorithm, a mutual coherence time of 0.5 s is successfully achieved, yielding a spectral resolution exceeding 7.2 GHz. This work fills a gap in high-energy dissipative soliton dual-comb sources at 1550 nm and provides an ideal solution for low-cost, high-sensitivity dual-comb spectroscopy requiring both broad bandwidth and high speed.

Ising machines are emerging as a powerful physical alternative to digital processors for solving combinatorial optimization problems. Among them, spatial photonic Ising machines (SPIMs) offer compact, room-temperature hardware with inherently parallel, energy efficient, single-shot optical evaluation of the Ising Hamiltonian. However, scalability has been fundamentally limited by optical aberrations and non-uniform illumination, which corrupt phase-based spin encoding and distort coupling representation, forcing operation to a restricted spatial light modulator (SLM) region. Here we introduce a high-precision full-aperture calibration scheme that overcomes these constraints. By implementing wavefront retrieval and correction with < λ/40 accuracy, we restore faithful phase encoding across the entire SLM area. Furthermore, we introduce a novel interaction-normalization method, which compensates for amplitude curvature and enables uniform coupling representation. Together, these advances establish a high-fidelity, full-area SPIM architecture that unlocks true scalability and enables larger and more reliable photonic Ising computations.

Radiative defects in hexagonal boron nitride (hBN) are active in a broad spectral range from deep ultraviolet to near-infrared wavelengths. Representatives of these defects act as bright single photon sources, spin-1 systems, and multiproperty atomic-scale sensors. They are predominantly investigated in bulk hBN films, where defects are decoupled from surface and interfacial effects. Here, we demonstrate a novel class of surface defects optically active in the green/yellow visible spectral range, which exhibit photophysical properties distinct from their bulk counterparts. High-power resonant laser illumination quenched the emission from the ensemble of such defects, which was attributed to a light-driven structural reconfiguration. The quenched defects were found to recover their emissive capabilities via a thermal cycling process, revealing an activation energy of 24.5 meV for the structural transition. Alternatively, permanent quenching of the defects was triggered by surface chemistry, involving lithiation-enabled attachment of functional groups. These mechanisms were utilized to realize negative-contrast direct laser writing, designing arbitrary geometric emissive patterns on demand in a microscopic configuration. The surface-active radiative centers in hBN appear particularly attractive for exploring environmental sensitivity, surface science, and coupling to photonic structures or electronic devices by taking unique advantage of the two-dimensional characteristics of the host lattice.

Background: Human behavior shapes infectious disease dynamics, yet its integration into transmission models remains fragmented. Recent epidemics, particularly COVID-19, highlight the need for models capturing adaptation to perceived risk, social influence, and policy signals. This review synthesizes post-2020 models incorporating behavioral adaptation, examines their theoretical grounding, and evaluates how behavioral constructs modify transmission, vaccination, and compliance. Methods: Following PRISMA guidelines, we searched Scopus and PubMed (2020-2025), screening 1,274 records with citation chaining. We extracted data on disease context, country, modeling framework, behavioral mechanisms (prevalence-dependent, policy/media, imitation/social learning), and psychosocial constructs (personal threat, coping appraisal, barriers, social norms, cues to action). A total of 216 studies met inclusion criteria. Results: COVID-19 accounted for 73% of studies. Most used compartmental ODE models (81%) and focused on theoretical or U.S. settings. Behavioral change was mainly reactive: 47% applied prevalence-dependent feedback, 25% included awareness/media dynamics, and 19% relied on exogenous policy triggers. Game-theoretic or social learning approaches were rare (less or equal than 5%). Behavioral effects primarily modified contact or transmission rates (91%). Psychosocial constructs were unevenly represented: cues to action (n=159) and personal threat (n=145) dominated, whereas coping appraisal (n=82), barriers (n=36), and social norms (n=25) were less common. Conclusions: We propose a taxonomy structured by behavioral drivers, social scale, and memory to clarify dominant paradigms and their empirical basis. Mapping models to psychosocial constructs provides guidance for more theory-informed and data grounded-integration of behavioral processes in epidemiological modeling.

There is growing evidence that independently trained AI systems come to represent the world in the same way. In other words, independently trained embeddings from text, vision, audio, and neural signals share an underlying geometry. We call this the Representational Alignment Hypothesis (RAH) and investigate evidence for and consequences of this claim. The evidence is of two kinds: (i) internal structure comparison techniques, such as representational similarity analysis and topological data analysis, reveal matching relational patterns across modalities without explicit mapping; and (ii) methods based on cross-modal embedding alignment, which learn mappings between representation spaces, show that simple linear transformations can bring different embedding spaces into close correspondence, suggesting near-isomorphism. Taken together, the evidence suggests that, even after controlling for trivial commonalities inherent in standard data preprocessing and embedding procedures, a robust structural correspondence persists, hinting at an underlying organizational principle. Some have argued that this result shows that the shared structure is getting at a fundamental, Platonic level of reality. We argue that this conclusion is unjustified. Moreover, we aim to give the idea an alternative philosophical home, rooted in contemporary metasemantics (i.e., theories of what makes a representation and what makes something meaningful) and responses to the symbol grounding problem. We conclude by considering the scope of the RAH and proposing new ways of distinguishing semantic structures that are genuinely invariant from those that inevitably arise due to the fact that all our data is generated under human-specific conditions on Earth.

The vast majority of biological sequences encode unknown functions and bear little resemblance to experimentally characterized proteins, limiting both our understanding of biology and our ability to harness functional potential for the bioeconomy. Predicting enzyme function from sequence remains a central challenge in computational biology, complicated by low sequence diversity and imbalanced label support in publicly available datasets. Models trained on these data can overestimate performance and fail to generalize. To address this, we introduce GRIMM (Genetic stRatification for Inference in Molecular Modeling), a benchmark for enzyme function prediction that employs genetic stratification: sequences are clustered by similarity and clusters are assigned exclusively to training, validation, or test sets. This ensures that sequences from the same cluster do not appear in multiple partitions. GRIMM produces multiple test sets: a closed-set test with the same label distribution as training (Test-1) and an open-set test containing novel labels (Test-2), serving as a realistic out-of-distribution proxy for discovering novel enzyme functions. While demonstrated on enzymes, this approach is generalizable to any sequence-based classification task where inputs can be clustered by similarity. By formalizing a splitting strategy often used implicitly, GRIMM provides a unified and reproducible framework for closed- and open-set evaluation. The method is lightweight, requiring only sequence clustering and label annotations, and can be adapted to different similarity thresholds, data scales, and biological tasks. GRIMM enables more realistic evaluation of functional prediction models on both familiar and unseen classes and establishes a benchmark that more faithfully assesses model performance and generalizability.

Gene Regulatory Networks(GRNs) with feedback are essential components of many cellular processes and may exhibit oscillatory behavior. Analyzing such systems becomes increasingly complex as the number of components increases. Since gene regulation often involves a small number of molecules, fluctuations are inevitable. Therefore, it is important to understand how fluctuations affect the oscillatory dynamics of cellular processes, as this will allow comprehension of the mechanisms that enable cellular functions to remain even in the presence of fluctuations or, failing that, to determine the limit of fluctuations that permits various cellular functions. In this study, we investigated the conditions under which GRNs with feedback and intrinsic fluctuations exhibit oscillatory behavior. Our focus was on developing a procedure that would be both manageable and practical, even for extensive regulatory networks, that is, those comprising numerous nodes. Using the second-moment approach, we described the stochastic dynamics through a set of ordinary differential equations for the mean concentration and its second central moment. The system can attain either a stable equilibrium or oscillatory behavior, depending on its scale and, consequently, the intensity of fluctuations. To illustrate the procedure, we analyzed two relevant systems: a repressilator with three nodes and a system with five nodes, both incorporating intrinsic fluctuations. In both cases, it was observed that for very small systems, which therefore exhibit significant fluctuations, oscillatory behavior is inhibited. The procedure presented here for analyzing the stability of oscillations under fluctuations enables the determination of the critical minimum size of GRNs at which intrinsic fluctuations do not eliminate their cyclical behavior.

The loss of biodiversity due to the likely widespread extinction of species in the near future is a focus of current concern in conservation biology. One approach to measure the impact of this extinction is based on the predicted loss of phylogenetic diversity. These predictions have become a focus of the Zoological Society of London's 'EDGE2' program for quantifying biodiversity loss and involves considering the HED (heightened evolutionary distinctiveness) and HEDGE (heightened evolutionary distinctiveness and globally endangered) indices. Here, we show how to generalise the HED(GE) indices by expanding their application to more general settings (to phylogenetic networks, to feature diversity on discrete traits, and to arbitrary biodiversity measures). We provide a simple and explicit description of the mean and variance of such measures, and illustrate our results by an application to the phylogeny of all 27 extant Crocodilians. We also derive various equalities for feature diversity, and an inequality if species extinction rates are correlated with feature types.

The rates at which individuals memorize and forget environmental information strongly influence their movement paths and long-term space use. To understand how these cognitive time scales shape population-level patterns, we propose and analyze a nonlocal population model with a cognitive map. The population density moves by a Fokker--Planck type diffusion driven by a cognitive map that stores a habitat quality information nonlocally. The map is updated through local presence with learning and forgetting rates, and we consider both truncated and normalized perception kernels. We first study the movement-only system without growth. We show that finite perceptual range generates spatial heterogeneity in the cognitive map even in nearly homogeneous habitats, and we prove a lingering phenomenon on unimodal landscapes: for the fixed high learning rate, the peak density near the best location is maximized at an intermediate forgetting rate. We then couple cognitive diffusion to logistic growth. We establish local well-posedness and persistence by proving instability of the extinction equilibrium and the existence of a positive steady state, with uniqueness under an explicit condition on the motility function. Numerical simulations show that lingering persists under logistic growth and reveal a trade-off between the lingering and total population size, since near the strongest-lingering regime the total mass can fall below the total resource, in contrast to classical random diffusive--logistic models.

Understanding the dynamic nature of brain connectivity is critical for elucidating neural processing, behavior, and brain disorders. Traditional approaches such as sliding-window correlation (SWC) characterize time-varying undirected associations but do not resolve directional interactions, limiting inference about time-resolved information flow in brain networks. We introduce sliding-window prediction correlation (SWpC), which embeds a directional linear time-invariant (LTI) model within each sliding window to estimate time-varying directed functional connectivity (FC). SWpC yields two complementary descriptors of directed interactions: a strength measure (prediction correlation) and a duration measure (window-wise duration of information transfer). Using concurrent local field potential (LFP) and fMRI BOLD recordings from rat somatosensory cortices, we demonstrate stable directionality estimates in both LFP band-limited power and BOLD. Using Human Connectome Project (HCP) motor task fMRI, SWpC detects significant task-evoked changes in directed FC strength and duration and shows higher sensitivity than SWC for identifying task-evoked connectivity differences. Finally, in post-concussion vestibular dysfunction (PCVD), SWpC reveals reproducible vestibular-multisensory brain-state shifts and improves healthy-control vs subacute patient (HC-ST) discrimination using state-derived features. Together, these results show that SWpC provides biologically interpretable, time-resolved directed connectivity patterns across multimodal validation and clinical application settings, supporting both basic and translational neuroscience.

In symmetric private information retrieval (SPIR), a user communicates with multiple servers to retrieve from them a message in a database, while not revealing the message index to any individual server (user privacy), and learning no additional information about the database (database privacy). We study the problem of SPIR on graph-replicated database systems, where each node of the graph represents a server and each link represents a message. Each message is replicated at exactly two servers; those at which the link representing the message is incident. To ensure database privacy, the servers share a set of common randomness, independent of the database and the user's desired message index. We study two cases of common randomness distribution to the servers: i) graph-replicated common randomness, and ii) fully-replicated common randomness. Given a graph-replicated database system, in i), we assign one randomness variable independently to every pair of servers sharing a message, while in ii), we assign an identical set of randomness variable to all servers, irrespective of the underlying graph. In both settings, our goal is to characterize the SPIR capacity, i.e., the maximum number of desired message symbols retrieved per downloaded symbol, and quantify the minimum amount of common randomness required to achieve the capacity. To this goal, in setting i), we derive a general lower bound on the SPIR capacity, and show it to be tight for path and regular graphs through a matching converse. Moreover, we establish that the minimum size of common randomness required for SPIR is equal to the message size. In setting ii), the SPIR capacity improves over the first, more restrictive setting. We show this through capacity lower bounds for a class of graphs, by constructing SPIR schemes from PIR schemes.

We present the RIS-VSign system, an active reconfigurable intelligent surface (RIS)-assisted multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) framework for vital signs extraction under an integrated sensing and communication (ISAC) model. The system consists of two stages: the phase selector of RIS and the extraction of respiration rate. To mitigate synchronization-induced common phase drifts, the difference of Möbius transformation (DMT) is integrated into the deep learning framework, named DMTNet, to jointly configure multiple active RIS elements. Notably, the training data are generated in simulation without collecting real-world measurements, and the resulting phase selector is validated experimentally. For sensing, multi-antenna measurements are fused by the DC-offset calibration and the DeepMining-MMV processing with CA-CFAR detection and Newton's refinements. Prototype experiments indicate that active RIS deployment improves respiration detectability while simultaneously enabling higher-order modulation; without RIS, respiration detection is unreliable and only lower-order modulation is supported.

This paper presents a Failure-Aware Access Point Selection (FAAS) method aimed at improving hardware resilience in cell-free massive MIMO (CF-mMIMO) networks. FAAS selects APs for each user by jointly considering channel strength and the failure probability of each AP. A tunable parameter \(α\in [0,1]\) scales these failure probabilities to model different levels of network stress. We evaluate resilience using two key metrics: the minimum-user spectral efficiency, which captures worst-case user performance, and the outage probability, defined as the fraction of users left without any active APs. Simulation results show that FAAS maintains significantly better performance under failure conditions compared to failure-agnostic clustering. At high failure levels, FAAS reduces outage by over 85\% and improves worst-case user rates. These results confirm that FAAS is a practical and efficient solution for building more reliable CF-mMIMO networks.

In this letter, we develop a continuous fluid antenna (FA) framework for uplink channel estimation in cell-free massive multiple-input and multiple-output (CF-mMIMO) systems. By modeling the wireless channel as a spatially correlated Gaussian random field, channel estimation is formulated as a Gaussian process (GP) regression problem with motion-constrained spatial sampling. Closed-form expressions for the linear minimum mean squared error (LMMSE) estimator and the corresponding estimation error are derived. A fundamental comparison with discrete port-based architectures is established under identical position constraints, showing that continuous FA sampling achieves equal or lower estimation error for any finite pilot budget, with strict improvement for non-degenerate spatial correlation models. Numerical results validate the analysis and show the performance gains of continuous FA sampling over discrete baselines.

Ambient backscatter communication (AmBC) enables ultra-low-power connectivity by allowing passive backscatter devices (BDs) to convey information through reflection of ambient signals. However, the cascaded AmBC channel suffers from severe double path loss and multiplicative fading, while accurate channel state information (CSI) acquisition is highly challenging due to the weak backscattered signal and the resource-limited nature of BDs. To address these challenges, this paper considers an AmBC system in which the reader is equipped with a pixel-based fluid antenna system (FAS). By dynamically selecting one antenna position from a dense set of pixels within a compact aperture, the FAS-enabled reader exploits spatial diversity through measurement-driven port selection, without requiring explicit CSI acquisition or multiple RF chains. The intrinsic rate-energy tradeoff at the BD is also incorporated by jointly optimizing the backscatter modulation coefficient under an energy harvesting (EH) neutrality constraint. To efficiently solve this problem, a particle swarm optimization (PSO)-based framework is developed to jointly determine the FAS port selection and modulation coefficient on an optimize-then-average (OTA) basis. Simulation results show that the proposed scheme significantly improves the achievable rate compared with conventional single-antenna readers, with gains preserved under imperfect observations, stringent EH constraints, and different pixel spacings.

Channel measurements in MIMO systems hinge on precise synchronization. While methods for time and frequency synchronization are well established, maintaining real-time phase coherence remains an open requirement for many MIMO systems. Phase coherence in MIMO systems is crucial for beamforming in digital arrays and enables precise parameter estimates such as Angle-of-Arrival/Departure. This work presents and validates a simple local real-time phase calibration method for a digital array. We compare two different approaches, instantaneous and smoothed calibration, to determine the optimal interval between synchronization procedures. To quantitatively assess calibration performance, we use two metrics: the average beamforming power loss and the RMS cycle-to-cycle jitter. Our results indicate that both approaches for phase calibration are effective and yield RMS of jitter in the 2.1 ps to 124 fs range for different SDR models. This level of precision enables coherent transmission on commonly available SDR platforms, allowing investigation on advanced MIMO techniques and transmit beamforming in practical testbeds.

Modulo sampling is a promising technology to preserve amplitude information that exceeds the observable range of analog-to-digital converters during the digitization of analog signals. Since conventional methods typically reconstruct the original signal by estimating the differences of the residual signal and computing their cumulative sum, each estimation error inevitably propagates through subsequent time samples. In this paper, to eliminate this error-propagation problem, we propose an algorithm that reconstructs the residual signal directly. The proposed method takes advantage of the high-frequency characteristics of the modulo samples and the sparsity of both the residual signal and its difference. Simulation results show that the proposed method reconstructs the original signal more accurately than a conventional method based on the differences of the residual signal.

This paper investigates the design of integrated sensing and communication (ISAC) systems assisted by simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs), which act as multi-functional programmable metasurfaces capable of supporting concurrent communication and sensing within a unified architecture. We propose a two-stage ISAC protocol, in which the preparation phase performs direction estimation for outdoor users located in the reflection space, while maintaining communication with both outdoor and indoor users in the transmission space. The subsequent communication phase exploits the estimated directions to enhance information transfer. The directions of outdoor users are modeled as Gaussian random variables to capture estimation uncertainty, and the corresponding average communication performance is incorporated into the design. Building on this framework, we formulate a performance-balanced optimization problem that maximizes the communication sum-rate while guaranteeing the required sensing accuracy, jointly determining the beamforming vectors at the base station (BS), the STAR-RIS transmission and reflection coefficients, and the metasurface partition between energy-splitting and transmit-only modes. The physical constraints of STAR-RIS elements and the required sensing performance are explicitly enforced. To address the non-convex nature of the problem, we combine fractional programming, Lagrangian dual reformulation, and successive convex approximation. The binary metasurface partition is ultimately recovered via continuous relaxation followed by projection-based binarization. Numerical results demonstrate that the proposed design achieves an effective trade-off between sensing accuracy and communication throughput, by significantly outperforming conventional STAR-RIS-aided ISAC schemes.

Hybrid received signal strength (RSS)-angle of arrival (AoA)-based positioning offers low-cost distance estimation and high-resolution angular measurements. Still, it comes at a cost of inherent nonlinearities, geometry-dependent noise, and suboptimal weighting in conventional linear estimators that might limit accuracy. In this paper, we propose a neural network-based approach using a multilayer perceptron (MLP) to directly map RSS-AoA measurements to 3D positions, capturing nonlinear relationships that are difficult to model with traditional methods. We evaluate the impact of input representation by comparing networks trained on raw measurements versus preprocessed features derived from a linearization method. Simulation results show that the learning-based approach consistently outperforms existing linear methods under RSS noise across all noise levels, and matches or surpasses state-of-the-art performance under increasing AoA noise. Furthermore, preprocessing measurements using the linearization method provides a clear advantage over raw data, demonstrating the benefit of geometry-aware feature extraction.

This paper investigates whether large-scale GNSS spoofing activity can be inferred from maritime Automatic Identification System (AIS) position reports. A data-processing framework, called SeaSpoofFinder, available here: seaspooffinder.github.io/ais_data, was developed to ingest and post-process global AIS streams and to detect candidate anomalies through a two-stage procedure. In Stage 1, implausible position jumps are identified using kinematic and data-quality filters; in Stage 2, events are retained only when multiple vessels exhibit spatially consistent source and target clustering, thereby reducing false positives from single-vessel artifacts. The resulting final potential spoofing events (FPSEs) reveal recurrent patterns in several regions, including the Baltic Sea, the Black Sea, Murmansk, Moscow, and the Haifa area, with affected footprints that can span large maritime areas. The analysis also highlights recurring non-spoofing artifacts (e.g., back-to-port jumps and data gaps) that can still pass heuristic filters in dense traffic regions. These results indicate that AIS-based monitoring can provide useful evidence for identifying and characterizing potential spoofing activity at scale, while emphasizing that AIS-only evidence does not provide definitive attribution.

A pinching antennas (PAs)-aided integrated sensing and multicast communication framework is proposed. In this framework, the communication performance is measured by the multicast rate considering max-min fairness. Moreover, the sensing performance is quantified by the Bayesian Cramér-Rao bound (BCRB), where a Gauss-Hermite quadrature-based approach is proposed to compute the Bayesian Fisher information matrix. Based on these metrics, PA placement is optimized under three criteria: communications-centric (C-C), sensing-centric (S-C), and Pareto-optimal designs. These designs are investigated in two scenarios: the single-PA case and the multi-PA case. 1) For the single-PA case, a closed-form solution is derived for the location of the C-C transmit PA, while the S-C design yields optimal transmit and receive PA placements that are symmetric about the target location. Leveraging this geometric insight, the Pareto-optimal design is solved by enforcing this PA placement symmetry, thereby reducing the joint transmit and receive PA placement to the transmit PA optimization. 2) For the general multi-PA case, the PA placements constitute a highly non-convex optimization problem. To solve this, an element-wise alternating optimization-based method is proposed to sequentially optimize all PA placements for the S-C design, and is further incorporated into an augmented Lagrangian (AL) framework and a rate-profile formulation to solve the C-C and Pareto-optimal design problems, respectively. Numerical results show that: i) PASS substantially outperforms fixed-antenna baselines in both multicast rate and sensing accuracy; ii) the multicasting gain becomes more pronounced as the user density increases; and iii) the sensing accuracy improves with the number of deployed PAs.

Vibration from an erroneous disturbance harms the manufactured components and lowers the output quality of an FDM printer. For moving machinery, vibration analysis and control are crucial. Additive manufacturing is the basis of 3D printing, which utilizes mechanical movement of the extruder to fabricate objects, and faults occur due to unwanted vibrations. Therefore, it is vital to examine the vibration patterns of a 3D printer. In this work, we observe these parameters of an FDM printer, exemplified by the MakerBot Method X. To analyze the system, it is necessary to understand the motion it generates and select appropriate sensors to detect those motions. The sensor measurement values can be used to determine the condition of the printer. We used an accelerometer and an acoustic sensor to measure the vibration and sound produced by the printer. The outputs from these sensors were examined individually. The findings show that vibration occurs at relatively low levels during continuous motion because it mainly appears at component transition edges. Due to abrupt acceleration and deceleration during zigzag motion, vibration reaches its peak.

The reliability and quality of 3D printing processes are critically dependent on the timely detection of mechanical faults. Traditional monitoring methods often rely on visual inspection and hardware sensors, which can be both costly and limited in scope. This paper explores a scalable and contactless method for the use of real-time audio signal analysis for detecting mechanical faults in 3D printers. By capturing and classifying acoustic emissions during the printing process, we aim to identify common faults such as nozzle clogging, filament breakage, pully skipping and various other mechanical faults. Utilizing Convolutional neural networks, we implement algorithms capable of real-time audio classification to detect these faults promptly. Our methodology involves conducting a series of controlled experiments to gather audio data, followed by the application of advanced machine learning models for fault detection. Additionally, we review existing literature on audio-based fault detection in manufacturing and 3D printing to contextualize our research within the broader field. Preliminary results demonstrate that audio signals, when analyzed with machine learning techniques, provide a reliable and cost-effective means of enhancing real-time fault detection.

Additive manufacturing, particularly fused deposition modeling, is transforming modern production by enabling rapid prototyping and complex part fabrication. However, its layer-by-layer process remains vulnerable to faults such as nozzle clogging, filament runout, and layer misalignment, which compromise print quality and reliability. Traditional inspection methods are costly, time-intensive, and often limited to post-process analysis, making them unsuitable for real-time intervention. In this current study, the authors developed a novel, low-cost, and portable faultdetection system that leverages multimodal sensor fusion and artificial intelligence for real-time monitoring in FDM-based 3D printing. The system integrates acoustic, vibration, and thermal sensing into a non-intrusive architecture, capturing complementary data streams that reflect both mechanical and process-related anomalies. Acoustic and thermal sensors operate in a fully contactless manner, while the vibration sensor requires minimal attachment such that it will not interfere with printer hardware, thereby preserving portability and ease of deployment. The multimodal signals are processed into spectrograms and time-frequency features, which are classified using convolutional neural networks for intelligent fault detection. The proposed system advances Industry 4.0 objectives by offering an affordable, scalable, and practical monitoring solution that improves faultdetection accuracy, reduces waste, and supports sustainable, adaptive manufacturing.

Reconfigurable intelligent surfaces (RISs) offer programmable control of radio propagation for future wireless systems. For configuration, geometry-driven analytical approaches are appealing for their simplicity and real-time operation, but their performance in challenging environments such as industrial halls with dense multipath and metallic scattering is not well established. To this end, we present a measurement-based evaluation of geometry-driven RIS beam steering in a large industrial hall using a 5 GHz RIS prototype. A novel RIS configuration is proposed in which four patch antennas are mounted in close proximity in front of the RIS to steer the incident field and enable controlled reflection. For this setup, analytically computed, quantized configurations are implemented. Two-dimensional received power maps from two measurement areas reveal consistent, spatially selective focusing. Configurations optimized near the receiver produce clear power maxima, while steering to offset locations triggers a rapid 20-30 dB reduction. With increasing RIS-receiver distance, elevation selectivity broadens due to finite-aperture and geometric constraints, while azimuth steering remains robust. These results confirm the practical viability of geometry-driven RIS beam steering in industrial environments and support its use for spatial field control and localization under non-ideal propagation.

Integrating artificial intelligence (AI) into wireless channel modeling requires large, accurate, and physically consistent datasets derived from real measurements. Such datasets are essential for training and validating models that learn spatio-temporal channel behavior across frequencies and environments. NYUSIM, introduced by NYU WIRELESS in 2016, generates realistic spatio-temporal channel data using extensive outdoor and indoor measurements between 28 and 142 GHz. To improve scalability and support 6G research, we migrated the complete NYUSIM framework from MATLAB to Python, and are incorporating new statistical model generation capabilities from extensive field measurements in the new 6G upper mid-band spectrum at 6.75 GHz (FR1(C)) and 16.95 GHz (FR3) [1]. The NYUSIM Python also incorporates a 3D antenna data format, referred to as Ant3D, which is a standardized, full-sphere format for defining canonical, commercial, or measured antenna patterns for any statistical or site-specific ray tracing modeling tool. Migration from MATLAB to Python was rigorously validated through Kolmogorov-Smirnov (K-S) tests, moment analysis, and end-to-end testing with unified randomness control, confirming statistical consistency and reproduction of spatio-temporal channel statistics, including spatial consistency with the open-source MATLAB NYUSIM v4.0 implementation. The NYUSIM Python version is designed to integrate with modern AI workflows and enable large-scale parallel data generation, establishing a robust, verified, and extensible foundation for future AI-enabled channel modeling.

It is known that waves generated by ambient noise sources and recorded by passive receivers can be used to image the reflectivities of an unknown medium. However, reconstructing the reflectivity of the medium from partial boundary measurements remains a challenging problem, particularly when the background wave speed is unknown. In this paper, we investigate passive correlation-based imaging in the daylight configuration, where uncontrolled noise sources illuminate the medium and only ambient fields are recorded by a sensor array. We first analyze daylight migration for a point reflector embedded in a homogeneous background. By introducing a searching wave speed into the migration functional, we derive an explicit characterization of the deterministic shift and defocusing effects induced by wave-speed mismatch. We show that the maximum of the envelope of the resulting functional provides a reliable estimator of the true wave speed. We then extend the analysis to a random medium with correlation length smaller than the wavelength. Leveraging the shift formula obtained in the homogeneous case, we introduce a virtual guide star that remains fixed under migration with different searching speeds. This property enables an effective wave-speed estimation strategy based on spatial averaging around the virtual guide star. For both homogeneous and random media, we establish resolution analyses for the proposed wave-speed estimators. Numerical experiments are conducted to validate the theoretical result.

In this paper we consider physics-informed detection of terrain material change in radar imagery (e.g., shifts in permittivity, roughness or moisture). We propose a lightweight electromagnetic (EM) forward model to simulate bi-temporal single-look complex (SLC) images from labelled material maps. On these data, we derive physics-aware feature stacks that include interferometric coherence, and evaluate unsupervised detectors: Reed-Xiaoli (RX)/Local-RX with robust scatter (Tyler's M-estimator), Coherent Change Detection (CCD), and a compact convolutional auto-encoder. Monte Carlo experiments sweep dielectric/roughness/moisture changes, number of looks and clutter regimes (gamma vs K-family) at fixed probability of false alarm. Results on synthetic but physically grounded scenes show that coherence and robust covariance markedly improve anomaly detection of material changes; a simple score-level fusion achieves the best F1 in heavy-tailed clutter.

Reliable detection and tracking in active sonar require accurate and efficient learning of the acoustic multipath background environment. Conventionally, background learning is performed after transforming measurements into the range-Doppler domain, a step that is computationally expensive and can obscure phase-coherent structure useful for monitoring and tracking. This paper proposes a framework for learning and tracking the multipath background directly in the raw measurement domain. Starting from a wideband Doppler linearization of the impulse response of a time-varying multipath channel, a state-space model with a heteroscedastic measurement equation is derived. This model enables channel tracking using an extended Kalman filter (EKF), and unknown model parameters are learned from the marginalized likelihood. The statistical adequacy of the proposed models is assessed via a p-value significance test. Finally, this paper integrates the learned channel model into a sequential likelihood-ratio test for target detection. BELLHOP-based simulations show that the proposed model better captures channel dynamics induced by sea-surface fluctuations and transmitter and receiver drift, yielding more reliable detection in time-varying shallow-water environments

We study joint transmit-waveform and receive-filter design for a multi-user downlink integrated sensing and communication (ISAC) system under practical constant-modulus and similarity constraints. We cast the design as a unified multi-objective program that balances communication sum rate and sensing signal-to-interference-plus-noise ratio (SINR). To address this, we introduce an efficient algorithm that use consensus alternating direction method of multipliers (ADMM) framework to alternately update the transmit waveform and radar filter. The proposed method effectively handles the non-convex fractional sensing's SINR formulation and ensures fast convergence. Simulation results demonstrate that the proposed approach achieves better trade-offs between communication sum rate and sensing's SINR compared to existing benchmark schemes.

This paper addresses the problem of adaptive codebook (CB) selection for downlink (DL) precoder quantization in channel state information (CSI) reporting. The accuracy of precoder quantization depends on propagation conditions, requiring independent parameter adaptation for each user equipment (UE). To enable optimal CB selection, this paper proposes UE-assisted CB selection at the base station (BS) using reported by the UE statistical channel properties across time, frequency, and spatial domains. The reported assistance information serves as input to a neural network (NN), which predicts the quantization accuracy of various CB types for each served user. The predicted accuracy is then used to select the optimal CB while considering the associated CSI reporting overhead and precoding performance. System-level simulations demonstrate that the proposed approach reduces total CSI overhead while maintaining the target system throughput performance.

Today we design wireless networks using mathematical models that govern communication in different propagation environments. We rely on measurement campaigns to deliver parametrized propagation models, and on the 3GPP standards process to optimize model-based performance, but as wireless networks become more complex this model-based approach is losing ground. Mobile Network Operators (MNOs) are counting on Artificial Intelligence (AI) to transform wireless by increasing spectral efficiency, reducing signaling overhead, and enabling continuous network innovation through software upgrades. They may also be interested in new use cases like integrated sensing and communications (ISAC). All we need is an AI-native physical layer, so why not simply tailor the offline AI algorithms that have revolutionized image and natural language processing to the wireless domain? We argue that these algorithms rely on off-line training that is precluded by the sub-millisecond speeds at which the wireless interference environment changes. We present an alternative architecture, a universal neural receiver based on convolution, which governs transmit and receive signal processing of any signal in any part of the wireless spectrum. Our neural receiver is designed to invert convolution, and we separate the question of which convolution to invert from the actual deconvolution. The neural network that performs deconvolution is very simple, and we configure this network by setting weights based on domain knowledge. By telling our neural network what we know, we avoid extensive offline training. By developing a universal receiver, we hope to simplify discussions about the proper choice of waveform for different use cases in the international standards. Since the receiver architecture is largely independent of technologies introduced at the base station, we hope to increase the rate of innovation in wireless.

This paper develops a tractable analytical channel model for first-hitting-time molecular communication systems under time-varying drift. While existing studies of nonstationary transport rely primarily on numerical solutions of advection--diffusion equations or parametric impulse-response fitting, they do not provide a closed-form description of trajectory-level arrival dynamics at absorbing boundaries. By adopting a change-of-measure formulation, we reveal a structural decomposition of the first-hitting-time density into a cumulative-drift displacement term and a stochastic boundary-flux modulation factor. This leads to an explicit analytical expression for the Corrected-Inverse-Gaussian (C-IG) density, extending the classical IG model to strongly nonstationary drift conditions while preserving constant-complexity evaluation. High-precision Monte Carlo simulations under both smooth pulsatile and abrupt switching drift profiles confirm that the proposed model accurately captures complex transport phenomena, including phase modulation, multi-pulse dispersion, and transient backflow. The resulting framework provides a physics-informed, computationally efficient channel model suitable for system-level analysis and receiver design in dynamic biological and molecular communication environments.

Intellicise (Intelligent and Concise) wireless network is the main direction of the evolution of future mobile communication systems, a perspective now widely acknowledged across academia and industry. As a key technology within it, Agentic AI has garnered growing attention due to its advanced cognitive capabilities, enabled through continuous perception-memory-reasoning-action cycles. This paper first analyses the unique advantages that Agentic AI introduces to intellicise wireless networks. We then propose a structured taxonomy for Agentic AI-enhanced secure intellicise wireless networks. Building on this framework, we identify emerging security and privacy challenges introduced by Agentic AI and summarize targeted strategies to address these vulnerabilities. A case study further demonstrates Agentic AI's efficacy in defending against intelligent eavesdropping attacks. Finally, we outline key open research directions to guide future exploration in this field.

Matrix approximation methods have successfully produced efficient, low-complexity approximate transforms for the discrete cosine transforms and the discrete Fourier transforms. For the DFT case, literature archives approximations operating at small power-of-two blocklenghts, such as \{8, 16, 32\}, or at large blocklengths, such as 1024, which are obtained by means of the Cooley-Tukey-based approximation relying on the small-blocklength approximate transforms. Cooley-Tukey-based approximations inherit the intermediate multiplications by twiddled factors which are usually not approximated; otherwise the effected error propagation would prevent the overall good performance of the approximation. In this context, the prime factor algorithm can furnish the necessary framework for deriving fully multiplierless DFT approximations. We introduced an approximation method based on small prime-sized DFT approximations which entirely eliminates intermediate multiplication steps and prevents internal error propagation. To demonstrate the proposed method, we design a fully multiplierless 1023-point DFT approximation based on 3-, 11- and 31-point DFT approximations. The performance evaluation according to popular metrics showed that the proposed approximations not only presented a significantly lower arithmetic complexity but also resulted in smaller approximation error measurements when compared to competing methods.

Integrated sensing and communication (ISAC) has emerged as a pivotal technology for next-generation wireless networks, enabling simultaneous data transmission and environmental sensing. However, existing ISAC systems face fundamental limitations in achieving high-resolution sensing while maintaining robust communication security and spectral efficiency. This paper introduces a transformative approach leveraging stacked intelligent metasurfaces (SIM) to overcome these challenges. We propose a multi-functional SIM-assisted system that jointly optimizes communication secrecy and sensing accuracy through a novel layered optimization framework. Our solution employs a multi-objective optimization formulation that balances secrecy rate maximization with sensing error minimization under practical hardware constraints. The proposed layered block coordinate descent algorithm efficiently coordinates sensing configuration, secure beamforming, communication metasurface optimization, and resource allocation while ensuring robustness to channel uncertainties. Extensive simulations demonstrate significant performance gains over conventional approaches, achieving 32-61\% improvement in sensing accuracy and 15-35\% enhancement in secrecy rates while maintaining computational efficiency. This work establishes a new paradigm for secure and high-precision multi-functional wireless systems.

Three-dimensional (3D) ultrasound promises various medical applications for abdominal, obstetrics, and cardiovascular imaging. However, ultrasound matrix arrays have extremely high element counts limiting their field of view (FOV). This work seeks to demonstrate an increased field-of-view using a reduced element count array design. The approach is to increase the element size and use advanced beamformers to maintain image quality. The delay and sum (DAS), Null Subtraction Imaging (NSI), directional coherence factor (DCF), and Minimum Variance (MV) beamformers were compared. K-wave simulations of the 3D point-spread functions (PSF) of NSI, DCF, and MV display reduced side lobes and narrowed main lobes compared to DAS. Experiments were conducted using a multiplexed 1024-element matrix array on a Verasonics 256 system. Elements were electronically coupled to imitate a larger pitch and element size. Then, a virtual large aperture was created by using a positioning system to collect data in sections with the matrix array. High-quality images were obtained using a coupling factor of two, doubling the FOV while maintaining the same element count in the virtual large aperture as the original matrix array. The NSI beamformer demonstrated the best resolution performance in simulations and on the large aperture, maintaining the same resolution as uncoupled DAS for coupling factors up to 4. Our results demonstrate how larger matrix arrays could be constructed with larger elements, with resolution maintained by advanced beamformers.

Multi-static backscatter networks (BNs) are strong candidates for joint communication and localization in the ambient IoT paradigm for 6G. Enabling real-time localization in large-scale multi-static deployments with thousands of devices require highly efficient algorithms for estimating key parameters such as range and angle of arrival (AoA), and for fusing these parameters into location estimates. We propose two low-complexity algorithms, Joint Range-Angle Clustering (JRAC) and Stage-wise Range-Angle Estimation (SRAE). Both deliver range and angle estimation accuracy comparable to FFT- and subspace-based baselines while significantly reducing the computation. We then introduce two real-time localization algorithms that fuse the estimated ranges and AoAs: a maximum-likelihood (ML) method solved via gradient search and an iterative re-weighted least squares (IRLS) method. Both achieve localization accuracy comparable to ML-based brute force search albeit with far lower complexity. Experiments on a real-world large-scale multi-static testbed with 4 illuminators, 1 multi-antenna receiver, and 100 tags show that JRAC and SRAE reduce runtime by up to 40X and IRLS achieves up to 500X reduction over ML-based brute force search without degrading localization accuracy. The proposed methods achieve 3 m median localization error across all 100 tags in a sub-6GHz band with 40 MHz bandwidth. These results demonstrate that multi-static range-angle estimation and localization algorithms can make real-time, scalable backscatter localization practical for next-generation ambient IoT networks.

This work presents the demonstration of lattice filters based on laterally excited bulk acoustic resonators (XBARs). Two filter implementations, namely direct lattice and layout-balanced lattice topologies, are designed and fabricated in periodically poled piezoelectric film (P3F) thin-film lithium niobate (TFLN). By leveraging the strong electromechanical coupling of XBARs in P3F TFLN together with the inherently wideband nature of the lattice topology, 3-dB fractional bandwidths (FBWs) of 27.42\% and 39.11\% and low insertion losses (ILs) of 0.88 dB and 0.96 dB are achieved at approximately 20 GHz for the direct and layout-balanced lattice filters, respectively, under conjugate matching. Notably, all prototypes feature compact footprints smaller than 1.3 mm\textsuperscript{2}. These results highlight the potential of XBAR-based lattice architectures to enable low-loss, wideband acoustic filters for compact, high-performance RF front ends in next-generation wireless communication and sensing systems, while also identifying key challenges and directions for further optimization.

In this paper, we present a detailed convergence analysis of a recently developed approximate Newton-type fully distributed optimization method for smooth, strongly convex local loss functions, called Network-GIANT, which has been empirically illustrated to show faster linear convergence properties while having the same communication complexity (per iteration) as its first order distributed counterparts. By using consensus based parameter updates, and a local Hessian based descent direction at the individual nodes with gradient tracking, we first explicitly characterize a global linear convergence rate for Network-GIANT, which can be computed as the spectral radius of a $3 \times 3$ matrix dependent on the Lipschitz continuity ($L$) and strong convexity ($μ$) parameters of the objective functions, and the spectral norm ($σ$) of the underlying undirected graph represented by a doubly stochastic consensus matrix. We provide an explicit bound on the step size parameter $η$, below which this spectral radius is guaranteed to be less than $1$. Furthermore, we derive a mixed linear-quadratic inequality based upper bound for the optimality gap norm, which allows us to conclude that, under small step size values, asymptotically, as the algorithm approaches the global optimum, it achieves a locally linear convergence rate of $1-η(1 -\fracγμ)$ for Network-GIANT, provided the Hessian approximation error $γ$ (between the harmonic mean of the local Hessians and the global hessian (the arithmetic mean of the local Hessians) is smaller than $μ$. This asymptotically linear convergence rate of $\approx 1-η$ explains the faster convergence rate of Network-GIANT for the first time. Numerical experiments are carried out with a reduced CovType dataset for binary logistic regression over a variety of graphs to illustrate the above theoretical results.

While synthetic aperture radar is widely adopted to provide high-resolution imaging at long distances using small arrays, the concept of coherent synthetic aperture communication (SAC) has not yet been explored. This article introduces the principles of SAC for direct satellite-to-device uplink, showcasing precise direction-of-arrival estimation for user equipment (UE) devices, facilitating spatial signal separation, localization, and easing link budget constraints. Simulations for a low Earth orbit satellite at 600 km orbit and two UE devices performing orthogonal frequency-division multiplexing-based transmission with polar coding at 3.5 GHz demonstrate block error rates below 0.1 with transmission powers as low as -10 dBm, even under strong interference when UE devices are resolved but fall on each other's strongest angular sidelobe. These results validate the ability of the proposed scheme to address mutual interference and stringent power limitations, paving the way for massive Internet of Things connectivity in non-terrestrial networks.

The Dirac operator provides a unified framework for processing signals defined over different order topological domains, such as node and edge signals. Its eigenmodes define a spectral representation that inherently captures cross-domain interactions, in contrast to conventional Hodge-Laplacian eigenmodes that operate within a single topological dimension. In this paper, we compare the two alternatives in terms of the distortion/sparsity trade-off and we show how an overcomplete basis built concatenating the two dictionaries can provide better performance with respect to each approach. Then, we propose a parameterized nonredundant transform whose eigenmodes incorporate a mode-specific mass parameter that captures the interplay between node and edge modes. Interestingly, we show that learning the mass parameters from data makes the proposed transform able to achieve the best distortion-sparsity tradeoff with respect to both complete and overcomplete bases.

In this work, we propose a method for computing centroids, or barycenters, in the spectral Wasserstein-2 metric for sets of power spectral densities, where the barycenters are restricted to belong to the set of all-pole spectra with a certain model order. This may be interpreted as finding an autoregressive representative for sets of second-order stationary Gaussian processes. While Wasserstein, or optimal transport, barycenters have been successfully used earlier in problems of spectral estimation and clustering, the resulting barycenters are non-parametric and the complexity of representing and storing them depends on, e.g., the choice of discretization grid. In contrast, the herein proposed method yields compact, low-dimensional, and interpretable spectral centroids that can be used in downstream tasks. Computing the all-pole centroids corresponds to solving a non-convex optimization problem in the model parameters, and we present a gradient descent scheme for addressing this. Although convergence to a globally optimal point cannot be guaranteed, the sub-optimality of the obtained centroids can be quantified. The proposed method is illustrated on a problem of phoneme classification.

Magnetic induction (MI) enables communication in RF-denied environments (underground, underwater, in-body), where the medium conductivity imprints a deterministic signature on the channel. This letter derives a closed-form Cramér--Rao bound (CRB) for the joint estimation of range and medium conductivity from MI pilot observations in an integrated sensing and communication (ISAC) framework. The Fisher information matrix reveals that the joint estimation penalty converges to 3\,dB in the near-field regime, meaning conductivity sensing adds at most a factor-of-two loss in ranging precision. Monte Carlo maximum-likelihood simulations confirm that the CRB is achievable under practical operating conditions.

This paper investigates the problem of high-precision target localization in integrated sensing and communication (ISAC) systems, where the target is sensed via both a direct path and a reconfigurable intelligent surface (RIS)-assisted reflection path. We first develop a sequential matched-filter estimator to acquire coarse angular parameters, followed by a range recovery process based on subcarrier phase differences. Subsequently, we formulate the target localization problem as a non-linear least squares optimization, using the coarse estimates to initialize the target's position coordinates. To solve this efficiently, we introduce a fast iterative refinement algorithm tailored for RIS-aided ISAC environments. Recognizing that the signal model involves both linear path gains and non-linear geometric dependencies, we exploit the separable least-squares structure to decouple these parameters. Furthermore, we propose a modified Levenberg algorithm with an approximation strategy, which enables low-cost parameter updates without necessitating repeated evaluations of the full non-linear model. Simulation results show that the proposed refinement method achieves accuracy comparable to conventional approaches, while significantly reducing algorithmic complexity.

AS-ISTA (Architecture Searched-Iterative Shrinkage Thresholding Algorithm) and AS-FISTA (AS-Fast ISTA) are compressed sensing algorithms introducing structural parameters to ISTA and FISTA to enable architecture search within the iterative process. The structural parameters are determined using deep unfolding, but this approach requires training data and the large overhead of training time. In this paper, we propose HGD-AS-ISTA (Hypergradient Descent-AS-ISTA) and HGD-AS-FISTA that use hypergradient descent, which is an online hyperparameter optimization method, to determine the structural parameters. Experimental results show that the proposed method improves performance of the conventional ISTA/FISTA while avoiding the need for re-training when the environment changes.

Several applications of medical ultrasound can benefit from a larger imaging field of view (FOV). This study is aimed at increasing the FOV of linear array probes by increasing the element size rather than the element count. To investigate larger FOV, this study used coupled elements to imitate a larger element size. The effects of coupling on array beam patterns are examined with Fourier transforms of elements. The effects of coupling on resolution, contrast, and speckle signal-to-noise ratio are examined through phantom images and in-vivo images of a rabbit tumor reconstructed with plane-wave compounding. Furthermore, a positioning system was used to acquire data from a virtual large aperture with 120 mm FOV and 128 elements, collected in sections with a single probe. This study also investigates the Null Subtraction Imaging (NSI), Sign Coherence Factor (SCF), and Minimum Variance (MV) beamformers for regaining resolution lost by an increased F-number with large elements. The MV beamformer, while the most computationally expensive, was best for improving resolution without increasing speckle variance, decreasing Full-Width at Half-Max (FWHM) estimates of wire targets from 0.78 mm with DAS on a 2.5 wavelength element size to 0.54 mm with MV on a 5 wavelength element size.

Integrating massive multiple-input multiple-output (mMIMO) systems with intelligent reflecting surfaces (IRS) presents a promising paradigm for enhancing physical-layer security (PLS) in wireless communications. However, deploying high-resolution quantizers in large-scale mMIMO arrays, along with numerous IRS elements, leads to substantial hardware complexity. To address these challenges, this paper proposes a cost-effective PLS design for IRS-assisted mMIMO systems by employing one-bit digital-to-analog converters (DACs). The focus is on jointly optimizing one-bit quantized precoding at the transmitter and constant-modulus phase shifts at the IRS to maximize the secrecy rate. This leads to a highly non-convex fractional secrecy rate maximization (SRM) problem. To efficiently solve this problem, two algorithms are proposed: (1) the WMMSE-PDD algorithm, which reformulates the SRM problem into a sequence of non-fractional programs with auxiliary variables using the weighted minimum mean-square error (WMMSE) method and solves them via the penalty dual decomposition (PDD) approach, achieving superior secrecy performance; and (2) the exact penalty product Riemannian gradient descent (EPPRGD) algorithm, which transforms the SRM problem into an unconstrained optimization over a product Riemannian manifold, eliminating auxiliary variables and enabling faster convergence with a slight trade-off in secrecy performance. Both algorithms provide analytical solutions at each iteration and are proven to converge to Karush-Kuhn-Tucker (KKT) points. Simulation results confirm the effectiveness of the proposed methods and highlight their respective advantages.

3D Gaussian Splatting (3DGS) has emerged as a powerful approach for novel view synthesis. However, the number of Gaussian primitives often grows substantially during training as finer scene details are reconstructed, leading to increased memory and storage costs. Recent coarse-to-fine strategies regulate Gaussian growth by modulating the frequency content of the ground-truth images. In particular, AutoOpti3DGS employs the learnable Discrete Wavelet Transform (DWT) to enable data-adaptive frequency modulation. Nevertheless, its modulation depth is limited by the 1-level DWT, and jointly optimizing wavelet regularization with 3D reconstruction introduces gradient competition that promotes excessive Gaussian densification. In this paper, we propose a multi-level DWT-based frequency modulation framework for 3DGS. By recursively decomposing the low-frequency subband, we construct a deeper curriculum that provides progressively coarser supervision during early training, consistently reducing Gaussian counts. Furthermore, we show that the modulation can be performed using only a single scaling parameter, rather than learning the full 2-tap high-pass filter. Experimental results on standard benchmarks demonstrate that our method further reduces Gaussian counts while maintaining competitive rendering quality.

This paper investigates secure downlink transmission in a UAV-assisted reconfigurable intelligent surface (RIS)-enabled multiuser multiple-input single-output network, where legitimate information-harvesting receivers coexist with untrusted energy-harvesting receivers (UEHRs) capable of eavesdropping. A UAV-mounted RIS provides blockage mitigation and passive beamforming, while the base station employs zero-forcing precoding for multiuser interference suppression. Due to limited feedback from UEHRs, their channel state information (CSI) is imperfect, leading to a worst-case secrecy energy efficiency (WCSEE) maximization problem. We jointly optimize the UAV horizontal position, RIS phase shifts, and transmit power allocation under both perfect and imperfect CSI, considering discrete RIS phases, UAV mobility, and energy-harvesting constraints. The resulting problem is highly nonconvex due to coupled channel geometry, robustness requirements, and discrete variables. To address this challenge, we propose a soft actor-critic (SAC)-based deep reinforcement learning framework that learns WCSEE-maximizing policies through interaction with the wireless environment. As a structured benchmark, a successive convex approximation (SCA) approach is developed for the perfect CSI case with continuous RIS phases. Simulation results show that the proposed SAC method achieves up to 28% and 16% secrecy energy efficiency gains over SCA and deep deterministic policy gradient baselines, respectively, while demonstrating superior robustness to CSI uncertainty and stable performance across varying transmit power levels and RIS sizes.

Signal processing has played, and continues to play, a fundamental role in the evolution of modern localization technologies. Localization using spatial variations in the Earth's magnetic field is no exception. It relies on signal-processing methods for statistical state inference, magnetic-field modeling, and sensor calibration. Contemporary localization techniques based on spatial variations in the magnetic field can provide decimeter-level indoor localization accuracy and outdoor localization accuracy on par with strategic-grade inertial navigation systems. This article provides a broad, high-level overview of current signal-processing principles and open research challenges in localization using spatial variations in the Earth's magnetic field. The aim is to provide the reader with an understanding of the similarities and differences among existing key technologies from a statistical signal-processing perspective. To that end, existing key technologies will be presented within a common parametric signal-model framework compatible with well-established statistical inference methods.

The detection of interictal epileptiform discharge (IED) is crucial for the diagnosis of epilepsy, but automated methods often lack interpretability. This study proposes IED-RAG, an explainable multimodal framework for joint IED detection and report generation. Our approach employs a dual-encoder to extract electrophysiological and semantic features, aligned via contrastive learning in a shared EEG-text embedding space. During inference, clinically relevant EEG-text pairs are retrieved from a vector database as explicit evidence to condition a large language model (LLM) for the generation of evidence-based reports. Evaluated on a private dataset from Wuhan Children's Hospital and the public TUH EEG Events Corpus (TUEV), the framework achieved balanced accuracies of 89.17\% and 71.38\%, with BLEU scores of 89.61\% and 64.14\%, respectively. The results demonstrate that retrieval of explicit evidence enhances both diagnostic performance and clinical interpretability compared to standard black-box methods.

Wireless communication systems exhibit structural and functional similarities to neural networks: signals propagate through cascaded elements, interact with the environment, and undergo transformations. Building upon this perspective, we introduce a unified paradigm, termed \textit{wireless physical neural networks (WPNNs)}, in which components of a wireless network, such as transceivers, relays, backscatter, and intelligent surfaces, are interpreted as computational layers within a learning architecture. By treating the wireless propagation environment and network elements as differentiable operators, new opportunities arise for joint communication-computation designs, where system optimization can be achieved through learning-based methods applied directly to the physical network. This approach may operate independently of, or in conjunction with, conventional digital neural layers, enabling hybrid communication learning pipelines. In the article, we outline representative architectures that embody this viewpoint and discuss the algorithmic and training considerations required to leverage the wireless medium as a computational resource. Through numerical examples, we highlight the potential performance gains in processing, adaptability, efficiency, and end-to-end optimization, demonstrating the promise of reconfiguring wireless systems as learning networks in next-generation communication frameworks.

Extra-large apertures, high carrier frequencies, and integrated sensing and communications (ISAC) are pushing array processing into the Fresnel region, where spherical wavefronts induce a range-dependent phase across the aperture. This curvature breaks the Fourier/Vandermonde structure behind classical subspace methods, and it is especially limiting with hybrid front-ends that provide only a small number of pilot measurements. Consequently, practical systems need continuous angle resolution and joint angle-range inference where many near-field approaches still rely on costly 2D gridding. We show that convexity can meet curvature via a lifted, gridless superresolution framework for near-field measurements. The key is a Bessel-Vandermonde factorization of the Fresnel-phase manifold that exposes a hidden Vandermonde structure in angle while isolating the range dependence into a compact coefficient map. Building on this, we introduce a lifting that maps each range bin and continuous angle to a structured rank-one atom, converting the nonlinear near-field model into a linear inverse problem over a row-sparse matrix. Recovery is posed as atomic-norm minimization and an explicit dual characterization via bounded trigonometric polynomials yields certificate-based localization that super-resolves off-grid angles and identifies active range bins. Simulations with strongly undersampled hybrid observations validate reliable joint angle-range recovery for next-generation wireless and ISAC systems.

Recent advances in deep learning (DL)-based joint source-channel coding (JSCC) have enabled efficient semantic communication in dynamic wireless environments. Among these approaches, vector quantization (VQ)-based JSCC effectively maps high-dimensional semantic feature vectors into compact codeword indices for digital modulation. However, existing methods, including universal JSCC (uJSCC), rely on fixed, modulation-specific encoders, decoders, and codebooks, limiting adaptability to fine-grained SNR variations. We propose an extended universal JSCC (euJSCC) framework that achieves SNR- and modulation-adaptive transmission within a single model. euJSCC employs a hypernetwork-based normalization layer for fine-grained feature vector normalization and a dynamic codebook generation (DCG) network that refines modulation-specific base codebooks according to block-wise SNR. To handle block fading channels, which consist of multiple coherence blocks, an inner-outer encoder-decoder architecture is adopted, where the outer encoder and decoder capture long-term channel statistics, and the inner encoder and decoder refine feature vectors to align with block-wise codebooks. A two-phase training strategy, i.e., pretraining on AWGN channels followed by finetuning on block fading channels, ensures stable convergence. Experiments on image transmission demonstrate that euJSCC consistently outperforms state-of-the-art channel-adaptive digital JSCC schemes under both block fading and AWGN channels.

The Received Signal Strength Indicator (RSSI) is widely available on commodity WiFi devices but is commonly regarded as too coarse for fine-grained sensing. This paper revisits its sensing potential and presents WiRSSI, a bistatic WiFi sensing framework for passive human tracking using only RSSI measurements. WiRSSI adopts a 1Tx-3Rx configuration and is readily extensible to Multiple-Input Multiple-Output (MIMO) deployments. We first reveal how CSI power implicitly encodes phase-related information and how this relationship carries over to RSSI, showing that RSSI preserves exploitable Doppler, Angle-of-Arrival (AoA), and delay cues associated with human motion. WiRSSI then extracts Doppler-AoA features via a 2D Fast Fourier Transform and infers delay from amplitude-only information in the absence of subcarrier-level phase. The estimated AoA and delay are then mapped to Cartesian coordinates and denoised to recover motion trajectories. Experiments in practical environments show that WiRSSI achieves median XY localization errors of 0.905 m, 0.784 m, and 0.785 m for elliptical, linear, and rectangular trajectories, respectively. In comparison, a representative CSI-based method attains median errors of 0.574 m, 0.599 m, and 0.514 m, corresponding to an average accuracy gap of 0.26 m. These results demonstrate that, despite its lower resolution, RSSI can support practical passive sensing and offers a low-cost alternative to CSI-based WiFi sensing.

Quasi-static human activities such as lying, standing or sitting produce very low Doppler shifts and highly spread radar signatures, making them difficult to detect with conventional constant-false-alarm rate (CFAR) detectors tuned for point targets. Moreover, privacy concerns and low lighting conditions limit the use of cameras in long-term care (LTC) facilities. This paper proposes a lightweight, non-visual image-based method for robust quasi-static human presence detection using a low-cost 60 GHz FMCW radar. On a dataset covering five semi-static activities, the proposed method improves average detection accuracy from 68.3% for Cell-Averaging CFAR (CA-CFAR) and 78.8% for Order-Statistics CFAR (OS-CFAR) to 93.24% for Subject 1, from 51.3%, 68.3% to 92.3% for Subject 2, and 57.72%, 69.94% to 94.82% for Subject 3, respectively. Finally, we benchmarked all three detectors across all activities on a Raspberry Pi 4B using a shared Range-Angle (RA) preprocessing pipeline. The proposed algorithm obtains an average 8.2 ms per frame, resulting in over 120 frames per second (FPS) and a 74 times speed-up over OS-CFAR. These results demonstrate that simple image-based processing can provide robust and deployable quasi-static human sensing in cluttered indoor environments.

Accurate cascaded channel state information is pivotal for extremely large-scale intelligent reflecting surfaces (XL-IRS) in next-generation wireless networks. However, the large XL-IRS aperture induces spherical wavefront propagation due to near-field (NF) effects, complicating cascaded channel estimation. Conventional dictionary-based methods suffer from cumulative quantization errors and high complexity, especially in uniform planar array (UPA) systems. To address these issues, we first propose a tensor modelization method for NF cascaded channels by exploiting the tensor product among the horizontal and vertical response vectors of the UPA-structured base station (BS) and the incident-reflective array response vector of the IRS. This structure leverages spatial characteristics, enabling independent estimation of factor matrices to improve efficiency. Meanwhile, to avoid quantization errors, we propose an off-grid cascaded channel estimation framework based on sparse Tucker decomposition. Specifically, we model the received signal as a Tucker tensor, where the sparse core tensor captures path gain-delay terms and three factor matrices are spanned by BS and NF IRS array responses. We then formulate a sparse core tensor minimization problem with tri-modal log-sum sparsity constraints to tackle the NP-hard challenge. Finally, the method is accelerated via higher-order singular value decomposition preprocessing, combined with majorization-minimization and a tailored tensor over-relaxation fast iterative shrinkage-thresholding technique. We derive the Cramér-Rao lower bound and conduct convergence analysis. Simulations show the proposed scheme achieves a 13.6 dB improvement in normalized mean square error over benchmarks with significantly reduced runtime.

Recent advances in satellite and communication technologies have significantly improved geographical information and monitoring systems. Global System for Mobile Communications (GSM) and Global Navigation Satellite System (GNSS) technologies, which rely on electromagnetic signals transmitted from satellites and base stations, have long been utilized for geolocation applications. However, signal attenuation due to environmental conditions or intentional interference such as jamming may lead to severe degradation or complete loss of positioning capability. In such GNSS-denied environments, landmark extraction becomes critical for the navigation of unmanned aerial vehicles (UAVs) used in monitoring applications. By processing images captured from onboard UAV cameras, reliable visual landmarks can be identified to enable navigation without GNSS support. In this study, a convolution-based deep learning approach is proposed for the extraction of appropriate landmarks, and its effectiveness is examined.

Off-grid targets whose Doppler (or angle) does not lie on the discrete processing grid can severely degrade classical normalized matched-filter (NMF) detectors: even at high SNR, the detection probability may saturate at operationally relevant low false-alarm rates. A principled remedy is the continuous-parameter GLRT, which maximizes a normalized correlation over the parameter domain; however, dense scanning increases test-time cost and remains sensitive to covariance mismatch through whitening. We propose DopplerGLRTNet, an amortized off-grid GLRT: a lightweight regressor predicts the continuous Doppler within a resolution cell from the whitened observation, and the detector outputs a single GLRT/NMF-like score given by the normalized matched-filter energy at the predicted Doppler. Monte Carlo simulations in Gaussian and compound-Gaussian clutter show that DopplerGLRTNet mitigates off-grid saturation, approaches dense-scan performance at a fraction of its cost, and improves robustness to covariance estimation at the same empirically calibrated Pfa.

We investigate bound states of a non-relativistic scalar particle in a three-dimensional helically twisted (torsional) geometry, considering both the free case and the presence of external radial interactions. The dynamics is described by the Schrödinger equation on a curved spatial background and, when included, by minimal coupling to a magnetic vector potential incorporating an Aharonov--Bohm flux. After separation of variables, the problem reduces to a one-dimensional radial eigenvalue equation governed by an effective potential that combines torsion-induced Coulomb-like and centrifugal-like structures with magnetic/flux-dependent terms and optional model interactions. Because closed-form analytic solutions are not reliable over the parameter ranges required for systematic scans, we compute spectra and eigenfunctions numerically by formulating the radial equation as a self-adjoint Sturm--Liouville problem and solving it with a finite-difference discretization on a truncated radial domain, with explicit convergence control. We analyze four representative scenarios: (i) no external potential, (ii) Cornell-type confinement, (iii) Kratzer-type interaction, and (iv) the small-oscillation regime around the minimum of a Morse potential. We present systematic trends of the low-lying levels as functions of the torsion parameter, magnetic field, and azimuthal sector, and we show that geometric couplings alone can produce effective confinement even in the absence of an external interaction.

We study scalar waves on subextremal Kerr-de Sitter spacetimes in a compact slow-rotation regime and at a fixed overtone index. Working initially at a fixed cosmological constant $Λ>0$ and uniformly for $(M,a)$ in a compact slow-rotation set, using the meromorphic/Fredholm framework for quasinormal modes and a semiclassical equatorial labeling proved in a companion paper, we establish a quantitative two-mode dominance theorem in an equatorial high-frequency package: after exact azimuthal reduction, microlocal equatorial localization, and analytic pole selection by entire localization weights constructed from equatorial pseudopoles, the $k=\pm\ell$ sector signals are each governed by a single quasinormal exponential, up to an explicitly controlled tail and an $\mathcal O(\ell^{-\infty})$ contribution from all other poles. We then develop a fully deterministic frequency-extraction stability estimate based on time-shift invariance, and combine it with the two-mode dominance result and the companion paper's inverse stability theorem to obtain an explicit parameter bias bound for ringdown-based recovery of $(M,a)$. Finally, using the companion paper's three-parameter inverse theorem and a damping observable based on the scaled imaginary part of one equatorial mode, we propagate the same deterministic error chain to a local bias bound for recovery of $(M,a,Λ)$ on compact parameter sets with $|a|$ bounded away from $0$. As a further consequence, we obtain a localized pseudospectral stability statement for the equatorial resolvent package, quantifying how large microlocalized resolvent norms enforce proximity to the labeled equatorial poles. The resulting estimates clarify the conditioning mechanisms (start time, window length, shift step, and detector nondegeneracy) and provide a rigorous PDE-to-data interface for high-frequency black-hole spectroscopy.

Intermodal quantum key distribution at telecom wavelengths provides a hybrid interface between fiber connections and free-space links, both essential for the realization of scalable and interoperable quantum networks. Although demonstrated over short-range free-space links, long-distance implementations of intermodal quantum key distribution remain challenging, due to turbulence-induced wavefront aberrations which limit efficient single-mode fiber coupling at the optical receiver. Here, we demonstrate a real-time intermodal quantum key distribution field trial over an 18 km free-space link, connecting a remote terminal to an urban optical ground station equipped with a 40 cm-class telescope. An adaptive optics system, implementing direct wavefront sensing and high-order aberration correction, enables efficient single-mode fiber coupling and allows secure key generation of 200 bit/s using a compact state analyzer equipped with room-temperature detectors. We further validate through experimental data a turbulence-based model for predicting fiber coupling efficiency, providing practical design guidelines for future intermodal quantum networks.

We consider the amplification of bosonic interactions through parametric control that implements squeezing along orthogonal quadratures. We show that bosonic interactions described by certain classes of quadratic and quartic Hamiltonians can be enhanced in this way while simultaneously overcoming noise and decoherence. In general, the amplification method enhances both desired and undesired interactions present in the system. Depending on the case, however, detrimental processes can be less amplified than the desired couplings. We leverage this observation to improve the fidelity for preparing Bell-type entangled states between two bosonic modes in the presence of noise and losses. We also investigate noise models for which the protocol either fails or partially achieves a loss-tolerant state preparation speedup. Our work facilitates faster preparation of complex quantum states and implementation of entangling gates in the presence of decoherence mechanisms.

Standard Variational Quantum Circuits (VQCs) struggle to scale to high-dimensional data due to the ``curse of dimensionality,'' which manifests as exponential simulation costs ($\mathcal{O}(2^d)$) and untrainable Barren Plateaus. Existing solutions often bypass this by relying on classical neural networks for feature compression, obscuring the true quantum capability. In this work, we propose the \textbf{Multi-Layer Fully-Connected VQC (FC-VQC)}, a modular architecture that performs \textbf{end-to-end quantum learning} without trainable classical encoders. By restricting local Hilbert space dimensions while enabling global feature interaction via structured block mixing, our framework achieves \textbf{linear scalability $\mathcal{O}(d)$}. We empirically validate this approach on standard benchmarks and a high-dimensional industrial task: \textbf{300-asset Option Portfolio Pricing}. In this regime, the FC-VQC breaks the ``Classical Ceiling,'' outperforming state-of-the-art Gradient Boosting baselines (XGBoost/CatBoost) while exhibiting \textbf{$\approx 17\times$ greater parameter efficiency} than Deep Neural Networks. These results provide concrete evidence that pure, modular quantum architectures can effectively learn industrial-scale feature spaces that are intractable for monolithic ansatzes.

As quantum technologies mature, telecommunication operators have a clear opportunity to unlock and scale new services by providing the connectivity layer that links quantum computers, sensors, clocks, and other quantum devices. Realizing this opportunity requires demonstrating quantum networking protocols, including quantum teleportation, under real-world conditions on existing telecom infrastructure. In this work, we demonstrate quantum teleportation over Deutsche Telekom's metropolitan fiber testbed in Berlin using commercial components deployed at the telecom datacenter. A local Bell-state measurement between 795 nm photons from a weak coherent source and from a bichromatic warm-atom entangled photon source enables conditional state transfer onto an O-band photon, which is transmitted through a 30-km field-deployed fiber loop under real-world environmental conditions. The teleported state is reconstructed after propagation via state tomography, achieving an average teleportation fidelity of 90\% on the deployed link. System performance is evaluated in both the absence and the presence of co-propagating C-band classical traffic within the same fiber, demonstrating compatibility with wavelength-division multiplexed telecom infrastructure carrying live data channels.

We investigate universal entanglement signatures of mixed-state phases obtained by decohering pure-state topological order (TO), focusing on topological corrections to logarithmic entanglement negativity and mutual information: topological entanglement negativity (TEN) and topological mutual information (TMI). For Abelian TOs under decoherence, we develop a replica field-theory framework based on a doubled-state construction that relates TEN and TMI to the quantum dimensions of domain-wall defects between decoherence-induced topological boundary conditions, yielding general expressions in the strong-decoherence regime. We further compute TEN and TMI exactly for decohered $G$-graded string-net states, including cases with non-Abelian anyons. We interpret the results within the strong one-form-symmetry framework for mixed-state TOs: TMI probes the total quantum dimension of the emergent premodular anyon theory, whereas TEN detects only its modular part.

We extend recent algebro-geometric results for coupled cluster theory of quantum many-body systems to the truncation varieties arising from the doubles approximation (CCD), focusing on the first genuinely nonlinear doubles regime of four electrons. Since this doubles truncation variety does not coincide with previously studied varieties, we initiate a systematic investigation of its basic algebro-geometric invariants. Combining theoretical and numerical results, we show that for $4$ electrons on $n\leq 12$ orbitals, the CCD truncation variety is a complete intersection of degree $2^{\binom{n-4}{4}}$. Using representation-theoretic arguments, we uncover a Pfaffian structure governing the quadratic relations that define the truncation variety for any $n$, and show that an exact tensor product factorization holds in a distinguished limit of disconnected doubles. We connect these structural results to the computation of the beryllium insertion into molecular hydrogen ({Be$\cdots$H$_2$ $\to$ H--Be--H}), a small but challenging bond formation process where multiconfigurational effects become pronounced.

Over an arbitrary commutative ring $R$, we develop a theory of quantum cellular automata. We then use algebraic K-theory to construct a space $\mathbf{Q}(X)$ of quantum cellular automata (QCA) on a given metric space $X$. In most cases of interest, $π_0 \mathbf{Q}(X)$ classifies QCA up to quantum circuits and stabilization. Notably, the QCA spaces are related by homotopy equivalences $\mathbf{Q}(*) \simeq Ω^n \mathbf{Q}(\mathbb{Z}^n)$ for all $n$, which shows that the classification of QCA on Euclidean lattices is given by an $Ω$-spectrum indexed by the dimension $n$. As a corollary, we also obtain a non-connective delooping of the K-theory of Azumaya $R$-algebras, which may be of independent interests. We also include a section leading to the $Ω$-spectrum for QCA over $C^*$-algebras with unitary circuits.

X-ray reflection spectroscopy is a powerful tool for testing the Kerr hypothesis and probing the strong gravity regime around accreting black holes. Most tests of General Relativity (GR) assume that the spacetime around a black hole is circular, meaning the metric possesses a specific symmetry structure common to the Kerr solution. However, deviations from circularity are predicted by various modified gravity theories and non-vacuum General Relativity solutions. In this work, we test a specific non-circular metric constructed based on a locality principle, where the deviation from the Kerr spacetime is driven by the local spacetime curvature. To accurately model the reflection spectrum in this background, we implement a relativistic ray-tracing code in horizon-penetrating (ingoing Kerr) coordinates, which are favored for their ability to avoid introducing curvature singularities at the horizon in non-circular spacetimes. We apply this model to the high-quality \textit{NuSTAR} spectrum of the Galactic black hole binary EXO 1846--031. Our spectral analysis reveals a source with a high inclination angle ($ι\approx 76^{\circ}$) and a near-extremal spin parameter ($a_* \approx 0.98$). While we identify a global minimum in the parameter space suggesting a non-zero deformation ($\ell_{\mathrm{NP}} \approx 0.12$), the 99\% confidence interval fully encompasses the Kerr limit ($\ell_{\mathrm{NP}}=0$). We conclude that the current X-ray reflection data for EXO 1846--031 are consistent with the Kerr hypothesis. This work demonstrates the feasibility of using X-ray reflection spectroscopy to constrain non-circular metrics and establishes a framework for future tests.

Measurements of the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) neutrino mixing parameters have entered a precision era, enabling increasingly stringent tests of neutrino oscillations. Within the framework of quantum estimation theory, we investigate whether flavor measurements, the only observables currently accessible experimentally, are optimal for extracting the oscillation parameters. We compute the Quantum Fisher Information (QFI) and the classical Fisher Information (FI) associated with ideal flavor projections for all oscillation parameters, considering accelerator muon (anti)neutrino and reactor electron antineutrino beams propagating in vacuum. Two main results emerge. First, flavor measurements saturate the QFI at the first oscillation maximum for $θ_{13}$, $θ_{23}$, and $θ_{12}$, demonstrating their information-theoretic optimality for these parameters. In contrast, they are far from optimal for $δ_{CP}$. In particular, only a small fraction of the available information on $δ_{CP}$ is extracted at the first maximum; the sensitivity improves at the second maximum, in line with the strategy of ESS$ν$SB, a planned facility. Second, the QFI associated with $δ_{CP}$ is approximately one order of magnitude smaller than that of the mixing angles, indicating that the neutrino state intrinsically encodes less information about CP violation. Nevertheless, this quantum bound lies well below current experimental uncertainties, implying that the present precision on $δ_{CP}$ is not fundamentally limited. Our results provide a quantitative framework to disentangle fundamental from practical limitations and establish a benchmark for optimizing future neutrino facilities.

We study deterministic port based teleportation in the presence of noise affecting both the entangled resource state and the measurement process. We focus on a physically motivated model in which each Bell pair constituting the resource interacts with an identical local environment, corresponding to independently distributed entangled links. Two noisy scenarios are analyzed: one with decoherence acting solely on the resource state and ideal measurements, and another with noisy, noise adapted measurements optimised for the given noise model. In the first case, we derive an analytical lower bound and later a closed-form expression for the entanglement fidelity of the teleportation channel and analyze its asymptotic behaviour. In the second, we combine semi analytical and numerical methods. Surprisingly, we find that noise-adapted measurements perform worse than the noiseless ones. To connect the abstract noise description with microscopic physics, we embed the protocol in a spin boson model and investigate the influence of bath memory and temperature on the teleportation fidelity, highlighting qualitative differences between different environments.

We critically examine proposals for the so-called warp-drive spacetimes and classify these models based on their various restrictions within the framework of General Relativity. We then provide a summary of general formalism for each class, and in the process, we highlight some misconceptions, misunderstandings, and errors in the literature that have been used to support claims about the physicality and feasibility of these models. On the way, we prove several new no-go theorems. Our analysis shows that when the principles of General Relativity are applied correctly, most claims regarding physical warp drives must be reassessed, and it becomes highly challenging to justify or support the viability of such models, not merely due to the violation of energy conditions.

The Casimir-Polder force is analyzed when an atom is moving at a constant velocity relative to a collection of translationally invariant macroscopic bodies with generic shapes and compositions. The interaction is described within an approach that accurately treats the atom-field coupling and accounts for the backaction from the environment onto the moving particle. Previously overlooked aspects are uncovered and linked to the nonequilibrium and nonconservative nature of the interaction. Specifically, we examine a behavior that can be understood by characterizing the underlying physical processes in terms of a motional-induced effective temperature. This phenomenon shares similarities with the Fulling-Davies-Unruh effect, opening new perspectives for the understanding of nonequilibrium physics at work in the system.

Sources of quantum correlated photons pairs bridging the 3um-4um Mid-infrared (MIR) band and Telecom/Near-Infrared/Visible band are of high importance for quantum technologies. Spontaneous Parametric Down Conversion is generally used for realizing such sources, but requires costly implementation platforms with reduced versatility. Here, we explore the potentialities of Spontaneous Four-Wave Mixing (SFWM) in all-solid Silicon On Insulator (SOI) waveguides thanks to an experimentally validated model and propose designs ensuring the production of correlated photon pairs bridging the 3um-4um Mid-infrared band and Telecom C-band. Choosing a pump with a wavelength in the range 2100nm-2210nm and a pulse duration of 5ps, we quantitatively performed simulations targeting a probability of photon pair generation per pulse of 0.05, and we found realistic conditions of utilization (2cm-length straight waveguides, intra-modal Four Wave Mixing with the fundamental TE00 mode) with a pump peak power in between 9.2mW and 32mW. A first design (wCOM) reaches a signal wavelength as high as 3.905um, which is situated in an atmospheric transparency window, while maintaining an idler in the Telecom C-band, making it of high interest for atmospheric Quantum Key Distribution. Two other designs wCH4 and wNO2 aim precise CH4 and NO2 gas sensing with a signal wavelength of 3265nm and 3461nm respectively. In terms of signal/idler wavelength separation, wCOM attains the value of 2364nm which is well above the current record of ~1125nm obtained in quantum regime with SFWM in all-solid SOI waveguides.

Motivated by recent developments in the theory of gravitation, we revisit the idea of topological variations, originally introduced by Wheeler and Hawking, from a rigorous perspective. Starting from a localized version of the Einstein-Hilbert variational principle, we encode the key aspects of the variational procedure in the form of a topology on a suitable space of variational configurations with low Sobolev regularity. This structure is the final topology with respect to the admissible variational maps and naturally lends itself to generalizations. We rigorously introduce two distinct types of topological variations, corresponding to the infinitesimal addition of disconnected components and to infinitesimal surgeries, both motivated by related physical concepts. Using tools from the theory of Sobolev spaces and precise asymptotics, we establish dimensional obstructions for the continuity and differentiability of the Einstein-Hilbert action with respect to these variations, and show that in the extended variational framework the action does not admit critical points in dimension $n=4$, while higher dimensions are free of this problem. Finally, we demonstrate the non-trivial effect of higher order curvature terms on the critical dimension.

The strong-field process of high-harmonic generation (HHG) has, in recent years, been treated from a quantum optical perspective in the emerging research area of strong-field quantum optics. These investigations show that HHG radiation is, in general, in a nonclassical state of light. However, the quantum optical treatment of HHG from topological nontrivial materials is missing. Here, we aim to address this gap in current knowledge and consider the quantum optical HHG response from the Su-Schrieffer-Heeger model, a finite chain of atoms with both a topologically trivial and nontrivial insulating phase, the latter supporting edge states. We find that HHG from both topological phases is squeezed at the band-gap frequency. Interestingly, while the harmonic spectrum discriminates the two topological phases of the system, the degree of squeezing only discriminates the phases for smaller chain lengths. We attribute this difference to a relative increase in overlap between bulk and edge states in the topological nontrivial phase for smaller systems. Our findings reveal how the strength of dipole couplings governs the nonclassical HHG response and define new research questions on topologically protected generation of quantum light in strong-field physics.

Black holes are the natural arena for exploring the interplay between gravity and thermodynamics. Although the association between black hole mechanics and black hole thermodynamics is well-established, the comprehensive geometric formulation of thermodynamic variables deserves further investigation. In this work, both Newman-Penrose (NP) and Geroch-Held-Penrose (GHP) formalisms are considered within the framework of horizon thermodynamics. We show that the NP formalism reformulates the horizon condition as the Penrose-Rindler equation. In this context, a Smarr-like formula for stationary black holes is recovered from the Penrose-Rindler equation reinterpreted as a horizon equilibrium of pressures, which includes a pressure associated with the horizon rotation. A complete geometric reformulation of this reinterpretation of the Penrose-Rindler equation evaluated at the horizon is developed within the GHP formalism. The GHP approach further inspires the introduction of the horizon-averaged matter pressure and its conjugate volume, thereby enabling a quasi-local realization of the Smarr-like formula for stationary black holes. This geometric formulation clarifies the connection between horizon dynamics and thermodynamics and offers a unified setting for extending black hole thermodynamics beyond spherical symmetry.

We define nonlocal predictability as how well one observer can predict another's measurement outcomes without classical communication, given full knowledge of the shared quantum state and measurement settings. The local bound on nonlocal predictability is defined as the maximum probability with which one observer can correctly predict the other's measurement outcome prior to measurement. We show that product states always meet this bound, while all pure entangled states and some classically correlated states can exceed it. This demonstrates a nonlocal phenomenon since the predictability of measurement outcomes increases after the measurement. Perfect nonlocal predictability for arbitrary projective measurements occurs only for maximally entangled states among all pure states, underscoring their special role. Comparing pure entangled states with their dephased versions, we find that dephasing on one subsystem can enhance nonlocal predictability for a broad class of states and measurements - a counterintuitive, noise-induced advantage that vanishes for maximally entangled states under any projective measurement.

Preparing subradiant steady states of collectively emitting quantum two-level emitters (TLEs) is hindered by their dark, weakly interacting nature. Existing approaches rely on patterned driving, local control, or structured environments. We propose a platform-independent protocol based on projective measurements on a single TLE. For permutation-symmetric ensembles, a single measurement yields appreciable occupation of single-excitation subradiant steady states. For generic arrays, repeated measurements on one emitter drive the unmeasured TLEs into a nearly pure state with large overlap with the subradiant Dicke subspace.

In this paper, we investigate the enhancement of delocalization and coin-position entanglement in asymmetric discrete-time quantum walks (DTQWs). The asymmetry results from asymmetric coin operations, asymmetric initial states, and asymmetric polarization-dependent losses. By varying these asymmetry factors, the inverse participation ratio and entanglement entropy of the walker are numerically calculated for different coin and loss parameters, both for symmetric and asymmetric initial states. We then experimentally implement a 16-step asymmetric DTQW using a time-multiplexing fiber loop structure. By choosing an asymmetric initial state, both coin-position entanglement and delocalization are simultaneously enhanced under specific coin parameters. Moreover, we observe that with finite asymmetric polarization-dependent loss, the photon probability on the left side decreases significantly, while that on the right side increases and becomes more localized. Interestingly, under specific coin parameters, the entanglement and delocalization exhibit improved robustness against polarization-dependent loss. These results demonstrate that the DTQWs constitute an ideal platform for investigating photonic delocalization and hybrid entanglement.

Feynman famously asserted that interference is the only real mystery in quantum mechanics (QM). It is concluded that the reason for this mystery, and thereby the related mysteries of complementarity, non-commutativity of observables, the uncertainty principle and violation of Bell's equality, is that the axioms of QM depend vitally on the continuum nature of Hilbert Space, deemed unphysical. We develop a theory of quantum physics - Rational Quantum Mechanics (RaQM) - in which Hilbert Space is gravitationally discretised. The key to solving the mysteries of QM in RaQM is a number-theoretic property of the cosine function, concealed in QM when angles range over the continuum. This number-theoretic property describes mathematically the utter indivisibility of the quantum world and implies that the laws of physics are profoundly holistic. We contrast holism with nonlocality. In theories which embrace the continuum, the violation of Bell's inequality requires the laws of physics to be either nonlocal or not realistic; both incomprehensible concepts. By contrast, holism, as embodied in Mach's Principle or in the fractal geometry of a chaotic attractor, is neither incomprehensible nor unphysical. As part of this, we solve the deepest mystery of all; why nature makes use of complex numbers.

We consider the Casimir force in configurations with magnetic metal plates and analyze the reasons why the predictions of the Lifshitz theory using the dielectric permittivity of the Drude model are inconsistent with the measurement data. For this purpose, the contributions of the electromagnetic waves with the transverse magnetic and transverse electric polarizations to the Casimir force are computed using the Lifshitz theory expressed in terms of the pure imaginary Matsubara frequencies. Furthermore, the fractions of the evanescent and propagating waves in these contributions are found using an equivalent formulation of the Lifshitz theory along the real frequency axis. All computations are performed for Au-Ni and Ni-Ni plates using the Drude model and the experimentally consistent plasma model over the separation region from 0.5 to 6~mum, where the total force value is determined by conduction electrons. It is shown that the transverse magnetic contribution to the Casimir force does not depend on the used model of the dielectric permittivity, so that the total difference between the predictions of the Lifshitz theory using the Drude model and the measurement data is determined by the transverse electric contribution. In doing so, as opposed to the case of nonmagnetic metals, both fractions of the evanescent and propagating waves in this contribution depend on the model of the dielectric permittivity used in computations, whereas the magnetic properties of the plate metal influence the Casimir force solely through the fraction of propagating waves in the transverse electric contribution. The issue of a more adequate theoretical description of the electromagnetic response of magnetic metals is discussed.

In this work, we present a logical formalism for reasoning about quantum systems in finite dimension. Contrary to the usual approach in quantum logic, our formalism is based classical first-order logic, which allows us to use the tools of model theory in our study. In particular, we show that our formal theory is complete, meaning that it entirely determines the behaviour of quantum systems. Moreover, we provide a characterization of the models of our formal theory, thus providing new insights in the study of hidden variable models of quantum theory.

Quantum stochastic processes are widely used in describing open quantum systems and in the context of quantum foundations. Physically relevant quantum stochastic processes driven by multiplicative colored noise are generically non-Markovian and analytically intractable. Further, their Markovian limits are generically inequivalent when using either the Ito or Stratonovich conventions for the same quantum stochastic processes. We introduce a quantum noise homogenization scheme that temporally coarse-grains non-Markovian, colored-noise driven quantum stochastic processes and connects them to their effective white-noise (Markovian) limits. Our approach uses a novel phase-space augmentation that maps the non-Markovian dynamics into a higher dimensional Markovian system and then applies a controlled perturbative coarse-graining scheme in the characteristic time scales of the noise. This allows an explicit analytical algorithm to derive effective Markovian generators with renormalized coefficients and enables imposing various physical constraints on them. We thus resolve the Ito-Stratonovich ambiguity for multiplicative colored noise driven quantum stochastic processes, wherein we show that their consistent Markovian limit corresponds to the Stratonovich convention with renormalized coefficients as well as Ito correction terms. By assuming their Markovian limit unravels completely positive, trace-preserving maps, we further characterize a physically relevant family of non-Markovian quantum stochastic processes driven by multiplicative colored noise.

Deviations from the area law of the horizon entropy, in the cosmological setup, are known to lead to modified Friedmann equations governing the evolution of the universe. In this work, we propose that such modifications need not be introduced phenomenologically but can emerge dynamically from stochastic fluctuations of the spacetime metric. We consider a Friedmann-Robertson-Walker (FRW) universe perturbed by a conformal, time-dependent noise factor, whose ensemble average vanishes, leaving the mean background geometry unchanged. By averaging the Einstein equations to second order in the fluctuation amplitude, we derive a modified Friedmann equation that includes an effective correction term. This correction is shown to be equivalent to the general expression obtained from an arbitrary deformation of the entropy-area relation. By specifying the statistical properties, particularly the variance of the conformal noise, we successfully reproduce the Friedmann equation corrections associated with several well-known generalized entropy frameworks, including Rényi, (dual) Kaniadakis, Barrow, logarithmic, and MOND inspired hypergeometric entropies. Our results suggest that deviations from the area law can be interpreted as the macroscopic, coarse-grained imprint of unresolved, microscopic stochastic degrees of freedom in spacetime.

A transient violation of the null energy condition (NEC) during inflation provides a novel mechanism for producing primordial black holes (PBHs) and stochastic gravitational wave (GW) backgrounds. In this work, we extend previous studies by computing the GW contributions from both the ringdown phase of PBH formation and subsequent binary mergers. Our results show that this scenario produces a rich, multi-component GW spectrum consisting of primordial GWs, scalar-induced GWs, and GW emissions from PBH ringdown and binary mergers. We demonstrate that these correlated signatures across different frequency bands provide a novel and powerful avenue to probe or constrain NEC violation during inflation through future multi-band GW observations.

Darwinian evolution requires (i) heritable records, (ii) repeatable copying with variation, and (iii) routine irreversibility. Categorical quantum mechanics (CQM) makes precise why ``copy'' and ``delete'' are not generic quantum operations: they exist only for a realized \emph{classical data} sector (a preferred basis/observable; a commutative structure). Decoherence explains how a pointer basis can be selected dynamically, but it does not by itself select a unique outcome. This motivates a neutral presentation of the main ontological options (unique-history, decohered multiplicity, agent-relative facticity, and a stochastic foundation with variable diffusion). We also note the relevance of the ``agency constraint'' argued by Adlam-McQueen-Waegell: in a strictly coherent, basis-unselected ``purely quantum'' regime, minimal agency fails due to no-cloning and linearity, which sharpens the role of classical resources for record-based processes. Extended Wigner's Friend scenarios then serve as a stress test, since they treat ``friends'' simultaneously as coherent quantum systems and as agents possessing stable records. Finally, a stochastic-mechanics foundation (with variable diffusion) offers a continuous bridge between quantum and classical regimes, and suggests a principled way to implement measurement update as conditioning plus a time-symmetric minimal-change rule.

We derive two model-independent results for spacetimes with globally bounded tidal fields. These are operational resolution scales of the local-inertial approximation and tidal dynamics; no spacetime discreteness is implied. Given an invariant bound $λ_{\max}\leλ_{\rm bound}$ on the electric Riemann eigenvalues $E_{ij}\equiv R_{\hat{0}i\hat{0}j}$ along freely falling worldlines, we prove (i)~a rigorous upper bound on accumulated geodesic deviation through any bounded curvature interior, controlled by $τ_*\equivλ_{\max}^{-1/2}$, and (ii)~the existence of a critical wavenumber $k_*\simτ_*^{-1}$ separating adiabatic from non-adiabatic perturbation transfer through high-curvature epochs, with Bogoliubov coefficients exponentially suppressed for $k\,τ_*\gg 1$. Both results depend only on the tidal bound (and, for mode transfer, on a mild timescale assumption for the curvature-driven effective potential) and are otherwise insensitive to metric details. For preparation, we collect the standard operational consequences of bounded curvature, including the accuracy-dependent local-inertial domain $L_{\rm LI}(\varepsilon)\sim\sqrt{\varepsilon}\, λ_{\max}^{-1/2}$ and, for conformally flat cores in four dimensions, the benchmark ratio $τ_*/L_*=24^{1/4}$ with $L_*\equiv K_{\max}^{-1/4}$. We quantify the robustness of this coefficient under departures from maximal symmetry via the Weyl-to-Kretschmann ratio $ε_C$. The general framework is validated numerically in the extremal Hayward geometry.

Entanglement is a central and subtle feature of quantum theory, whose structure and operational behavior can change dramatically when additional physical constraints, such as symmetries or superselection rules, are imposed. Such constraints can give rise to striking and counter-intuitive phenomena, including local broadcasting of entangled states and failures of entanglement monogamy. These effects naturally arise in tomographically nonlocal theories (like real quantum theory, twirled worlds, or fermionic quantum theory), where composite systems possess holistic degrees of freedom that are inaccessible to local measurements. In this work, we study entanglement in such theories within the framework of generalized probabilistic theories. We show that the failure of tomographic locality leads to two qualitatively distinct forms of entanglement, which we term $\textit{tomographically-local}$ entanglement and $\textit{tomographically-nonlocal}$ entanglement. We analyze the operational consequences of this distinction, proving that tomographically-nonlocal entanglement is useless for Bell nonlocality, steering, and teleportation, but sufficient for dense coding and perfectly secure data hiding. This framework clarifies the origin of several previously puzzling features of entanglement that arise when tomographic locality fails, as can happen even in quantum theory when one considers fermions or fundamental superselection rules.

We present a new family of multi-centered rotating black hole solutions in 5D vacuum Einstein gravity, providing explicit examples of cohomogeneity-three spacetimes. It is well known that, in the presence of two commuting Killing vector fields, the theory reduces to 3D gravity coupled to an $SL(3,\mathbb{R})$ nonlinear sigma model with five scalar fields. We show that the scalar fields of the extremal Myers-Perry solution can be expressed in terms of two harmonic functions on 3D flat space, and that promoting these functions to include multiple sources yields explicit multi-centered extremal Myers-Perry black holes located at arbitrary positions. Each center forms a smooth $S^3$ Killing horizon, provided that the rotation parameters satisfy $|j_i|<1/2$. We further demonstrate that all curvature singularities are hidden behind the horizons and that no closed timelike curves arise on or outside the horizons. The solutions are asymptotically locally Minkowski in the sense that constant-time hypersurfaces are asymptotically locally Euclidean (ALE). As a concrete example, we consider a binary configuration, examine its rod structure, and demonstrate the absence of conical singularities between the two black holes, indicating that they are supported by an intermediate bubble region separating them.

The Event Horizon Telescope (EHT) image of M87* provides a direct test of strong-field gravity, measuring an angular shadow diameter $θ_d = 42 \pm 3~μ\mathrm{as}$ and a circularity deviation $ΔC \leq 0.1$. Such observations allow quantitative tests of the Kerr paradigm and of possible deviations from the no-hair theorem. In scalar-tensor extensions of gravity, black holes may possess primary scalar hair, introducing an additional independent parameter beyond mass and spin. In this work, we construct rotating black hole solutions with primary scalar hair in beyond Horndeski gravity and analyze their photon regions and shadow formation. We show that the scalar hair parameter $Q$ induces characteristic modifications of the shadow, and in particular negative $Q$ enlarges the shadow and reduces its oblateness, while positive $Q$ shrinks and enhances its distortion. Modeling M87* within this framework and imposing the EHT bounds on $θ_d$ and $ΔC$, we determine the viable $(a,Q)$ parameter space. We find that current observations do not exclude rotating black holes with primary scalar hair, although the allowed region is significantly restricted for $Q>0$. Finally, the scalar-hair-induced deviations are of order $\mathcal{O}(μ\mathrm{as})$, placing them near the sensitivity threshold of present instruments and within reach of next-generation horizon-scale imaging.

Is it possible to induce an effective generalized uncertainty principle (GUP) emerging from geometry and reinterpret the gravitational GUP as the effective uncertainty relation induced by microscopic horizon geometry? More broadly, is it possible to develop a notion of quantum gravity emerging from geometry? We will give a positive answer, but with important caveats.

In conventional lasers, the emitters are typically incoherent, radiating photons independently; in superradiant lasers, many coherent emitters radiate photons collectively, but they essentially do not interact with each other. Here, we present the concept of quantum many-body lasers, in which the emitters interact coherently and radiate collectively. In this proof-of-concept study, we consider a cavity coupled to many pumped spin-1/2 emitters with all-to-all interaction. We find that the squeezing induced by the coherent many-body interaction can be transferred from the spins to photons through superradiant lasing. This work illustrates the concept of using a pumped quantum many-body system to generate bright quantum light with quantum correlations beyond conventional optical coherence, which can facilitate quantum technologies and the study of nonlinear optics in the quantum realm.

Many attempts to introduce fundamental nonlocality into quantum (or classical) field theory are based on the assumption that exponentials of the d'Alembertian are positive-definite, so that these operators can be employed without engendering the Ostrogradskian instability associated with higher derivative Lagrangians. {\bf This assumption is false.} Working in the simple context of a 1-dimensional, point particle $q(t)$, I demonstrate that the equation $\exp[T^2 \tfrac{d^2}{dt^2}] q(t) = 0$ has an infinite number of rapidly oscillating, exponentially rising and falling solutions. This infinite kernel is in one-to-one correspondence with the ability to specify ``initial value data'' {\it arbitrarily} over {\it any} finite interval $t_1 < t < t_2$.

The performance of the Quantum Approximate Optimization Algorithm (QAOA) is closely tied to the structure of the dynamical Lie algebra (DLA) generated by its Hamiltonians, which determines both its expressivity and trainability. In this work, we show that classical symmetries can be systematically exploited as a design principle for QAOA. Focusing on the MaxCut problem with global bit-flip symmetry, we analyze reduced QAOA instances obtained by fixing a single variable and study how this choice affects the associated DLAs. We show that the structure of the DLAs can change dramatically depending on which variable is held fixed. In particular, we construct explicit examples where the dimension collapses from exponential to quadratic, uncovering phenomena that do not appear in the original formulation. Numerical experiments on asymmetric graphs indicate that such reductions often produce DLAs of much smaller dimension, suggesting improved trainability. We also prove that any graph can be embedded into a slightly larger one (requiring only quadratic overhead) such that the standard reduced DLA coincides with the free reduced DLA, in most cases implying exponential dimension and irreducibility on the Hilbert space for reduced QAOA instances. These results establish symmetry-aware reduction as a principled tool for designing expressive and potentially trainable QAOA circuits.

In this work, we reformulate the relativistic BDNK (Bemfica-Disconzi-Noronha-Kovtun) diffusion equation in flux-conservative form, and solve the resulting equations in $(1+1)$D using both a second-order Kurganov-Tadmor finite volume scheme and physics-informed neural networks (PINNs). In particular, we introduce the SA-PINN-ACTO framework, which combines the self-adaptive PINN technique with an exact enforcement of initial and periodic boundary conditions through an algebraic transform of the network's raw output, allowing the network to focus solely on minimizing the PDE residual. We test both approaches on smooth and discontinuous initial data, for both trivial and dynamically evolving velocity and temperature BDNK backgrounds, and for two characteristic speeds. The SA-PINN-ACTO method matches the converged Kurganov-Tadmor solutions for smooth profiles, while for discontinuous profiles the errors increase, reflecting an expected limitation of PINNs near sharp gradients.

Spacetime singularities, in the sense that curvature invariants are infinite at some point or region, are thought to be impossible to observe, and must be hidden within an event horizon. This conjecture is called Cosmic Censorship (CC), and was formulated by Penrose. Here we review another type of CC where spacetime singularities are causally disconnected from the universe, because the throat of a wormhole ``sucks in'' the geodesics and prevents them from making contact with the singularity. In this work, we present a series of exact solutions to the Einstein--Maxwell--Dilaton equations that feature a ring singularity; that is, the curvature invariants are singular in this ring, but the ring is causally disconnected from the universe so that no geodesics can touch it. This extension of CC is called Wormhole Cosmic Censorship.

Fidelity-based quantum kernels provide a direct interface between quantum feature maps and classical kernel methods, but they can exhibit exponential concentration: with increasing system size or circuit expressivity, the Gram matrix approaches the identity and suppresses informative similarity structure. We present an empirical study of two mitigation strategies implemented in Qiskit: (i) local (patch-wise) kernels that aggregate subsystem similarities, and (ii) multi-scale kernels that mix local and global similarity across patch granularities. We benchmark baseline, local, and multi-scale kernels under matched preprocessing, splits, and SVM protocols on several tabular datasets, sweeping the feature dimension $d\in\{4,6,\dots,20\}$. We report concentration diagnostics based on off-diagonal kernel statistics, spectral richness via effective rank, and centered alignment with labels. Across datasets, local and multi-scale constructions consistently mitigate concentration and yield richer kernel spectra relative to the global fidelity baseline, while the impact on classification accuracy depends on the dataset and dimension.

Lie algebras provide a useful framework for theoretical analysis in quantum machine learning, particularly in hybrid quantum-classical learning. From the viewpoint of function approximation, expectation values of parameterized quantum circuits can be viewed as trigonometric polynomials whose accessible Fourier modes are determined by the spectra of the generators. In this study, we describe: (1) a minimax lower bound on the $ L^{2} $-approximation error over a Sobolev ball when the circuit's effective frequency set is contained in a radius-$K$ ball, which yields a scaling law of the form $ Ω(K^{\frac{d}{2} - r}) $ for $ r > \frac{d}{2} $ (assuming the target function belongs to the Sobolev space $ W_2^{r}(\mathbb{T}^{d}) $), and we also derive a Jackson-type upper bound on the approximation error of quantum circuits under Sobolev regularity of the target function, expressed in terms of an effective bandwidth determined by generator spectral gaps; (2) a generator-selection rule motivated by enlarging the effective frequency set via non-commuting generators; and (3) a simple heuristic metric based on the trace component of generators, aimed at characterizing training behaviors related to barren plateaus. Simulation experiments on toy problems illustrate the practical implications of the frequency-spectrum perspective and the proposed heuristics.

The string-inspired running vacuum model (StRVM) of inflation is based on a Chern-Simons (CS) gravity effective action, in which the only four-spacetime-derivative-order term is a gravitational anomalous CS Pontryagin density coupled to an axion. In this work, we revisit curvature-squared string-inspired effective actions, from the point of view of appropriate local field redefinitions, leaving the perturbative string scattering matrices invariant. We require simultaneously unitarity and torsion interpretation of the field strength of the Kalb-Ramond antisymmetric tensor, features characterising the (3+1)-dimensional StRVM Cosmology. Unlike the higher dimensional case, the above feature is possible in the context of (3+1)-dimensional spacetimes, obtained after string compactification. We demonstrate that the unitarity and torsion-interpretation requirements lead to a single-type of extra four-derivative terms in the effective gravitational action, not discussed in the previous literature of StRVM, which however is shown to be subleading by many orders of magnitude, compared to the terms of the StRVM framework. Hence, its presence has no practical implications for the relevant inflationary (and, hence, postinflationary) physics of the StRVM. This demonstrates the phenomenological completeness of the StRVM cosmological scenario, which is thus fully embeddable in the UV complete (quantum-gravity compatible) string theory framework.

In a number of physically relevant contexts, a quantum system interacting with a decohering environment is simultaneously subjected to time-dependent controls and its dynamics is thus described by a time-dependent Lindblad master equation. Of particular interest in such systems is to understand the circumstances in which, despite the ability to apply time-dependent controls, they lose information about their initial state exponentially with time i.e., their dynamics are exponentially contractive. While there exists an extensive framework to study contractivity for time-independent Lindbladians, their time-dependent counterparts are far less well understood. In this paper, we study the contractivity of Lindbladians, which have a fixed dissipator (describing the interaction with an environment), but with a time-dependent driving Hamiltonian. We establish exponential contractivity in the limit of sufficiently small or sufficiently slow drives together with explicit examples showing that, even when the fixed dissipator is exponentially contractive by itself, a sufficiently large or a sufficiently fast Hamiltonian can result in non-contractive dynamics. Furthermore, we provide a number of sufficient conditions on the fixed dissipator that imply exponential contractivity independently of the Hamiltonian. These sufficient conditions allow us to completely characterize Hamiltonian-independent contractivity for unital dissipators and for two-level systems.

Waveguide quantum electrodynamics platforms have emerged as promising candidates for exploring and implementing non-Markovian quantum phenomena. In this work, we investigate the formation and dynamics of superpositions of bound states in a cavity array waveguide coupled to two spatially separated quantum emitters. By tuning the system parameters, we show that spontaneous emission can drive the system into non-local equilibrium states in which both photonic and emitter populations exhibit persistent oscillations. These states arise from the coexistence of bound states embedded in the energy continuum and bound states outside it, leading to hybrid oscillatory modes. We analytically derive the conditions required for the emergence of these states, numerically simulate their formation through spontaneous emission, and predict their long-time behaviour. Our results demonstrate that such bound-state superpositions enable the generation of emitter-emitter interaction through free evolution, while supporting oscillatory breathing modes of the photon density between the emitters.

Current architectures for neutral-atom arrays utilize devices such as acousto-optic deflectors (AODs) and spatial light modulators (SLMs) to multiplex a single classical control line into N qubit control lines. Dynamic control is speed-limited by the response time of AODs, and geometrically constrained to respect a product structure, limiting motion to row-by-row or column-by-column moves. We propose an optical rastering device that can produce any 2D pattern, not limited to grids, at 1 MHz refresh rates. We demonstrate a design with a resolution of 40 x 40 that can be further scaled up to 100 x 100 to match existing and future neutral atom devices. The ability to simultaneously transport atomic qubits in arbitrary directions will enhance qubit connectivity, enable more efficient circuits, and may have broader applications ranging from LiDAR to fluorescence microscopy.

We implement the Lights Out problem on a 2D grid and on Mobius ladder graphs and evaluate the performance of Grover's search on real quantum hardware. We use two instances using 9 and 16 qubits, and implement them on publicly available quantum hardware by IBM and IQM. Our experiments show improvements in IBM hardware between the Heron r1 and Heron r2 generations, highlighting progress in IBM hardware during the 2023-2024 period. The Lights Out circuits produced output distributions close to uniform on IQM devices. To diagnose device limitations, we additionally ran a small Grover SAT baseline, finding that IQM Garnet performs more reliably than other tested IQM devices. We also observed that QPUs of the same manufacturing revision can differ significantly in performance (a newer device is not guaranteed to be better), and that calibration has a significant impact on the performance of quantum devices, so the choice of device strongly depends on calibration quality.

Power-law inflation with $a(t) \propto t^m$ is conceptually simple and predicts a scalar tilt $n_s = 1 - 2/m$ compatible with CMB data, but in four-dimensional Einstein gravity it typically yields a tensor-to-scalar ratio $r = 16/m$ that is too large to satisfy current bounds. We show that a minimal extension based on fractional scalar-field cosmology resolves this tension. Introducing a fractional order $α\neq 1$ generates non-local (memory) corrections in the Friedmann and Klein-Gordon dynamics that suppress $r$ while keeping $n_s$ essentially unchanged. We derive an explicit mapping $α(n,m)$ and recover the standard power-law limit as $α\to 1$. For observationally favored values $α\approx 0.8$-$0.9$ in four dimensions we obtain $n_s \approx 0.965$ and $r \lesssim 0.04$, bringing power-law inflation into agreement with data. The scalar potential follows self-consistently as an exponential, and a dynamical-systems analysis shows the fractional power-law solutions form stable inflationary attractors over the viable parameter range. These results establish fractional power-law inflation as a predictive and testable framework, with clear targets for forthcoming CMB polarization measurements.

We study Gaussian continuous tensor network states (GCTNS) - a finitely-parameterized subclass of Gaussian states admitting an interpretation as continuum limits of discrete tensor network states. We show that, at short distance, GCTNS correspond to free Lifshitz vacua, establishing a connection between certain entanglement properties of the two. Two schemes to approximate ground states of (free) bosonic field theories using GCTNS are presented: rational approximants to the exact dispersion relation and Trotterized imaginary-time evolution. We apply them to Klein-Gordon theory and characterize the resulting approximations, identifying the energy scales at which deviations from the target theory appear. These results provide a simple and analytically controlled setting to assess the strengths and limitations of GCTNS as variational ansätze for relativistic quantum fields.

We study a sequential coherent side-channel model in which an adversarial probe qubit interacts with a target qubit during a hidden gate sequence. Repeating the same hidden sequence for $N$ shots yields an empirical full-correlation record: the joint histogram $\widehat{P}_g(b)$ over probe bit-strings $b\in\{0,1\}^k$, which is a sufficient statistic for classical post-processing under identically and independently distributed (i.i.d.) shots but grows exponentially with circuit depth. We first describe this sequential probe framework in a coupling- and measurement-agnostic form, emphasizing the scaling of the observation space and why exact analytic distinguishability becomes intractable with circuit depth. We then specialize to a representative instantiation (a controlled-rotation probe coupling with fixed projective readout and a commuting $R_x$ gate alphabet) where we (i) derive a depth-dependent leakage envelope whose maximizer predicts a "Goldilocks" coupling band as a function of depth, and (ii) provide an operational decoder, via machine learning, a single parameter-conditioned map from $\widehat{P}_g$ to Alice's per-step gate labels, generalizing across coupling and noise settings without retraining. Experiments over broad coupling and noise grids show that strict sequence recovery concentrates near the predicted coupling band and degrades predictably under decoherence and finite-shot estimation.

Niobium Titanium nitride (NbTiN) based superconducting nanowire single photon detectors (SNSPDs) are known for their high performance across a wide spectral range, from the X-ray to the mid-infrared. Nonetheless, fabrication challenges and performance degradation attributable to surface oxidation and lack of uniformity in films thinner than 5 nm remain a significant barrier for achieving high-quality detectors. In this work, we study the influence of a Silicon capping layer on film properties and on the performance of SNSPDs. A Silicon capping layer effectively suppresses oxidation and increases the superconducting transition temperature. This enables superconductivity in films as thin as 3 nm at 3 K, increases critical current in patterned nanowires and significantly extends the saturation plateau from the visible to the near infrared (up to 2050 nm): These detectors maintain sub-50 ps timing jitter, even for nanowires as wide as 250 nm and with detection areas of 20x20μm2. Our results establish that thinner films protected by a capping layer allow for the fabrication of wider wires, decreasing nanofabrication challenges and extending the operating temperature range for efficient single photon detection.

Tensor network methods leverage the limited entanglement of quantum states to efficiently simulate many-body systems. Alternatively, Clifford circuits provide a framework for handling highly entangled stabilizer states, which have low magic and are thus also classically tractable. Clifford tensor networks combine the benefits of both approaches, exploiting Clifford circuits to reduce the classical complexity of the tensor network description of states, with promising effects on simulation approaches. We study the disentangling power of Clifford transformations acting on tensor networks, with a particular emphasis on entanglement cooling strategies. We identify regimes where exact or heuristic Clifford disentanglers are effective, explain the link between the two approaches, and characterize their breakdown as non-Clifford resources accumulate. Additionally, we prove that, beyond stabilizer settings, no Clifford operation can universally disentangle even a single qubit from an arbitrary non-Clifford rotation. Our results clarify both the capabilities and fundamental limitations of Clifford-based simulation methods.

Approximating the dynamics given by a complex many-body Hamiltonian with a simpler effective model lies at the interface of quantum Hamiltonian learning and quantum simulation. In this context, quantum generative adversarial networks (QGANs) have been shown to outperform standard Trotter-based approximations. However, their performance is often hindered by training plateaus and local minima that become increasingly severe with system size. To overcome these limitations, we propose an entanglement-assisted learning strategy that couples a single randomly initialized auxiliary qubit to the learning system at an intermediate stage of the training process. The interplay between randomization and entanglement significantly enhances the learning performance of the protocol.

Supermassive black hole growth plausibly occurs via runaway astrophysical black hole mergers in nuclear star clusters that form intermediate mass black hole seeds at high redshifts. Such a model of Little Red Dots yields an order-of-magnitude higher rate of tidal disruption events than that of black hole captures. Our prediction, normalised to our proposed resolution of SMBH seeding, yields detectable TDE rates at high redshift. The resulting dense gas cocoons generate the nuclei of LRDs, each incorporating a central massive black-hole-star, with comparable masses in gas, stars, and massive black hole within a scale of around a parsec as inferred from the various spectral signatures.

We consider the recent proposal of steady state coherences in a single qubit in the case of a composite system-bath interaction. Based on a field theoretical approach we reanalyse the issue within a Redfield description. We find that the Redfield approach in accordance with a recent proposal yields steady state coherences but also violates the properties of a quantum map yielding 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 investigate the emergence of cosmic hysteresis in cyclic and bouncing cosmologies within the framework of reconstructed $f(T)$ gravity. In contrast to curvature-based modifications of General Relativity, teleparallel gravity attributes gravitation to spacetime torsion encoded in the torsion scalar $T$. By reconstructing viable $f(T)$ functions corresponding to analytically prescribed nonsingular bouncing scale factors and coupling the geometry to a minimally interacting canonical scalar field, we demonstrate that asymmetric scalar field dynamics between expansion and contraction phases give rise to a non-vanishing thermodynamic work integral $\oint p_φ\, dV$ over complete cycles. This hysteresis manifests as closed loops in the $(w_φ,a)$ plane, signifying thermodynamic memory and irreversibility. We derive the modified Friedmann equations, establish exact bounce and turnaround conditions, and discuss the implications of torsion-induced hysteresis for the cosmological arrow of time. Our results confirm that cosmic hysteresis is a generic feature of cyclic universes in modified gravity, extending beyond curvature-based theories.

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.

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.

Event horizons are a defining feature of black holes. Consequently, there have been many efforts to probe their existence in astrophysical black hole candidates, spanning ten orders of magnitude in mass. Nevertheless, horizons remain an obstacle to unifying general relativity and quantum mechanics, most sharply presented by the information paradox. This has motivated a proliferation of horizonless alternatives (black hole foils) that avoid event horizons and are therefore benign. We show that for typical accreting astronomical targets, largely independent of a foil's underlying microphysics, a horizonless compact surface will generically be ensconced within an optically thick, scattering dominated baryonic settling layer that efficiently reprocesses the kinetic energy of infalling matter into observable thermal emission. The emergent photosphere luminosity is driven toward the accretion-powered equilibrium value and is only weakly sensitive to the foil redshift. These atmospheres are convectively stable and naturally imply that the emitting photosphere forms at modest redshift even when the surface redshift is extreme. Moreover, local gas-surface interaction provides a microphysical lower bound on the effective base temperature, insulating the atmosphere from arbitrarily cold foils. The unknown properties of the foil enter only through local boundary conditions controlling baryon processing and thermal coupling at the surface, making the solutions broadly applicable to horizonless alternatives that do not invoke significant additional nonlocal interactions. Thus, under minimal assumptions (GR exterior and local surface interactions), horizonless foils are generically observationally exposed: the absence of a thermal photosphere directly constrains or rules out broad classes of such models.

We present an experimental characterization of coexistence-induced interference in QKD transmission arising from SpRS and FWM, validating a comprehensive semi-analytical model for accurate noise estimation. Experimental results show good agreement with theoretical predictions.

We investigate whether the spinor field can be eliminated from the Maxwell-Dirac equations by differential algebraic methods. A generic truncated power series solution is constructed, the prolonged system of the Maxwell-Dirac electrodynamics is linearized about the solution, and the ranks of the associated coefficient matrices are computed. The results indicate that, generically, the spinor components are uniquely determined by the electromagnetic field and its derivatives. This strongly suggests that differential-algebraic elimination of the spinor field is possible.

We investigate quantum-inspired tensor networks (QTNs) for approximating flow maps of hydrodynamic partial differential equations (PDEs). Motivated by the effective low-rank structure that emerges after tensorization of discretized transport and diffusion dynamics, we encode PDE states as matrix product states (MPS) and represent the evolution operator as a structured low-rank matrix product operator (MPO) in tensor-train form (e.g., arising from finite-difference discretizations assembled in MPO form). The MPO is applied directly in MPS form, and rank growth is controlled via canonicalization and SVD-based truncation after each step. We provide theoretical context through standard matrix product properties, including exact MPS representability bounds, local optimality of SVD truncation, and a Lipschitz-type multi-step error propagation estimate. Experiments on one- and two-dimensional linear advection-diffusion and nonlinear viscous Burgers equations demonstrate accurate short-horizon prediction, favorable scaling in smooth diffusive regimes, and error growth in nonlinear multi-step predictions.

I discuss the physical basis of classical mechanics, such as expressed commonly using the framework of Newton's Principia. Newton's formulation of the laws of motion is seen to have quite a few ambiguities and shortcomings. Therefore I offer an alternative set of laws, based in particular on ideas of his contemporaries Huygens and Leibniz with a crucial addition by Ludwig Lange, which avoids the problems with Newton's formulation. It is shown that from these laws of motion all the usual results of classical mechanics, as it concerns the motion of idealized point masses, can be rederived. The application of these principles to relativistic point particles is discussed.

We continue the study of flux tubes in confining gauge theories placed in a rigid AdS background, focusing on the three-dimensional case. Our analysis is performed in the large-radius regime, where effective string theory provides a good approximation of the dynamics. Using a combination of techniques, primarily the analytic transcendentality ansatz bootstrap, we compute observables up to two-loop order in the expansion in powers of the string length over the AdS radius, which constitutes the main result of this work. Finally, we employ Padé resummations to explore the possible compatibility of our results with a smooth interpolation of observables between large-radius AdS and small-radius AdS, in which gauge theory is weakly coupled.

We study the interaction of a kaon with the $f_1(1285)$ resonance, assuming that the $f_1(1285)$ is a molecular state generated by the $K \bar K^*, \bar K K^*$ interaction, evaluating the scattering amplitude, the scattering length and effective range of the $K f_1$ system. The scattering amplitude develops a resonant structure approximately $10$ MeV below the $K f_1$ threshold, with a width of around $15$ MeV. The corresponding correlation function has the distinctive shape of a system with a bound state close to threshold. We also show that the interaction of the $K f_1$ system is differs significantly from the one obtained assuming that the $f_1(1285)$ is an elementary particle. This provides motivation to continue the search for these observables, already initiated by the measurement of the $p f_1(1285)$ correlation function by the ALICE collaboration.

This study investigates how target geometry and material influence pion and muon production from an 8 GeV proton beam, in support of target-system design for a muon collider demonstrator. A 2 m long, 0.7 m radius solenoid with a 5 T peak magnetic field is used to capture secondary particles, with the target positioned at its center. We examine how variations in target radius, length, and material affect secondary-beam yield and emittance at the solenoid exit. In parallel, we evaluate temperature rise within the target to assess material limitations and guide future work on thermal and structural survivability. The results provide initial intuition for optimizing both particle yield and target durability in muon collider front-end systems.

We consider the phenomenon of string breaking in the context of the spatial Wilson loops using the gauge/string duality. In particular, we discuss the impact of light flavors on the pseudopotential. We also introduce the notion of the spatial string breaking distance and estimate it for $SU(3)$ gauge theory in the temperature range $0\,\text{-}\,3\,T_c$.

We develop the theoretical model that describes dynamic non-equilibrium effects of external inertial and axion fields in a system of particles with spin. The possibility of using the spin density and the current density of non-relativistic quantum particle systems for the detection of the hypothetical axion-like dark matter is discussed. The resulting closed system of dynamic equations encompasses the continuity equation, the momentum balance equation, and the spin density evolution equation, accounting for the influence of the spin-rotation coupling and the external axion fields. The new formalism opens up new perspectives for an experimental search of dark matter axions.

We perform a new global QCD analysis of unpolarized parton distribution functions (PDFs) in the nucleon from proton, deuteron and $A=3$ data, including recent measurements of $^3$He/$D$ and $^3$H/$D$ cross section ratios from the MARATHON experiment at Jefferson Lab. Simultaneously inferring the PDFs and nucleon off-shell corrections allows both to be determined consistently, without theoretical assumptions about the isospin dependence of nuclear effects. The analysis provides strong evidence for the need of nucleon off-shell corrections to describe the $A=3$ data, and suggests the presence of both isoscalar and isovector nuclear effects in $A \leq 3$ nuclei. We find that the extracted EMC ratios of nuclear to nucleon structure functions for $A=2$ and 3 differ from those naively extrapolated from heavy nuclei down to low $A$.

Higher-Form Symmetries (HFS) of a closed bosonic M2-brane formulated on a compactified target space $\mathcal{M}_9 \times T^2$ are investigated. We show that there is an obstruction to the gauging of these global symmetries in the presence of background fields, a mixed 't~Hooft anomaly. Its cancellation is obtained by the inflow term constructed in terms of gauge fields which are flat connections on a $U(1)$-principal bundle and a torsion $\mathcal{G}_1^{\nabla_c}$-gerbe on the M2-brane worldvolume. The effect of these gauge structures together with non trivial \textit{winding} embedding maps ensures the breaking of the continuous HFS $U(1)$ symmetry to a discrete subgroup and a worldvolume flux condition on the M2-brane. A Wilson surface, identified with the holonomy Hol$_\nabla$ one of the Gerbe structures, the flat $\mathcal{G}_1^{\nabla_c}$-gerbe, is naturally introduced as the topological operator characterizing the M2-brane. The resulting topological operators realize discrete symmetries associated with the \textit{winding} and the flux/\textit{monopole} sectors, and their operator algebra is well-defined: the \textit{monopole} operator acts non trivially on a \textit{vortex-dressed} operator, while the winding operator acts on the pullback of the Wilson surface.

We present a lattice-QCD determination of the $Λ\to p$ vector and axial-vector form factors, providing theoretical input for studies of the semileptonic decay $Λ\to p\ell\barν_\ell$. The calculation is carried out on a single gauge ensemble with physical light, strange, and charm quark masses and delivers a precise determination of the complete set of transition form factors, including second-class contributions. Using these form factors, we compute decay rates for both the electronic and muonic channels, as well as their ratio, which offers a sensitive test of lepton-flavor universality and possible non-standard scalar or tensor interactions. This decay mode provides a theoretically well-controlled avenue for extracting the CKM matrix element $|V_{us}|$ from the baryon sector. Our estimate of $|V_{us}|$ is obtained by combining our recent lattice-QCD results with recent measurements of the relevant branching fraction reported by BESIII and LHCb.

In this work, we present a systematic study of the event characteristics and physics of neutrino-antineutrino discrimination associated with atmospheric neutrino charged-current interactions in large liquid scintillator detectors. This study encompasses the primary neutrino interactions, the sequential second interactions of final-state particles, and the final neutron captures. We carefully investigate the properties of final-state charged leptons and hadrons, providing distinct distributions of inelasticity and captured neutron multiplicity for both neutrino and antineutrino interactions. These distributions are employed to assess the quantitative performance of neutrino-antineutrino discrimination. Our findings lay the groundwork for atmospheric neutrino oscillation studies in large liquid scintillator detectors, particularly in the determination of neutrino mass ordering.

Refracton of particles (nucleons, nuclei, $γ$-quanta) in matter with polarized protons (nuclei) results in revealing coherent quasi-optical phenomenon of nuclear spin precession of particles (nuclei) in the pseudomagnetic field of matter with polarized spins and the phenomenon of birefringence of particles (nuclei) with spin $S \ge 1$. These phenomena can be observed and studied at complex NuclotronM-NICA. The similar effects for $γ$-quanta could be observed at LINAC accelerator. Quasi-optical coherent phenomena of spin rotation and dichroism are not caused by strong interactions only, the T-odd P-odd, T-odd P-even, T-even P-odd interactions also contribute. Limits for the value of these contributions at energies available at complex NuclotronM-NICA can be obtained by investigating all these phenomena. When studying polarized particles collisions, it is necessary to consider possible influences of quasi-optical phenomena of spin rotation and spin dichroism caused by nuclear precession and birefringence.

We present a detailed study of momentum broadening for high-energy partons traversing the Quark-Gluon Plasma (QGP), extending the Gyulassy-Levai-Vitev (GLV) formalism to include both all-path-length (APL) and sub-eikonal corrections. Traditional GLV calculations rely on the large separation distance and large formation time approximations, which are valid for large systems, but whose applicability in small systems such as pp and p/dA may fail. We derive analytic expressions for the momentum broadening distributions to first order in the opacity expansion, and perform a numerical investigation to quantify their impact. Our results show that the APL result reduces the low-momentum broadening and the sub-eikonal correction enhances the high-momentum broadening. The combined APL and sub eikonal correction show that the sub-eikonal correction mitigates the effect of the APL correction.

The effective potential for a scalar theory with $λφ^4$ interaction, coupled to a massless fermion through Yukawa interaction is calculated by summing over infinite number of two particle irreducible (2PI) diagrams of two different types and a 2PI diagram of a third type using Cornwall, Jackiw and Tomboulis (CJT) method. There is an inversion symmetry present in the effective potential under $φ\rightarrow -φ$. At large coupling it exhibits minima below the zero potential line and on either side of the maximum at $φ=0$. The fermion acquires a mass at any non-trivial, positive minimum breaking the inversion symmetry of the vacuum.

A search for the decay $B^+\to D_s^+ K^+K^-$ has been reported by the LHCb Collaboration using $pp$ collision data corresponding to an integrated luminosity of $4.8\,\mathrm{fb}^{-1}$, collected at center-of-mass energies of 7, 8, and 13~TeV, in which no amplitude analysis of the $K^+K^-$ subsystem was performed. In this work, we study the resonant contributions to the decay $B^+\to D_s^+ K^+K^-$ within the perturbative QCD (PQCD) factorization framework. Contributions from the $S$-wave resonances $f_0(980)$, $f_0(1370)$, and $f_0(1500)$, the $P$-wave resonance $φ(1020)$, and the $D$-wave resonances $f_2(1270)$ and $f_2(1525)$ are taken into account. By introducing the corresponding two-meson distribution amplitudes for the $K^+K^-$ system, we perform a complete perturbative analysis of the quasi-two-body decays $B^+\to D_s^+(R\to)K^+K^-$, where $R$ denotes an intermediate resonance, and present the first PQCD predictions for the associated branching fractions. Using the narrow-width approximation, we further extract the branching fractions of the corresponding two-body decays $B^+\to D_s^+R$. Our results are consistent with the available experimental measurements and previous theoretical studies. Finally, we find that direct CP asymmetries vanish for these quasi-two-body decays within the Standard Model, so that any experimentally observed nonzero CP asymmetry would constitute a clear signal of physics beyond the Standard Model.

We construct affine charts of a smooth projective toric variety which contain its nonnegative points, and which admit a closed embedding into the total coordinate space of Cox's quotient construction. We show that such positive charts arise from smooth subcones of the nef cone. To each positive chart we associate an algebraic moment map, the fibers of which are the critical points of a monomial function in Cox coordinates. This work provides a toric framework for the theory of $u$-equations in positive geometry.

We consider the V-A Fermi interaction and apply an earlier developed method for summing up the leading asymptotics for scattering amplitudes in non-renormalizable theories. We consider the amplitude of fermion-antifermion scattering and derive the corresponding RG equation that sums the leading logarithmic contributions just like in renormalizable models. Numerical solution of this equation in the asymptotic regime $s\sim t\sim u \sim E^2 \to \infty$ leads to amplitude logarithmically decreasing with energy, thus restoring the unitarity violated at the tree level.

I study the event-by-event fluctuation and rapidity decorrelation of the mean transverse momentum $\spt$, which has recently been proposed as a sensitive probe of the equation of state at finite baryon density. The investigation reveals that, in a baryon-rich medium, the event-by-event fluctuation of the mean transverse momentum is driven by the combined effects of energy-density and net-baryon-density fluctuations. Consequently, the rapidity dependence of this observable provides a promising handle to probe the three-dimensional structure of both energy and baryon density profiles. Previous studies have shown that $\spt$ decorrelation along rapidity is largely insensitive to shear and bulk viscosity; however, its dependence on baryon diffusion, another key transport coefficient in baryonic matter, has not been explored. I find that baryon diffusion has a negligible impact, establishing this observable as a robust probe of the equation of state. Furthermore, I present predictions for identified hadrons and observe a pronounced splitting in the rapidity decorrelation of mean transverse momentum between protons and antiprotons, indicating different transverse flow dynamics for baryons and antibaryons.

This report presents the design, characterization, and application of a high-sensitivity optical detection system based on plastic scintillators coupled to Multi-Pixel Photon Counters (MPPCs). The primary objective was to evaluate the performance of MPPCs (Silicon Photomultipliers) as robust, low-voltage alternatives to traditional photomultiplier tubes for detecting faint scintillation light. The optoelectronic properties of the sensors were analyzed, including single-photoelectron gain calibration and dark count rate measurements, to optimize the signal-to-noise ratio. By embedding wavelength-shifting fibers to enhance light collection efficiency, the system was configured into a three-fold coincidence telescope. The angular distribution of the cosmic ray muon flux was measured to validate the detector's stability and geometric acceptance. Fitting the experimental data to a $\bm{\cos^n(θ)}$ distribution yielded an angular exponent of $\bm{n = 1.44 \pm 0.06}$, consistent with literature values. These results demonstrate the efficacy of the MPPC-scintillator coupling for precise photon counting and timing applications in high-energy physics instrumentation.

We study Higgs-boson pair production in the Standard Model Effective Field Theory (SMEFT) up to dimension six and in the Higgs Effective Field Theory (HEFT) at leading order in the effective theory expansion, and assess which description is appropriate in concrete UV scenarios. Motivated by "Loryon"-inspired models, we compare the Higgs pair production cross sections predicted by the full models to their SMEFT and HEFT counterparts. We identify regimes in which the two EFTs provide comparable descriptions, and clarify the limits required for their couplings to match. We also find that, for parts of parameter space in some of these models, HEFT can reproduce Higgs pair production more accurately than SMEFT, highlighting di-Higgs measurements as a potential probe of non-linear electroweak dynamics.

We study error correcting codes that construct the Narain lattices of heterotic strings as code lattices. We identify, in both $E_8\times E_8$ and Spin$(32)/Z_2$ heterotic strings, a pair of a binary code and a set of the corresponding metric, B field, and background gauge field, such that the lattice constructed from the binary code by Construction A coincides with the Narain lattice. We also construct heterotic Narain lattices using codes over $F_3$ and $F_5$ by Construction A${}_C$ and "Construction A${}_g$" with $g=SU(5)$, respectively. As a bi-product, we also clarify the relationship between codes that construct Euclidean even self-dual lattices and NSR-fermions, where the $Z_2$ inversion structure of the generator matrices plays a significant role.

Studies of the electromagnetic production of strange quarks began in the 1950s as something of a curiosity that puzzled experimentalists and theorists alike. As the datasets increased, concomitant advances in theoretical models were realized. A paradigm shift occurred in the 1990s with the development of second-generation facilities at ELSA, MAMI, SPring-8, and JLab, which brought nuclear physics experiments forward by orders of magnitude in counting statistics compared to the first-generation efforts. This was an utter boon to strangeness physics investigations, and to date, more than 50 dedicated experiments in kaon photo- and electroproduction have been completed at facilities around the world, leading to a host of experimental observables that have enabled significant advances in the exploration of strongly interacting systems that decay via $s\bar{s}$ quark pair creation. This review was designed to provide the first-ever in-depth overview of both the experimental and theoretical progress in the field of the electromagnetic production of strangeness. This work looks back over 70 years of past developments, discusses ongoing work and near-term plans, and details future possibilities being considered for third-generation facilities. Throughout this work, the primary impacts of these explorations are highlighted, along with connections to a wide range of related phenomenological applications. An important goal of this review is to provide a complete, self-contained guide into this field prepared at a level that is relevant for both new and seasoned scientists, whether experimentalists, phenomenologists, or theorists, to better understand what has been accomplished by so many dedicated folks-each building on what has come before-and to appreciate the exciting future potential for continued studies in this area. A more complete abstract is provided in the paper.

Average of exponential ${\rm Tr}_R e^X$, i.e. of a group rather than an algebra character, in Gaussian matrix model is known to be an amusing generalization of Schur polynomial, where time variables are substituted by traces of products of non-commuting matrices ${\rm Tr} \left(\prod_i A_{k_i}\right)$ and are thus labeled by weak compositions. The entries of matrices $A_k$ are made from extended Laguerre polynomials, what introduces additional difficulties. We describe the generic sum rules, which express arbitrary traces through convolutions of a single Laguerre polynomial $L_{N-1}^1(z_{k_i})$, what is a considerable simplification.

In this paper we present the first application of the MiNLO method to the calculation of QED NLO corrections to the production of a neutral vector boson in Drell-Yan processes. We consider only the case of initial-state radiation, when the Z boson decays into neutrinos. We illustrate the abelianization procedure of the MiNLO formulae and discuss the impact that it has on the differential cross section. We then propose a variant of the MiNLO formulae in order to circumvent some of the problems that arise when dealing with QED emissions. Since this is a case study, we use ad-hoc parton distribution functions and a larger value of the electromagnetic coupling constant, in order to emphasize potential discrepancies with respect to the expected behavior of the MiNLO formulae. We quantify the uncertainties connected with the proposed method, also for a the physical value of the electromagnetic coupling. The study presented here is a necessary first step towards incorporating full electroweak effects into the MiNNLOPS framework.

We investigate the effects of the tensor polarizability of a nucleus on thebound-state energy levels, and obtain a general formula for the contribution of the tensor polarizability to the energy levels in two-body bound systems. In particular, it is demonstrated that the tensor polarizability leads to mixing between states with different orbital angular momenta. The effect of tensor polarizability is evaluated for the hyperfine-structure components of P states and for the mixing of S and D states in muonic deuterium.

We explore the sensitivity of the SNO experiment to light dark matter particles $χ$ with spin-dependent interactions with nucleons. We show that the pair-production of MeV scale dark matter is possible in heavy water (CANDU) reactors via ${\rm D}(n,χ\barχ)^3{\rm He}$, and calculate the expected rate within the simplest models of $χ$-nucleon interactions. %Heavy water nuclear reactors serve as an excellent production method for spin-dependent dark matter. Owing to a sizable $Q$-value for this reaction, a large fraction of DM particles produced this way are above the threshold for deuteron disintegration, ${\rm D}(χ,χ)np$, which adds to the SNO neutral current signal. Evaluating the CANDU-to-SNO scheme for the production and detection of DM, we derive novel constraints for the $χ$-nucleon spin-dependent cross sections, showing that cross sections above $σ_{χp} \sim 10^{-33}\,{\rm cm}^{2}$ are generally excluded if $m_χ\leq1.5$\,MeV. An isospin-mirror reaction will occur in the Sun, and for the kinematically allowed region it excludes a portion of parameter space with cross sections on the order $10^{-37}\,{\rm cm}^{2}$. We also evaluate the potential sensitivity of small ``near" detectors placed in close proximity to a CANDU reactor to search for a coherent nuclear recoil, finding subdominant sensitivity.

The time-dependent modulation of the event rate in dark matter direct detection experiments, arising from the motion of the Earth with respect to the Galactic rest frame, is a distinctive signature whose observation is crucial for claiming a discovery of dark matter. While annual modulation has been well studied for decades, daily modulation due to the Earth's rotation has attracted increased attention recently due to the identification of anisotropic solid-state detector materials that yield a direction-dependent scattering rate without sacrificing the overall rate. We perform a statistical analysis of daily modulation in dark matter scattering experiments, with the goal of maximizing the statistical significance of a modulating signal in the presence of an unknown background rate, which may be either flat (non-modulating), or modulating over a 24-hour period with a known or unknown phase. In the background-dominated regime, we find that the discovery significance scales as $f_\text{RMS} \sqrt{T}$, where $T$ is the total exposure time and $f_\text{RMS}$ is the root-mean-square modulation amplitude; in particular, the significance continues to improve with exposure rather than saturating due to systematic uncertainties in the background rate. Using anisotropic trans-stilbene detectors for sub-GeV dark matter as a benchmark example, we provide prescriptions for optimizing the significance for a given total detector mass and location. In an example analysis using three detectors, optimizing the detector orientations can reduce the required exposure by a factor of $\sim 5$ for a desired discovery or exclusion significance, even after profiling over an unknown modulating background phase.

Modern scientific instruments operate under increasingly extreme constraints on bandwidth, latency, and power. Inference at the sensor edge determines experimental data collection efficiency by deciding which information to save for further analysis. Particle tracking detectors at the Large Hadron Collider exemplify this challenge: pixelated silicon sensors generate rich spatiotemporal ionization patterns, yet most of this information is discarded due to data-rate limitations. Concurrently, advancements in co-design tools provide rapid turn-around for incorporating machine learning into application-specific integrated circuits, motivating designs for particle detectors with new integrated technologies. We demonstrate that neural networks embedded in the front-end electronics can infer charged-particle kinematic parameters from a single silicon layer. We regress hit positions and incident angles with calibrated uncertainties, while satisfying stringent constraints on numerical precision, latency, and silicon area. Our results establish a path toward probabilistic inference directly at the edge, opening new opportunities for intelligent sensing in high-rate scientific instruments.

We study constraints imposed by four-dimensional unitarity (formalised as graded unitarity in recent work by the first author) on possible ${\mathcal W}_3$ vertex algebras arising from four-dimensions via the SCFT/VOA correspondence. Under the assumption that the $\mathfrak{R}$-filtration is a weight-based filtration with respect to the usual strong generators of the vertex algebra, we demonstrate that all values of the central charge other than those of the $(3,q+4)$ minimal models are incompatible with four-dimensional unitarity. These algebras are precisely the ones that are realised by performing principal Drinfel'd--Sokolov reduction to boundary-admissible $\mathfrak{sl}_3$ affine current algebras; those affine algebras were singled out by a similar graded unitarity analysis in \cite{ArabiArdehali:2025fad}. Furthermore, these particular vertex algebras are known to be associated with the $(A_2,A_q)$ Argyres--Douglas theories.

Quark portal interactions $qqqN$ with a light singlet fermion $N$ make baryon number testable through missing transverse energy (MET). We find that present LHC data constrain scales up to 10 TeV (MET plus jet) and 15 TeV (MET plus top). With increasing mass, or larger portal couplings, $N$ becomes less long-lived, and gives clean displaced vertex signatures, which encourage dedicated searches. We also narrow down viable mesogenesis models with color triplet scalars to a mass range $\sim y \, (2-3) \, \text{TeV}$ and couplings $y$ to $b$- and second generation quarks, a window that can be scrutinized by HL-LHC. Interactions also induce rare decays of type baryon (meson) to meson (baryon) plus invisible, which complement high-$p_T$ searches and can prove baryon number violation. We explore charm decays $Λ_c \to (π,K) + \mathrm{invisible}$. Their branching ratios are subject to sizable hadronic uncertainties and require high luminosity flavor facilities such as a Tera-Z facility (FCC-ee, CEPC). Branching ratios of top quarks into one or two $b$-jets plus $N$ can reach few$\times 10^{-6}$.

The recent light-ion collision programme at RHIC and the LHC provides a unique opportunity to investigate the onset of quark-gluon plasma formation and parton energy loss in small systems. A quantitative interpretation of emerging jet quenching measurements requires precise control over cold nuclear matter (CNM) effects, which modify hard-process cross sections independently of any hot-medium dynamics. In this work, we present a comprehensive set of perturbative QCD baseline calculations for nuclear modification factors ($R_{AA}$) in proton-oxygen (pO), oxygen-oxygen (OO) and neon-neon (NeNe) collisions at LHC energies. The study includes charged hadron, neutral pion, prompt photon, and electroweak-boson production computed at next-to-leading order using a broad set of recent nuclear parton distribution functions (nPDFs). We demonstrate that CNM effects alone can induce sizeable suppressions in light-ion systems, with large associated nPDF uncertainties that currently limit the quantitative extraction of parton energy loss. To address this limitation, we explore a range of multi-cross-section ratios in which CNM effects and their uncertainties largely cancel. In particular, ratios of neutral pion $R_{OO}$ to prompt photon $R_{OO}$ or charged hadron $R_{OO}$ to $R_{pO}^2$ provide theoretically robust observables with substantially reduced nPDF uncertainties, thereby enhancing sensitivity to possible energy-loss signatures.

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.

Motivated by observational evidence from JWST and theoretical results from cosmological simulations, we use a simple parametric, phenomenological model to test to what extent bursty star formation with standard Initial Mass Function, no continuous star formation, no mergers, \mr{and no dust} can account for the observed properties in the $M_{UV}$ vs $M_*$ plane of galaxies at redshifts $z>5$. We find that the simplest model that fits the data has a quiescence period between bursts $Δt \sim 100$~Myrs and the stellar mass in each galaxy grows linearly as a function of time from $z=12$ to $z=5$ (i.e., repeated bursts in each galaxy produce approximately equal mass in stars). The distribution of burst masses across different galaxies follows a power-law $dN/dM_* \propto M_*^α$ with slope $α\sim -2$. At $z>9-10$ the observed galaxy population typically had only one or two bursts of stars formation, hence the observed stellar masses at these redshifts (reaching $M_* \sim 10^{10}$~M$_\odot$), roughly represent the distribution of masses formed in one burst.

The nearly scale-invariant primordial power spectrum provides the standard initial conditions for cosmological perturbations. However, the largest scales remain only weakly constrained by CMB observations, leaving room for deviations such as an infrared (IR) cut-off. This possibility is further motivated by the persistence of large-scale CMB anomalies, most notably the low quadrupole power. In this work, we revisit several broad classes of phenomenologically motivated IR cut-off scenarios using parametrised functional forms of the primordial power spectrum. We confront these models with the latest CMB, BAO, and supernova data and derive updated constraints on the cut-off scale and associated features. Our results remain consistent with earlier studies, showing that although such models suppress power at low multipoles, the improvement in fit is marginal and does not overcome the associated parameter penalties. We therefore find no statistically significant evidence favouring IR cut-off models over the standard power-law spectrum with current data. We further explore the interplay between IR cut-off features and a possible increase in the reionisation optical depth, motivated by the recent CMB-BAO tension highlighted by DESI DR2 within the $Λ$CDM framework. We show that the additional freedom introduced by large-scale suppression is generally insufficient to support a substantial increase in optical depth, owing to the weak statistical preference for suppressed large-scale temperature power. Finally, we examine the implications of IR cut-off models for large-scale CMB anomalies by analysing the corresponding anomaly statistics within a Bayesian framework.

The redshifted 21-cm signal emitted by neutral Hydrogen (HI) is a promising probe to understand the evolution of the topology of ionized regions during the Epoch of Reionization (EoR). The topology of ionized regions allows us to infer the nature and properties of ionizing sources, i.e., early galaxies and AGNs. Traditional Fourier statistics, such as the power spectrum, help us quantify the strength of fluctuations in this field at different length scales but do not preserve its phase information. Analyzing the 21-cm brightness temperature field in the image domain retains its non-Gaussian characteristics and morphological information. One such approach is to track the coalescence of multiple ionized regions to form one contiguous ionized region spanning the universe. This is referred to as percolation, and its onset is quantified by a sharp rise in the value of the Largest Cluster Statistic (LCS) approaching unity. In this work, we carry out a percolation analysis of 21-cm brightness temperature fields by studying the redshift evolution of the LCS along a lightcone to distinguish between several simulated reionization scenarios. We have extended previous results on reionization model comparison from the analysis of coeval 21-cm maps to understand how the lightcone effect biases the observed percolation behavior and affects the distinguishability of the source models. We estimate the LCS of subvolumes of different sizes in the 21-cm lightcone maps and study their redshift evolution for different reionization scenarios using a moving volume approach. We find that the percolation transition inferred from a lightcone approaches that from the coeval box as we increase the bandwidth of the moving volume in all but one reionization scenario.

Strongly lensed supernovae (SNe) provide a powerful way to study cosmology, SNe and galaxies. Modelling the lens system is key to extracting astrophysical and cosmological information. We present adaptive-optics-assisted high-resolution images of SN Winny (SN 2025wny) in the J and K filters obtained with the Large Binocular Telescope (LBT). The LBT imaging confirms the presence of a fifth point source, whose colour is consistent with that of the other SN images at similar phases, while lens modelling robustly supports its interpretation as an additional image of SN~Winny. We measure the positions of the five SN images with uncertainties varying between 1 and 14 milliarcseconds. We build the first mass models using lenstronomy and GLEE, and explore three classes of mass models for the two lens galaxies G1 and G2. The optimal model class of the three is a singular isothermal ellipsoid for G1, a singular isothermal sphere for G2, and an external shear. We infer the enclosed masses within the Einstein radius as 4.61^{+0.06}_{-0.04} \times 10^{11}\,M_\odot for G1 and 1.01\pm0.02 \times 10^{11}\,M_\odot for G2. The lensing configuration by the two lens galaxies can produce two additional magnified SN images beyond the five observed ones; the exclusion of such model configurations further constrains the lens model parameters. Our model fits to the observed image positions with an RMS of ~0.0012" - 0.0025", within the observed positional uncertainties. The predicted magnifications of the multiple images vary between ~1.6 (for the faintest fifth image E) to ~10 (for the brightest image A). The predicted relative lensing magnifications of the multiple images do not match that of the observed within 2σuncertainties. The differences in the relative magnifications could be due to millilensing/microlensing. Our mass models form the basis for future analyses of this unique system. (abridged)

Cosmological vector fields are central to many early-Universe phenomena, including inflationary dynamics, primordial magnetogenesis, and dark-matter scenarios. However, constructing models able to generate cosmological magnetic fields while avoiding strong coupling, backreaction, and cosmic microwave background constraints remains challenging. We study a novel mechanism in which brief non--slow-roll phases during inflation amplify primordial magnetic fields at small scales, while maintaining theoretical consistency and observational viability. We incorporate parity-violating interactions in the vector sector and demonstrate, for the first time in a non--slow-roll framework, that chirality can significantly boost magnetic-field amplitudes and imprint distinctive polarization-dependent spectral features. We complement detailed numerical computations with an analytical treatment yielding compact expressions for chiral vector mode functions that reproduce the main spectral properties. We then develop a systematic formalism to evaluate the stochastic gravitational-wave background naturally induced at second order by these amplified fields, identifying both an intensity component and a circularly polarized contribution with characteristic frequency profiles. We discuss detection prospects with future multiband gravitational-wave observatories, showing that chiral signatures could provide a distinctive observational probe. Our results introduce new avenues for enhancing primordial magnetic fields and their associated gravitational-wave signals, opening promising possibilities for their future detection and interpretation, both with cosmological and gravitational wave probes.

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

White dwarfs (WDs), the evolutionary endpoints of most stars, can form through both single-star and binary channels. While single-star evolutionary models enable reliable WD age estimates, binary evolution introduces interactions that can accelerate WD formation and result in a variety of exotic WDs, which may exhibit strong magnetic fields, rapid rotation, or even serve as potential gravitational wave sources. Such systems offer valuable insights into magnetic field generation, angular momentum evolution, and compact object physics. Star clusters, with their approximately coeval populations, allow precise age determination of member WDs. If a WD's total age derived from single-star evolution exceeds that of its host cluster, it likely indicates a binary origin. In this study, we use \textit{Gaia} 5D astrometry to identify 439 WD candidates in 117 open clusters, with 244 likely formed via binary evolution. We discuss the possibility of dynamical ejection for WDs meeting only 2D (proper motion space) membership criteria. Spectroscopic observations further reveal a subset with strong magnetic fields and rapid rotation, supporting their binary evolutionary origin.

The nearby quasar, PG 1211+143, has one of the prototype examples of an ultra fast outflow (UFO), as seen in several past XMM-Newton and Chandra observations. In December 2024, PG 1211+143 was observed simultaneously with XRISM Resolve and XMM-Newton, allowing both the Fe K and soft X-ray outflows to be examined at high resolution simultaneously. The Resolve spectrum revealed a forest of Fe K band absorption lines from the UFO (Mizumoto et al. 2026), comprising of up to six discrete velocity components ranging from $v/c=-0.074$ to $v/c=-0.40$. Here we present the simultaneous XMM-Newton RGS (Reflection Grating Spectrometer) spectrum, where three lower ionization counterparts of the Fe K velocity zones are observed; at $v/c=-0.074, -0.12$ and $-0.33$. The soft X-ray absorbers tend to be somewhat less ionized than their Fe K counterparts, with their opacity mainly arising from Fe L shell lines and highly ionized Oxygen. From comparing the Resolve and RGS absorbers, we show that the outflow can be parameterized with a density profile varying with radius as $r^{-5/3}$, while the lower ionization zones likely originate from denser clumps of gas. Pure electron scattering appears insufficient to provide enough thrust to power the wind, unless sufficient low ionization gas capable of radiative line driving exists outside of the line of sight. Overall, PG 1211+143 provides further evidence for the clumpy nature of accretion disk winds, as was recently revealed in the quasar PDS 456 with XRISM.

The Euclid galaxy survey will provide unprecedented constraints on cosmology, but achieving unbiased results will require an optimal characterisation and mitigation of systematic effects. Among these, the intrinsic alignments (IA) of galaxies are one of the dominant contaminants of the weak lensing (WL) and galaxy-galaxy lensing (GGL) probes. In this work, we assess IA modelling choices for Euclid DR1 3x2pt analyses by comparing the performance of the two most commonly used IA models, nonlinear alignment (NLA) and tidal alignment tidal torquing (TATT), along with several variations. Our analyses combine three perspectives: i) the constraining power on the IA and cosmological parameters for each IA model, ii) the bias that results when the IA analysis model differs from the model used to generate the synthetic data vector, and iii) the degeneracies between IA and photometric redshift (photo-z) nuisance parameters. Among the IA models analysed, the redshift-dependent TATT model (zTATT) provides the most flexible description of IA, with a similar constraining power compared to simpler IA models, making it a well-motivated choice for Euclid DR1 3x2pt analyses.

The merger of a He white dwarf (WD) and a CO WD is the favored formation channel for R Coronae Borealis (RCB) stars. These stars exhibit ${^{16}}\mathrm{O}/{^{18}}\mathrm{O}$ ratios that are orders of magnitude lower than the solar value. However, it is not fully understood whether such low ${^{16}}\mathrm{O}/{^{18}}\mathrm{O}$ ratios can be achieved in WD merger remnants for the predicted lifetime of RCB stars of around $10^4\,\mathrm{years}$. In this work, we perform detailed nucleosynthesis calculations of a 3D magnetohydrodynamical simulation of a merger of a $0.3\,M_\odot$ He WD and a $0.6\,M_\odot$ CO WD for $4000\,\mathrm{s}$ at which point a steady state in temperature and density is reached. From this point, we follow several radial zones to study the long-term production of ${^{18}}\mathrm{O}$ and its variability throughout the burning region. We find that the asymmetric merger process leaves an imprint on the distribution of the abundances at the end of our hydrodynamic simulation. During the long-term evolution up to $100\,\mathrm{years}$, we observe ${^{16}}\mathrm{O}/{^{18}}\mathrm{O}$ ratios of order of unity, although the timescale on which ${^{18}}\mathrm{O}$ is destroyed again is highly location dependent. Importantly, our calculations suggest that in the outer layers of the burning shell, the dominant production channel is $^{14}\mathrm{C}(α,γ)^{18}\mathrm{O}$ instead of the commonly considered $^{14}\mathrm{N}(α,γ)^{18}\mathrm{F}(β^+)^{18}\mathrm{O}$ reaction, whereby the former can be sustained for longer periods of time. Furthermore, these outer regions do not reach the conditions necessary for fast $α$-captures in ${^{18}}\mathrm{O}$ to ${^{22}}\mathrm{Ne}$, thus being favorable to maintaining a low ${^{16}}\mathrm{O}/{^{18}}\mathrm{O}$ ratio.

We present the results of observations of the radio halo in the Coma galaxy cluster at 4.8 GHz performed with the Sardinia Radio Telescope. The radio halo in this cluster is detected for the first time at this frequency with a statistical significance higher than $3σ$. After the removal of the Radio Frequency Interference and of the discrete sources contribution, and after the correction for the Sunyaev-Zel'dovich effect, we estimate a flux density of $61\pm11$ mJy, higher than the value previously reported in literature at this frequency. By using the value we obtained, it is possible to estimate an integrated spectral index between 4.8 and 6.6 GHz of $α\sim1.17$, where $F(ν)\propto ν^{-α}$, indicating a possible higher-frequency slowdown of the spectral steepening observed between 1.4 and 4.8 GHz. Such a spectral behavior is compatible with turbulent re-acceleration if the seed electrons have a spectrum extending up to high energies, as in the case of continuous injection by hadronic interactions or dark matter annihilation. We also report the detection at 4.8 GHz of a polarized spot inside the halo, without an evident counterpart, already detected at 6.6 GHz.

We present an in-depth Chandra X-ray analysis of the galaxy cluster SPT-CL J0417$-$4748 (hereafter SPT J0417), at z = 0.58, with a focus on its thermodynamic properties and the apparent absence of central star formation. Utilizing a total Chandra exposure of 103 ks, we find that the large-scale X-ray morphology is consistent with a dynamically relaxed, cool-core system. The intracluster medium (ICM) shows a central density of 0.08+/-0.01 cm^{-3}, a central pseudo-entropy of 26^{+6}_{-5} keVcm^{2} and a central cooling time of 515^{+96}_{-75} Myr, values typical of massive cool-core clusters. Despite these conditions, no evidence of recent or ongoing star formation is detected in the brightest cluster galaxy (BCG). Spectral energy distribution (SED) fitting of DES photometry indicates that the bulk of the stellar population formed at z~1.25, with no significant star formation over the past ~3 Gyr, while optical spectra from Magellan show no [O II] emission. Complementary ASKAP radio and Spitzer infrared data indicate a lack of strong current AGN activity in the BCG. SPT J0417 exemplifies massive, relaxed, cool-core clusters in which cooling and star formation appear almost completely quenched, providing valuable insights into how AGN feedback regulates the long-term thermal balance of the intracluster medium.

The Orion Nebula Cluster (ONC) provides the closest example of star and planet formation in highly irradiated environments. It is a key region to study how UV radiation from massive stars drive mass-loss in protoplanetary disks through photoevaporation. Far-UV photons (6<E<13.6 eV) heat up the gas of the disk, forming a photodissociation region (PDRs). We use the NIRCam images from the PDRs4All program combined with those of the GTO program 1256 to extract key information on ONC disks. Specifically, the radii of the disks observed in silhouette against the bright background , the presence and positions of the dissociation fronts (DFs), the presence and positions of ionization fronts (IFs), intensities of Pa $α$ lines, and their near IR SEDs. From those, we construct a typology for ONC disks: Type I show an IF and DF nearly merged at the disk surface; Type II have their DFs at the disk surface and IFs at a distance of several 10s of astronomical units from the disk; and Type III also have their DF at the disk surface, but show no IF. For all disks, we find that PAH emission traces the PDR. We establish that the SEDs of candidate JuMBOs observed as part of the PDRs4All program are similar to the SEDs of Type III ONC disks except for JuMBO24, which is of Type I or II. A detailed look at this SED shows it is compatible with a young low mass binary star with an unresolved ionized disk : a microproplyd binary. We observe that the disk radius of ONC disks increases with increasing projected distance to the ionizing source interpreted as evidence of the truncation of the disks by the photoevaporation. The disk radii measured at IR wavelengths appear larger than at millimeter wavelengths, interpreted as evidence of the dust radial segregation within the disks. The thermal pressure within the PDRs of ONC disk increases with the FUV radiation field, but with a flatter slope.

We analyze an XRISM/Resolve observation of NGC1068, focusing on the Fe K$α$ and Fe K$β$ fluorescent lines and on the Fe XXV and Fe XXVI emission complexes. Line centroid energies, intrinsic widths, flux ratios, and constraints on the Compton shoulder are derived through local spectral fitting, and compared with atomic calculations and theoretical predictions. The centroid energies of the Fe K$α$ and Fe K$β$ lines tightly constrain the emitting material to be neutral or near-neutral. The observed Fe K$β$/K$α$ ratio, together with the stringent upper limit on the Compton shoulder ($\lesssim$8--11% of the core flux), disfavour reflection dominated by a homogeneous, classical Compton-thick medium, indicating that most of the neutral Fe K$α$ emission arises in optically thin or moderately Compton-thick gas. The Fe XXV and Fe XXVI emission lines exhibit remarkably large velocity widths, of several thousand km~s$^{-1}$. These broad profiles closely resemble the integrated optical and infrared [O III] and [O IV] lines associated with the large-scale biconical outflow, and are naturally interpreted as the X-ray signature of a more highly ionized, faster, and more spatially confined phase of the same outflow. The iron-K emission of NGC1068 reveals a stratified circumnuclear environment in which neutral and highly ionized components arise in physically distinct regions. The neutral Fe K fluorescence originates predominantly in optically thin or mildly Compton-thick material, despite the persistently Compton-thick line-of-sight obscuration, indicating a geometrically complex cold reprocessor. The highly ionized iron emission lines trace a fast component consistent with a warm bipolar outflow on parsec scales, whose large velocities and inferred energetics suggest that it may represent an efficient channel for feedback in a heavily obscured Seyfert galaxy.

We report multiwavelength observations of the highly offset (~22.4 kpc) SN 2024abvb, the sixth Type Icn supernova to date. With a peak magnitude of Mr = -19.55 +/- 0.11 mag, it is among the most luminous in the existing sample and shows similar colours and decline rates to other SNe Icn. The early optical spectra show a blue continuum with narrow C II features (vFWHM ~ 2000 km s^-1), consistent with a typical wind velocity of a Wolf-Rayet star. The absence of C III lambda 5696 emission at the time of explosion is consistent with a Type Ibn supernova; however, the lack of narrow He lines in both the optical and near-infrared spectra supports a SNe Icn classification. Unlike the majority of SNe Icn, we do not detect broad features in the late-time (7-21 days relative to o-band peak) spectral phase of SN 2024abvb. Semi-analytical modelling of the light curves shows that it can be reproduced by ~2.6 Msun of SN ejecta interacting with ~0.3 Msun of circumstellar material (CSM), both larger than other SNe Icn but consistent with rapidly evolving SNe Ibn. The metallicity at the SN location is significantly lower than the global metallicity of its host galaxy, suggesting that line-driven mass loss required to strip the progenitor of its H and He envelopes was likely inefficient. We estimate the star-formation-rate history at the location of SN 2024abvb and find that it lies at the bottom ~5th percentile among SESNe hosts, inconsistent with a Wolf-Rayet progenitor. Based on its spectral features, local and host environment properties, and host-galaxy offset, we favour an 8-10 Msun star stripped by a compact companion as the progenitor, with a sufficient runaway velocity to reach the observed offset.

We study the advection effect of the Galactic wind on the local cosmic ray spectra. The spectral hardening from a few hundred GV and softening from a few TV are reproduced by a velocity profile with a maximum velocity of $\sim 700~\mbox{km}~ \mbox{s}^{-1}$ without introducing a break in the power-law dependence of the diffusion coefficient. Additionally, we find that a hard CR spectrum below $\sim$ TV with an index of $\sim 2$ at an altitude $\sim 3$-$5$ kpc from the Galactic disk. This hard spectrum is favorable for the gamma-ray spectrum of the Fermi bubbles. With the obtained CR fluxes, we discuss the matter circulation in our Galaxy with the wind. While the wind has an essential role in maintaining the metal abundance in the disk, the production rate of Beryllium, which originates from CR spallation, is so low that the ratio Be/O in the halo should be larger than that in the disk gas.

We present the first detection of the 3-Point Correlation Function (3PCF) Baryon Acoustic Oscillation (BAO) signal from the DESI Data Release 1 (DR1) sample of Luminous Red Galaxies (LRGs), which contains over 2.1 million galaxies. Our analysis is based on a tree-level redshift-space bispectrum template, which is then transformed to position space using the Fast Fourier Transform on Logarithmic scales (FFTLog) algorithm. We detect the BAO feature with a significance of approximately $8.1σ$ using the EZmock covariance matrix and $8.5σ$ using the analytical covariance matrix, for the full LRG redshift range ($0.4<z<1.1$), denoted as the $z_{\rm full}$ sample. We use the Abacus altMTL mocks, the most precise DESI DR1 mock catalogs currently available, to validate our model. We find that our model fits the mocks well, with a small offset of $0.6\%$ in the recovered BAO scale, which we treat as a systematic error due to modeling. We measure the angle-averaged distance, $D_{\rm V}(z = 0.68)/r_{\rm d} = 15.88 \pm 0.27$ ($1.72\%$ precision) when using the covariance matrix estimated from EZmocks and $D_{\rm V}(z = 0.68)/r_{\rm d} = 15.72 \pm 0.18$ ($1.12\%$ precision) when using the analytical Gaussian covariance matrix. Our results show excellent agreement with the DESI DR1 2PCF BAO measurements as well. We also explore several other ways to estimate the error and find between $1.7$--$2.2\%$ precision on the BAO scale from the EZmock covariance matrix and between $1.1$--$1.5\%$ precision from the analytical covariance matrix. This work represents the first detection of the BAO feature in the DESI 3PCF, establishing its ability to probe the expansion history of the Universe with future DESI 3PCF measurements.

In this work, we investigate the effects of the growth rate scale dependence in the Symmetron modified gravity (MG) model on cosmic structure formation and we analyze the redshift-space distortion (RSD) multipoles, comparing with the Hu-Sawicki $f(R)$ model (specifically the F6) and the standard $Λ$CDM model. The analysis employs a scale-dependent growth equation and utilizes the fk-PT perturbation theory approach, implemented in the FOLPS-nu code, to compute the full 1-loop power spectrum multipoles, in particular, the monopole and quadrupole ($\ell=0,2$, respectively). The results show that at redshift $z=0$, the monopole of both MG models is suppressed compared to $Λ$CDM, with the Symmetron being closer to the standard model, while the quadrupole presents the opposite behavior. To validate the pipeline, we use General Relativity (GR) mock catalogs (EZMocks), since suitable Symmetron simulations are not available. The main result is that the Markov Chain Monte Carlo (MCMC) analysis successfully recovers the expected GR limit (i.e., $β_0 \approx 0$) from the Symmetron model when applied to this mock data, confirming the viability of our methodology for cosmological inference. Then, we conclude that the pipeline is prepared to test MG models against current and near-future galaxy surveys.

We present the discovery and analysis of three microlensing planets identified through brief positive anomalies on the wings of their light curves. The events, KMT-2021-BLG-0852, KMT-2024-BLG-2005, and KMT-2025-BLG-0481, were detected in high-cadence survey data from the KMTNet, OGLE, MOA, and PRIME collaborations. The anomaly morphologies are consistent with major-image perturbations induced by planetary-mass companions located near the peripheral caustic. A systematic exploration of model degeneracies, including binary-source scenarios, higher mass-ratio binary lenses, and the inner--outer caustic degeneracy, firmly establishes the planetary origin of each signal. Measurements of the angular Einstein radius and event timescale, combined with Bayesian priors from a Galactic model, yield the physical parameters of each system. The hosts are low-mass stars (0.12--0.75~$M_\odot$), while the companions are Saturn-mass planets (0.16--0.59 $M_{\rm J}$) projected at separations of 1.1--7.8 au, placing them beyond the snowline of their hosts. These results demonstrate the capability of microlensing to detect and characterize cold giant planets around low-mass stars at kpc distances, populating the critical transition region between ice giants and gas giants.

The first stars likely formed from pristine clouds, marking a transformative epoch after the dark ages by initiating reionisation and synthesising the first heavy elements. Among these, low-mass Population III stars are of particular interest, as their long lifespans raise the possibility that some may survive to the present day in the Milky Way's stellar halo or satellite dwarfs. As the first paper in a series, we present hydrodynamic evolutionary models for 0.7 - 1 MSun stars evolved up to the white dwarf phase, utilising the MESA software instrument. We systematically vary mass-loss efficiencies, convective transport, and overshooting prescriptions, thereby mapping how uncertain physics influences nucleosynthetic yields; surface enrichment, including nitrogen-rich post-main sequence stars arising from convective shell mergers; remnant properties, such as low-mass helium or carbon-oxygen white dwarfs (M_WD ~ 0.45-0.55 MSun) and transient UV-bright phases; and potential observational signatures, including neutrino emission during shell mergers and helium flashes. These models establish a predictive framework for identifying surviving Pop III stars and their descendants, providing both evolutionary and observational constraints that were previously unexplored.

The nature of dark matter and dark energy are among the central questions in cosmology. Strong gravitational lenses with multiple source planes provide a geometric probe of cosmology: the ratio of deflection angles at different redshifts depends only on angular-diameter distances, constraining the matter density $Ω_m$ and the dark energy equation of state $w$. However, constraints from this technique have historically lagged behind those from the CMB, SNe Ia, and BAO. In this work, we present new cosmological constraints from the Carousel Lens, a cluster-scale lens with more than 40 extended images from 11 spectroscopically confirmed sources. Its relaxed core and rich set of extended images behind the main halo make it particularly suitable for cosmological inference. Using the GIGA-Lens pipeline, we construct a pixel-level lens model including six HST-detected sources and four mass components. From this model, we obtain $w$CDM constraints of $Ω_m = 0.34^{+0.16}_{-0.13}$ and $w = -1.31^{+0.35}_{-0.32}$ from the Carousel Lens alone, accounting for both statistical and systematic uncertainties. We further project that including four additional known higher-redshift sources, assuming similar fractional uncertainties, could improve the constraining power by ~80%, bringing the precision close to that of the CMB and SNe Ia. For an evolving dark energy model ($w_0w_a$CDM), the Carousel Lens alone yields constraints comparable to the CMB, providing an independent and complementary probe alongside SN Ia and BAO. While currently systematic uncertainties dominate, which we quantify through simulations, our results demonstrate that relaxed multi-source-plane cluster lenses can deliver competitive cosmological constraints. Further improvements are expected from reductions in systematics and from incorporating higher-redshift sources (known and new) with high-resolution imaging.

Mid-infrared observations of planet-forming disks reveal a wide diversity in molecular spectra. Carbon and oxygen abundances play a central role in setting the chemical environment of the inner disk and the spectral appearance. We aim to systematically explore how variations in elemental carbon and oxygen abundances affect the mid-infrared spectra of planet-forming disks, and to identify robust mid-infrared molecular diagnostics of C/H, O/H, and the C/O ratio. Using the thermochemical disk code ProDiMo and the line radiative transfer code FLiTs, we construct a grid of 25 models with varying carbon and oxygen abundances, covering a broad range of C/O ratios. We analyze the resulting mid-infrared molecular emission, including species such as $\rm H_2O$, $\rm CO$, $\rm CO_2$, $\rm C_2H_2$, $\rm OH$. We find that the mid-infrared molecular spectra are highly sensitive not only to the C/O ratio, but also to the absolute abundances of carbon and oxygen. Despite the same disk structure and C/O ratios, molecular fluxes (e.g., $\rm C_2H_2$, $\rm CO_2$) vary by more than an order of magnitude. This variation stems from the differences in excitation conditions and emitting regions caused by the elemental abundances of oxygen and carbon. We identify diagnostic molecular flux ratios - such as $\rm CO_2$/$\rm H_2O$ and $\rm H_2O$/$\rm C_2H_2$ - that can serve as tracers of C/H and O/H respectively. By combining these diagnostics, we demonstrate a method to infer the underlying C/O ratio. Our model grid provides a framework for interpreting mid-infrared molecular emission from disks, allowing estimates of elemental abundances if the disk properties and structure are known. Comparisons with recent JWST observations suggest that a variety in C and O abundances is seen in a sample of T Tauri disks, possibly shaped by disk transport processes and the presence of gaps.

Jellyfish galaxies, which exhibit tails of gas opposite to their direction of motion, are a galaxy population showcasing the most extreme effects of ram pressure stripping (RPS). We present the emission line properties of a preliminary sample of five jellyfish galaxies in the Coma cluster, observed with the WEAVE Large-IFU as part of the Coma Legacy IFU Survey (CLIFS). When complete, CLIFS will form a sample of 29 jellyfish galaxies in Coma, selected based on the presence of one-sided tails in the radio continuum, enabling a comprehensive picture of the effects of ram pressure on galaxies in the Coma cluster. We extract emission line properties and confirm consistency between disk fluxes measured from WEAVE and MaNGA for galaxies with overlapping disk coverage between surveys. Comparing resolved radio and H$α$-based star formation rates, we find that, in contrast to the disk, the dominant source of tail emission is not star formation. We find evidence for diffuse ionized gas excited by RPS-driven shocks in the tails, as indicated by: (1) LINER-like tail emission with the [OI]/H$α$ BPT diagnostic; (2) enhanced [OII]/H$α$ ratios in the tails relative to the disks; and (3) similarly elevated emission line velocities and velocity dispersions in the tails with respect to the disks. These results demonstrate that ram-pressure-driven shocks dominate the ionized emission in jellyfish galaxy tails.

Stellar microlensing is a powerful tool for probing dark matter in the form of planetary and stellar mass compact objects (COs), in particular primordial black holes (PBHs). Under standard assumptions, current observations exclude COs in the mass range $10^{-11} \lesssim M/M_{\odot} \lesssim 10^{4}$ making up all of the dark matter. We provide an overview, aimed at theorists working on PBHs, of the history, theory, observational status, and future prospects of the field.

Recent observations indicate that in some Milky Way globular clusters (GCs) pristine red giant branch (RGB) stars appear more centrally concentrated than enriched ones. This contradicts most multiple stellar population (MSP) formation scenarios, which predict that the enriched (second) population (2P) should initially be more concentrated than the pristine (first) population (1P). Previous MOCCA Monte Carlo simulations suggested that this apparent overconcentration is a transient effect arising in clusters that have lost a large fraction of their initial mass and host an active black hole subsystem (BHS), and is visible only when RGB stars are used as tracers. In this letter, we test this interpretation using tailored NBODY6++GPU models evolved with direct N-body simulations, providing an independent validation that does not rely on a statistical treatment of relaxation. We performed direct N-body simulations with the NBODY6++GPU code, adopting initial conditions designed to reproduce the dynamical regime relevant to the proposed mechanism. The simulations include updated stellar and binary evolution, dynamical interactions, and the Galactic tidal field, enabling a direct comparison with MOCCA results. The simulations confirm that the spatial distributions and kinematics inferred from RGB stars can be strongly affected by stochastic fluctuations and interactions with the BHS. Preferential ejection of 2P RGB and their progenitors from the cluster center leads to a transient apparent overconcentration of 1P RGB stars, in agreement with earlier MOCCA predictions.

We investigate the growth of central supermassive black holes in galaxies, aiming to distinguish between gas accretion versus BH merger-driven growth modes. By performing and analysing cosmological hydrodynamical simulations of $(50 ~ {\rm Mpc})^3$ comoving boxes, we also study how the BH feedback parameters affect the coevolution between SMBHs and their host galaxies. Starting as $10^5 M_{\odot}$ seeds, we find that the BHs grow initially via BH mergers to $\sim 10^7 M_{\odot}$. Gas accretion onto the BHs is initially low, then increases with time, and reaches the Eddington rate after $7-9$ Gyrs. The BHs then undergo very fast growth via efficient gas accretion over a period of $600 - 700$ Myr, when the BH mass increases $10^2 - 10^3$ times, causing their predominant growth from $10^7 M_{\odot}$ to $(10^9 - 10^{10}) M_{\odot}$. Taking into account the cosmological gas inflows and outflows, SMBHs do not grow to more than $10^{10} M_{\odot}$ by $z=0$, because of gas depletion from galaxy centers driven by AGN feedback. In terms of SMBH - host galaxy coevolution along the $M_{\rm BH} - M_{\star}$ relation, we find that they initially lie below and thereby move upward toward the relation. We make some physical implications of the growth of high-$z$ Little Red Dots recently observed by JWST: the normal-mass SMBHs had predominantly undergone BH merger driven evolution, whereas the overmassive BHs underwent periods of Eddington-limited or super-Eddington bursts of gas accretion.

We present the third data release of the LOFAR Two-metre Sky Survey (LoTSS-DR3). The survey images cover 88% of the northern sky and were created from 12,950 hrs of data (18.6 PB) accumulated over 10.5 years. The images were produced through direction-independent and direction-dependent calibration pipelines that correct for instrumental effects as well as spatially and temporally varying ionospheric distortions. In our 120-168 MHz continuum mosaic images with an angular resolution of 6 arcsec (9 arcsec below declination 10$^\circ$) we catalogue 13,667,877 sources, formed from 16,943,656 Gaussian components. The scatter in the astrometric precision approximately follows the expected noise-like behaviour but with an additional systematic component of at least 0.24 arcsec that is likely due to calibration imperfections. The random flux density scale error is 6%, while the systematic offset was previously shown to be within 2%. The median sensitivity of our mosaics is 92$μ$Jy beam$^{-1}$. Completeness simulations, accounting for realistic source models, time- and bandwidth-smearing effects, and astrometric errors, indicate that we detect more than 95% of compact sources with integrated flux densities exceeding 9 times the local root mean square (RMS) noise. However, the recovered source counts in a particular integrated flux density bin do not match the injected counts until flux densities exceed 45 times the local RMS noise. The Euclidean-normalised differential source counts derived from the survey constrain the radio source population over five orders of magnitude and are in good agreement with previous deep and wide-area surveys. All data products are publicly available, including catalogues, individual-field Stokes I, Q, U, and V images, mosaicked Stokes I images, and $uv$ data with associated direction-dependent calibration solutions.

Synchrotron radiation plays a central role in astrophysical and high-energy processes. Its spectral description involves the synchrotron function, defined by a non-trivial integral of modified Bessel functions and commonly evaluated through numerical methods or dedicated approximations. In this work, we obtain a compact analytical representation of the synchrotron function using the Method of Brackets, which yields systematically controllable asymptotic expansions in both the small- and large-argument regimes. The resulting expressions accurately reproduce numerical integration and make the analytic structure of the function explicit. Our results provide an efficient alternative to repeated numerical evaluations and facilitate applications requiring fast and controlled approximations.

Deep JWST images of AS1063 reveals tens of thousands of globular clusters in the galaxy cluster AS1063. When compared with the lensing model based on the same JWST data, the distribution of globular clusters traces closely the distribution of lensing mass (mostly composed of dark matter). Interestingly, both the distributions of dark matter and globular clusters have large central cores. However the size of the core in the distribution of globular clusters is about half the size the core of the dark matter distribution. We argue that the standard cold dark matter and fuzzy dark matter models struggle to explain these large cores. Meanwhile, the self interacting dark matter with a velocity dependent cross section, combined with core stalling, offers a natural explanation to the existence of these cores if $σ_{\rm SI}\approx 0.3$ cm$^2$ g$^{-1}$ for galaxy cluster halos. But we also discuss how the lack of hydrodynamical N-body simulations capable of resolving globular clusters in galaxy cluster scale halos, hinders the possibility of ruling out the standard non-collisional dark matter scenario. Future high-resolution hydrodynamical simulations of galaxy clusters, with several trillion particles, and containing over a hundred thousand globular clusters, can provide the insight needed to transform the epistemic nature of dark matter into an ontological one

Particle-in-Cell (PIC) methods have achieved widespread recognition as simple and flexible approaches to model collisionless plasma physics in fully kinetic simulations of astrophysical environments. However, in many situations the standard PIC algorithm must be extended to include macroscopic effects in microscale simulations. For plasmas subjected to shearing or expansion, shearing-box and expanding-box methods can be incorporated into PIC to account for these global effects. For plasmas subjected to local acceleration in confined regions of space, a leaky-box method can allow closed-box PIC simulations to account for particle escape from the accelerator region. In this work, we review and improve methods to include shearing, expansion, and escape in PIC simulations. We provide the numerical details of how Maxwell's equations and the particle equations of motion are solved in each case, and introduce generalized Boris-like particle pushers to solve the momentum equation in the presence of extra forces. This work is intended to serve as a comprehensive reference for the implementation of shearing-box, expanding-box, and leaky-box algorithms in PIC.

We investigate how pulsar timing array (PTA) measurements of the nanoHertz gravitational-wave background (GWB) can constrain models for the growth history of supermassive black holes (SMBHs) and how active galactic nucleus (AGN) and stellar feedback models can affect GWB predictions. Feedback regulates supermassive black hole (SMBH) growth, altering the black hole mass function (BHMF). Using BHMFs drawn from multiple cosmological simulation suites including IllustrisTNG, MillenniumTNG, Simba, and CAMELS, and combining these with a quasar-based SMBH binary population framework, we predict the resulting GWB amplitude under a range of different stellar and AGN feedback prescriptions. We find that the choice of both stellar and AGN feedback models alters the high-mass end of the BHMF and changes the predicted GWB amplitude by up to a factor of 2 for the fiducial simulations and a factor 10 for extreme feedback variations in CAMELS. Models with inefficient or absent AGN feedback produce abundant SMBHs and yield GWB amplitudes consistent with PTA data, yet fail in producing realistic galaxies. Fiducial models of AGN and stellar feedback suppress SMBH growth too much and under-predict the expected signal, an effect which could possibly be mitigated by more realistic black hole seeding and growth prescriptions. The mismatch between the GWB amplitudes predicted by cosmological simulations and those observed by PTAs suggests that SMBH growth is more efficient or occurs earlier than captured by current models. This demonstrates that PTA measurements provide a powerful new probe of feedback physics and the SMBH population.

We present a comprehensive photometric and kinematic study of the open clusters HSC 25, HSC 37, and HSC 2878, located in the innermost regions of the Galactic disc. Utilizing data from Gaia DR3 and the UPMASK membership algorithm, we identify 44, 55, and 112 most probable members for HSC 25, HSC 37, and HSC 2878, respectively. The mean proper-motion components are obtained as (-5.901 +/- 0.41, -6.213 +/- 0.40), (-3.231 +/- 0.56, -4.564 +/- 0.47), and (-3.830 +/- 0.51, -5.198 +/- 0.44) mas yr^-1 for HSC 25, HSC 37, and HSC 2878, respectively. The open clusters span a broad range of evolutionary stages, with estimated ages of log(t/yr) = 8.38 +/- 0.08, 7.04 +/- 0.09, and 9.04 +/- 0.09, and corresponding heliocentric distances of 7.36 +/- 0.37, 6.79 +/- 0.18, and 6.17 +/- 0.22 kpc. The obtained metallicities are 0.0388 +/- 0.0039, 0.0259 +/- 0.0028, and 0.0209 +/- 0.0023, respectively. Total mass estimates are 135, 755, and 204 solar masses, respectively, highlighting notable differences in stellar content across the clusters. An analysis of dynamical relaxation times suggests that HSC 25 and HSC 2878 are dynamically evolved, whereas the much younger HSC 37 is still in an early phase of dynamical evolution. The high space velocities and orbital parameters of these clusters reveal significant deviations from typical disc kinematics. HSC 25 and HSC 37 exhibit eccentric orbits and small perigalactic distances, consistent with dynamically heated or accreted origins within the Galactic bulge. In contrast, HSC 2878's relaxed, planar orbit suggests in situ bulge membership despite its age. These findings point toward a heterogeneous dynamical origin for the clusters, with implications for star formation and evolution in the inner Milky Way.

Little Red Dots (LRDs), among the most enigmatic high-redshift discoveries by JWST, are commonly believed to be powered by accreting supermassive black holes. Here, we explore the possibility that these sources are globular clusters in formation, with rest-frame UV arising from a very young stellar population and rest-frame optical from a short-lived supermassive ($>10^4$ M$_\odot$) star. The spectral profiles of LRDs are broadly consistent with this scenario, though the observed temperatures and bolometric luminosities favor emission reprocessed by optically thick, continuum-driven winds not fully captured by current models. The LRD $z\sim5-7$ UV luminosity function naturally evolves, under standard evolutionary and mass-loss prescriptions, into a present-day mass function with a turnover at $\log_{10}(M_\ast$/$M_\odot)=5.3$ and an exponential cutoff at high masses, consistent with local globular-cluster populations. We estimate the total present-day number density of LRDs formed across all redshifts to be $\approx0.3$ Mpc$^{-3}$, similar to local globular clusters. The observed LRD redshift range matches the age distribution of metal-poor globular clusters, without current LRD counterparts to the metal-rich population. If LRDs are globular clusters in formation, we predict chemical abundance patterns characteristic of multiple stellar populations, including enhanced He and N, and potential Na-O and Al-Mg anti-correlations. These results offer a local perspective to explore this surprisingly abundant population of distant sources, and a potential new window into extreme stellar astrophysics in the early Universe.

In interstellar medium studies, separating ordered and random velocity or magnetic fields is essential for interpreting turbulence in both simulations and observations. We investigate how the choice of averaging scale affects the measurement and characterization of magnetic and kinetic turbulence in Milky Way-sized disk galaxies with different initial magnetic morphologies. We analyze two magnetohydrodynamic simulations of isolated disk galaxies, one initialized with a toroidal magnetic field (Model T) and the other with a random field (Model R). Using spherical filtering, we decompose the magnetic and velocity fields into mean and fluctuating components while varying the averaging scale, and examine their energy ratios and power spectra as functions of time and averaging radius. Both models develop ordered and turbulent magnetic structures, whose relative strengths vary strongly with the averaging radius. The power spectra of the velocity and magnetic mean fields steepen with increasing smoothing scale, tracing the transition from coherent to turbulent regimes. The turbulent kinetic energy dominates over the magnetic counterpart, though the latter remains dynamically significant. Very importantly, we find that these ratios depend more strongly on the averaging radius than on the initial conditions of the magnetic field. The characterization of turbulence as strong or weak, meaning whether or not fluctuations dominate over the mean depends sensitively on the chosen averaging scale, rather than being an intrinsic property of the system. The strong dependence of the turbulence fractions on the averaging scale has direct implications for magnetic field estimates obtained from observational methods such as the Davis-Chandrasekhar-Fermi technique emphasizing the need for careful scale selection when interpreting magnetic and kinetic turbulence in galaxies.

We measure the velocity anisotropy profile $β(r)$ of different galaxy cluster member populations by analysing the stacked projected phase space of nine massive ($M_\mathrm{200c}>7\times10^{14}$ M$_\odot$) galaxy clusters at intermediate redshifts ($0.18 < z < 0.45$). We select our sample of galaxy clusters by choosing the most round and virialised objects among the targets of the CLASH-VLT spectroscopic program, which offers a large spectral database. Complementary MUSE observations on most of these clusters allowed us to identify an unprecedented number of cluster members, strongly enhancing the precision of our measurement with respect to previous studies. Our sample of cluster members is divided in four classes: red and blue by colour, and high- and low-mass by stellar mass. We employ two parallel techniques, namely the MAMPOSSt method (parametric in $β(r)$) and the Jeans equation inversion (non parametric in $β(r)$). The results from both techniques are found in agreement for any given cluster member population, and suggest that the orbital anisotropy in galaxy clusters grows from the centre (where $β\approx 0.2-0.4$) to the virial radius ($β\gtrsim 0.8$), and it is similar for the different cluster member populations. We also find an interesting dynamical feature in the Jeans inversion results, that is a drop in $β(r)$ at a distance of $\sim 250$ kpc from the cluster centre. We provide robust anisotropy estimates by exploring a highly significant number of model combinations: 72 with MAMPOSSt (varying the mass, surface number density, $β(r)$ model, and galaxy population) and 18 (varying total mass model and galaxy population) in the Jeans inversion. Such an extensive investigation of the $β(r)$ profile in galaxy clusters is a wide basis for future studies on cluster dynamical masses and cluster cosmology in the era of large spectroscopic surveys

Fast radio bursts (FRBs) are bright, energetic, radio pulses of extragalactic origin. A dichotomy has emerged in the population: some produce repeat bursts, while the majority do not. Most repeating sources only show rare repetitions, and none have been studied extensively over the wide bandwidths necessary to disentangle the physical processes that produce emission from distortions to bursts caused by intervening ionised gas. Here we present radio observations of the most active repeating source, FRB 20240114A. Using an ultrawideband receiving system, we have detected 5526 repetitions, revealing an extreme spectral and temporal variability in the burst emission. The bursts exhibit longer-term broadband variations in central emission frequency over multiple months, and narrowband bursts that have correlations in central frequencies on time scales of milliseconds to minutes. The spectral and temporal properties are consistent with the source undergoing magnification by foreground plasma lenses, potentially embedded in a turbulent circumsource medium. This extreme example highlights the role of plasma lenses in the observed properties of burst emission and can explain the diversity in activity and energetics of the entire FRB population.

We demonstrate that magnetospheric interactions between merging neutron stars (NSs) generate dual-jetted current outflows, analogous to the Alfvén wings observed during planetary interactions in the Solar System. Using 3D relativistic MHD simulations, we model the interaction as a conducting sphere moving through a highly magnetized plasma of the companion's magnetosphere. Unusually, the interaction operates in a regime that is relativistic yet sub-Alfvénic. Electromagnetic draping amplifies magnetic fields in a narrow layer near the stellar surface, leading to the generation of electric currents along the local magnetic field. The generation of beamed outflows enhances the instantaneous power of the pulsar-like radio and high-energy emission, produces spin/orbital modulations, and is likely to lead to observable precursor emission preceding the main gravitational wave event.