arXiv Daily Digest - 2026-01-28

The performance evaluation remains a complex challenge in audio separation, and existing evaluation metrics are often misaligned with human perception, course-grained, relying on ground truth signals. On the other hand, subjective listening tests remain the gold standard for real-world evaluation, but they are expensive, time-consuming, and difficult to scale. This paper addresses the growing need for automated systems capable of evaluating audio separation without human intervention. The proposed evaluation metric, SAM Audio Judge (SAJ), is a multimodal fine-grained reference-free objective metric, which shows highly alignment with human perceptions. SAJ supports three audio domains (speech, music and general sound events) and three prompt inputs (text, visual and span), covering four different dimensions of evaluation (recall, percision, faithfulness, and overall). SAM Audio Judge also shows potential applications in data filtering, pseudo-labeling large datasets and reranking in audio separation models. We release our code and pre-trained models at: https://github.com/facebookresearch/sam-audio.

Knowledge editing emerges as a crucial technique for efficiently correcting incorrect or outdated knowledge in large language models (LLM). Existing editing methods for unimodal LLM rely on a rigid parameter-to-output mapping, which causes causal-underfit and causal-overfit in cascaded reasoning for Multimodal LLM (MLLM). In this paper, we reformulate MLLM editing as an out-of-distribution (OOD) generalization problem, where the goal is to discern semantic shift with factual shift and thus achieve robust editing among diverse cross-modal prompting. The key challenge of this OOD problem lies in identifying invariant causal trajectories that generalize accurately while suppressing spurious correlations. To address it, we propose ODEdit, a plug-and-play invariant learning based framework that optimizes the tripartite OOD risk objective to simultaneously enhance editing reliability, locality, and generality.We further introduce an edit trajectory invariant learning method, which integrates a total variation penalty into the risk minimization objective to stabilize edit trajectories against environmental variations. Theoretical analysis and extensive experiments demonstrate the effectiveness of ODEdit.

Repository-level code completion remains a challenging task for existing code large language models (code LLMs) due to their limited understanding of repository-specific context and domain knowledge. While retrieval-augmented generation (RAG) approaches have shown promise by retrieving relevant code snippets as cross-file context, they suffer from two fundamental problems: misalignment between the query and the target code in the retrieval process, and the inability of existing retrieval methods to effectively utilize the inference information. To address these challenges, we propose AlignCoder, a repository-level code completion framework that introduces a query enhancement mechanism and a reinforcement learning based retriever training method. Our approach generates multiple candidate completions to construct an enhanced query that bridges the semantic gap between the initial query and the target code. Additionally, we employ reinforcement learning to train an AlignRetriever that learns to leverage inference information in the enhanced query for more accurate retrieval. We evaluate AlignCoder on two widely-used benchmarks (CrossCodeEval and RepoEval) across five backbone code LLMs, demonstrating an 18.1% improvement in EM score compared to baselines on the CrossCodeEval benchmark. The results show that our framework achieves superior performance and exhibits high generalizability across various code LLMs and programming languages.

The increasing scale and complexity of modern model parameters underscore the importance of pre-trained models. However, deployment often demands architectures of varying sizes, exposing limitations of conventional pre-training and fine-tuning. To address this, we propose SWEET, a self-supervised framework that performs constraint-based pre-training to enable scalable initialization in vision tasks. Instead of pre-training a fixed-size model, we learn a shared weight template and size-specific weight scalers under Tucker-based factorization, which promotes modularity and supports flexible adaptation to architectures with varying depths and widths. Target models are subsequently initialized by composing and reweighting the template through lightweight weight scalers, whose parameters can be efficiently learned from minimal training data. To further enhance flexibility in width expansion, we introduce width-wise stochastic scaling, which regularizes the template along width-related dimensions and encourages robust, width-invariant representations for improved cross-width generalization. Extensive experiments on \textsc{classification}, \textsc{detection}, \textsc{segmentation} and \textsc{generation} tasks demonstrate the state-of-the-art performance of SWEET for initializing variable-sized vision models.

Generative AI plays an increasing role during software engineering activities to make them, e.g., more efficient or provide better quality. However, it is often unclear how much benefit LLMs really provide. We concentrate on software architects and investigated how an LLM-supported evaluation of architecture documents can support software architects to improve such artefacts. In the context of a research project where a digital marketplace is developed and digital solutions should be analyzed, we used different LLMs to analyze the quality of architecture documents and compared the results with evaluations from software architects. We found out that the quality of the artifact has a strong influence on the quality of the LLM, i.e., the better the quality of the architecture document was, the more consistent were the LLM-based evaluation and the human expert evaluation. While using LLMs in this architecture task is promising, our results showed inconsistencies that need further analyses before generalizing them.

Post-training quantization (PTQ) enables effective model compression while preserving relatively high accuracy. Current weight-only PTQ methods primarily focus on the challenging sub-3-bit regime, where approaches often suffer significant accuracy degradation, typically requiring fine-tuning to achieve competitive performance. In this work, we revisit the fundamental characteristics of weight quantization and analyze the challenges in quantizing the residual matrix under low-rank approximation. We propose LoPRo, a novel fine-tuning-free PTQ algorithm that enhances residual matrix quantization by applying block-wise permutation and Walsh-Hadamard transformations to rotate columns of similar importance, while explicitly preserving the quantization accuracy of the most salient column blocks. Furthermore, we introduce a mixed-precision fast low-rank decomposition based on rank-1 sketch (R1SVD) to further minimize quantization costs. Experiments demonstrate that LoPRo outperforms existing fine-tuning-free PTQ methods at both 2-bit and 3-bit quantization, achieving accuracy comparable to fine-tuning baselines. Specifically, LoPRo achieves state-of-the-art quantization accuracy on LLaMA-2 and LLaMA-3 series models while delivering up to a 4$\times$ speedup. In the MoE model Mixtral-8x7B, LoPRo completes quantization within 2.5 hours, simultaneously reducing perplexity by 0.4$\downarrow$ and improving accuracy by 8\%$\uparrow$. Moreover, compared to other low-rank quantization methods, LoPRo achieves superior accuracy with a significantly lower rank, while maintaining high inference efficiency and minimal additional latency.

Ambitious decarbonisation targets are catalysing growth in orders of new offshore wind farms. For these newly commissioned plants to run, accurate power forecasts are needed from the onset. These allow grid stability, good reserve management and efficient energy trading. Despite machine learning models having strong performances, they tend to require large volumes of site-specific data that new farms do not yet have. To overcome this data scarcity, we propose a novel transfer learning framework that clusters power output according to covariate meteorological features. Rather than training a single, general-purpose model, we thus forecast with an ensemble of expert models, each trained on a cluster. As these pre-trained models each specialise in a distinct weather pattern, they adapt efficiently to new sites and capture transferable, climate-dependent dynamics. Through the expert models' built-in calibration to seasonal and meteorological variability, we remove the industry-standard requirement of local measurements over a year. Our contributions are two-fold - we propose this novel framework and comprehensively evaluate it on eight offshore wind farms, achieving accurate cross-domain forecasting with under five months of site-specific data. Our experiments achieve a MAE of 3.52\%, providing empirical verification that reliable forecasts do not require a full annual cycle. Beyond power forecasting, this climate-aware transfer learning method opens new opportunities for offshore wind applications such as early-stage wind resource assessment, where reducing data requirements can significantly accelerate project development whilst effectively mitigating its inherent risks.

The present benchmarks for testing the audio modality of multimodal large language models concentrate on testing various audio tasks such as speaker diarization or gender identification in isolation. Whether a multimodal model can answer the questions that require reasoning skills to combine audio tasks of different categories, cannot be verified with their use. To address this issue, we propose Audio Reasoning Tasks (ART), a new benchmark for assessing the ability of multimodal models to solve problems that require reasoning over audio signal.

Federated Learning (FL) enables collaborative training of Large Language Models (LLMs) across distributed data sources while preserving privacy. However, when federated LLMs are deployed in critical applications, it remains unclear which client(s) contributed to specific generated responses, hindering debugging, malicious client identification, fair reward allocation, and trust verification. We present ProToken, a novel Provenance methodology for Token-level attribution in federated LLMs that addresses client attribution during autoregressive text generation while maintaining FL privacy constraints. ProToken leverages two key insights to enable provenance at each token: (1) transformer architectures concentrate task-specific signals in later blocks, enabling strategic layer selection for computational tractability, and (2) gradient-based relevance weighting filters out irrelevant neural activations, focusing attribution on neurons that directly influence token generation. We evaluate ProToken across 16 configurations spanning four LLM architectures (Gemma, Llama, Qwen, SmolLM) and four domains (medical, financial, mathematical, coding). ProToken achieves 98% average attribution accuracy in correctly localizing responsible client(s), and maintains high accuracy when the number of clients are scaled, validating its practical viability for real-world deployment settings.

Data augmentation is a crucial tool in time series forecasting, especially for deep learning architectures that require a large training sample size to generalize effectively. However, extensive datasets are not always available in real-world scenarios. Although many data augmentation methods exist, their limitations include the use of transformations that do not adequately preserve data properties. This paper introduces Grasynda, a novel graph-based approach for synthetic time series generation that: (1) converts univariate time series into a network structure using a graph representation, where each state is a node and each transition is represented as a directed edge; and (2) encodes their temporal dynamics in a transition probability matrix. We performed an extensive evaluation of Grasynda as a data augmentation method for time series forecasting. We use three neural network variations on six benchmark datasets. The results indicate that Grasynda consistently outperforms other time series data augmentation methods, including ones used in state-of-the-art time series foundation models. The method and all experiments are publicly available.

We present SynCABEL (Synthetic Contextualized Augmentation for Biomedical Entity Linking), a framework that addresses a central bottleneck in supervised biomedical entity linking (BEL): the scarcity of expert-annotated training data. SynCABEL leverages large language models to generate context-rich synthetic training examples for all candidate concepts in a target knowledge base, providing broad supervision without manual annotation. We demonstrate that SynCABEL, when combined with decoder-only models and guided inference establish new state-of-the-art results across three widely used multilingual benchmarks: MedMentions for English, QUAERO for French, and SPACCC for Spanish. Evaluating data efficiency, we show that SynCABEL reaches the performance of full human supervision using up to 60% less annotated data, substantially reducing reliance on labor-intensive and costly expert labeling. Finally, acknowledging that standard evaluation based on exact code matching often underestimates clinically valid predictions due to ontology redundancy, we introduce an LLM-as-a-judge protocol. This analysis reveals that SynCABEL significantly improves the rate of clinically valid predictions. Our synthetic datasets, models, and code are released to support reproducibility and future research.

Continual learning for pre-trained vision-language models requires balancing three competing objectives: retaining pre-trained knowledge, preserving knowledge from a sequence of learned tasks, and maintaining the plasticity to acquire new knowledge. This paper presents a simple but effective approach called KeepLoRA to effectively balance these objectives. We first analyze the knowledge retention mechanism within the model parameter space and find that general knowledge is mainly encoded in the principal subspace, while task-specific knowledge is encoded in the residual subspace. Motivated by this finding, KeepLoRA learns new tasks by restricting LoRA parameter updates in the residual subspace to prevent interfering with previously learned capabilities. Specifically, we infuse knowledge for a new task by projecting its gradient onto a subspace orthogonal to both the principal subspace of pre-trained model and the dominant directions of previous task features. Our theoretical and empirical analyses confirm that KeepLoRA balances the three objectives and achieves state-of-the-art performance. The implementation code is available at https://github.com/MaolinLuo/KeepLoRA.

Diffusion Language Models (DLMs) have emerged as a compelling alternative to autoregressive approaches, enabling parallel text generation with competitive performance. Despite these advantages, there is a critical instability in DLMs: the moving sink phenomenon. Our analysis indicates that sink tokens exhibit low-norm representations in the Transformer's value space, and that the moving sink phenomenon serves as a protective mechanism in DLMs to prevent excessive information mixing. However, their unpredictable positions across diffusion steps undermine inference robustness. To resolve this, we propose a simple but effective extra sink token implemented via a modified attention mask. Specifically, we introduce a special token constrained to attend solely to itself, while remaining globally visible to all other tokens. Experimental results demonstrate that introducing a single extra token stabilizes attention sinks, substantially improving model performance. Crucially, further analysis confirms that the effectiveness of this token is independent of its position and characterized by negligible semantic content, validating its role as a robust and dedicated structural sink.

In recent years, applications such as real-time simulations, autonomous systems, and video games increasingly demand the processing of complex geometric models under stringent time constraints. Traditional geometric algorithms, including the convex hull, are subject to these challenges. A common approach to improve performance is scaling computational resources, which often results in higher energy consumption. Given the growing global concern regarding sustainable use of energy, this becomes a critical limitation. This work presents a 3D preprocessing filter for the convex hull algorithm using ray tracing and tensor core technologies. The filter builds a delimiter polyhedron based on Manhattan distances that discards points from the original set. The filter is evaluated on two point distributions: uniform and sphere. Experimental results show that the proposed filter, combined with convex hull construction, accelerates the computation of the 3D convex hull by up to $200 \times$ with respect to a CPU parallel implementation. This research demonstrates that geometric algorithms can be accelerated through massive parallelism while maintaining efficient energy utilization. Beyond execution time and speedup evaluation, we also analyze GPU energy consumption, showing that the proposed preprocessing filter not only reduces the computational workload but also achieves performance gains with controlled energy usage. These results highlight the dual benefit of the method in terms of both speed and energy efficiency, reinforcing its applicability in modern high-performance scenarios.

Decidability or complexity issues about the consistency problem for description logics with concrete domains have already been analysed with tableaux-based or type elimination methods. Concrete domains in ontologies are essential to consider concrete objects and predefined relations. In this work, we expose an automata-based approach leading to the optimal upper bound EXPTIME, that is designed by enriching the transitions with symbolic constraints. We show that the nonemptiness problem for such automata belongs to EXPTIME if the concrete domains satisfy a few simple properties. Then, we provide a reduction from the consistency problem for ontologies, yielding EXPTIME-membership.Thanks to the expressivity of constraint automata, the results are extended to additional ingredients such as inverse roles, functional role names and constraint assertions, while maintaining EXPTIME-membership, which illustrates the robustness of the approach

Reviewer assignment is increasingly critical yet challenging in the LLM era, where rapid topic shifts render many pre-2023 benchmarks outdated and where proxy signals poorly reflect true reviewer familiarity. We address this evaluation bottleneck by introducing LR-bench, a high-fidelity, up-to-date benchmark curated from 2024-2025 AI/NLP manuscripts with five-level self-assessed familiarity ratings collected via a large-scale email survey, yielding 1055 expert-annotated paper-reviewer-score annotations. We further propose RATE, a reviewer-centric ranking framework that distills each reviewer's recent publications into compact keyword-based profiles and fine-tunes an embedding model with weak preference supervision constructed from heuristic retrieval signals, enabling matching each manuscript against a reviewer profile directly. Across LR-bench and the CMU gold-standard dataset, our approach consistently achieves state-of-the-art performance, outperforming strong embedding baselines by a clear margin. We release LR-bench at https://huggingface.co/datasets/Gnociew/LR-bench, and a GitHub repository at https://github.com/Gnociew/RATE-Reviewer-Assign.

Vulnerability identifiers such as CVE, CWE, and GHSA are standardised references to known software security issues, yet their use in practice is not well understood. This paper compares vulnerability ID use in GitHub pull requests authored by autonomous agents, bots, and human developers. Using the AIDev pop dataset and an augmented set of pull requests from the same repositories, we analyse who mentions vulnerability identifiers and where they appear. Bots account for around 69.1% of all mentions, usually adding few identifiers in pull request descriptions, while human and agent mentions are rarer but span more locations. Qualitative analysis shows that bots mainly reference identifiers in automated dependency updates and audits, whereas humans and agents use them to support fixes, maintenance, and discussion.

Quantum cloud platforms remain fundamentally non-virtualised: despite rapid hardware scaling, each user program still monopolises an entire quantum processor, preventing resource sharing, economic scalability, and quality-of-service differentiation. Existing Quantum Virtual Machine (QVM) designs attempt spatial multiplexing through topology-specific or template-based partitioning, but these approaches are brittle under hardware heterogeneity, calibration drift, and transient defects, which dominate real quantum devices. We present DynQ, the first dynamic, topology-agnostic Quantum Virtual Machine that virtualises quantum hardware using quality-weighted community detection. Instead of imposing fixed geometric regions, DynQ models a quantum processor as a weighted graph derived from live calibration data and automatically discovers execution regions that maximise internal gate quality while minimising inter-region coupling. This operationalises the classical virtualisation principle of high cohesion and low coupling in a quantum-native setting, producing execution regions that are connectivity-efficient, noise-aware, and resilient to crosstalk and defects. We evaluate DynQ across five IBM Quantum backends using calibration-derived noise simulation and on two production devices, comparing against state-of-the-art QVM and standard compilation baselines. On hardware with pronounced spatial quality variation, DynQ achieves up to 19.1 percent higher fidelity and 45.1 percent lower output error. When transient hardware defects cause baseline executions to fail completely, DynQ adapts dynamically and achieves over 86 percent fidelity. By transforming calibrated device graphs into adaptive virtual hardware abstractions, DynQ decouples quantum programs from fragile physical layouts and enables reliable, high-utilisation quantum cloud services.

Rising publication pressure and the routine use of generative AI tools are reshaping how software engineering research is produced, assessed, and taught. While these developments promise efficiency, they also raise concerns about skill degradation, responsibility, and trust in scholarly outputs. This vision paper employs Design Fiction as a methodological lens to examine how such concerns might materialise if current practices persist. Drawing on themes reported in a recent community survey, we construct a speculative artifact situated in a near future research setting. The fiction is used as an analytical device rather than a forecast, enabling reflection on how automated assistance might impede domain knowledge competence, verification, and mentoring practices. By presenting an intentionally unsettling scenario, the paper invites discussion on how the software engineering research community in the future will define proficiency, allocate responsibility, and support learning.

Real-world reinforcement learning often faces environment drift, but most existing methods rely on static entropy coefficients/target entropy, causing over-exploration during stable periods and under-exploration after drift (thus slow recovery), and leaving unanswered the principled question of how exploration intensity should scale with drift magnitude. We prove that entropy scheduling under non-stationarity can be reduced to a one-dimensional, round-by-round trade-off, faster tracking of the optimal solution after drift vs. avoiding gratuitous randomness when the environment is stable, so exploration strength can be driven by measurable online drift signals. Building on this, we propose AES (Adaptive Entropy Scheduling), which adaptively adjusts the entropy coefficient/temperature online using observable drift proxies during training, requiring almost no structural changes and incurring minimal overhead. Across 4 algorithm variants, 12 tasks, and 4 drift modes, AES significantly reduces the fraction of performance degradation caused by drift and accelerates recovery after abrupt changes.

Heuristic functions are essential to the performance of tree search algorithms such as A*, where their accuracy and efficiency directly impact search outcomes. Traditionally, such heuristics are handcrafted, requiring significant expertise. Recent advances in large language models (LLMs) and evolutionary frameworks have opened the door to automating heuristic design. In this paper, we extend the Evolution of Heuristics (EoH) framework to investigate the automated generation of guiding heuristics for A* search. We introduce a novel domain-agnostic prompt augmentation strategy that includes the A* code into the prompt to leverage in-context learning, named Algorithmic - Contextual EoH (A-CEoH). To evaluate the effectiveness of A-CeoH, we study two problem domains: the Unit-Load Pre-Marshalling Problem (UPMP), a niche problem from warehouse logistics, and the classical sliding puzzle problem (SPP). Our computational experiments show that A-CEoH can significantly improve the quality of the generated heuristics and even outperform expert-designed heuristics.

Large reasoning models (LRMs) aim to solve diverse and complex problems through structured reasoning. Recent advances in group-based policy optimization methods have shown promise in enabling stable advantage estimation without reliance on process-level annotations. However, these methods rely on advantage gaps induced by high-quality samples within the same batch, which makes the training process fragile and inefficient when intra-group advantages collapse under challenging tasks. To address these problems, we propose a reinforcement learning mechanism named \emph{\textbf{R^3}} that along three directions: (1) a \emph{cross-context \underline{\textbf{R}}eplay} strategy that maintains the intra-group advantage by recalling valuable examples from historical trajectories of the same query, (2) an \emph{in-context self-\underline{\textbf{R}}eflection} mechanism enabling models to refine outputs by leveraging past failures, and (3) a \emph{structural entropy \underline{\textbf{R}}anking reward}, which assigns relative rewards to truncated or failed samples by ranking responses based on token-level entropy patterns, capturing both local exploration and global stability. We implement our method on Deepseek-R1-Distill-Qwen-1.5B and train it on the DeepscaleR-40k in the math domain. Experiments demonstrate our method achieves SoTA performance on several math benchmarks, representing significant improvements and fewer reasoning tokens over the base models. Code and model will be released.

Differential privacy (DP) provides formal protection for sensitive data but typically incurs substantial losses in diagnostic performance. Model initialization has emerged as a critical factor in mitigating this degradation, yet the role of modern self-supervised learning under full-model DP remains poorly understood. Here, we present a large-scale evaluation of initialization strategies for differentially private medical image analysis, using chest radiograph classification as a representative benchmark with more than 800,000 images. Using state-of-the-art ConvNeXt models trained with DP-SGD across realistic privacy regimes, we compare non-domain-specific supervised ImageNet initialization, non-domain-specific self-supervised DINOv3 initialization, and domain-specific supervised pretraining on MIMIC-CXR, the largest publicly available chest radiograph dataset. Evaluations are conducted across five external datasets spanning diverse institutions and acquisition settings. We show that DINOv3 initialization consistently improves diagnostic utility relative to ImageNet initialization under DP, but remains inferior to domain-specific supervised pretraining, which achieves performance closest to non-private baselines. We further demonstrate that initialization choice strongly influences demographic fairness, cross-dataset generalization, and robustness to data scale and model capacity under privacy constraints. The results establish initialization strategy as a central determinant of utility, fairness, and generalization in differentially private medical imaging.

Key Information Extraction (KIE) from visually-rich documents (VrDs) is a critical task, for which recent Large Language Models (LLMs) and Multi-Modal Large Language Models (MLLMs) have demonstrated strong potential. However, their reliance on autoregressive inference, which generates outputs sequentially, creates a significant efficiency bottleneck, especially as KIE tasks often involve extracting multiple, semantically independent fields. To overcome this limitation, we introduce PIP: a Parallel Inference Paradigm for KIE. Our approach reformulates the problem by using "[mask]" tokens as placeholders for all target values, enabling their simultaneous generation in a single forward pass. To facilitate this paradigm, we develop a tailored mask pre-training strategy and construct large-scale supervised datasets. Experimental results show that our PIP-models achieve a 5-36x inference speedup with negligible performance degradation compared to traditional autoregressive base models. By substantially improving efficiency while maintaining high accuracy, PIP paves the way for scalable and practical real-world KIE solutions.

Safe exploration is a key requirement for reinforcement learning (RL) agents to learn and adapt online, beyond controlled (e.g. simulated) environments. In this work, we tackle this challenge by utilizing suboptimal yet conservative policies (e.g., obtained from offline data or simulators) as priors. Our approach, SOOPER, uses probabilistic dynamics models to optimistically explore, yet pessimistically fall back to the conservative policy prior if needed. We prove that SOOPER guarantees safety throughout learning, and establish convergence to an optimal policy by bounding its cumulative regret. Extensive experiments on key safe RL benchmarks and real-world hardware demonstrate that SOOPER is scalable, outperforms the state-of-the-art and validate our theoretical guarantees in practice.

In large language models built upon the Transformer architecture, recent studies have shown that inter-head interaction can enhance attention performance. Motivated by this, we propose Multi-head Explicit Attention (MEA), a simple yet effective attention variant that explicitly models cross-head interaction. MEA consists of two key components: a Head-level Linear Composition (HLC) module that separately applies learnable linear combinations to the key and value vectors across heads, thereby enabling rich inter-head communication; and a head-level Group Normalization layer that aligns the statistical properties of the recombined heads. MEA shows strong robustness in pretraining, which allows the use of larger learning rates that lead to faster convergence, ultimately resulting in lower validation loss and improved performance across a range of tasks. Furthermore, we explore the parameter efficiency of MEA by reducing the number of attention heads and leveraging HLC to reconstruct them using low-rank "virtual heads". This enables a practical key-value cache compression strategy that reduces KV-cache memory usage by 50% with negligible performance loss on knowledge-intensive and scientific reasoning tasks, and only a 3.59% accuracy drop for Olympiad-level mathematical benchmarks.

Emerging 6G networks rely on complex cross-layer optimization, yet manually translating high-level intents into mathematical formulations remains a bottleneck. While Large Language Models (LLMs) offer promise, monolithic approaches often lack sufficient domain grounding, constraint awareness, and verification capabilities. To address this, we present ComAgent, a multi-LLM agentic AI framework. ComAgent employs a closed-loop Perception-Planning-Action-Reflection cycle, coordinating specialized agents for literature search, coding, and scoring to autonomously generate solver-ready formulations and reproducible simulations. By iteratively decomposing problems and self-correcting errors, the framework effectively bridges the gap between user intent and execution. Evaluations demonstrate that ComAgent achieves expert-comparable performance in complex beamforming optimization and outperforms monolithic LLMs across diverse wireless tasks, highlighting its potential for automating design in emerging wireless networks.

Recent advances in video-audio (V-A) understanding and generation have increasingly relied on joint V-A embeddings, which serve as the foundation for tasks such as cross-modal retrieval and generation. While prior methods like CAVP effectively model semantic and temporal correspondences between modalities using contrastive objectives, their performance remains suboptimal. A key limitation is the insufficient modeling of the dense, multi-scale nature of both video and audio signals, correspondences often span fine- to coarse-grained spatial-temporal structures, which are underutilized in existing frameworks. To this end, we propose GMS-CAVP, a novel framework that combines Multi-Scale Video-Audio Alignment and Multi-Scale Spatial-Temporal Diffusion-based pretraining objectives to enhance V-A correspondence modeling. First, GMS-CAVP introduces a multi-scale contrastive learning strategy that captures semantic and temporal relations across varying granularities. Second, we go beyond traditional contrastive learning by incorporating a diffusion-based generative objective, enabling modality translation and synthesis between video and audio. This unified discriminative-generative formulation facilitates deeper cross-modal understanding and paves the way for high-fidelity generation. Extensive experiments on VGGSound, AudioSet, and Panda70M demonstrate that GMS-CAVP outperforms previous methods in generation and retrieval.

Recent work has shown that integrating large language models (LLMs) with theorem provers (TPs) in neuro-symbolic pipelines helps with entailment verification and proof-guided refinement of explanations for natural language inference (NLI). However, scaling such refinement to naturalistic NLI remains difficult: long, syntactically rich inputs and deep multi-step arguments amplify autoformalisation errors, where a single local mismatch can invalidate the proof. Moreover, current methods often handle failures via costly global regeneration due to the difficulty of localising the responsible span or step from prover diagnostics. Aiming to address these problems, we propose a decompose-and-formalise framework that (i) decomposes premise-hypothesis pairs into an entailment tree of atomic steps, (ii) verifies the tree bottom-up to isolate failures to specific nodes, and (iii) performs local diagnostic-guided refinement instead of regenerating the whole explanation. Moreover, to improve faithfulness of autoformalisation, we introduce $θ$-substitution in an event-based logical form to enforce consistent argument-role bindings. Across a range of reasoning tasks using five LLM backbones, our method achieves the highest explanation verification rates, improving over the state-of-the-art by 26.2%, 21.7%, 21.6% and 48.9%, while reducing refinement iterations and runtime and preserving strong NLI accuracy.

While InfoNCE powers modern contrastive learning, its geometric mechanisms remain under-characterized beyond the canonical alignment--uniformity decomposition. We present a measure-theoretic framework that models learning as the evolution of representation measures on a fixed embedding manifold. By establishing value and gradient consistency in the large-batch limit, we bridge the stochastic objective to explicit deterministic energy landscapes, uncovering a fundamental geometric bifurcation between the unimodal and multimodal regimes. In the unimodal setting, the intrinsic landscape is strictly convex with a unique Gibbs equilibrium; here, entropy acts merely as a tie-breaker, clarifying "uniformity" as a constrained expansion within the alignment basin. In contrast, the symmetric multimodal objective contains a persistent negative symmetric divergence term that remains even after kernel sharpening. We show that this term induces barrier-driven co-adaptation, enforcing a population-level modality gap as a structural geometric necessity rather than an initialization artifact. Our results shift the analytical lens from pointwise discrimination to population geometry, offering a principled basis for diagnosing and controlling distributional misalignment.

The deployment of Artificial Intelligence in high-risk domains, such as finance and healthcare, necessitates models that are both fair and transparent. While regulatory frameworks, including the EU's AI Act, mandate bias mitigation, they are deliberately vague about the definition of bias. In line with existing research, we argue that true fairness requires addressing bias at the intersections of protected groups. We propose a unified framework that leverages Mixed-Integer Optimization (MIO) to train intersectionally fair and intrinsically interpretable classifiers. We prove the equivalence of two measures of intersectional fairness (MSD and SPSF) in detecting the most unfair subgroup and empirically demonstrate that our MIO-based algorithm improves performance in finding bias. We train high-performing, interpretable classifiers that bound intersectional bias below an acceptable threshold, offering a robust solution for regulated industries and beyond.

Adapting Large Language Models (LLMs) to specialized domains without human-annotated data is a crucial yet formidable challenge. Widely adopted knowledge distillation methods often devolve into coarse-grained mimicry, where the student model inefficiently targets its own weaknesses and risks inheriting the teacher's reasoning flaws. This exposes a critical pedagogical dilemma: how to devise a reliable curriculum when the teacher itself is not an infallible expert. Our work resolves this by capitalizing on a key insight: while LLMs may exhibit fallibility in complex, holistic reasoning, they often exhibit high fidelity on focused, atomic sub-problems. Based on this, we propose Divergence-Guided Reasoning Curriculum (DGRC), which constructs a learning path from atomic knowledge to reasoning chains by dynamically deriving two complementary curricula from disagreements in reasoning pathways. When a student and teacher produce conflicting results, DGRC directs the teacher to perform a diagnostic analysis: it analyzes both reasoning paths to formulate atomic queries that target the specific points of divergence, and then self-answers these queries to create high-confidence atomic question-answer pairs. These pairs then serve a dual purpose: (1) providing an atomic curriculum to rectify the student's knowledge gaps, and (2) serving as factual criteria to filter the teacher's original reasoning chains, yielding a verified CoT curriculum that teaches the student how to integrate atomic knowledge into complete reasoning paths. Experiments across the medical and legal domains on student models of various sizes demonstrate the effectiveness of our DGRC framework. Notably, our method achieves a 7.76% relative improvement for the 1.5B student model in the medical domain over strong unlabeled baseline.

Interactive recommender systems can dynamically adapt to user feedback, but often suffer from content homogeneity and filter bubble effects due to overfitting short-term user preferences. While recent efforts aim to improve content diversity, they predominantly operate in static or one-shot settings, neglecting the long-term evolution of user interests. Reinforcement learning provides a principled framework for optimizing long-term user satisfaction by modeling sequential decision-making processes. However, its application in recommendation is hindered by sparse, long-tailed user-item interactions and limited semantic planning capabilities. In this work, we propose LLM-Enhanced Reinforcement Learning (LERL), a novel hierarchical recommendation framework that integrates the semantic planning power of LLM with the fine-grained adaptability of RL. LERL consists of a high-level LLM-based planner that selects semantically diverse content categories, and a low-level RL policy that recommends personalized items within the selected semantic space. This hierarchical design narrows the action space, enhances planning efficiency, and mitigates overexposure to redundant content. Extensive experiments on real-world datasets demonstrate that LERL significantly improves long-term user satisfaction when compared with state-of-the-art baselines. The implementation of LERL is available at https://anonymous.4open.science/r/code3-18D3/.

LLM-based agents are becoming central to software engineering tasks, yet evaluating them remains fragmented and largely model-centric. Existing studies overlook how architectural components, such as planners, memory, and tool routers, shape agent behavior, limiting diagnostic power. We propose a lightweight, architecture-informed approach that links agent components to their observable behaviors and to the metrics capable of evaluating them. Our method clarifies what to measure and why, and we illustrate its application through real world agents, enabling more targeted, transparent, and actionable evaluation of LLM-based agents.

Deep research has emerged as a transformative capability for autonomous agents, empowering Large Language Models to navigate complex, open-ended tasks. However, realizing its full potential is hindered by critical limitations, including escalating contextual noise in long-horizon tasks, fragility leading to cascading errors, and a lack of modular extensibility. To address these challenges, we introduce Yunque DeepResearch, a hierarchical, modular, and robust framework. The architecture is characterized by three key components: (1) a centralized Multi-Agent Orchestration System that routes subtasks to an Atomic Capability Pool of tools and specialized sub-agents; (2) a Dynamic Context Management mechanism that structures completed sub-goals into semantic summaries to mitigate information overload; and (3) a proactive Supervisor Module that ensures resilience through active anomaly detection and context pruning. Yunque DeepResearch achieves state-of-the-art performance across a range of agentic deep research benchmarks, including GAIA, BrowseComp, BrowseComp-ZH, and Humanity's Last Exam. We open-source the framework, reproducible implementations, and application cases to empower the community.

Code localization constitutes a key bottleneck in automated software development pipelines. While concurrent tool execution can enhance discovery speed, current agents demonstrate a 34.9\% redundant invocation rate, which negates parallelism benefits. We propose \textbf{FuseSearch}, reformulating parallel code localization as a \textbf{joint quality-efficiency optimization} task. Through defining \textbf{tool efficiency} -- the ratio of unique information gain to invocation count -- we utilize a two-phase SFT and RL training approach for learning adaptive parallel strategies. Different from fixed-breadth approaches, FuseSearch dynamically modulates search breadth according to task context, evolving from exploration phases to refinement stages. Evaluated on SWE-bench Verified, FuseSearch-4B achieves SOTA-level performance (84.7\% file-level and 56.4\% function-level $F_1$ scores) with 93.6\% speedup, utilizing 67.7\% fewer turns and 68.9\% fewer tokens. Results indicate that efficiency-aware training naturally improves quality through eliminating noisy redundant signals, enabling high-performance cost-effective localization agents.

We address the intrinsic dimensionality (ID) of high-dimensional trajectories, comprising $n_s = 4\,000\,000$ data points, of the Fermi-Pasta-Ulam-Tsingou (FPUT) $β$ model with $N = 32$ oscillators. To this end, a deep autoencoder (DAE) model is employed to infer the ID in the weakly nonlinear regime ($β\lesssim 1$). We find that the trajectories lie on a nonlinear manifold of dimension $m^{\ast} = 2$ embedded in a $64$-dimensional phase space. The DAE further reveals that this dimensionality increases to $m^{\ast} = 3$ at $β= 1.1$, coinciding with a symmetry breaking transition, in which additional energy modes with even wave numbers $k = 2, 4$ become excited. Finally, we discuss the limitations of the linear approach based on principal component analysis (PCA), which fails to capture the underlying structure of the data and therefore yields unreliable results in most cases.

Foundation models (FMs) unlock unprecedented multimodal and multitask intelligence, yet their cloud-centric deployment precludes real-time responsiveness and compromises user privacy. Meanwhile, monolithic execution at the edge remains infeasible under stringent resource limits and uncertain network dynamics. To bridge this gap, we propose a microservice-based FM inference framework that exploits the intrinsic functional asymmetry between heavyweight core services and agile light services. Our two-tier deployment strategy ensures robust Quality of Service (QoS) under resource contention. Specifically, core services are placed statically via a long-term network-aware integer program with sparsity constraints to form a fault-tolerant backbone. On the other hand, light services are orchestrated dynamically by a low-complexity online controller that integrates effective capacity theory with Lyapunov optimization, providing probabilistic latency guarantees under real-time workload fluctuations. Simulations demonstrate that our framework achieves over 84% average on-time task completion with moderate deployment costs and maintains strong robustness as the system load scales.

Quality diversity (QD) is a branch of evolutionary computation that seeks high-quality and behaviorally diverse solutions to a problem. While adversarial problems are common, classical QD cannot be easily applied to them, as both the fitness and the behavior depend on the opposing solutions. Recently, Generational Adversarial MAP-Elites (GAME) has been proposed to coevolve both sides of an adversarial problem by alternating the execution of a multi-task QD algorithm against previous elites, called tasks. The original algorithm selects new tasks based on a behavioral criterion, which may lead to undesired dynamics due to inter-side dependencies. In addition, comparing sets of solutions cannot be done directly using classical QD measures due to side dependencies. In this paper, we (1) use an inter-variants tournament to compare the sets of solutions, ensuring a fair comparison, with 6 measures of quality and diversity, and (2) propose two tournament-informed task selection methods to promote higher quality and diversity at each generation. We evaluate the variants across three adversarial problems: Pong, a Cat-and-mouse game, and a Pursuers-and-evaders game. We show that the tournament-informed task selection method leads to higher adversarial quality and diversity. We hope that this work will help further advance adversarial quality diversity. Code, videos, and supplementary material are available at https://github.com/Timothee-ANNE/GAME_tournament_informed.

Public olfaction datasets are small and fragmented across single molecules and mixtures, limiting learning of generalizable odor representations. Recent works either learn single-molecule embeddings or address mixtures via similarity or pairwise label prediction, leaving representations separate and unaligned. In this work, we propose AROMMA, a framework that learns a unified embedding space for single molecules and two-molecule mixtures. Each molecule is encoded by a chemical foundation model and the mixtures are composed by an attention-based aggregator, ensuring both permutation invariance and asymmetric molecular interactions. We further align odor descriptor sets using knowledge distillation and class-aware pseudo-labeling to enrich missing mixture annotations. AROMMA achieves state-of-the-art performance in both single-molecule and molecule-pair datasets, with up to 19.1% AUROC improvement, demonstrating a robust generalization in two domains.

We demonstrate that gauge equivariant diffusion models can accurately model the physics of non-Abelian lattice gauge theory using the Metropolis-adjusted annealed Langevin algorithm (MAALA), as exemplified by computations in two-dimensional U(2) and SU(2) gauge theories. Our network architecture is based on lattice gauge equivariant convolutional neural networks (L-CNNs), which respect local and global symmetries on the lattice. Models are trained on a single ensemble generated using a traditional Monte Carlo method. By studying Wilson loops of various size as well as the topological susceptibility, we find that the diffusion approach generalizes remarkably well to larger inverse couplings and lattice sizes with negligible loss of accuracy while retaining moderately high acceptance rates.

Recent deep learning models increasingly rely on depth without structural guarantees on the validity of intermediate representations, rendering early stopping and adaptive computation ill-posed. We address this limitation by formulating a structural requirement for state-space model's scale-consistent latent dynamics across iterative refinement, and derive Fractal of Stationary Transformations (FROST), which enforces a self-similar representation manifold through a fractal inductive bias. Under this geometry, intermediate states correspond to different resolutions of a shared representation, and we provide a geometric analysis establishing contraction and stable convergence across iterations. As a consequence of this scale-consistent structure, halting naturally admits a ranking-based formulation driven by intrinsic feature quality rather than extrinsic objectives. Controlled experiments on ImageNet-100 empirically verify the predicted scale-consistent behavior, showing that adaptive efficiency emerges from the aligned latent geometry.

Software architecture is inherently knowledge-centric. The architectural knowledge is distributed across heterogeneous software artifacts such as requirements documents, design diagrams, code, and documentation, making it difficult for developers to access and utilize this knowledge effectively. Moreover, as systems evolve, inconsistencies frequently emerge between these artifacts, leading to architectural erosion and impeding maintenance activities. We envision an automated pipeline that systematically extracts architectural knowledge from diverse artifacts, links them, identifies and resolves inconsistencies, and consolidates this knowledge into a structured knowledge base. This knowledge base enables critical activities such as architecture conformance checking and change impact analysis, while supporting natural language question-answering to improve access to architectural knowledge. To realize this vision, we plan to develop specialized extractors for different artifact types, design a unified knowledge representation schema, implement consistency checking mechanisms, and integrate retrieval-augmented generation techniques for conversational knowledge access.

Real-world physical systems are inherently complex, often involving the coupling of multiple physics, making their simulation both highly valuable and challenging. Many mainstream approaches face challenges when dealing with decoupled data. Besides, they also suffer from low efficiency and fidelity in strongly coupled spatio-temporal physical systems. Here we propose GenCP, a novel and elegant generative paradigm for coupled multiphysics simulation. By formulating coupled-physics modeling as a probability modeling problem, our key innovation is to integrate probability density evolution in generative modeling with iterative multiphysics coupling, thereby enabling training on data from decoupled simulation and inferring coupled physics during sampling. We also utilize operator-splitting theory in the space of probability evolution to establish error controllability guarantees for this "conditional-to-joint" sampling scheme. We evaluate our paradigm on a synthetic setting and three challenging multi-physics scenarios to demonstrate both principled insight and superior application performance of GenCP. Code is available at this repo: github.com/AI4Science-WestlakeU/GenCP.

Speech separation (SS) has advanced significantly with neural network-based methods, showing improved performance on signal-level metrics. However, these methods often struggle to maintain speech intelligibility in the separated signals, which can negatively affect the performance of downstream tasks such as speech recognition. In this work, we propose SLM-SS, a novel approach that applies speech language models to SS, aiming to enhance the intelligibility and coherence of the separated signals. We frame SS as discrete multi-codebook sequence generation, using Encoder-Decoder models to map quantized speech mixtures to target tokens. In addition to the autoregressive modeling strategy, we introduce a non-autoregressive model to improve decoding efficiency for residual tokens. Experimental results on the LibriMix dataset demonstrate that our approach shows significantly better preservation of speech intelligibility, leading to improved linguistic consistency in a variety of downstream tasks compared to existing approaches.

Benchmarks are important tools to track progress in the development of Large Language Models (LLMs), yet inaccuracies in datasets and evaluation methods consistently undermine their effectiveness. Here, we present Omni-MATH-2, a manually revised version of the Omni-MATH dataset comprising a clean, exact-answer subset ($n{=}4181$) and a tagged, non-standard subset ($n{=}247$). Each problem was audited to ensure LaTeX compilability, solvability and verifiability, which involved adding missing figures or information, labeling problems requiring a proof, estimation or image, and removing clutter. This process significantly reduces dataset-induced noise, thereby providing a more precise assessment of model performance. The annotated dataset also allows us to evaluate judge-induced noise by comparing GPT-5 mini with the original Omni-Judge, revealing substantial discrepancies between judges on both the clean and tagged problem subsets. Expert annotations reveal that Omni-Judge is wrong in $96.4\%$ of the judge disagreements, indicating its inability to differentiate between models' abilities, even well before saturation of the benchmark occurs. As problems become more challenging, we find that increasingly competent judges become essential in order to prevent judge errors from masking genuine differences between models. Finally, neither judge identifies the present failure modes for the subset of tagged problems, demonstrating that dataset quality and judge reliability are both critical to develop accurate benchmarks of model performance.

Purifying sour water is essential for reducing emissions, minimizing corrosion risks, enabling the reuse of treated water in industrial or domestic applications, and ultimately lowering operational costs. Moreover, automating the purification process helps reduce the risk of worker harm by limiting human involvement. Crude oil contains acidic components such as hydrogen sulfide, carbon dioxide, and other chemical compounds. During processing, these substances are partially released into sour water. If not properly treated, sour water poses serious environmental threats and accelerates the corrosion of pipelines and equipment. This paper presents a fuzzy expert system, combined with a custom-generated digital twin, developed from a documented industrial process to maintain key parameters at desired levels by mimicking human reasoning. The control strategy is designed to be simple and intuitive, allowing junior or non-expert personnel to interact with the system effectively. The digital twin was developed using Honeywell UniSim Design R492 to simulate real industrial behavior accurately. Valve dynamics were modeled through system identification in MATLAB, and real-time data exchange between the simulator and controller was established using OPC DA. The fuzzy controller applies split-range control to two valves and was tested under 21 different initial pressure conditions using five distinct defuzzification strategies, resulting in a total of 105 unique test scenarios. System performance was evaluated using both error-based metrics (MSE, RMSE, MAE, IAE, ISE, ITAE) and dynamic response metrics, including overshoot, undershoot, rise time, fall time, settling time, and steady-state error. A web-based simulation interface was developed in Python using the Streamlit framework. Although demonstrated here for sour water treatment, the proposed fuzzy expert system is general-purpose.

We introduce a novel motion capture system that reconstructs full-body 3D motion using only sparse pairwise distance (PWD) measurements from body-mounted(UWB) sensors. Using time-of-flight ranging between wireless nodes, our method eliminates the need for external cameras, enabling robust operation in uncontrolled and outdoor environments. Unlike traditional optical or inertial systems, our approach is shape-invariant and resilient to environmental constraints such as lighting and magnetic interference. At the core of our system is Wild-Poser (WiP for short), a compact, real-time Transformer-based architecture that directly predicts 3D joint positions from noisy or corrupted PWD measurements, which can later be used for joint rotation reconstruction via learned methods. WiP generalizes across subjects of varying morphologies, including non-human species, without requiring individual body measurements or shape fitting. Operating in real time, WiP achieves low joint position error and demonstrates accurate 3D motion reconstruction for both human and animal subjects in-the-wild. Our empirical analysis highlights its potential for scalable, low-cost, and general purpose motion capture in real-world settings.

In the era of explosive growth in academic literature, the burden of literature review on scholars are increasing. Proactively recommending academic papers that align with scholars' literature needs in the research process has become one of the crucial pathways to enhance research efficiency and stimulate innovative thinking. Current academic paper recommendation systems primarily focus on broad and coarse-grained suggestions based on general topic or field similarities. While these systems effectively identify related literature, they fall short in addressing scholars' more specific and fine-grained needs, such as locating papers that utilize particular research methods, or tackle distinct research tasks within the same topic. To meet the diverse and specific literature needs of scholars in the research process, this paper proposes a novel academic paper recommendation method. This approach embeds multidimensional information by integrating new types of fine-grained knowledge entities, title and abstract of document, and citation data. Recommendations are then generated by calculating the similarity between combined paper vectors. The proposed recommendation method was evaluated using the STM-KG dataset, a knowledge graph that incorporates scientific concepts derived from papers across ten distinct domains. The experimental results indicate that our method outperforms baseline models, achieving an average precision of 27.3% among the top 50 recommendations. This represents an improvement of 6.7% over existing approaches.

Large Language Models (LLMs) have recently empowered agentic frameworks to exhibit advanced reasoning and planning capabilities. However, their integration in robotic control pipelines remains limited in two aspects: (1) prior \ac{llm}-based approaches often lack modular, agentic execution mechanisms, limiting their ability to plan, reflect on outcomes, and revise actions in a closed-loop manner; and (2) existing benchmarks for manipulation tasks focus on low-level control and do not systematically evaluate multistep reasoning and linguistic variation. In this paper, we propose Agentic LLM for Robot Manipulation (ALRM), an LLM-driven agentic framework for robotic manipulation. ALRM integrates policy generation with agentic execution through a ReAct-style reasoning loop, supporting two complementary modes: Code-asPolicy (CaP) for direct executable control code generation, and Tool-as-Policy (TaP) for iterative planning and tool-based action execution. To enable systematic evaluation, we also introduce a novel simulation benchmark comprising 56 tasks across multiple environments, capturing linguistically diverse instructions. Experiments with ten LLMs demonstrate that ALRM provides a scalable, interpretable, and modular approach for bridging natural language reasoning with reliable robotic execution. Results reveal Claude-4.1-Opus as the top closed-source model and Falcon-H1-7B as the top open-source model under CaP.

Since their inception, artificial neural networks have relied on manually designed architectures and inductive biases to better adapt to data and tasks. With the rise of deep learning and the expansion of parameter spaces, they have begun to exhibit brain-like functional behaviors. Nevertheless, artificial neural networks remain fundamentally different from biological neural systems in structural organization, learning mechanisms, and evolutionary pathways. From the perspective of neuroscience, we rethink the formation and evolution of intelligence and proposes a new neural network paradigm, Brain-like Neural Network (BNN). We further present the first instantiation of a BNN termed LuminaNet that operates without convolutions or self-attention and is capable of autonomously modifying its architecture. We conduct extensive experiments demonstrating that LuminaNet can achieve self-evolution through dynamic architectural changes. On the CIFAR-10, LuminaNet achieves top-1 accuracy improvements of 11.19%, 5.46% over LeNet-5 and AlexNet, respectively, outperforming MLP-Mixer, ResMLP, and DeiT-Tiny among MLP/ViT architectures. On the TinyStories text generation task, LuminaNet attains a perplexity of 8.4, comparable to a single-layer GPT-2-style Transformer, while reducing computational cost by approximately 25% and peak memory usage by nearly 50%. Code and interactive structures are available at https://github.com/aaroncomo/LuminaNet.

Large vision-language models (LVLMs) exhibit remarkable capabilities in cross-modal tasks but face significant safety challenges, which undermine their reliability in real-world applications. Efforts have been made to build LVLM safety evaluation benchmarks to uncover their vulnerability. However, existing benchmarks are hindered by their labor-intensive construction process, static complexity, and limited discriminative power. Thus, they may fail to keep pace with rapidly evolving models and emerging risks. To address these limitations, we propose VLSafetyBencher, the first automated system for LVLM safety benchmarking. VLSafetyBencher introduces four collaborative agents: Data Preprocessing, Generation, Augmentation, and Selection agents to construct and select high-quality samples. Experiments validates that VLSafetyBencher can construct high-quality safety benchmarks within one week at a minimal cost. The generated benchmark effectively distinguish safety, with a safety rate disparity of 70% between the most and least safe models.

Fine-tuning Large Language Models (LLMs) with downstream data is often considered time-consuming and expensive. Structured pruning methods are primarily employed to improve the inference efficiency of pre-trained models. Meanwhile, they often require additional time and memory for training, knowledge distillation, structure search, and other strategies, making efficient model fine-tuning challenging to achieve. To simultaneously enhance the training and inference efficiency of downstream task fine-tuning, we introduce GradPruner, which can prune layers of LLMs guided by gradients in the early stages of fine-tuning. GradPruner uses the cumulative gradients of each parameter during the initial phase of fine-tuning to compute the Initial Gradient Information Accumulation Matrix (IGIA-Matrix) to assess the importance of layers and perform pruning. We sparsify the pruned layers based on the IGIA-Matrix and merge them with the remaining layers. Only elements with the same sign are merged to reduce interference from sign variations. We conducted extensive experiments on two LLMs across eight downstream datasets. Including medical, financial, and general benchmark tasks. The results demonstrate that GradPruner has achieved a parameter reduction of 40% with only a 0.99% decrease in accuracy. Our code is publicly available.

Synthetic neuroimaging data can mitigate critical limitations of real-world datasets, including the scarcity of rare phenotypes, domain shifts across scanners, and insufficient longitudinal coverage. However, existing generative models largely rely on weak conditioning signals, such as labels or text, which lack anatomical grounding and often produce biologically implausible outputs. To this end, we introduce Cor2Vox, a cortex-grounded generative framework for brain magnetic resonance image (MRI) synthesis that ties image generation to continuous structural priors of the cerebral cortex. It leverages high-resolution cortical surfaces to guide a 3D shape-to-image Brownian bridge diffusion process, enabling topologically faithful synthesis and precise control over underlying anatomies. To support the generation of new, realistic brain shapes, we developed a large-scale statistical shape model of cortical morphology derived from over 33,000 UK Biobank scans. We validated the fidelity of Cor2Vox based on traditional image quality metrics, advanced cortical surface reconstruction, and whole-brain segmentation quality, outperforming many baseline methods. Across three applications, namely (i) anatomically consistent synthesis, (ii) simulation of progressive gray matter atrophy, and (iii) harmonization of in-house frontotemporal dementia scans with public datasets, Cor2Vox preserved fine-grained cortical morphology at the sub-voxel level, exhibiting remarkable robustness to variations in cortical geometry and disease phenotype without retraining.

High-quality evaluation benchmarks are pivotal for deploying Large Language Models (LLMs) in Automated Code Review (ACR). However, existing benchmarks suffer from two critical limitations: first, the lack of multi-language support in repository-level contexts, which restricts the generalizability of evaluation results; second, the reliance on noisy, incomplete ground truth derived from raw Pull Request (PR) comments, which constrains the scope of issue detection. To address these challenges, we introduce AACR-Bench a comprehensive benchmark that provides full cross-file context across multiple programming languages. Unlike traditional datasets, AACR-Bench employs an "AI-assisted, Expert-verified" annotation pipeline to uncover latent defects often overlooked in original PRs, resulting in a 285\% increase in defect coverage. Extensive evaluations of mainstream LLMs on AACR-Bench reveal that previous assessments may have either misjudged or only partially captured model capabilities due to data limitations. Our work establishes a more rigorous standard for ACR evaluation and offers new insights on LLM based ACR, i.e., the granularity/level of context and the choice of retrieval methods significantly impact ACR performance, and this influence varies depending on the LLM, programming language, and the LLM usage paradigm e.g., whether an Agent architecture is employed. The code, data, and other artifacts of our evaluation set are available at https://github.com/alibaba/aacr-bench .

Fact-checking remains a demanding and time-consuming task, still largely dependent on manual verification and unable to match the rapid spread of misinformation online. This is particularly important because debunking false information typically takes longer to reach consumers than the misinformation itself; accelerating corrections through automation can therefore help counter it more effectively. Although many organizations perform manual fact-checking, this approach is difficult to scale given the growing volume of digital content. These limitations have motivated interest in automating fact-checking, where identifying claims is a crucial first step. However, progress has been uneven across languages, with English dominating due to abundant annotated data. Portuguese, like other languages, still lacks accessible, licensed datasets, limiting research, NLP developments and applications. In this paper, we introduce ClaimPT, a dataset of European Portuguese news articles annotated for factual claims, comprising 1,308 articles and 6,875 individual annotations. Unlike most existing resources based on social media or parliamentary transcripts, ClaimPT focuses on journalistic content, collected through a partnership with LUSA, the Portuguese News Agency. To ensure annotation quality, two trained annotators labeled each article, with a curator validating all annotations according to a newly proposed scheme. We also provide baseline models for claim detection, establishing initial benchmarks and enabling future NLP and IR applications. By releasing ClaimPT, we aim to advance research on low-resource fact-checking and enhance understanding of misinformation in news media.

Safety-aligned LLMs suffer from two failure modes: jailbreak (answering harmful inputs) and over-refusal (declining benign queries). Existing vector steering methods adjust the magnitude of answer vectors, but this creates a fundamental trade-off -- reducing jailbreak increases over-refusal and vice versa. We identify the root cause: LLMs encode the decision to answer (answer vector $v_a$) and the judgment of input safety (benign vector $v_b$) as nearly orthogonal directions, treating them as independent processes. We propose LLM-VA, which aligns $v_a$ with $v_b$ through closed-form weight updates, making the model's willingness to answer causally dependent on its safety assessment -- without fine-tuning or architectural changes. Our method identifies vectors at each layer using SVMs, selects safety-relevant layers, and iteratively aligns vectors via minimum-norm weight modifications. Experiments on 12 LLMs demonstrate that LLM-VA achieves 11.45% higher F1 than the best baseline while preserving 95.92% utility, and automatically adapts to each model's safety bias without manual tuning. Code and models are available at https://hotbento.github.io/LLM-VA-Web/.

The calibration of simulators for complex social systems aims to identify the optimal parameter that drives the output of the simulator best matching the target data observed from the system. As many social systems may change internally over time, calibration naturally becomes an online task, requiring parameters to be updated continuously to maintain the simulator's fidelity. In this work, the online setting is first formulated as a dynamic optimization problem (DOP), requiring the search for a sequence of optimal parameters that fit the simulator to real system changes. However, in contrast to traditional DOP formulations, online calibration explicitly incorporates the observational data as the driver of environmental dynamics. Due to this fundamental difference, existing Evolutionary Dynamic Optimization (EDO) methods, despite being extensively studied for black-box DOPs, are ill-equipped to handle such a scenario. As a result, online calibration problems constitute a new set of challenging DOPs. Here, we propose to explicitly learn the posterior distributions of the parameters and the observational data, thereby facilitating both change detection and environmental adaptation of existing EDOs for this scenario. We thus present a pretrained posterior model for implementation, and fine-tune it during the optimization. Extensive tests on both economic and financial simulators verify that the posterior distribution strongly promotes EDOs in such DOPs widely existed in social science.

Injury occurrence in football poses significant challenges for athletes and teams, carrying personal, competitive, and financial consequences. While machine learning has been applied to injury prediction before, existing approaches often rely on static pre-season data and binary outcomes, limiting their real-world utility. This study investigates the feasibility of using a DeepHit neural network to forecast time-to-injury from longitudinal athlete monitoring data, while providing interpretable predictions. The analysis utilised the publicly available SoccerMon dataset, containing two seasons of training, match, and wellness records from elite female footballers. Data was pre-processed through cleaning, feature engineering, and the application of three imputation strategies. Baseline models (Random Forest, XGBoost, Logistic Regression) were optimised via grid search for benchmarking, while the DeepHit model, implemented with a multilayer perceptron backbone, was evaluated using chronological and leave-one-player-out (LOPO) validation. DeepHit achieved a concordance index of 0.762, outperforming baseline models and delivering individualised, time-varying risk estimates. Shapley Additive Explanations (SHAP) identified clinically relevant predictors consistent with established risk factors, enhancing interpretability. Overall, this study provides a novel proof of concept: survival modelling with DeepHit shows strong potential to advance injury forecasting in football, offering accurate, explainable, and actionable insights for injury prevention across competitive levels.

Informed down-sampling (IDS) is known to improve performance in symbolic regression when combined with various selection strategies, especially tournament selection. However, recent work found that IDS's gains are not consistent across all problems. Our analysis reveals that IDS performance is worse for problems containing outliers. IDS systematically favors including outliers in subsets which pushes GP towards finding solutions that overfit to outliers. To address this, we introduce ROIDS (Robust Outlier-Aware Informed Down-Sampling), which excludes potential outliers from the sampling process of IDS. With ROIDS it is possible to keep the advantages of IDS without overfitting to outliers and to compete on a wide range of benchmark problems. This is also reflected in our experiments in which ROIDS shows the desired behavior on all studied benchmark problems. ROIDS consistently outperforms IDS on synthetic problems with added outliers as well as on a wide range of complex real-world problems, surpassing IDS on over 80% of the real-world benchmark problems. Moreover, compared to all studied baseline approaches, ROIDS achieves the best average rank across all tested benchmark problems. This robust behavior makes ROIDS a reliable down-sampling method for selection in symbolic regression, especially when outliers may be included in the data set.

We investigate the expressive power of state space models (SSM), which have recently emerged as a potential alternative to transformer architectures in large language models. Building on recent work, we analyse SSM expressiveness through fragments and extensions of linear temporal logic over finite traces. Our results show that the expressive capabilities of SSM vary substantially depending on the underlying gating mechanism. We further distinguish between SSM operating over fixed-width arithmetic (quantised models), whose expressive power remains within regular languages, and SSM with unbounded precision, which can capture counting properties and non-regular languages. In addition, we provide a systematic comparison between these different SSM variants and known results on transformers, thereby clarifying how the two architectures relate in terms of expressive power.

Incorporating demonstration data into reinforcement learning (RL) can greatly accelerate learning, but existing approaches often assume demonstrations are optimal and fully aligned with the target task. In practice, demonstrations are frequently sparse, suboptimal, or misaligned, which can degrade performance when these demonstrations are integrated into RL. We propose Adaptive Policy Composition (APC), a hierarchical model that adaptively composes multiple data-driven Normalizing Flow (NF) priors. Instead of enforcing strict adherence to the priors, APC estimates each prior's applicability to the target task while leveraging them for exploration. Moreover, APC either refines useful priors, or sidesteps misaligned ones when necessary to optimize downstream reward. Across diverse benchmarks, APC accelerates learning when demonstrations are aligned, remains robust under severe misalignment, and leverages suboptimal demonstrations to bootstrap exploration while avoiding performance degradation caused by overly strict adherence to suboptimal demonstrations.

Recent advances in LLM-based ASR connect frozen speech encoders with Large Language Models (LLMs) via lightweight projectors. While effective in monolingual settings, a single projector struggles to capture the diverse acoustic-to-semantic mappings required for multilingual ASR. To address this, we propose SMEAR-MoE, a stabilized Mixture-of-Experts projector that ensures dense gradient flow to all experts, preventing expert collapse while enabling cross-lingual sharing. We systematically compare monolithic, static multi-projector, and dynamic MoE designs across four Indic languages (Hindi, Marathi, Tamil, Telugu). Our SMEAR-MoE achieves strong performance, delivering upto a 7.6% relative WER reduction over the single-projector baseline, while maintaining comparable runtime efficiency. Analysis of expert routing further shows linguistically meaningful specialization, with related languages sharing experts. These results demonstrate that stable multi-expert projectors are key to scalable and robust multilingual ASR.

Graph neural networks (GNNs) are widely believed to excel at node representation learning through trainable neighborhood aggregations. We challenge this view by introducing Fixed Aggregation Features (FAFs), a training-free approach that transforms graph learning tasks into tabular problems. This simple shift enables the use of well-established tabular methods, offering strong interpretability and the flexibility to deploy diverse classifiers. Across 14 benchmarks, well-tuned multilayer perceptrons trained on FAFs rival or outperform state-of-the-art GNNs and graph transformers on 12 tasks -- often using only mean aggregation. The only exceptions are the Roman Empire and Minesweeper datasets, which typically require unusually deep GNNs. To explain the theoretical possibility of non-trainable aggregations, we connect our findings to Kolmogorov-Arnold representations and discuss when mean aggregation can be sufficient. In conclusion, our results call for (i) richer benchmarks benefiting from learning diverse neighborhood aggregations, (ii) strong tabular baselines as standard, and (iii) employing and advancing tabular models for graph data to gain new insights into related tasks.

Deep Neural Networks remain inherently vulnerable to backdoor attacks. Traditional test-time defenses largely operate under the paradigm of internal diagnosis methods like model repairing or input robustness, yet these approaches are often fragile under advanced attacks as they remain entangled with the victim model's corrupted parameters. We propose a paradigm shift from Internal Diagnosis to External Semantic Auditing, arguing that effective defense requires decoupling safety from the victim model via an independent, semantically grounded auditor. To this end, we present a framework harnessing Universal Vision-Language Models (VLMs) as evolving semantic gatekeepers. We introduce PRISM (Prototype Refinement & Inspection via Statistical Monitoring), which overcomes the domain gap of general VLMs through two key mechanisms: a Hybrid VLM Teacher that dynamically refines visual prototypes online, and an Adaptive Router powered by statistical margin monitoring to calibrate gating thresholds in real-time. Extensive evaluation across 17 datasets and 11 attack types demonstrates that PRISM achieves state-of-the-art performance, suppressing Attack Success Rate to <1% on CIFAR-10 while improving clean accuracy, establishing a new standard for model-agnostic, externalized security.

Claim verification is a core component of automated fact-checking systems, aimed at determining the truthfulness of a statement by assessing it against reliable evidence sources such as documents or knowledge bases. This work presents KG-CRAFT, a method that improves automatic claim verification by leveraging large language models (LLMs) augmented with contrastive questions grounded in a knowledge graph. KG-CRAFT first constructs a knowledge graph from claims and associated reports, then formulates contextually relevant contrastive questions based on the knowledge graph structure. These questions guide the distillation of evidence-based reports, which are synthesised into a concise summary that is used for veracity assessment by LLMs. Extensive evaluations on two real-world datasets (LIAR-RAW and RAWFC) demonstrate that our method achieves a new state-of-the-art in predictive performance. Comprehensive analyses validate in detail the effectiveness of our knowledge graph-based contrastive reasoning approach in improving LLMs' fact-checking capabilities.

The automation of analog integrated circuit (IC) design remains a longstanding challenge, primarily due to the intricate interdependencies among physical layout, parasitic effects, and circuit-level performance. These interactions impose complex constraints that are difficult to accurately capture and optimize using conventional design methodologies. Although recent advances in machine learning (ML) have shown promise in automating specific stages of the analog design flow, the development of holistic, end-to-end frameworks that integrate these stages and iteratively refine layouts using post-layout, parasitic-aware performance feedback is still in its early stages. Furthermore, progress in this direction is hindered by the limited availability of open, high-quality datasets tailored to the analog domain, restricting both the benchmarking and the generalizability of ML-based techniques. To address these limitations, we present OSIRIS, a scalable dataset generation pipeline for analog IC design. OSIRIS systematically explores the design space of analog circuits while producing comprehensive performance metrics and metadata, thereby enabling ML-driven research in electronic design automation (EDA). In addition, we release a dataset consisting of 87,100 circuit variations generated with OSIRIS, accompanied by a reinforcement learning (RL)-based baseline method that exploits OSIRIS for analog design optimization.

Sustainable monetization of Large Language Models (LLMs) remains a critical open challenge. Traditional search advertising, which relies on static keywords, fails to capture the fleeting, context-dependent user intents--the specific information, goods, or services a user seeks--embedded in conversational flows. Beyond the standard goal of social welfare maximization, effective LLM advertising imposes additional requirements on contextual coherence (ensuring ads align semantically with transient user intents) and computational efficiency (avoiding user interaction latency), as well as adherence to ethical and regulatory standards, including preserving privacy and ensuring explicit ad disclosure. Although various recent solutions have explored bidding on token-level and query-level, both categories of approaches generally fail to holistically satisfy this multifaceted set of constraints. We propose a practical framework that resolves these tensions through two decoupling strategies. First, we decouple ad insertion from response generation to ensure safety and explicit disclosure. Second, we decouple bidding from specific user queries by using ``genres'' (high-level semantic clusters) as a proxy. This allows advertisers to bid on stable categories rather than sensitive real-time response, reducing computational burden and privacy risks. We demonstrate that applying the VCG auction mechanism to this genre-based framework yields approximately dominant strategy incentive compatibility (DSIC) and individual rationality (IR), as well as approximately optimal social welfare, while maintaining high computational efficiency. Finally, we introduce an "LLM-as-a-Judge" metric to estimate contextual coherence. Our experiments show that this metric correlates strongly with human ratings (Spearman's $ρ\approx 0.66$), outperforming 80% of individual human evaluators.

Humanoid control often leverages motion priors from human demonstrations to encourage natural behaviors. However, such demonstrations are frequently suboptimal or misaligned with robotic tasks due to embodiment differences, retargeting errors, and task-irrelevant variations, causing naïve imitation to degrade task performance. Conversely, task-only reinforcement learning admits many task-optimal solutions, often resulting in unnatural or unstable motions. This exposes a fundamental limitation of linear reward mixing in adversarial imitation learning. We propose \emph{Task-Centric Motion Priors} (TCMP), a task-priority adversarial imitation framework that treats imitation as a conditional regularizer rather than a co-equal objective. TCMP maximizes task improvement while incorporating imitation signals only when they are compatible with task progress, yielding an adaptive, geometry-aware update that preserves task-feasible descent and suppresses harmful imitation under misalignment. We provide theoretical analysis of gradient conflict and task-priority stationary points, and validate our claims through humanoid control experiments demonstrating robust task performance with consistent motion style under noisy demonstrations.

Automatic post-editing (APE) aims to refine machine translations by correcting residual errors. Although recent large language models (LLMs) demonstrate strong translation capabilities, their effectiveness for APE--especially under document-level context--remains insufficiently understood. We present a systematic comparison of proprietary and open-weight LLMs under a naive document-level prompting setup, analyzing APE quality, contextual behavior, robustness, and efficiency. Our results show that proprietary LLMs achieve near human-level APE quality even with simple one-shot prompting, regardless of whether document context is provided. While these models exhibit higher robustness to data poisoning attacks than open-weight counterparts, this robustness also reveals a limitation: they largely fail to exploit document-level context for contextual error correction. Furthermore, standard automatic metrics do not reliably reflect these qualitative improvements, highlighting the continued necessity of human evaluation. Despite their strong performance, the substantial cost and latency overheads of proprietary LLMs render them impractical for real-world APE deployment. Overall, our findings elucidate both the promise and current limitations of LLM-based document-aware APE, and point toward the need for more efficient long-context modeling approaches for translation refinement.

Synthetic simulation data and real-world human data provide scalable alternatives to circumvent the prohibitive costs of robot data collection. However, these sources suffer from the sim-to-real visual gap and the human-to-robot embodiment gap, respectively, which limits the policy's generalization to real-world scenarios. In this work, we identify a natural yet underexplored complementarity between these sources: simulation offers the robot action that human data lacks, while human data provides the real-world observation that simulation struggles to render. Motivated by this insight, we present SimHum, a co-training framework to simultaneously extract kinematic prior from simulated robot actions and visual prior from real-world human observations. Based on the two complementary priors, we achieve data-efficient and generalizable robotic manipulation in real-world tasks. Empirically, SimHum outperforms the baseline by up to $\mathbf{40\%}$ under the same data collection budget, and achieves a $\mathbf{62.5\%}$ OOD success with only 80 real data, outperforming the real only baseline by $7.1\times$. Videos and additional information can be found at \href{https://kaipengfang.github.io/sim-and-human}{project website}.

Within the domain of large language models, reinforcement fine-tuning algorithms necessitate the generation of a complete reasoning trajectory beginning from the input query, which incurs significant computational overhead during the rollout phase of training. To address this issue, we analyze the impact of different segments of the reasoning path on the correctness of the final result and, based on these insights, propose Reinforcement Fine-Tuning with Partial Reasoning Optimization (RPO), a plug-and-play reinforcement fine-tuning algorithm. Unlike traditional reinforcement fine-tuning algorithms that generate full reasoning paths, RPO trains the model by generating suffixes of the reasoning path using experience cache. During the rollout phase of training, RPO reduces token generation in this phase by approximately 95%, greatly lowering the theoretical time overhead. Compared with full-path reinforcement fine-tuning algorithms, RPO reduces the training time of the 1.5B model by 90% and the 7B model by 72%. At the same time, it can be integrated with typical algorithms such as GRPO and DAPO, enabling them to achieve training acceleration while maintaining performance comparable to the original algorithms. Our code is open-sourced at https://github.com/yhz5613813/RPO.

Obstructions such as raindrops, fences, or dust degrade captured images, especially when mechanical cleaning is infeasible. Conventional solutions to obstructions rely on a bulky compound optics array or computational inpainting, which compromise compactness or fidelity. Metalenses composed of subwavelength meta-atoms promise compact imaging, but simultaneous achievement of broadband and obstruction-free imaging remains a challenge, since a metalens that images distant scenes across a broadband spectrum cannot properly defocus near-depth occlusions. Here, we introduce a learned split-spectrum metalens that enables broadband obstruction-free imaging. Our approach divides the spectrum of each RGB channel into pass and stop bands with multi-band spectral filtering and learns the metalens to focus light from far objects through pass bands, while filtering focused near-depth light through stop bands. This optical signal is further enhanced using a neural network. Our learned split-spectrum metalens achieves broadband and obstruction-free imaging with relative PSNR gains of 32.29% and improves object detection and semantic segmentation accuracies with absolute gains of +13.54% mAP, +48.45% IoU, and +20.35% mIoU over a conventional hyperbolic design. This promises robust obstruction-free sensing and vision for space-constrained systems, such as mobile robots, drones, and endoscopes.

Production LLM deployments serve diverse workloads where cost and quality requirements vary by customer tier, time of day, and query criticality. Model serving systems accept latency SLOs directly. LLM routers do not. They force operators to tune parameters offline and guess what accuracy might result. The relationship between parameters and outcomes is indirect, non-monotonic, and dataset-dependent. Operators need to specify accuracy targets, not infer them from opaque settings. We present PROTEUS (Polymorphic Router for Operational Target Enforcement with Unified SLA), a router that accepts accuracy targets tau as runtime input. PROTEUS uses Lagrangian dual control. A learned dual variable lambda tracks constraint violations during training and conditions the policy network. This lets the router translate specified tau values into routing decisions that satisfy them. A single trained model serves the full accuracy spectrum without retraining.We evaluate on RouterBench (11 models, 405K queries) and SPROUT (14 models, 45K queries). PROTEUS achieves consistent floor compliance where accuracy meets or exceeds tau. The target-response correlation reaches 0.97 to 0.98. The closest baseline, OmniRouter, meets floors only 22% of the time despite also using Lagrangian optimization. PROTEUS operates across tau in [0.85, 0.95] from a single model. On RouterBench it achieves 90.1% accuracy, within 1.3% of oracle. On SPROUT it achieves 94.0% accuracy, within 4.6% of oracle. Cost savings reach 89.8% versus the best fixed model.

The Muon optimizer has recently attracted attention due to its orthogonalized first-order updates, and a deeper theoretical understanding of its convergence behavior is essential for guiding practical applications; however, existing convergence guarantees are either coarse or obtained under restrictive analytical settings. In this work, we establish sharper convergence guarantees for the Muon optimizer through a direct and simplified analysis that does not rely on restrictive assumptions on the update rule. Our results improve upon existing bounds by achieving faster convergence rates while covering a broader class of problem settings. These findings provide a more accurate theoretical characterization of Muon and offer insights applicable to a broader class of orthogonalized first-order methods.

Recent speech modeling relies on explicit attributes such as pitch, content, and speaker identity, but these alone cannot capture the full richness of natural speech. We introduce RT-MAE, a novel masked autoencoder framework that augments the supervised attributes-based modeling with unsupervised residual trainable tokens, designed to encode the information not explained by explicit labeled factors (e.g., timbre variations, noise, emotion etc). Experiments show that RT-MAE improves reconstruction quality, preserving content and speaker similarity while enhancing expressivity. We further demonstrate its applicability to speech enhancement, removing noise at inference while maintaining controllability and naturalness.

Real-world Vehicle Routing Problems (RWVRPs) require solving complex, sequence-dependent challenges at scale with constraints such as delivery time window, replenishment or recharging stops, asymmetric travel cost, etc. While recent neural methods achieve strong results on large-scale classical VRP benchmarks, they struggle to address RWVRPs because their strategies overlook sequence dependencies and underutilize edge-level information, which are precisely the characteristics that define the complexity of RWVRPs. We present SEAFormer, a novel transformer that incorporates both node-level and edge-level information in decision-making through two key innovations. First, our Clustered Proximity Attention (CPA) exploits locality-aware clustering to reduce the complexity of attention from $O(n^2)$ to $O(n)$ while preserving global perspective, allowing SEAFormer to efficiently train on large instances. Second, our lightweight edge-aware module captures pairwise features through residual fusion, enabling effective incorporation of edge-based information and faster convergence. Extensive experiments across four RWVRP variants with various scales demonstrate that SEAFormer achieves superior results over state-of-the-art methods. Notably, SEAFormer is the first neural method to solve 1,000+ node RWVRPs effectively, while also achieving superior performance on classic VRPs, making it a versatile solution for both research benchmarks and real-world applications.

Domain Generalization (DG) is a critical area that focuses on developing models capable of performing well on data from unseen distributions, which is essential for real-world applications. Existing approaches primarily concentrate on learning domain-invariant features, which assume that a model robust to variations in the source domains will generalize well to unseen target domains. However, these approaches neglect a deeper analysis at the parameter level, which makes the model hard to explicitly differentiate between parameters sensitive to domain shifts and those robust, potentially hindering its overall ability to generalize. In order to address these limitations, we first build a covariance-based parameter sensitivity analysis framework to quantify the sensitivity of each parameter in a model to domain shifts. By computing the covariance of parameter gradients across multiple source domains, we can identify parameters that are more susceptible to domain variations, which serves as our theoretical foundation. Based on this, we propose Domain-Sensitive Parameter Regularization (DSP-Reg), a principled framework that guides model optimization by a soft regularization technique that encourages the model to rely more on domain-invariant parameters while suppressing those that are domain-specific. This approach provides a more granular control over the model's learning process, leading to improved robustness and generalization to unseen domains. Extensive experiments on benchmarks, such as PACS, VLCS, OfficeHome, and DomainNet, demonstrate that DSP-Reg outperforms state-of-the-art approaches, achieving an average accuracy of 66.7\% and surpassing all baselines.

As GenAI models are adopted to support software engineers and their development teams, understanding effective human-AI collaboration (HAIC) is increasingly important. Socio-emotional intelligence (SEI) enhances collaboration among human teammates, but its role in HAIC remains unclear. Current AI systems lack SEI capabilities that humans bring to teamwork, creating a potential gap in collaborative dynamics. In this study, we investigate how software practitioners perceive the socio-emotional gap in HAIC and what capabilities AI systems require for effective collaboration. Through semi-structured interviews with 10 practitioners, we examine how they think about collaborating with human versus AI teammates, focusing on their SEI expectations and the AI capabilities they envision. Results indicate that practitioners currently view AI models as intellectual teammates rather than social partners and expect fewer SEI attributes from them than from human teammates. However, they see the socio-emotional gap not as AIs failure to exhibit SEI traits, but as a functional gap in collaborative capabilities (AIs inability to negotiate responsibilities, adapt contextually, or maintain sustained partnerships). We introduce the concept of functional equivalents: technical capabilities (internal cognition, contextual intelligence, adaptive learning, and collaborative intelligence) that achieve collaborative outcomes comparable to human SEI attributes. Our findings suggest that effective collaboration with AI for SE tasks may benefit from functional design rather than replicating human SEI traits for SE tasks, thereby redefining collaboration as functional alignment.

Code comments serve a crucial role in software development for documenting functionality, clarifying design choices, and assisting with issue tracking. They capture developers' insights about the surrounding source code, serving as an essential resource for both human comprehension and automated analysis. Nevertheless, since comments are in natural language, they present challenges for machine-based code understanding. To address this, recent studies have applied natural language processing (NLP) and deep learning techniques to classify comments according to developers' intentions. However, existing datasets for this task suffer from size limitations and class imbalance, as they rely on manual annotations and may not accurately represent the distribution of comments in real-world codebases. To overcome this issue, we introduce new synthetic oversampling and augmentation techniques based on high-quality data generation to enhance the NLBSE'26 challenge datasets. Our Synthetic Quality Oversampling Technique and Augmentation Technique (Q-SYNTH) yield promising results, improving the base classifier by $2.56\%$.

This paper considers decentralized nonsmooth nonconvex optimization problem with Lipschitz continuous local functions. We propose an efficient stochastic first-order method with client sampling, achieving the $(δ,ε)$-Goldstein stationary point with the overall sample complexity of ${\mathcal O}(δ^{-1}ε^{-3})$, the computation rounds of ${\mathcal O}(δ^{-1}ε^{-3})$, and the communication rounds of ${\tilde{\mathcal O}}(γ^{-1/2}δ^{-1}ε^{-3})$, where $γ$ is the spectral gap of the mixing matrix for the network. Our results achieve the optimal sample complexity and the sharper communication complexity than existing methods. We also extend our ideas to zeroth-order optimization. Moreover, the numerical experiments show the empirical advantage of our methods.

Using multiple open-access models trained on public datasets, we developed Tri-Reader, a comprehensive, freely available pipeline that integrates lung segmentation, nodule detection, and malignancy classification into a unified tri-stage workflow. The pipeline is designed to prioritize sensitivity while reducing the candidate burden for annotators. To ensure accuracy and generalizability across diverse practices, we evaluated Tri-Reader on multiple internal and external datasets as compared with expert annotations and dataset-provided reference standards.

This paper considers the problem of asynchronous stochastic nonconvex optimization with heavy-tailed gradient noise and arbitrarily heterogeneous computation times across workers. We propose an asynchronous normalized stochastic gradient descent algorithm with momentum. The analysis show that our method achieves the optimal time complexity under the assumption of bounded $p$th-order central moment with $p\in(1,2]$. We also provide numerical experiments to show the effectiveness of proposed method.

This paper presents the web-based platform Machine Learning with Bricks and an accompanying two-day course designed to teach machine learning concepts to students aged 12 to 17 through programming-free robotics activities. Machine Learning with Bricks is an open source platform and combines interactive visualizations with LEGO robotics to teach three core algorithms: KNN, linear regression, and Q-learning. Students learn by collecting data, training models, and interacting with robots via a web-based interface. Pre- and post-surveys with 14 students demonstrate significant improvements in conceptual understanding of machine learning algorithms, positive shifts in AI perception, high platform usability, and increased motivation for continued learning. This work demonstrates that tangible, visualization-based approaches can make machine learning concepts accessible and engaging for young learners while maintaining technical depth. The platform is freely available at https://learning-and-dynamics.github.io/ml-with-bricks/, with video tutorials guiding students through the experiments at https://youtube.com/playlist?list=PLx1grFu4zAcwfKKJZ1Ux4LwRqaePCOA2J.

Despite significant progress in alignment, large language models (LLMs) remain vulnerable to adversarial attacks that elicit harmful behaviors. Activation steering techniques offer a promising inference-time intervention approach, but existing methods suffer from critical limitations: activation addition requires careful coefficient tuning and is sensitive to layer-specific norm variations, while directional ablation provides only binary control. Recent work on Angular Steering introduces continuous control via rotation in a 2D subspace, but its practical implementation violates norm preservation, causing distribution shift and generation collapse, particularly in models below 7B parameters. We propose Selective Steering, which addresses these limitations through two key innovations: (1) a mathematically rigorous norm-preserving rotation formulation that maintains activation distribution integrity, and (2) discriminative layer selection that applies steering only where feature representations exhibit opposite-signed class alignment. Experiments across nine models demonstrate that Selective Steering achieves 5.5x higher attack success rates than prior methods while maintaining zero perplexity violations and approximately 100\% capability retention on standard benchmarks. Our approach provides a principled, efficient framework for controllable and stable LLM behavior modification. Code: https://github.com/knoveleng/steering

Fully Homomorphic Encryption (FHE) enables computations directly on encrypted data, but its high computational cost remains a significant barrier. Writing efficient FHE code is a complex task requiring cryptographic expertise, and finding the optimal sequence of program transformations is often intractable. In this paper, we propose CHEHAB RL, a novel framework that leverages deep reinforcement learning (RL) to automate FHE code optimization. Instead of relying on predefined heuristics or combinatorial search, our method trains an RL agent to learn an effective policy for applying a sequence of rewriting rules to automatically vectorize scalar FHE code while reducing instruction latency and noise growth. The proposed approach supports the optimization of both structured and unstructured code. To train the agent, we synthesize a diverse dataset of computations using a large language model (LLM). We integrate our proposed approach into the CHEHAB FHE compiler and evaluate it on a suite of benchmarks, comparing its performance against Coyote, a state-of-the-art vectorizing FHE compiler. The results show that our approach generates code that is $5.3\times$ faster in execution, accumulates $2.54\times$ less noise, while the compilation process itself is $27.9\times$ faster than Coyote (geometric means).

Modern data parallel (DP) training favors collective communication over parameter servers (PS) for its simplicity and efficiency under balanced workloads. However, the balanced workload assumption no longer holds in large language model (LLM) post-training due to the high variance in sequence lengths. Under imbalanced workloads, collective communication creates synchronization barriers, leading to under-utilization of devices with smaller workloads. This change in training dynamics calls for a revisit of the PS paradigm for its robustness to such imbalance. We propose \textbf{On-Demand Communication (ODC)}, which adapts PS into Fully Sharded Data Parallel (FSDP) by replacing collective all-gather and reduce-scatter with direct point-to-point communication. Compared to FSDP, ODC reduces the synchronization barrier from once per layer to once per minibatch and decouples the workload on each device so that faster workers are not stalled. It also enables simpler and more effective load balancing at the minibatch level. Across diverse LLM post-training tasks, ODC consistently improves device utilization and training throughput, achieving up to a 36\% speedup over standard FSDP. These results demonstrate that ODC is a superior fit for the prevalent imbalanced workloads in LLM post-training. Our implementation of ODC and integration with FSDP is open-sourced at https://github.com/sail-sg/odc.

We present a comprehensive approach for multiword expression (MWE) identification that combines binary token-level classification, linguistic feature integration, and data augmentation. Our DeBERTa-v3-large model achieves 69.8% F1 on the CoAM dataset, surpassing the best results (Qwen-72B, 57.8% F1) on this dataset by 12 points while using 165x fewer parameters. We achieve this performance by (1) reformulating detection as binary token-level START/END/INSIDE classification rather than span-based prediction, (2) incorporating NP chunking and dependency features that help discontinuous and NOUN-type MWEs identification, and (3) applying oversampling that addresses severe class imbalance in the training data. We confirm the generalization of our method on the STREUSLE dataset, achieving 78.9% F1. These results demonstrate that carefully designed smaller models can substantially outperform LLMs on structured NLP tasks, with important implications for resource-constrained deployments.

Imbalanced node classification is a critical challenge in graph learning, where most existing methods typically utilize Graph Neural Networks (GNNs) to learn node representations. These methods can be broadly categorized into the data-level and the algorithm-level. The former aims to synthesize minority-class nodes to mitigate quantity imbalance, while the latter tries to optimize the learning process to highlight minority classes. However, neither of them addresses the inherently imbalanced graph structure, which is a fundamental factor that incurs majority-class dominance and minority-class assimilation in GNNs. Our theoretical analysis further supports this critical insight. Therefore, we propose GraphSB (Graph Structural Balance), a novel framework that incorporates Structural Balance as a key strategy to address the underlying imbalanced graph structure before node synthesis. Structural Balance performs a two-stage structure optimization: Structure Enhancement that mines hard samples near decision boundaries through dual-view analysis and enhances connectivity for minority classes through adaptive augmentation, and Relation Diffusion that propagates the enhanced minority context while simultaneously capturing higher-order structural dependencies. Thus, GraphSB balances structural distribution before node synthesis, enabling more effective learning in GNNs. Extensive experiments demonstrate that GraphSB significantly outperforms the state-of-the-art methods. More importantly, the proposed Structural Balance can be seamlessly integrated into state-of-the-art methods as a simple plug-and-play module, increasing their accuracy by an average of 4.57%.

Traditional metrics like BLEU and BERTScore fail to capture semantic fidelity in generative text-to-text tasks. We adapt the Cross-Examination Framework (CEF) for a reference-free, multi-dimensional evaluation by treating the source and candidate as independent knowledge bases. CEF generates verifiable questions from each text and performs a cross-examination to derive three interpretable scores: Coverage, Conformity, and Consistency. Validated across translation, summarization and clinical note-generation, our framework identifies critical errors, such as content omissions and factual contradictions, missed by standard metrics. A key contribution is a systematic robustness analysis to select a stable judge model. Crucially, the strong correlation between our reference-free and with-reference modes validates CEF's reliability without gold references. Furthermore, human expert validation demonstrates that CEF mismatching questions align with meaning-altering semantic errors higher than with non-semantic errors, particularly excelling at identifying entity-based and relational distortions.

Estimating uncertainty in deep learning models is critical for reliable decision-making in high-stakes applications such as medical imaging. Prior research has established that the difference between an input sample and its reconstructed version produced by an auxiliary model can serve as a useful proxy for uncertainty. However, directly comparing reconstructions with the original input is degraded by information loss and sensitivity to superficial details, which limits its effectiveness. In this work, we propose Difference Reconstruction Uncertainty Estimation (DRUE), a method that mitigates this limitation by reconstructing inputs from two intermediate layers and measuring the discrepancy between their outputs as the uncertainty score. To evaluate uncertainty estimation in practice, we follow the widely used out-of-distribution (OOD) detection paradigm, where in-distribution (ID) training data are compared against datasets with increasing domain shift. Using glaucoma detection as the ID task, we demonstrate that DRUE consistently achieves superior AUC and AUPR across multiple OOD datasets, highlighting its robustness and reliability under distribution shift. This work provides a principled and effective framework for enhancing model reliability in uncertain environments.

Modern AI systems increasingly comprise multiple interconnected neural networks to tackle complex inference tasks. Testing such systems for robustness and safety entails significant challenges. Current state-of-the-art robustness testing techniques, whether black-box or white-box, have been proposed and implemented for single-network models and do not scale well to multi-network pipelines. We propose a modular robustness testing framework that applies a given set of perturbations to test data. Our testing framework supports (1) a component-wise system analysis to isolate errors and (2) reasoning about error propagation across the neural network modules. The testing framework is architecture and modality agnostic and can be applied across domains. We apply the framework to a real-world autonomous rail inspection system composed of multiple deep networks and successfully demonstrate how our approach enables fine-grained robustness analysis beyond conventional end-to-end metrics.

While model-based reinforcement learning (MBRL) improves sample efficiency by learning world models from raw observations, existing methods struggle to generalize across structurally similar scenes and remain vulnerable to spurious variations such as textures or color shifts. From a cognitive science perspective, humans segment continuous sensory streams into discrete events and rely on these key events for decision-making. Motivated by this principle, we propose the Event-Aware World Model (EAWM), a general framework that learns event-aware representations to streamline policy learning without requiring handcrafted labels. EAWM employs an automated event generator to derive events from raw observations and introduces a Generic Event Segmentor (GES) to identify event boundaries, which mark the start and end time of event segments. Through event prediction, the representation space is shaped to capture meaningful spatio-temporal transitions. Beyond this, we present a unified formulation of seemingly distinct world model architectures and show the broad applicability of our methods. Experiments on Atari 100K, Craftax 1M, and DeepMind Control 500K, DMC-GB2 500K demonstrate that EAWM consistently boosts the performance of strong MBRL baselines by 10%-45%, setting new state-of-the-art results across benchmarks. Our code is released at https://github.com/MarquisDarwin/EAWM.

Benchmark-based evaluation is the de facto standard for comparing large language models (LLMs). However, its reliability is increasingly threatened by test set contamination, where test samples or their close variants leak into training data and artificially inflate reported performance. To address this issue, prior work has explored two main lines of mitigation. One line attempts to identify and remove contaminated benchmark items before evaluation, but this inevitably alters the evaluation set itself and becomes unreliable when contamination is moderate or severe. The other line preserves the benchmark and instead suppresses contaminated behavior at evaluation time; however, such interventions often interfere with normal inference and lead to noticeable performance degradation on clean inputs. We propose DeconIEP, a decontamination framework that operates entirely during evaluation by applying small, bounded perturbations in the input embedding space. Guided by a relatively less-contaminated reference model, DeconIEP learns an instance-adaptive perturbation generator that steers the evaluated model away from memorization-driven shortcut pathways. Across multiple open-weight LLMs and benchmarks, extensive empirical results show that DeconIEP achieves strong decontamination effectiveness while incurring only minimal degradation in benign utility.

Clustering is a fundamental primitive in unsupervised learning. However, classical algorithms for $k$-clustering (such as $k$-median and $k$-means) assume access to exact pairwise distances -- an unrealistic requirement in many modern applications. We study clustering in the \emph{Rank-model (R-model)}, where access to distances is entirely replaced by a \emph{quadruplet oracle} that provides only relative distance comparisons. In practice, such an oracle can represent learned models or human feedback, and is expected to be noisy and entail an access cost. Given a metric space with $n$ input items, we design randomized algorithms that, using only a noisy quadruplet oracle, compute a set of $O(k \cdot \mathsf{polylog}(n))$ centers along with a mapping from the input items to the centers such that the clustering cost of the mapping is at most constant times the optimum $k$-clustering cost. Our method achieves a query complexity of $O(n\cdot k \cdot \mathsf{polylog}(n))$ for arbitrary metric spaces and improves to $O((n+k^2) \cdot \mathsf{polylog}(n))$ when the underlying metric has bounded doubling dimension. When the metric has bounded doubling dimension we can further improve the approximation from constant to $1+\varepsilon$, for any arbitrarily small constant $\varepsilon\in(0,1)$, while preserving the same asymptotic query complexity. Our framework demonstrates how noisy, low-cost oracles, such as those derived from large language models, can be systematically integrated into scalable clustering algorithms.

We present Innovator-VL, a scientific multimodal large language model designed to advance understanding and reasoning across diverse scientific domains while maintaining excellent performance on general vision tasks. Contrary to the trend of relying on massive domain-specific pretraining and opaque pipelines, our work demonstrates that principled training design and transparent methodology can yield strong scientific intelligence with substantially reduced data requirements. (i) First, we provide a fully transparent, end-to-end reproducible training pipeline, covering data collection, cleaning, preprocessing, supervised fine-tuning, reinforcement learning, and evaluation, along with detailed optimization recipes. This facilitates systematic extension by the community. (ii) Second, Innovator-VL exhibits remarkable data efficiency, achieving competitive performance on various scientific tasks using fewer than five million curated samples without large-scale pretraining. These results highlight that effective reasoning can be achieved through principled data selection rather than indiscriminate scaling. (iii) Third, Innovator-VL demonstrates strong generalization, achieving competitive performance on general vision, multimodal reasoning, and scientific benchmarks. This indicates that scientific alignment can be integrated into a unified model without compromising general-purpose capabilities. Our practices suggest that efficient, reproducible, and high-performing scientific multimodal models can be built even without large-scale data, providing a practical foundation for future research.

Quantization-aware training (QAT) is essential for deploying large models under strict memory and latency constraints, yet achieving stable and robust optimization at ultra-low bitwidths remains challenging. Common approaches based on the straight-through estimator (STE) or soft quantizers often suffer from gradient mismatch, instability, or high computational overhead. As such, we propose StableQAT, a unified and efficient QAT framework that stabilizes training in ultra low-bit settings via a novel, lightweight, and theoretically grounded surrogate for backpropagation derived from a discrete Fourier analysis of the rounding operator. StableQAT strictly generalizes STE as the latter arises as a special case of our more expressive surrogate family, yielding smooth, bounded, and inexpensive gradients that improve QAT training performance and stability across various hyperparameter choices. In experiments, StableQAT exhibits stable and efficient QAT at 2-4 bit regimes, demonstrating improved training stability, robustness, and superior performance with negligible training overhead against standard QAT techniques. Our code is available at https://github.com/microsoft/StableQAT.

Empirical software engineering research often depends on datasets of code repository artifacts, where sampling strategies are employed to enable large-scale analyses. The design and evaluation of these strategies are critical, as they directly influence the generalizability of research findings. However, sampling remains an underestimated aspect in software engineering research: we identify two main challenges related to (1) the design and representativeness of sampling approaches, and (2) the ability to reason about the implications of sampling decisions on generalizability. To address these challenges, we propose a Domain-Specific Language (DSL) to explicitly describe complex sampling strategies through composable sampling operators. This formalism supports both the specification and the reasoning about the generalizability of results based on the applied sampling strategies. We implement the DSL as a Python-based fluent API, and demonstrate how it facilitates representativeness reasoning using statistical indicators extracted from sampling workflows. We validate our approach through a case study of MSR papers involving code repository sampling. Our results show that the DSL can model the sampling strategies reported in recent literature.

Machine learning models have demonstrated strong performance in classifying network traffic and identifying Internet-of-Things (IoT) devices, enabling operators to discover and manage IoT assets at scale. However, many existing approaches rely on end-to-end supervised pipelines or task-specific fine-tuning, resulting in traffic representations that are tightly coupled to labeled datasets and deployment environments, which can limit generalizability. In this paper, we study the problem of learning generalizable traffic representations for IoT device identification. We design compact encoder architectures that learn per-flow embeddings from unlabeled IoT traffic and evaluate them using a frozen-encoder protocol with a simple supervised classifier. Our specific contributions are threefold. (1) We develop unsupervised encoder--decoder models that learn compact traffic representations from unlabeled IoT network flows and assess their quality through reconstruction-based analysis. (2) We show that these learned representations can be used effectively for IoT device-type classification using simple, lightweight classifiers trained on frozen embeddings. (3) We provide a systematic benchmarking study against the state-of-the-art pretrained traffic encoders, showing that larger models do not necessarily yield more robust representations for IoT traffic. Using more than 18 million real IoT traffic flows collected across multiple years and deployment environments, we learn traffic representations from unlabeled data and evaluate device-type classification on disjoint labeled subsets, achieving macro F1-scores exceeding 0.9 for device-type classification and demonstrating robustness under cross-environment deployment.

Accurate depth estimation is fundamental to 3D perception in autonomous driving, supporting tasks such as detection, tracking, and motion planning. However, monocular camera-based 3D detection suffers from depth ambiguity and reduced robustness under challenging conditions. Radar provides complementary advantages such as resilience to poor lighting and adverse weather, but its sparsity and low resolution limit its direct use in detection frameworks. This motivates the need for effective Radar-camera fusion with improved preprocessing and depth estimation strategies. We propose an end-to-end framework that enhances monocular 3D object detection through two key components. First, we introduce InstaRadar, an instance segmentation-guided expansion method that leverages pre-trained segmentation masks to enhance Radar density and semantic alignment, producing a more structured representation. InstaRadar achieves state-of-the-art results in Radar-guided depth estimation, showing its effectiveness in generating high-quality depth features. Second, we integrate the pre-trained RCDPT into the BEVDepth framework as a replacement for its depth module. With InstaRadar-enhanced inputs, the RCDPT integration consistently improves 3D detection performance. Overall, these components yield steady gains over the baseline BEVDepth model, demonstrating the effectiveness of InstaRadar and the advantage of explicit depth supervision in 3D object detection. Although the framework lags behind Radar-camera fusion models that directly extract BEV features, since Radar serves only as guidance rather than an independent feature stream, this limitation highlights potential for improvement. Future work will extend InstaRadar to point cloud-like representations and integrate a dedicated Radar branch with temporal cues for enhanced BEV fusion.

The Schrodinger Bridge and Bass (SBB) formulation, which jointly controls drift and volatility, is an established extension of the classical Schrodinger Bridge (SB). Building on this framework, we introduce LightSBB-M, an algorithm that computes the optimal SBB transport plan in only a few iterations. The method exploits a dual representation of the SBB objective to obtain analytic expressions for the optimal drift and volatility, and it incorporates a tunable parameter beta greater than zero that interpolates between pure drift (the Schrodinger Bridge) and pure volatility (Bass martingale transport). We show that LightSBB-M achieves the lowest 2-Wasserstein distance on synthetic datasets against state-of-the-art SB and diffusion baselines with up to 32 percent improvement. We also illustrate the generative capability of the framework on an unpaired image-to-image translation task (adult to child faces in FFHQ). These findings demonstrate that LightSBB-M provides a scalable, high-fidelity SBB solver that outperforms existing SB and diffusion baselines across both synthetic and real-world generative tasks. The code is available at https://github.com/alexouadi/LightSBB-M.

As large language models become integral to agentic artificial intelligence systems, their energy demands during inference may pose significant sustainability challenges. This study investigates whether deploying smaller-scale language models can reduce energy consumption without compromising responsiveness and output quality in a multi-agent, real-world environments. We conduct a comparative analysis across language models of varying scales to quantify trade-offs between efficiency and performance. Results show that smaller open-weights models can lower energy usage while preserving task quality. Building on these findings, we propose practical guidelines for sustainable artificial intelligence design, including optimal batch size configuration and computation resource allocation. These insights offer actionable strategies for developing scalable, environmentally responsible artificial intelligence systems.

Mobile agents have made progress toward reliable smartphone automation, yet performance in complex applications remains limited by incomplete knowledge and weak generalization to unseen environments. We introduce a curiosity driven knowledge retrieval framework that formalizes uncertainty during execution as a curiosity score. When this score exceeds a threshold, the system retrieves external information from documentation, code repositories, and historical trajectories. Retrieved content is organized into structured AppCards, which encode functional semantics, parameter conventions, interface mappings, and interaction patterns. During execution, an enhanced agent selectively integrates relevant AppCards into its reasoning process, thereby compensating for knowledge blind spots and improving planning reliability. Evaluation on the AndroidWorld benchmark shows consistent improvements across backbones, with an average gain of six percentage points and a new state of the art success rate of 88.8\% when combined with GPT-5. Analysis indicates that AppCards are particularly effective for multi step and cross application tasks, while improvements depend on the backbone model. Case studies further confirm that AppCards reduce ambiguity, shorten exploration, and support stable execution trajectories. Task trajectories are publicly available at https://lisalsj.github.io/Droidrun-appcard/.

Prompting techniques such as Chain-of-Thought (CoT) and Program-of-Thought (PoT) improve LLM mathematical reasoning by structuring intermediate steps in natural language or code. However, applied mathematics problems in domains like finance, physics, and cryptography often require recalling or deriving governing equations, a step that current approaches do not explicitly leverage. We propose Formula-One Prompting (F-1), a two-phase approach that uses mathematical equations as an intermediate representation before adaptive solving. F-1 first formulates governing equations from problem descriptions, then selects a solving strategy among CoT, PoT, or direct computation based on the generated equations, all within a single LLM call. Results across five models and four benchmarks show F-1 outperforms CoT by +5.76% and PoT by +8.42% on average. Crucially, gains are largest in applied domains: +13.30% on FinanceMath over CoT, and within OlympiadBench, larger gains on physics (+2.55%) than pure math (+0.44%). This demonstrates that F-1 is more effective than CoT in applied mathematics problems.

We introduce contextual queueing bandits, a new context-aware framework for scheduling while simultaneously learning unknown service rates. Individual jobs carry heterogeneous contextual features, based on which the agent chooses a job and matches it with a server to maximize the departure rate. The service/departure rate is governed by a logistic model of the contextual feature with an unknown server-specific parameter. To evaluate the performance of a policy, we consider queue length regret, defined as the difference in queue length between the policy and the optimal policy. The main challenge in the analysis is that the lists of remaining job features in the queue may differ under our policy versus the optimal policy for a given time step, since they may process jobs in different orders. To address this, we propose the idea of policy-switching queues equipped with a sophisticated coupling argument. This leads to a novel queue length regret decomposition framework, allowing us to understand the short-term effect of choosing a suboptimal job-server pair and its long-term effect on queue state differences. We show that our algorithm, CQB-$\varepsilon$, achieves a regret upper bound of $\widetilde{\mathcal{O}}(T^{-1/4})$. We also consider the setting of adversarially chosen contexts, for which our second algorithm, CQB-Opt, achieves a regret upper bound of $\mathcal{O}(\log^2 T)$. Lastly, we provide experimental results that validate our theoretical findings.

Accurately predicting procurement lead time (PLT) remains a challenge in engineered-to-order industries such as shipbuilding and plant construction, where delays in a single key component can disrupt project timelines. In shipyards, pipe spools are critical components; installed deep within hull blocks soon after steel erection, any delay in their procurement can halt all downstream tasks. Recognizing their importance, existing studies predict PLT using the static physical attributes of pipe spools. However, procurement is inherently a dynamic, multi-stakeholder business process involving a continuous sequence of internal and external events at the shipyard, factors often overlooked in traditional approaches. To address this issue, this paper proposes a novel framework that combines event logs, dataset records of the procurement events, with static attributes to predict PLT. The temporal attributes of each event are extracted to reflect the continuity and temporal context of the process. Subsequently, a deep sequential neural network combined with a multi-layered perceptron is employed to integrate these static and dynamic features, enabling the model to capture both structural and contextual information in procurement. Comparative experiments are conducted using real-world pipe spool procurement data from a globally renowned South Korean shipbuilding corporation. Three tasks are evaluated, which are production, post-processing, and procurement lead time prediction. The results show a 22.6% to 50.4% improvement in prediction performance in terms of mean absolute error over the best-performing existing approaches across the three tasks. These findings indicate the value of considering procurement process information for more accurate PLT prediction.

Large language models are increasingly deployed as multi-agent systems, where specialized roles communicate and collaborate through structured interactions to solve complex tasks that often exceed the capacity of a single agent. However, most existing systems still rely on a fixed role library and an execution-frozen interaction topology, a rigid design choice that frequently leads to task mismatch, prevents timely adaptation when new evidence emerges during reasoning, and further inflates inference cost. We introduce MetaGen, a training-free framework that adapts both the role space and the collaboration topology at inference time, without updating base model weights. MetaGen generates and rewrites query-conditioned role specifications to maintain a controllable dynamic role pool, then instantiates a constrained execution graph around a minimal backbone. During execution, it iteratively updates role prompts and adjusts structural decisions using lightweight feedback signals. Experiments on code generation and multi-step reasoning benchmarks show that MetaGen improves the accuracy and cost tradeoff over strong multi-agent baselines.

While prior work has examined the generation capabilities of Agentic AI systems, little is known about how reviewers respond to AI-authored code in practice. In this paper, we present a large-scale empirical study of code review dynamics in agent-generated PRs. Using a curated subset of the AIDev dataset, we analyze 19,450 inline review comments spanning 3,177 agent-authored PRs from real-world GitHub repositories. We first derive a taxonomy of 12 review comment themes using topic modeling combined with large language model (LLM)-assisted semantic clustering and consolidation. According to this taxonomy, we then investigate whether zero-shot prompts to LLM can reliably annotate review comments. Our evaluation against human annotations shows that open-source LLM achieves reasonably high exact match (78.63%), macro F1 score (0.78), and substantial agreement with human annotators at the review comment level. At the PR level, the LLM also correctly identifies the dominant review theme with 78% Top-1 accuracy and achieves an average Jaccard similarity of 0.76, indicating strong alignment with human judgments. Applying this annotation pipeline at scale, we find that apart from functional correctness and logical changes, reviews of agent-authored PRs predominantly focus on documentation gaps, refactoring needs, styling and formatting issues, with testing and security-related concerns. These findings suggest that while AI agents can accelerate code production, there remain gaps requiring targeted human review oversight.

Electronic Health Records (EHRs) provide crucial information for clinical decision-making. However, their high-dimensionality, heterogeneity, and sparsity make clinical prediction challenging. Large Language Models (LLMs) allowed progress towards addressing this challenge by leveraging parametric medical knowledge to enhance EHR data for clinical prediction tasks. Despite the significant achievements made so far, most of the existing approaches are fundamentally task-agnostic in the sense that they deploy LLMs as EHR encoders or EHR completion modules without fully integrating signals from the prediction tasks. This naturally hinders task performance accuracy. In this work, we propose Rewrite-To-Predict (ReToP), an LLM-based framework that addresses this limitation through an end-to-end training of an EHR rewriter and a clinical predictor. To cope with the lack of EHR rewrite training data, we generate synthetic pseudo-labels using clinical-driven feature selection strategies to create diverse patient rewrites for fine-tuning the EHR rewriter. ReToP aligns the rewriter with prediction objectives using a novel Classifier Supervised Contribution (CSC) score that enables the EHR rewriter to generate clinically relevant rewrites that directly enhance prediction. Our ReToP framework surpasses strong baseline models across three clinical tasks on MIMIC-IV. Moreover, the analysis of ReToP shows its generalizability to unseen datasets and tasks with minimal fine-tuning while preserving faithful rewrites and emphasizing task-relevant predictive features.

Diffusion models achieve remarkable generation quality, yet face a fundamental challenge known as memorization, where generated samples can replicate training samples exactly. We develop a theoretical framework to explain this phenomenon by showing that the empirical score function (the score function corresponding to the empirical distribution) is a weighted sum of the score functions of Gaussian distributions, in which the weights are sharp softmax functions. This structure causes individual training samples to dominate the score function, resulting in sampling collapse. In practice, approximating the empirical score function with a neural network can partially alleviate this issue and improve generalization. Our theoretical framework explains why: In training, the neural network learns a smoother approximation of the weighted sum, allowing the sampling process to be influenced by local manifolds rather than single points. Leveraging this insight, we propose two novel methods to further enhance generalization: (1) Noise Unconditioning enables each training sample to adaptively determine its score function weight to increase the effect of more training samples, thereby preventing single-point dominance and mitigating collapse. (2) Temperature Smoothing introduces an explicit parameter to control the smoothness. By increasing the temperature in the softmax weights, we naturally reduce the dominance of any single training sample and mitigate memorization. Experiments across multiple datasets validate our theoretical analysis and demonstrate the effectiveness of the proposed methods in improving generalization while maintaining high generation quality.

Stabilizing a dynamical system is a fundamental problem that serves as a cornerstone for many complex tasks in the field of control systems. The problem becomes challenging when the system model is unknown. Among the Reinforcement Learning (RL) algorithms that have been successfully applied to solve problems pertaining to unknown linear dynamical systems, the policy gradient (PG) method stands out due to its ease of implementation and can solve the problem in a model-free manner. However, most of the existing works on PG methods for unknown linear dynamical systems assume full-state feedback. In this paper, we take a step towards model-free learning for partially observable linear dynamical systems with output feedback and focus on the fundamental stabilization problem of the system. We propose an algorithmic framework that stretches the boundary of PG methods to the problem without global convergence guarantees. We show that by leveraging zeroth-order PG update based on system trajectories and its convergence to stationary points, the proposed algorithms return a stabilizing output feedback policy for discrete-time linear dynamical systems. We also explicitly characterize the sample complexity of our algorithm and verify the effectiveness of the algorithm using numerical examples.

Recent progress in Large Language Model (LLM) reasoning is increasingly driven by the refinement of post-training loss functions and alignment strategies. However, standard Reinforcement Learning (RL) paradigms like Group Relative Policy Optimization (GRPO) remain constrained by static uniformity: uniform prompt sampling and a fixed number of rollouts per prompt. For heterogeneous, heavy-tailed reasoning data, this creates structural inefficiencies that waste compute on already-solved patterns while under-training the long tail of hard problems. To address this, we propose Multi-Adversary Group Distributionally Robust Optimization (GDRO), an optimization-first framework that moves beyond uniform reasoning models by dynamically adapting the training distribution. We introduce an Online Difficulty Classifier that partitions prompts into dynamic pass@k difficulty groups. We then propose two independent GDRO games for post-training: (1) Prompt-GDRO, which employs an EMA-debiased multiplicative-weights bandit sampler to target the intensive difficulty margin and upweight persistently hard groups without frequency bias; and (2) Rollout-GDRO, which uses a shadow-price controller to reallocate rollouts across groups, maximizing gradient variance reduction on hard tasks under a fixed mean budget (compute-neutral). We provide no-regret guarantees for both controllers and additionally a variance-proxy analysis motivating a square-root optimal rollout allocation for Rollout-GDRO. We validate our framework on the DAPO 14.1k dataset using Qwen3-Base models. Prompt-GDRO and Rollout-GDRO achieve average relative gains of +10.6% and +10.1%, respectively, in pass@8 accuracy across 1.7B, 4B, and 8B scales compared to the GRPO baseline. Qualitative analysis shows an emergent curriculum: the adversaries shift resources to the evolving reasoning frontier, enhancing the reasoning model's performance.

Speculative decoding is an effective and lossless approach for accelerating LLM inference. However, existing widely adopted model-based draft designs, such as EAGLE3, improve accuracy at the cost of multi-step autoregressive inference, resulting in high drafting latency and ultimately rendering the drafting stage itself a performance bottleneck. Inspired by diffusion-based large language models (dLLMs), we propose DART, which leverages parallel generation to reduce drafting latency. DART predicts logits for multiple future masked positions in parallel within a single forward pass based on hidden states of the target model, thereby eliminating autoregressive rollouts in the draft model while preserving a lightweight design. Based on these parallel logit predictions, we further introduce an efficient tree pruning algorithm that constructs high-quality draft token trees with N-gram-enforced semantic continuity. DART substantially reduces draft-stage overhead while preserving high draft accuracy, leading to significantly improved end-to-end decoding speed. Experimental results demonstrate that DART achieves a 2.03x--3.44x wall-clock time speedup across multiple datasets, surpassing EAGLE3 by 30% on average and offering a practical speculative decoding framework. Code is released at https://github.com/fvliang/DART.

Recommender systems (RS) aim to retrieve a small set of items that best match individual user preferences. Naturally, RS place primary emphasis on the quality of the Top-$K$ results rather than performance across the entire item set. However, estimating Top-$K$ accuracy (e.g., Precision@$K$, Recall@$K$) requires determining the ranking positions of items, which imposes substantial computational overhead and poses significant challenges for optimization. In addition, RS often suffer from distribution shifts due to evolving user preferences or data biases, further complicating the task. To address these issues, we propose Talos, a loss function that is specifically designed to optimize the Talos recommendation accuracy. Talos leverages a quantile technique that replaces the complex ranking-dependent operations into simpler comparisons between predicted scores and learned score thresholds. We further develop a sampling-based regression algorithm for efficient and accurate threshold estimation, and introduce a constraint term to maintain optimization stability by preventing score inflation. Additionally, we incorporate a tailored surrogate function to address discontinuity and enhance robustness against distribution shifts. Comprehensive theoretical analyzes and empirical experiments are conducted to demonstrate the effectiveness, efficiency, convergence, and distributional robustness of Talos. The code is available at https://github.com/cynthia-shengjia/WWW-2026-Talos.

Tactile memory, the ability to store and retrieve touch-based experience, is critical for contact-rich tasks such as key insertion under uncertainty. To replicate this capability, we introduce Tactile Memory with Soft Robot (TaMeSo-bot), a system that integrates a soft wrist with tactile retrieval-based control to enable safe and robust manipulation. The soft wrist allows safe contact exploration during data collection, while tactile memory reuses past demonstrations via retrieval for flexible adaptation to unseen scenarios. The core of this system is the Masked Tactile Trajectory Transformer (MAT$^\text{3}$), which jointly models spatiotemporal interactions between robot actions, distributed tactile feedback, force-torque measurements, and proprioceptive signals. Through masked-token prediction, MAT$^\text{3}$ learns rich spatiotemporal representations by inferring missing sensory information from context, autonomously extracting task-relevant features without explicit subtask segmentation. We validate our approach on peg-in-hole tasks with diverse pegs and conditions in real-robot experiments. Our extensive evaluation demonstrates that MAT$^\text{3}$ achieves higher success rates than the baselines over all conditions and shows remarkable capability to adapt to unseen pegs and conditions.

Riddles are concise linguistic puzzles that describe an object or idea through indirect, figurative, or playful clues. They are a longstanding form of creative expression, requiring the solver to interpret hints, recognize patterns, and draw inferences to identify the answers. In this work, we introduce a simple pipeline for creating and evaluating analogy-based riddles. The system includes a triples creator that builds structured facts about a concept, a semantic mapper that selects attributes useful for analogy, a stylized generator that turns them into riddle clues, and a validator that collects all possible answers the riddle could point to. We use this validator to study whether large language models can recover the full answer set for different riddle types. Our case study shows that while models often guess the main intended answer, they frequently miss other valid interpretations. This highlights the value of riddles as a lightweight tool for examining reasoning coverage and ambiguity handling in language models.

Accurate dialogue description in audiovisual video captioning is crucial for downstream understanding and generation tasks. However, existing models generally struggle to produce faithful dialogue descriptions within audiovisual captions. To mitigate this limitation, we propose DiaDem, a powerful audiovisual video captioning model capable of generating captions with more precise dialogue descriptions while maintaining strong overall performance. We first synthesize a high-quality dataset for SFT, then employ a difficulty-partitioned two-stage GRPO strategy to further enhance dialogue descriptions. To enable systematic evaluation of dialogue description capabilities, we introduce DiaDemBench, a comprehensive benchmark designed to evaluate models across diverse dialogue scenarios, emphasizing both speaker attribution accuracy and utterance transcription fidelity in audiovisual captions. Extensive experiments on DiaDemBench reveal even commercial models still exhibit substantial room for improvement in dialogue-aware captioning. Notably, DiaDem not only outperforms the Gemini series in dialogue description accuracy but also achieves competitive performance on general audiovisual captioning benchmarks, demonstrating its overall effectiveness.

Large language models (LLMs) have made it remarkably easy to synthesize plausible source code from natural language prompts. While this accelerates software development and supports learning, it also raises new risks for academic integrity, authorship attribution, and responsible AI use. This paper investigates the problem of distinguishing human-written from machine-generated code by comparing two complementary approaches: feature-based detectors built from lightweight, interpretable stylometric and structural properties of code, and embedding-based detectors leveraging pretrained code encoders. Using a recent large-scale benchmark dataset of 600k human-written and AI-generated code samples, we find that feature-based models achieve strong performance (ROC-AUC 0.995, PR-AUC 0.995, F1 0.971), while embedding-based models with CodeBERT embeddings are also very competitive (ROC-AUC 0.994, PR-AUC 0.994, F1 0.965). Analysis shows that features tied to indentation and whitespace provide particularly discriminative cues, whereas embeddings capture deeper semantic patterns and yield slightly higher precision. These findings underscore the trade-offs between interpretability and generalization, offering practical guidance for deploying robust code-origin detection in academic and industrial contexts.

Split learning is a distributed training paradigm where a neural network is partitioned between clients and a server, which allows data to remain at the client while only intermediate activations are shared. Traditional split learning relies on end-to-end backpropagation across the client-server split point. This incurs a large communication overhead (i.e., forward activations and backward gradients need to be exchanged every iteration) and significant memory use (for storing activations and gradients). In this paper, we develop a beyond-backpropagation training method for split learning. In this approach, the client and server train their model partitions semi-independently, using local loss signals instead of propagated gradients. In particular, the client's network is augmented with a small auxiliary classifier at the split point to provide a local error signal, while the server trains on the client's transmitted activations using the true loss function. This decoupling removes the need to send backward gradients, which cuts communication costs roughly in half and also reduces memory overhead (as each side only stores local activations for its own backward pass). We evaluate our approach on CIFAR-10 and CIFAR-100. Our experiments show two key results. First, the proposed approach achieves performance on par with standard split learning that uses backpropagation. Second, it significantly reduces communication (of transmitting activations/gradient) by 50% and peak memory usage by up to 58%.

The discussion around AI-Engineering, that is, Software Engineering (SE) for AI-enabled Systems, cannot ignore a crucial class of software systems that are increasingly becoming AI-enhanced: Those used to enable or support the SE process, such as Computer-Aided SE (CASE) tools and Integrated Development Environments (IDEs). In this paper, we study the energy efficiency of these systems. As AI becomes seamlessly available in these tools and, in many cases, is active by default, we are entering a new era with significant implications for energy consumption patterns throughout the Software Development Lifecycle (SDLC). We focus on advanced Machine Learning (ML) capabilities provided by Large Language Models (LLMs). Our proposed approach combines Retrieval-Augmented Generation (RAG) with Prompt Engineering Techniques (PETs) to enhance both the quality and energy efficiency of LLM-based code generation. We present a comprehensive framework that measures real-time energy consumption and inference time across diverse model architectures ranging from 125M to 7B parameters, including GPT-2, CodeLlama, Qwen 2.5, and DeepSeek Coder. These LLMs, chosen for practical reasons, are sufficient to validate the core ideas and provide a proof of concept for more in-depth future analysis.

Modern mobile applications rely on hidden interactions--gestures without visual cues like long presses and swipes--to provide functionality without cluttering interfaces. While experienced users may discover these interactions through prior use or onboarding tutorials, their implicit nature makes them difficult for most users to uncover. Similarly, mobile agents--systems designed to automate tasks on mobile user interfaces, powered by vision language models (VLMs)--struggle to detect veiled interactions or determine actions for completing tasks. To address this challenge, we present GhostUI, a new dataset designed to enable the detection of hidden interactions in mobile applications. GhostUI provides before-and-after screenshots, simplified view hierarchies, gesture metadata, and task descriptions, allowing VLMs to better recognize concealed gestures and anticipate post-interaction states. Quantitative evaluations with VLMs show that models fine-tuned on GhostUI outperform baseline VLMs, particularly in predicting hidden interactions and inferring post-interaction screens, underscoring GhostUI's potential as a foundation for advancing mobile task automation.

Molecular representation learning aims to learn vector embeddings that capture molecular structure and geometry, thereby enabling property prediction and downstream scientific applications. In many AI for science tasks, labeled data are expensive to obtain and therefore limited in availability. Under the few-shot setting, models trained with scarce supervision often learn brittle structure-property relationships, resulting in substantially higher prediction errors and reduced generalization to unseen molecules. To address this limitation, we propose PCEvo, a path-consistent representation method that learns from virtual paths through dynamic structural evolution. PCEvo enumerates multiple chemically feasible edit paths between retrieved similar molecular pairs under topological dependency constraints. It transforms the labels of the two molecules into stepwise supervision along each virtual evolutionary path. It introduces a path-consistency objective that enforces prediction invariance across alternative paths connecting the same two molecules. Comprehensive experiments on the QM9 and MoleculeNet datasets demonstrate that PCEvo substantially improves the few-shot generalization performance of baseline methods. The code is available at https://anonymous.4open.science/r/PCEvo-4BF2.

Covariate-dependent uncertainty quantification in simulation-based inference is crucial for high-stakes decision-making but remains challenging due to the limitations of existing methods such as conformal prediction and classical bootstrap, which struggle with covariate-specific conditioning. We propose Efficient Quantile-Regression-Based Generative Metamodeling (E-QRGMM), a novel framework that accelerates the quantile-regression-based generative metamodeling (QRGMM) approach by integrating cubic Hermite interpolation with gradient estimation. Theoretically, we show that E-QRGMM preserves the convergence rate of the original QRGMM while reducing grid complexity from $O(n^{1/2})$ to $O(n^{1/5})$ for the majority of quantile levels, thereby substantially improving computational efficiency. Empirically, E-QRGMM achieves a superior trade-off between distributional accuracy and training speed compared to both QRGMM and other advanced deep generative models on synthetic and practical datasets. Moreover, by enabling bootstrap-based construction of confidence intervals for arbitrary estimands of interest, E-QRGMM provides a practical solution for covariate-dependent uncertainty quantification.

Time series anomaly detection is critical for supply chain management to take proactive operations, but faces challenges: classical unsupervised anomaly detection based on exploiting data patterns often yields results misaligned with business requirements and domain knowledge, while manual expert analysis cannot scale to millions of products in the supply chain. We propose a framework that leverages large language models (LLMs) to systematically encode human expertise into interpretable, logic-based rules for detecting anomaly patterns in supply chain time series data. Our approach operates in three stages: 1) LLM-based labeling of training data instructed by domain knowledge, 2) automated generation and iterative improvements of symbolic rules through LLM-driven optimization, and 3) rule augmentation with business-relevant anomaly categories supported by LLMs to enhance interpretability. The experiment results showcase that our approach outperforms the unsupervised learning methods in both detection accuracy and interpretability. Furthermore, compared to direct LLM deployment for time series anomaly detection, our approach provides consistent, deterministic results with low computational latency and cost, making it ideal for production deployment. The proposed framework thus demonstrates how LLMs can bridge the gap between scalable automation and expert-driven decision-making in operational settings.

Most existing memory-enhanced Large Language Model (LLM) approaches implicitly assume that memory validity can be established either through external evaluators that provide task-specific success signals or through internal model cognition, such as reflection, for editing memory entries. However, these assumptions often break down in practical environments with dynamic drifts. We propose the Global Verifier (GLOVE), a framework that introduces a new design dimension for LLM memory systems by establishing a relative notion of truth. Through active probing to detect inconsistencies between retrieved memories and fresh observations, GLOVE enables memory-environment realignment by verifying and updating memory without access to ground-truth supervision or strong reliance on model introspection. We evaluate GLOVE on diverse benchmarks spanning web navigation, planning, and control, augmented with controlled environmental drifts that introduce non-stationarity beyond the original benchmark settings. Our results show that GLOVE substantially improves agent success rates, suggesting a robust pathway to cognitive agents capable of self-evolving.

Hallucination detection is critical for deploying large language models (LLMs) in real-world applications. Existing hallucination detection methods achieve strong performance when the training and test data come from the same domain, but they suffer from poor cross-domain generalization. In this paper, we study an important yet overlooked problem, termed generalizable hallucination detection (GHD), which aims to train hallucination detectors on data from a single domain while ensuring robust performance across diverse related domains. In studying GHD, we simulate multi-turn dialogues following LLMs initial response and observe an interesting phenomenon: hallucination-initiated multi-turn dialogues universally exhibit larger uncertainty fluctuations than factual ones across different domains. Based on the phenomenon, we propose a new score SpikeScore, which quantifies abrupt fluctuations in multi-turn dialogues. Through both theoretical analysis and empirical validation, we demonstrate that SpikeScore achieves strong cross-domain separability between hallucinated and non-hallucinated responses. Experiments across multiple LLMs and benchmarks demonstrate that the SpikeScore-based detection method outperforms representative baselines in cross-domain generalization and surpasses advanced generalization-oriented methods, verifying the effectiveness of our method in cross-domain hallucination detection.

Deep neural networks (DNNs) have recently been applied to inverse scattering problems (ISPs) due to their strong nonlinear mapping capabilities. However, supervised DNN solvers require large-scale datasets, which limits their generalization in practical applications. Untrained neural networks (UNNs) address this issue by updating weights from measured electric fields and prior physical knowledge, but existing UNN solvers suffer from long inference time. To overcome these limitations, this paper proposes a contrast-source-based physics-driven neural network (CSPDNN), which predicts the induced current distribution to improve efficiency and incorporates an adaptive total variation loss for robust reconstruction under varying contrast and noise conditions. The improved imaging performance is validated through comprehensive numerical simulations and experimental data.

In this paper, we present the first comprehensive empirical study of specialized LLM-based detectors and compare them with traditional static analyzers at the project scale. Specifically, our study evaluates five latest and representative LLM-based methods and two traditional tools using: 1) an in-house benchmark of 222 known real-world vulnerabilities (C/C++ and Java) to assess detection capability, and 2) 24 active open-source projects, where we manually inspected 385 warnings to assess their practical usability and underlying root causes of failures. Our evaluation yields three key findings: First, while LLM-based detectors exhibit low recall on the in-house benchmark, they still uncover more unique vulnerabilities than traditional tools. Second, in open-source projects, both LLM-based and traditional tools generate substantial warnings but suffer from very high false discovery rates, hindering practical use. Our manual analysis further reveals shallow interprocedural reasoning and misidentified source/sink pairs as primary failure causes, with LLM-based tools exhibiting additional unique failures. Finally, LLM-based methods incurs substantial computational costs-hundreds of thousands to hundreds of millions of tokens and multi-hour to multi-day runtimes. Overall, our findings underscore critical limitations in the robustness, reliability, and scalability of current LLM-based detectors. We ultimately summarize a set of implications for future research toward more effective and practical project-scale vulnerability detection.

RNA inverse folding, designing sequences to form specific 3D structures, is critical for therapeutics, gene regulation, and synthetic biology. Current methods, focused on sequence recovery, struggle to address structural objectives like secondary structure consistency (SS), minimum free energy (MFE), and local distance difference test (LDDT), leading to suboptimal structural accuracy. To tackle this, we propose a reinforcement learning (RL) framework integrated with a latent diffusion model (LDM). Drawing inspiration from the success of diffusion models in RNA inverse folding, which adeptly model complex sequence-structure interactions, we develop an LDM incorporating pre-trained RNA-FM embeddings from a large-scale RNA model. These embeddings capture co-evolutionary patterns, markedly improving sequence recovery accuracy. However, existing approaches, including diffusion-based methods, cannot effectively handle non-differentiable structural objectives. By contrast, RL excels in this task by using policy-driven reward optimization to navigate complex, non-gradient-based objectives, offering a significant advantage over traditional methods. In summary, we propose the Step-wise Optimization of Latent Diffusion Model (SOLD), a novel RL framework that optimizes single-step noise without sampling the full diffusion trajectory, achieving efficient refinement of multiple structural objectives. Experimental results demonstrate SOLD surpasses its LDM baseline and state-of-the-art methods across all metrics, establishing a robust framework for RNA inverse folding with profound implications for biotechnological and therapeutic applications.

Large Language Models (LLMs) have recently demonstrated remarkable reasoning abilities, yet hallucinate on knowledge-intensive tasks. Retrieval-augmented generation (RAG) mitigates this issue by grounding answers in external sources, e.g., knowledge graphs (KGs). However, existing KG-based RAG approaches rely on semantics-unaware path sampling and are weakly aligned with KG reasoning objectives, which limits further accuracy gains. They also feed retrieved paths directly into the reasoner without organizing them into answer-centered reasoning paths, hindering small LLMs' ability to leverage the retrieved knowledge. Furthermore, prior works predominantly rely on large LLMs (e.g., ChatGPT/GPT-4) or assume backbones above 7B parameters, leaving sub-7B models underexplored. We address this gap with RPO-RAG, the first KG-based RAG framework specifically designed for small LLMs, to the best of our knowledge. RPO-RAG introduces three key innovations: (1) a query-path semantic sampling strategy that provides informative supervisory signals; (2) a relation-aware preference optimization that aligns training with intermediate KG reasoning signals (e.g., relation); and (3) an answer-centered prompt design that organizes entities and reasoning paths in an interpretable format. Extensive experiments on two benchmark Knowledge Graph Question Answering (KGQA) datasets, WebQSP and CWQ, demonstrate that RPO-RAG effectively bridges the performance gap between small and large language models. On WebQSP, it improves F1 by up to 8.8%, reflecting enhanced answer precision, while on CWQ it achieves new state-of-the-art results among models under 8B parameters in both Hit and F1. Overall, RPO-RAG substantially improves the reasoning capability of small LLMs, even under 3B parameters-highlighting their potential for resource-efficient and practical on-device KGQA applications.

Multimodal Large Language Models (MLLMs) show promise for general industrial quality inspection, but fall short in complex scenarios, such as Printed Circuit Board (PCB) inspection. PCB inspection poses unique challenges due to densely packed components, complex wiring structures, and subtle defect patterns that require specialized domain expertise. However, a high-quality, unified vision-language benchmark for quantitatively evaluating MLLMs across PCB inspection tasks remains absent, stemming not only from limited data availability but also from fragmented datasets and inconsistent standardization. To fill this gap, we propose UniPCB, the first unified vision-language benchmark for open-ended PCB quality inspection. UniPCB is built via a systematic pipeline that curates and standardizes data from disparate sources across three annotated scenarios. Furthermore, we introduce PCB-GPT, an MLLM trained on a new instruction dataset generated by this pipeline, utilizing a novel progressive curriculum that mimics the learning process of human experts. Evaluations on the UniPCB benchmark show that while existing MLLMs falter on domain-specific tasks, PCB-GPT establishes a new baseline. Notably, it more than doubles the performance on fine-grained defect localization compared to the strongest competitors, with significant advantages in localization and analysis. We will release the instruction data, benchmark, and model to facilitate future research.

Modern Recurrent Neural Networks (RNNs), such as RWKV, are distinguished by their powerful short-range modeling capabilities and efficient fixed-size states, which constitute a core advantage over standard Transformers. However, there is a significant lack of research into their internal state as an editable knowledge representation. To fill this gap, we first explore the representational properties of the RWKV state by proposing the DREAMSTATE framework. This framework utilizes a conditional Diffusion Transformer (DiT) to directly model the probability manifold of the state, enabling its generation and editing. The structural nature of this representation is validated through t-SNE visualizations and controlled generation experiments. After successfully uncovering and modeling the state's representational potential, we further propose a novel hybrid architecture that combines the local advantages of RNNs with global context adaptability. This architecture features a parallel DiT that processes a variable-length global context to dynamically generate and adjust the core recurrent module's WKV parameters, transforming the fixed recurrence mechanism into a context-aware dynamic function. Experiments demonstrate that this hybrid model can be trained stably via a multi-objective loss, validating its design feasibility. Our work not only opens a new research direction for RNN state representation but also provides a concrete architectural reference for future model design. The code is publicly available at: https://huggingface.co/2dgx41s/DreamState.

We study multi-objective optimization over probability distributions in Wasserstein space. Recently, Nguyen et al. (2025) introduced Multiple Wasserstein Gradient Descent (MWGraD) algorithm, which exploits the geometric structure of Wasserstein space to jointly optimize multiple objectives. Building on this approach, we propose an accelerated variant, A-MWGraD, inspired by Nesterov's acceleration. We analyze the continuous-time dynamics and establish convergence to weakly Pareto optimal points in probability space. Our theoretical results show that A-MWGraD achieves a convergence rate of O(1/t^2) for geodesically convex objectives and O(e^{-\sqrtβt}) for $β$-strongly geodesically convex objectives, improving upon the O(1/t) rate of MWGraD in the geodesically convex setting. We further introduce a practical kernel-based discretization for A-MWGraD and demonstrate through numerical experiments that it consistently outperforms MWGraD in convergence speed and sampling efficiency on multi-target sampling tasks.

In 2025, SE publishing faces an existential crisis of scalability. As our communities swell globally and integrate fast-moving methodologies like LLMs, traditional peer-review practices are collapsing under the strain. The "bureaucratic anomaly" of monolithic review has become mathematically unsustainable, creating a stochastic "lottery" that punishes novelty and exhausts researchers. This paper, written from the perspective of 2036, documents potential solutions. Here, the editors of ASE, EMSE, IST, JSS, TOSEM and TSE dream a collective dream of a brighter future. In summary first we stopped fighting (The Journal Alliance). Then we fixed the process (The Lottery / Unbundling / Fixing the Benchmark Graveyard). And then we fixed the culture (Cathedrals/Bazaars).

The emerging applications of next-generation wireless networks (e.g., immersive 3D communication, low-altitude networks, and integrated sensing and communication) necessitate high-fidelity environmental intelligence. 3D radio maps have emerged as a critical tool for this purpose, enabling spectrum-aware planning and environment-aware sensing by bridging the gap between physical environments and electromagnetic signal propagation. However, constructing accurate 3D radio maps requires fine-grained 3D geometric information and a profound understanding of electromagnetic wave propagation. Existing approaches typically treat optical and wireless knowledge as distinct modalities, failing to exploit the fundamental physical principles governing both light and electromagnetic propagation. To bridge this gap, we propose URF-GS, a unified radio-optical radiation field representation framework for accurate and generalizable 3D radio map construction based on 3D Gaussian splatting (3D-GS) and inverse rendering. By fusing visual and wireless sensing observations, URF-GS recovers scene geometry and material properties while accurately predicting radio signal behavior at arbitrary transmitter-receiver (Tx-Rx) configurations. Experimental results demonstrate that URF-GS achieves up to a 24.7% improvement in spatial spectrum prediction accuracy and a 10x increase in sample efficiency for 3D radio map construction compared with neural radiance field (NeRF)-based methods. This work establishes a foundation for next-generation wireless networks by integrating perception, interaction, and communication through holistic radiation field reconstruction.

Extracting actionable suggestions from customer reviews is essential for operational decision-making, yet these directives are often embedded within mixed-intent, unstructured text. Existing approaches either classify suggestion-bearing sentences or generate high-level summaries, but rarely isolate the precise improvement instructions businesses need. We evaluate a hybrid pipeline combining a high-recall RoBERTa classifier trained with a precision-recall surrogate to reduce unrecoverable false negatives with a controlled, instruction-tuned LLM for suggestion extraction, categorization, clustering, and summarization. Across real-world hospitality and food datasets, the hybrid system outperforms prompt-only, rule-based, and classifier-only baselines in extraction accuracy and cluster coherence. Human evaluations further confirm that the resulting suggestions and summaries are clear, faithful, and interpretable. Overall, our results show that hybrid reasoning architectures achieve meaningful improvements fine-grained actionable suggestion mining while highlighting challenges in domain adaptation and efficient local deployment.

Semantic associations such as the link between "bird" and "flew" are foundational for language modeling as they enable models to go beyond memorization and instead generalize and generate coherent text. Understanding how these associations are learned and represented in language models is essential for connecting deep learning with linguistic theory and developing a mechanistic foundation for large language models. In this work, we analyze how these associations emerge from natural language data in attention-based language models through the lens of training dynamics. By leveraging a leading-term approximation of the gradients, we develop closed-form expressions for the weights at early stages of training that explain how semantic associations first take shape. Through our analysis, we reveal that each set of weights of the transformer has closed-form expressions as simple compositions of three basis functions (bigram, token-interchangeability, and context mappings), reflecting the statistics of the text corpus and uncovering how each component of the transformer captures semantic associations based on these compositions. Experiments on real-world LLMs demonstrate that our theoretical weight characterizations closely match the learned weights, and qualitative analyses further show how our theorem shines light on interpreting the learned associations in transformers.

Designing enzymes with substrate-binding pockets is a critical challenge in protein engineering, as catalytic activity depends on the precise interaction between pockets and substrates. Currently, generative models dominate functional protein design but cannot model pocket-substrate interactions, which limits the generation of enzymes with precise catalytic environments. To address this issue, we propose EnzyPGM, a unified framework that jointly generates enzymes and substrate-binding pockets conditioned on functional priors and substrates, with a particular focus on learning accurate pocket-substrate interactions. At its core, EnzyPGM includes two main modules: a Residue-atom Bi-scale Attention (RBA) that jointly models intra-residue dependencies and fine-grained interactions between pocket residues and substrate atoms, and a Residue Function Fusion (RFF) that incorporates enzyme function priors into residue representations. Also, we curate EnzyPock, an enzyme-pocket dataset comprising 83,062 enzyme-substrate pairs across 1,036 four-level enzyme families. Extensive experiments demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock. Notably, EnzyPGM reduces the average binding energy of 0.47 kcal/mol over EnzyGen, showing its superior performance on substrate-specific enzyme design. The code and dataset will be released later.

Recent vision-language models have strong perceptual ability but their implicit reasoning is hard to explain and easily generates hallucinations on complex queries. Compositional methods improve interpretability, but most rely on a single agent or hand-crafted pipeline and cannot decide when to collaborate across complementary agents or compete among overlapping ones. We introduce MATA (Multi-Agent hierarchical Trainable Automaton), a multi-agent system presented as a hierarchical finite-state automaton for visual reasoning whose top-level transitions are chosen by a trainable hyper agent. Each agent corresponds to a state in the hyper automaton, and runs a small rule-based sub-automaton for reliable micro-control. All agents read and write a shared memory, yielding transparent execution history. To supervise the hyper agent's transition policy, we build transition-trajectory trees and transform to memory-to-next-state pairs, forming the MATA-SFT-90K dataset for supervised finetuning (SFT). The finetuned LLM as the transition policy understands the query and the capacity of agents, and it can efficiently choose the optimal agent to solve the task. Across multiple visual reasoning benchmarks, MATA achieves the state-of-the-art results compared with monolithic and compositional baselines. The code and dataset are available at https://github.com/ControlNet/MATA.

Vision-Language Models (VLMs) have shown strong multimodal reasoning capabilities on Visual-Question-Answering (VQA) benchmarks. However, their robustness against textual misinformation remains under-explored. While existing research has studied the effect of misinformation in text-only domains, it is not clear how VLMs arbitrate between contradictory information from different modalities. To bridge the gap, we first propose the CONTEXT-VQA (i.e., Conflicting Text) dataset, consisting of image-question pairs together with systematically generated persuasive prompts that deliberately conflict with visual evidence. Then, a thorough evaluation framework is designed and executed to benchmark the susceptibility of various models to these conflicting multimodal inputs. Comprehensive experiments over 11 state-of-the-art VLMs reveal that these models are indeed vulnerable to misleading textual prompts, often overriding clear visual evidence in favor of the conflicting text, and show an average performance drop of over 48.2% after only one round of persuasive conversation. Our findings highlight a critical limitation in current VLMs and underscore the need for improved robustness against textual manipulation.

Mobile GUI agents powered by large foundation models enable autonomous task execution, but frequent updates altering UI appearance and reorganizing workflows cause agents trained on historical data to fail. Despite surface changes, functional semantics and task intents remain fundamentally stable. Building on this insight, we introduce MAGNET, a memory-driven adaptive agent framework with dual-level memory: stationary memory linking diverse visual features to stable functional semantics for robust action grounding and procedural memory capturing stable task intents across varying workflows. We propose a dynamic memory evolution mechanism that continuously refines both memories by prioritizing frequently accessed knowledge. Online benchmark AndroidWorld evaluations show substantial improvements over baselines, while offline benchmarks confirm consistent gains under distribution shifts. These results validate that leveraging stable structures across interface changes improves agent performance and generalization in evolving software environments.

The integration of Large Language Models (LLMs) into recommendation systems has introduced unprecedented capabilities for natural language understanding, explanation generation, and conversational interactions. However, existing evaluation methodologies focus predominantly on traditional accuracy metrics, failing to capture the multifaceted human-centered qualities that determine the real-world user experience. We introduce \framework{} (\textbf{H}uman-centered \textbf{E}valuation for \textbf{L}LM-powered reco\textbf{M}menders), a comprehensive evaluation framework that systematically assesses LLM-powered recommender systems across five human-centered dimensions: \textit{Intent Alignment}, \textit{Explanation Quality}, \textit{Interaction Naturalness}, \textit{Trust \& Transparency}, and \textit{Fairness \& Diversity}. Through extensive experiments involving three state-of-the-art LLM-based recommenders (GPT-4, LLaMA-3.1, and P5) across three domains (movies, books, and restaurants), and rigorous evaluation by 12 domain experts using 847 recommendation scenarios, we demonstrate that \framework{} reveals critical quality dimensions invisible to traditional metrics. Our results show that while GPT-4 achieves superior explanation quality (4.21/5.0) and interaction naturalness (4.35/5.0), it exhibits a significant popularity bias (Gini coefficient 0.73) compared to traditional collaborative filtering (0.58). We release \framework{} as an open-source toolkit to advance human-centered evaluation practices in the recommender systems community.

Speaker-attributed automatic speech recognition (ASR) in multi-speaker environments remains a major challenge. While some approaches achieve strong performance when fine-tuned on specific domains, few systems generalize well across out-of-domain datasets. Our prior work, Diarization-Conditioned Whisper (DiCoW), leverages speaker diarization outputs as conditioning information and, with minimal fine-tuning, demonstrated strong multilingual and multi-domain performance. In this paper, we address a key limitation of DiCoW: ambiguity in Silence-Target-Non-target-Overlap (STNO) masks, where two or more fully overlapping speakers may have nearly identical conditioning despite differing transcriptions. We introduce SE-DiCoW (Self-Enrolled Diarization-Conditioned Whisper), which uses diarization output to locate an enrollment segment anywhere in the conversation where the target speaker is most active. This enrollment segment is used as fixed conditioning via cross-attention at each encoder layer. We further refine DiCoW with improved data segmentation, model initialization, and augmentation. Together, these advances yield substantial gains: SE-DiCoW reduces macro-averaged tcpWER by 52.4% relative to the original DiCoW on the EMMA MT-ASR benchmark.

Existing datasets for multimodal table understanding, such as MMTab, primarily provide short factual answers without explicit multi-step reasoning supervision. Models trained on these datasets often generate brief responses that offers insufficient accuracy and limited interpretability into how these models arrive at the final answer. We introduce CoReTab, a code-driven reasoning framework that produces scalable, interpretable, and automatically verifiable annotations by coupling multi-step reasoning with executable Python code. Using the CoReTab framework, we curate a dataset of 115K verified samples averaging 529 tokens per response and fine-tune open-source MLLMs through a three-stage pipeline. We evaluate the resulting model trained on CoReTab across 17 MMTab benchmarks spanning table question answering, fact verification, and table structure understanding. Our model achieves significant gains of +6.2%, +5.7%, and +25.6%, respectively, over MMTab-trained baselines, while producing transparent and verifiable reasoning traces. These results establish CoReTab as a robust and generalizable supervision framework for improving multi-step reasoning in multimodal table understanding.

We introduce TeMLM, a set of transparency-first release artifacts for clinical language models. TeMLM unifies provenance, data transparency, modeling transparency, and governance into a single, machine-checkable release bundle. We define an artifact suite (TeMLM-Card, TeMLM-Datasheet, TeMLM-Provenance) and a lightweight conformance checklist for repeatable auditing. We instantiate the artifacts on Technetium-I, a large-scale synthetic clinical NLP dataset with 498,000 notes, 7.74M PHI entity annotations across 10 types, and ICD-9-CM diagnosis labels, and report reference results for ProtactiniumBERT (about 100 million parameters) on PHI de-identification (token classification) and top-50 ICD-9 code extraction (multi-label classification). We emphasize that synthetic benchmarks are valuable for tooling and process validation, but models should be validated on real clinical data prior to deployment.

Risk disclosures in SEC filings describe potential adverse events but rarely quantify their likelihood, limiting their usefulness for probabilistic analysis. A central obstacle is the absence of large-scale, risk-level supervision linking disclosed risks to realized outcomes. We introduce a fully automated data generation pipeline that converts qualitative SEC risk disclosures into temporally grounded supervision using only public data. For each filing, the pipeline generates firm-specific, time-bounded risk queries from the Risk Factors section and labels them by automatically resolving outcomes against subsequent disclosures. Using this dataset of risk queries and outcomes grounded in SEC filings, we train a compact large language model to estimate the probability that a disclosed risk will materialize within a specified horizon. Despite its modest size, the resulting model substantially improves over pretrained and heuristic baselines, and outperforms frontier general-purpose models, including GPT-5, on probabilistic accuracy and calibration. More broadly, this work demonstrates that Foresight Learning enables scalable and fully automated training of domain-specific expert models using only raw, chronological, in-domain text -- without proprietary data, external corpora, or manual annotation. The resulting models achieve frontier-level performance while remaining deployable on a single GPU. This result suggests a general pathway for learning calibrated, decision-relevant signals from naturally occurring enterprise documents. To support transparency and reproducibility, we open-source the evaluation dataset used in this study. Evaluation Data: https://huggingface.co/datasets/LightningRodLabs/sec_risk_questions_test_set Data Generation Platform: https://lightningrod.ai/ SDK: https://github.com/lightning-rod-labs/lightningrod-python-sdk

Fairness is a central pillar of trustworthy machine learning, especially in domains where accuracy- or profit-driven optimization is insufficient. While most fairness research focuses on supervised learning, fairness in policy learning remains less explored. Because policy learning is interventional, it induces two distinct fairness targets: action fairness (equitable action assignments) and outcome fairness (equitable downstream consequences). Crucially, equalizing actions does not generally equalize outcomes when groups face different constraints or respond differently to the same action. We propose a novel double fairness learning (DFL) framework that explicitly manages the trade-off among three objectives: action fairness, outcome fairness, and value maximization. We integrate fairness directly into a multi-objective optimization problem for policy learning and employ a lexicographic weighted Tchebyshev method that recovers Pareto solutions beyond convex settings, with theoretical guarantees on the regret bounds. Our framework is flexible and accommodates various commonly used fairness notions. Extensive simulations demonstrate improved performance relative to competing methods. In applications to a motor third-party liability insurance dataset and an entrepreneurship training dataset, DFL substantially improves both action and outcome fairness while incurring only a modest reduction in overall value.

Inversion-free image editing using flow-based generative models challenges the prevailing inversion-based pipelines. However, existing approaches rely on fixed Gaussian noise to construct the source trajectory, leading to biased trajectory dynamics and causing structural degradation or quality loss. To address this, we introduce SNR-Edit, a training-free framework achieving faithful Latent Trajectory Correction via adaptive noise control. Mechanistically, SNR-Edit uses structure-aware noise rectification to inject segmentation constraints into the initial noise, anchoring the stochastic component of the source trajectory to the real image's implicit inversion position and reducing trajectory drift during source--target transport. This lightweight modification yields smoother latent trajectories and ensures high-fidelity structural preservation without requiring model tuning or inversion. Across SD3 and FLUX, evaluations on PIE-Bench and SNR-Bench show that SNR-Edit delivers performance on pixel-level metrics and VLM-based scoring, while adding only about 1s overhead per image.

Autoencoders have long been considered a nonlinear extension of Principal Component Analysis (PCA). Prior studies have demonstrated that linear autoencoders (LAEs) can recover the ordered, axis-aligned principal components of PCA by incorporating non-uniform $\ell_2$ regularization or by adjusting the loss function. However, these approaches become insufficient in the nonlinear setting, as the remaining variance cannot be properly captured independently of the nonlinear mapping. In this work, we propose a novel autoencoder framework that integrates non-uniform variance regularization with an isometric constraint. This design serves as a natural generalization of PCA, enabling the model to preserve key advantages, such as ordered representations and variance retention, while remaining effective for nonlinear dimensionality reduction tasks.

Sequential recommendation models are widely used in applications, yet they face stringent latency requirements. Mainstream models leverage the Transformer attention mechanism to improve performance, but its computational complexity grows with the sequence length, leading to a latency challenge for long sequences. Consequently, KV cache technology has recently been explored in sequential recommendation systems to reduce inference latency. However, KV cache introduces substantial storage overhead in sequential recommendation systems, which often have a large user base with potentially very long user history sequences. In this work, we observe that KV sequences across different users exhibit significant similarities, indicating the existence of collaborative signals in KV. Furthermore, we analyze the KV using singular value decomposition (SVD) and find that the information in KV can be divided into two parts: the majority of the information is shareable across users, while a small portion is user-specific. Motivated by this, we propose CollectiveKV, a cross-user KV sharing mechanism. It captures the information shared across users through a learnable global KV pool. During inference, each user retrieves high-dimensional shared KV from the pool and concatenates them with low-dimensional user-specific KV to obtain the final KV. Experiments on five sequential recommendation models and three datasets show that our method can compress the KV cache to only 0.8% of its original size, while maintaining or even enhancing model performance.

Link sign prediction on a signed graph is a task to determine whether the relationship represented by an edge is positive or negative. Since the presence of negative edges violates the graph homophily assumption that adjacent nodes are similar, regular graph methods have not been applicable without auxiliary structures to handle them. We aim to directly model the latent statistical dependency among edges with the Gaussian copula and its corresponding correlation matrix, extending CopulaGNN. However, a naive modeling of edge-edge relations is computationally intractable even for a graph with moderate scale. To address this, we propose to 1) represent the correlation matrix as a Gramian of edge embeddings, significantly reducing the number of parameters, and 2) reformulate the conditional probability distribution to dramatically reduce the inference cost. We theoretically verify scalability of our method by proving its linear convergence. Also, our extensive experiments demonstrate that it achieves significantly faster convergence than baselines, maintaining competitive prediction performance to the state-of-the-art models.

Sponge attacks increasingly threaten LLM systems by inducing excessive computation and DoS. Existing defenses either rely on statistical filters that fail on semantically meaningful attacks or use static LLM-based detectors that struggle to adapt as attack strategies evolve. We introduce SHIELD, a multi-agent, auto-healing defense framework centered on a three-stage Defense Agent that integrates semantic similarity retrieval, pattern matching, and LLM-based reasoning. Two auxiliary agents, a Knowledge Updating Agent and a Prompt Optimization Agent, form a closed self-healing loop, when an attack bypasses detection, the system updates an evolving knowledgebase, and refines defense instructions. Extensive experiments show that SHIELD consistently outperforms perplexity-based and standalone LLM defenses, achieving high F1 scores across both non-semantic and semantic sponge attacks, demonstrating the effectiveness of agentic self-healing against evolving resource-exhaustion threats.

While generative AI enables high-fidelity UI generation from text prompts, users struggle to articulate design intent and evaluate or refine results-creating gulfs of execution and evaluation. To understand the information needed for UI generation, we conducted a thematic analysis of UI prompting guidelines, identifying key design semantics and discovering that they are hierarchical and interdependent. Leveraging these findings, we developed a system that enables users to specify semantics, visualize relationships, and extract how semantics are reflected in generated UIs. By making semantics serve as an intermediate representation between human intent and AI output, our system bridges both gulfs by making requirements explicit and outcomes interpretable. A comparative user study suggests that our approach enhances users' perceived control over intent expression, outcome interpretation, and facilitates more predictable, iterative refinement. Our work demonstrates how explicit semantic representation enables systematic and explainable exploration of design possibilities in AI-driven UI design.

Automatically extracting workflows as procedural graphs from natural language is promising yet underexplored, demanding both structural validity and logical alignment. While recent large language models (LLMs) show potential for procedural graph extraction, they often produce ill-formed structures or misinterpret logical flows. We present \model{}, a multi-agent framework that formulates procedural graph extraction as a multi-round reasoning process with dedicated structural and logical refinement. The framework iterates through three stages: (1) a graph extraction phase with the graph builder agent, (2) a structural feedback phase in which a simulation agent diagnoses and explains structural defects, and (3) a logical feedback phase in which a semantic agent aligns semantics between flow logic and linguistic cues in the source text. Important feedback is prioritized and expressed in natural language, which is injected into subsequent prompts, enabling interpretable and controllable refinement. This modular design allows agents to target distinct error types without supervision or parameter updates. Experiments demonstrate that \model{} achieves substantial improvements in both structural correctness and logical consistency over strong baselines.

FaaS platforms rely on cluster managers like Kubernetes for resource management. Kubernetes is popular due to its state-centric APIs that decouple the control plane into modular controllers. However, to scale out a burst of FaaS instances, message passing becomes the primary bottleneck as controllers have to exchange extensive state through the API Server. Existing solutions opt for a clean-slate redesign of cluster managers, but at the expense of compatibility with existing ecosystem and substantial engineering effort. We present KUBEDIRECT, a Kubernetes-based cluster manager for FaaS. We find that there exists a common narrow waist across FaaS platform that allows us to achieve both efficiency and external compatibility. Our insight is that the sequential structure of the narrow waist obviates the need for a single source of truth, allowing us to bypass the API Server and perform direct message passing for efficiency. However, our approach introduces a set of ephemeral states across controllers, making it challenging to enforce end-to-end semantics due to the absence of centralized coordination. KUBEDIRECT employs a novel state management scheme that leverages the narrow waist as a hierarchical write-back cache, ensuring consistency and convergence to the desired state. KUBEDIRECT can seamlessly integrate with Kubernetes, adding ~150 LoC per controller. Experiments show that KUBEDIRECT reduces serving latency by 26.7x over Knative, and has similar performance as the state-of-the-art clean-slate platform Dirigent.

We analyze Muon as originally proposed and used in practice -- using the momentum orthogonalization with a few Newton-Schulz steps. The prior theoretical results replace this key step in Muon with an exact SVD-based polar factor. We prove that Muon with Newton-Schulz converges to a stationary point at the same rate as the SVD-polar idealization, up to a constant factor for a given number $q$ of Newton-Schulz steps. We further analyze this constant factor and prove that it converges to 1 doubly exponentially in $q$ and improves with the degree of the polynomial used in Newton-Schulz for approximating the orthogonalization direction. We also prove that Muon removes the typical square-root-of-rank loss compared to its vector-based counterpart, SGD with momentum. Our results explain why Muon with a few low-degree Newton-Schulz steps matches exact-polar (SVD) behavior at a much faster wall-clock time and explain how much momentum matrix orthogonalization via Newton-Schulz benefits over the vector-based optimizer. Overall, our theory justifies the practical Newton-Schulz design of Muon, narrowing its practice-theory gap.

Image geolocation aims to infer capture locations based on visual content. Fundamentally, this constitutes a reasoning process composed of \textit{hypothesis-verification cycles}, requiring models to possess both geospatial reasoning capabilities and the ability to verify evidence against geographic facts. Existing methods typically internalize location knowledge and reasoning patterns into static memory via supervised training or trajectory-based reinforcement fine-tuning. Consequently, these methods are prone to factual hallucinations and generalization bottlenecks in open-world settings or scenarios requiring dynamic knowledge. To address these challenges, we propose a Hierarchical Localization Agent, called LocationAgent. Our core philosophy is to retain hierarchical reasoning logic within the model while offloading the verification of geographic evidence to external tools. To implement hierarchical reasoning, we design the RER architecture (Reasoner-Executor-Recorder), which employs role separation and context compression to prevent the drifting problem in multi-step reasoning. For evidence verification, we construct a suite of clue exploration tools that provide diverse evidence to support location reasoning. Furthermore, to address data leakage and the scarcity of Chinese data in existing datasets, we introduce CCL-Bench (China City Location Bench), an image geolocation benchmark encompassing various scene granularities and difficulty levels. Extensive experiments demonstrate that LocationAgent significantly outperforms existing methods by at least 30\% in zero-shot settings.

Shuffling is a powerful way to amplify privacy of a local randomizer in private distributed data analysis, but existing analyses mostly treat the local differential privacy (DP) parameter $\varepsilon_0$ as the only knob and give generic upper bounds that can be loose and do not even characterize how shuffling amplifies privacy for basic mechanisms such as the Gaussian mechanism. We revisit the privacy blanket bound of Balle et al. (the blanket divergence) and develop an asymptotic analysis that applies to a broad class of local randomizers under mild regularity assumptions, without requiring pure local DP. Our key finding is that the leading term of the blanket divergence depends on the local mechanism only through a single scalar parameter $χ$, which we call the shuffle index. By applying this asymptotic analysis to both upper and lower bounds, we obtain a tight band for $δ_n$ in the shuffled mechanism's $(\varepsilon_n,δ_n)$-DP guarantee. Moreover, we derive a simple structural necessary and sufficient condition on the local randomizer under which the blanket-divergence-based upper and lower bounds coincide asymptotically. $k$-RR families with $k\ge3$ satisfy this condition, while for generalized Gaussian mechanisms the condition may not hold but the resulting band remains tight. Finally, we complement the asymptotic theory with an FFT-based algorithm for computing the blanket divergence at finite $n$, which offers rigorously controlled relative error and near-linear running time in $n$, providing a practical numerical analysis for shuffle DP.

Recent progress at the intersection of large language models (LLMs) and time series (TS) analysis has revealed both promise and fragility. While LLMs can reason over temporal structure given carefully engineered context, they often struggle with numeric fidelity, modality interference, and principled cross-modal integration. We present TS-Debate, a modality-specialized, collaborative multi-agent debate framework for zero-shot time series reasoning. TS-Debate assigns dedicated expert agents to textual context, visual patterns, and numerical signals, preceded by explicit domain knowledge elicitation, and coordinates their interaction via a structured debate protocol. Reviewer agents evaluate agent claims using a verification-conflict-calibration mechanism, supported by lightweight code execution and numerical lookup for programmatic verification. This architecture preserves modality fidelity, exposes conflicting evidence, and mitigates numeric hallucinations without task-specific fine-tuning. Across 20 tasks spanning three public benchmarks, TS-Debate achieves consistent and significant performance improvements over strong baselines, including standard multimodal debate in which all agents observe all inputs.

G protein-coupled receptors (GPCRs) govern diverse physiological processes and are central to modern pharmacology. Yet discovering GPCR modulators remains challenging because receptor activation often arises from complex allosteric effects rather than direct binding affinity, and conventional assays are slow, costly, and not optimized for capturing these dynamics. Here we present GPCR-Filter, a deep learning framework specifically developed for GPCR modulator discovery. We assembled a high-quality dataset of over 90,000 experimentally validated GPCR-ligand pairs, providing a robust foundation for training and evaluation. GPCR-Filter integrates the ESM-3 protein language model for high-fidelity GPCR sequence representations with graph neural networks that encode ligand structures, coupled through an attention-based fusion mechanism that learns receptor-ligand functional relationships. Across multiple evaluation settings, GPCR-Filter consistently outperforms state-of-the-art compound-protein interaction models and exhibits strong generalization to unseen receptors and ligands. Notably, the model successfully identified micromolar-level agonists of the 5-HT\textsubscript{1A} receptor with distinct chemical frameworks. These results establish GPCR-Filter as a scalable and effective computational approach for GPCR modulator discovery, advancing AI-assisted drug development for complex signaling systems.

Continuous Integration (CI) provides early feedback by automatically building software, but long build durations can hinder developer productivity. CI services offer caching mechanisms to speed up builds by reusing infrequently changing artifacts, yet little is known about how caching is adopted in practice and what challenges it entails. In this paper, we conduct a large-scale empirical study of CI caching in Travis CI, analyzing 513,384 builds from 1,279 GitHub projects. We find that only 30% of projects adopt CI caching, and early adoption is strongly associated with project maturity, such as more dependencies, more commits, and longer CI lifespans. To understand why many projects do not adopt caching, we submitted pull requests enabling caching in non-adopting projects, and nearly half were accepted or merged. Developer feedback suggests that non- or late adoption mainly stems from limited awareness of CI caching support. We also examine cache maintenance and identify five common activities, performed by 24% of cache-enabled projects. Although one-third of projects see substantial build-time reductions, cache uploads occur in 97% of builds, and 33% of projects contain stale cached artifacts. Finally, our analysis of reported caching issues shows developers mainly struggle with corrupted or outdated caches or request broader caching features. Overall, CI caching does not help all projects, needs ongoing maintenance, and is more complex in practice than many developers expect.

User behavior sequences in modern recommendation systems exhibit significant length heterogeneity, ranging from sparse short-term interactions to rich long-term histories. While longer sequences provide more context, we observe that increasing the maximum input sequence length in existing CTR models paradoxically degrades performance for short-sequence users due to attention polarization and length imbalance in training data. To address this, we propose LAIN(Length-Adaptive Interest Network), a plug-and-play framework that explicitly incorporates sequence length as a conditioning signal to balance long- and short-sequence modeling. LAIN consists of three lightweight components: a Spectral Length Encoder that maps length into continuous representations, Length-Conditioned Prompting that injects global contextual cues into both long- and short-term behavior branches, and Length-Modulated Attention that adaptively adjusts attention sharpness based on sequence length. Extensive experiments on three real-world benchmarks across five strong CTR backbones show that LAIN consistently improves overall performance, achieving up to 1.15% AUC gain and 2.25% log loss reduction. Notably, our method significantly improves accuracy for short-sequence users without sacrificing longsequence effectiveness. Our work offers a general, efficient, and deployable solution to mitigate length-induced bias in sequential recommendation.

The growing adoption of Apple Silicon for machine learning development has created demand for efficient inference solutions that leverage its unique unified memory architecture. However, existing tools either lack native optimization (PyTorch MPS) or focus solely on text models (llama.cpp), leaving multimodal workloads underserved. We present vllm-mlx, a framework for efficient LLM and MLLM inference on Apple Silicon built natively on MLX. For text models, we achieve 21% to 87% higher throughput than llama.cpp across models ranging from Qwen3-0.6B to Nemotron-30B, while providing continuous batching that scales to 4.3x aggregate throughput at 16 concurrent requests. For multimodal models, we introduce content-based prefix caching that eliminates redundant vision encoding by identifying identical images through content hashing, regardless of input format. Our evaluation on Apple M4 Max demonstrates throughput of up to 525 tokens per second on text models and 28x speedup on repeated image queries, reducing multimodal latency from 21.7 seconds to under 1 second. Video analysis with up to 64 frames achieves 24.7x cache speedup. We release our implementation as open source to support efficient inference on consumer Apple hardware.

Secure code review is critical at the pre-commit stage, where vulnerabilities must be caught early under tight latency and limited-context constraints. Existing SAST-based checks are noisy and often miss immature, context-dependent vulnerabilities, while standalone Large Language Models (LLMs) are constrained by context windows and lack explicit tool use. Agentic AI, which combine LLMs with autonomous decision-making, tool invocation, and code navigation, offer a promising alternative, but their effectiveness for pre-commit secure code review is not yet well understood. In this work, we introduce AgenticSCR, an agentic AI for secure code review for detecting immature vulnerabilities during the pre-commit stage, augmented by security-focused semantic memories. Using our own curated benchmark of immature vulnerabilities, tailored to the pre-commit secure code review, we empirically evaluate how accurate is our AgenticSCR for localizing, detecting, and explaining immature vulnerabilities. Our results show that AgenticSCR achieves at least 153% relatively higher percentage of correct code review comments than the static LLM-based baseline, and also substantially surpasses SAST tools. Moreover, AgenticSCR generates more correct comments in four out of five vulnerability types, consistently and significantly outperforming all other baselines. These findings highlight the importance of Agentic Secure Code Review, paving the way towards an emerging research area of immature vulnerability detection.

Amazon published its Frontier Model Safety Framework (FMSF) as part of the Paris AI summit, following which we presented a report on Amazon's Premier model. In this report, we present an evaluation of Nova 2.0 Lite. Nova 2.0 Lite was made generally available from amongst the Nova 2.0 series and is one of its most capable reasoning models. The model processes text, images, and video with a context length of up to 1M tokens, enabling analysis of large codebases, documents, and videos in a single prompt. We present a comprehensive evaluation of Nova 2.0 Lite's critical risk profile under the FMSF. Evaluations target three high-risk domains-Chemical, Biological, Radiological and Nuclear (CBRN), Offensive Cyber Operations, and Automated AI R&D-and combine automated benchmarks, expert red-teaming, and uplift studies to determine whether the model exceeds release thresholds. We summarize our methodology and report core findings. We will continue to enhance our safety evaluation and mitigation pipelines as new risks and capabilities associated with frontier models are identified.

This paper summarizes the opportunities of in-network collective operations (INC) for accelerated collective operations in AI workloads. We provide sufficient detail to make this important field accessible to non-experts in AI or networking, fostering a connection between these communities. Consider two types of INC: Edge-INC, where the system is implemented at the node level, and Core-INC, where the system is embedded within network switches. We outline the potential performance benefits as well as six key obstacles in the context of both Edge-INC and Core-INC that may hinder their adoption. Finally, we present a set of predictions for the future development and application of INC.

Improper exposure often leads to severe loss of details, color distortion, and reduced contrast. Exposure correction still faces two critical challenges: (1) the ignorance of object-wise regional semantic information causes the color shift artifacts; (2) real-world exposure images generally have no ground-truth labels, and its labeling entails massive manual editing. To tackle the challenges, we propose a new unsupervised semantic-aware exposure correction network. It contains an adaptive semantic-aware fusion module, which effectively fuses the semantic information extracted from a pre-trained Fast Segment Anything Model into a shared image feature space. Then the fused features are used by our multi-scale residual spatial mamba group to restore the details and adjust the exposure. To avoid manual editing, we propose a pseudo-ground truth generator guided by CLIP, which is fine-tuned to automatically identify exposure situations and instruct the tailored corrections. Also, we leverage the rich priors from the FastSAM and CLIP to develop a semantic-prompt consistency loss to enforce semantic consistency and image-prompt alignment for unsupervised training. Comprehensive experimental results illustrate the effectiveness of our method in correcting real-world exposure images and outperforms state-of-the-art unsupervised methods both numerically and visually.

The Bahnar people, an ethnic minority in Vietnam with a rich ancestral heritage, possess a language of immense cultural and historical significance. The government places a strong emphasis on preserving and promoting the Bahnaric language by making it accessible online and encouraging communication across generations. Recent advancements in artificial intelligence, such as Neural Machine Translation (NMT), have brought about a transformation in translation by improving accuracy and fluency. This, in turn, contributes to the revival of the language through educational efforts, communication, and documentation. Specifically, NMT is pivotal in enhancing accessibility for Bahnaric speakers, making information and content more readily available. Nevertheless, the translation of Vietnamese into Bahnaric faces practical challenges due to resource constraints, especially given the limited resources available for the Bahnaric language. To address this, we employ state-of-the-art techniques in NMT along with two augmentation strategies for domain-specific Vietnamese-Bahnaric translation task. Importantly, both approaches are flexible and can be used with various neural machine translation models. Additionally, they do not require complex data preprocessing steps, the training of additional systems, or the acquisition of extra data beyond the existing training parallel corpora.

Function call capabilities have become crucial for Large Language Models (LLMs), enabling them to interact more effectively with external tools and APIs. Existing methods for improving the function call capabilities of LLMs rely on data obtained either through manual annotation or automated generation by models, and use this data to finetune the LLMs. However, these methods often lack targeted design and are constrained by fixed patterns and data distributions, which limits their effectiveness in enhancing the generalization and robustness of function call LLMs. To address this limitation, we propose a novel adversarial data augmentation method that employs reinforcement learning to systematically identify and target the weaknesses of function call LLMs. Our training framework introduces a query model trained with reinforcement learning (RL) to generate adversarial queries that are specifically designed to challenge function call (FC) models. This approach adopts a zero sum game formulation, where the query model and the FC model engage in iterative alternating training. Overall, our method advances the development of more robust FC models and provides a systematic way to identify and correct weaknesses in the ability of LLMs to interact with external tools.

Recommendation systems must optimize multiple objectives while satisfying hard business constraints such as fairness and coverage. For example, an e-commerce platform may require every recommendation list to include items from multiple sellers and at least one newly listed product; violating such constraints--even once--is unacceptable in production. Prior work on multi-objective recommendation and recent LLM-based recommender agents largely treat constraints as soft penalties or focus on item scoring and interaction, leading to frequent violations in real-world deployments. How to leverage LLMs for coordinating constrained optimization in recommendation systems remains underexplored. We propose DualAgent-Rec, an LLM-coordinated dual-agent framework for constrained multi-objective e-commerce recommendation. The framework separates optimization into an Exploitation Agent that prioritizes accuracy under hard constraints and an Exploration Agent that promotes diversity through unconstrained Pareto search. An LLM-based coordinator adaptively allocates resources between agents based on optimization progress and constraint satisfaction, while an adaptive epsilon-relaxation mechanism guarantees feasibility of final solutions. Experiments on the Amazon Reviews 2023 dataset demonstrate that DualAgent-Rec achieves 100% constraint satisfaction and improves Pareto hypervolume by 4-6% over strong baselines, while maintaining competitive accuracy-diversity trade-offs. These results indicate that LLMs can act as effective orchestration agents for deployable and constraint-compliant recommendation systems.

Large Language Models (LLMs) are increasingly used to generate natural-language explanations in recommender systems, acting as explanation agents that reason over user behavior histories. While prior work has focused on explanation fluency and relevance under fixed inputs, the robustness of LLM-generated explanations to realistic user behavior noise remains largely unexplored. In real-world web platforms, interaction histories are inherently noisy due to accidental clicks, temporal inconsistencies, missing values, and evolving preferences, raising concerns about explanation stability and user trust. We present RobustExplain, the first systematic evaluation framework for measuring the robustness of LLM-generated recommendation explanations. RobustExplain introduces five realistic user behavior perturbations evaluated across multiple severity levels and a multi-dimensional robustness metric capturing semantic, keyword, structural, and length consistency. Our goal is to establish a principled, task-level evaluation framework and initial robustness baselines, rather than to provide a comprehensive leaderboard across all available LLMs. Experiments on four representative LLMs (7B--70B) show that current models exhibit only moderate robustness, with larger models achieving up to 8% higher stability. Our results establish the first robustness benchmarks for explanation agents and highlight robustness as a critical dimension for trustworthy, agent-driven recommender systems at web scale.

Emotional Talking Face synthesis is pivotal in multimedia and signal processing, yet existing 3D methods suffer from two critical challenges: poor audio-vision emotion alignment, manifested as difficult audio emotion extraction and inadequate control over emotional micro-expressions; and a one-size-fits-all multi-view fusion strategy that overlooks uncertainty and feature quality differences, undermining rendering quality. We propose UA-3DTalk, Uncertainty-Aware 3D Emotional Talking Face Synthesis with emotion prior distillation, which has three core modules: the Prior Extraction module disentangles audio into content-synchronized features for alignment and person-specific complementary features for individualization; the Emotion Distillation module introduces a multi-modal attention-weighted fusion mechanism and 4D Gaussian encoding with multi-resolution code-books, enabling fine-grained audio emotion extraction and precise control of emotional micro-expressions; the Uncertainty-based Deformation deploys uncertainty blocks to estimate view-specific aleatoric (input noise) and epistemic (model parameters) uncertainty, realizing adaptive multi-view fusion and incorporating a multi-head decoder for Gaussian primitive optimization to mitigate the limitations of uniform-weight fusion. Extensive experiments on regular and emotional datasets show UA-3DTalk outperforms state-of-the-art methods like DEGSTalk and EDTalk by 5.2% in E-FID for emotion alignment, 3.1% in SyncC for lip synchronization, and 0.015 in LPIPS for rendering quality. Project page: https://mrask999.github.io/UA-3DTalk

Machine learning systems engineering requires a deep understanding of framework internals. Yet most current education separates algorithms from systems. Students learn gradient descent without measuring memory usage, and attention mechanisms without profiling computational cost. This split leaves graduates unprepared to debug real production failures and widens the gap between machine learning research and reliable deployment. We present TinyTorch, a 20 module curriculum in which students implement the core components of PyTorch, including tensors, autograd, optimizers, and neural networks, entirely in pure Python. The curriculum is built around three pedagogical principles. Progressive disclosure gradually introduces complexity as students build confidence. Systems first integration embeds memory and performance awareness from the very beginning. Historical milestone validation guides students to recreate key breakthroughs, from the Perceptron in 1958 to modern Transformers, using only code they have written themselves. TinyTorch requires only a laptop with 4GB of RAM and no GPU, making machine learning systems education accessible worldwide. Its goal is to prepare the next generation of AI engineers, practitioners who understand not only what machine learning systems do, but why they work and how to make them scale. The curriculum is available as open source at mlsysbook.ai slash tinytorch.

Large Language Models (LLMs) for code generation boost productivity but frequently introduce Knowledge Conflicting Hallucinations (KCHs), subtle, semantic errors, such as non-existent API parameters, that evade linters and cause runtime failures. Existing mitigations like constrained decoding or non-deterministic LLM-in-the-loop repair are often unreliable for these errors. This paper investigates whether a deterministic, static-analysis framework can reliably detect \textit{and} auto-correct KCHs. We propose a post-processing framework that parses generated code into an Abstract Syntax Tree (AST) and validates it against a dynamically-generated Knowledge Base (KB) built via library introspection. This non-executing approach uses deterministic rules to find and fix both API and identifier-level conflicts. On a manually-curated dataset of 200 Python snippets, our framework detected KCHs with 100\% precision and 87.6\% recall (0.934 F1-score), and successfully auto-corrected 77.0\% of all identified hallucinations. Our findings demonstrate that this deterministic post-processing approach is a viable and reliable alternative to probabilistic repair, offering a clear path toward trustworthy code generation.

Graph data is informative to represent complex relationships such as transactions between accounts, communications between devices, and dependencies among machines or processes. Correspondingly, graph anomaly detection (GAD) plays a critical role in identifying anomalies across various domains, including finance, cybersecurity, manufacturing, etc. Facing the large-volume and multi-domain graph data, nascent efforts attempt to develop foundational generalist models capable of detecting anomalies in unseen graphs without retraining. To the best of our knowledge, the different feature semantics and dimensions of cross-domain graph data heavily hinder the development of the graph foundation model, leaving further in-depth continual learning and inference capabilities a quite open problem. Hence, we propose OWLEYE, a novel zero-shot GAD framework that learns transferable patterns of normal behavior from multiple graphs, with a threefold contribution. First, OWLEYE proposes a cross-domain feature alignment module to harmonize feature distributions, which preserves domain-specific semantics during alignment. Second, with aligned features, to enable continuous learning capabilities, OWLEYE designs the multi-domain multi-pattern dictionary learning to encode shared structural and attribute-based patterns. Third, for achieving the in-context learning ability, OWLEYE develops a truncated attention-based reconstruction module to robustly detect anomalies without requiring labeled data for unseen graph-structured data. Extensive experiments on real-world datasets demonstrate that OWLEYE achieves superior performance and generalizability compared to state-of-the-art baselines, establishing a strong foundation for scalable and label-efficient anomaly detection.

Reinforcement learning is increasingly used for code-centric tasks. These tasks include code generation, summarization, understanding, repair, testing, and optimization. This trend is growing faster with large language models and autonomous agents. A key challenge is how to design reward signals that make sense for software. In many RL problems, the reward is a clear number. In software, this is often not possible. The goal is rarely a single numeric objective. Instead, rewards are usually proxies. Common proxies check if the code compiles, passes tests, or satisfies quality metrics. Many reward designs have been proposed for code-related tasks. However, the work is scattered across areas and papers. There is no single survey that brings these approaches together and shows the full landscape of reward design for RL in software. In this survey, we provide the first systematic and comprehensive review of reward engineering for RL in software tasks. We focus on existing methods and techniques. We structure the literature along three complementary dimensions, summarizing the reward-design choices within each. We conclude with challenges and recommendations in the reward design space for SE tasks.

Vision--language models (VLMs) achieve strong performance on many multimodal benchmarks but remain brittle on spatial reasoning tasks that require aligning abstract overhead representations with egocentric views. We introduce m2sv, a scalable benchmark for map-to-street-view spatial reasoning that asks models to infer camera viewing direction by aligning a north-up overhead map with a Street View image captured at the same real-world intersection. We release m2sv-20k, a geographically diverse benchmark with controlled ambiguity, along with m2sv-sft-11k, a curated set of structured reasoning traces for supervised fine-tuning. Despite strong performance on existing multimodal benchmarks, the best evaluated VLM achieves only 65.2% accuracy on m2sv, far below the human baseline of 95%. While supervised fine-tuning and reinforcement learning yield consistent gains, cross-benchmark evaluations reveal limited transfer. Beyond aggregate accuracy, we systematically analyze difficulty in map-to-street-view reasoning using both structural signals and human effort, and conduct an extensive failure analysis of adapted open models. Our findings highlight persistent gaps in geometric alignment, evidence aggregation, and reasoning consistency, motivating future work on grounded spatial reasoning across viewpoints.

Recent advances in large language models have enabled mental health dialogue systems, yet existing approaches remain predominantly reactive, lacking systematic user state modeling for proactive therapeutic exploration. We introduce PsyProbe, a dialogue system designed for the exploration phase of counseling that systematically tracks user psychological states through the PPPPPI framework (Presenting, Predisposing, Precipitating, Perpetuating, Protective, Impact) augmented with cognitive error detection. PsyProbe combines State Builder for extracting structured psychological profiles, Memory Construction for tracking information gaps, Strategy Planner for Motivational Interviewing behavioral codes, and Response Generator with Question Ideation and Critic/Revision modules to generate contextually appropriate, proactive questions. We evaluate PsyProbe with 27 participants in real-world Korean counseling scenarios, including automatic evaluation across ablation modes, user evaluation, and expert evaluation by a certified counselor. The full PsyProbe model consistently outperforms baseline and ablation modes in automatic evaluation. User evaluation demonstrates significantly increased engagement intention and improved naturalness compared to baseline. Expert evaluation shows that PsyProbe substantially improves core issue understanding and achieves question rates comparable to professional counselors, validating the effectiveness of systematic state modeling and proactive questioning for therapeutic exploration.

Developing models capable of complex, multi-step reasoning is a central goal in artificial intelligence. While representing problems as graphs is a powerful approach, Graph Neural Networks (GNNs) are fundamentally constrained by their message-passing mechanism, which imposes a local bottleneck that limits global, holistic reasoning. We argue that dynamic programming (DP), which solves problems by iteratively refining a global state, offers a more powerful and suitable learning paradigm. We introduce FloydNet, a new architecture that embodies this principle. In contrast to local message passing, FloydNet maintains a global, all-pairs relationship tensor and learns a generalized DP operator to progressively refine it. This enables the model to develop a task-specific relational calculus, providing a principled framework for capturing long-range dependencies. Theoretically, we prove that FloydNet achieves 3-WL (2-FWL) expressive power, and its generalized form aligns with the k-FWL hierarchy. FloydNet demonstrates state-of-the-art performance across challenging domains: it achieves near-perfect scores (often >99\%) on the CLRS-30 algorithmic benchmark, finds exact optimal solutions for the general Traveling Salesman Problem (TSP) at rates significantly exceeding strong heuristics, and empirically matches the 3-WL test on the BREC benchmark. Our results establish this learned, DP-style refinement as a powerful and practical alternative to message passing for high-level graph reasoning.

Scaling modern deep learning workloads demands coordinated placement of data and compute across device meshes, memory hierarchies, and heterogeneous accelerators. We present Axe Layout, a hardware-aware abstraction that maps logical tensor coordinates to a multi-axis physical space via named axes. Axe unifies tiling, sharding, replication, and offsets across inter-device distribution and on-device layouts, enabling collective primitives to be expressed consistently from device meshes to threads. Building on Axe, we design a multi-granularity, distribution-aware DSL and compiler that composes thread-local control with collective operators in a single kernel. Experiments show that our unified approach can bring performance close to hand-tuned kernels on across latest GPU devices and multi-device environments and accelerator backends.

Neural physics solvers are increasingly used in scientific discovery, given their potential for rapid in silico insights into physical, materials, or biological systems and their long-time evolution. However, poor generalization beyond their training support limits exploration of novel designs and long-time horizon predictions. We introduce NOVA, a route to generalizable neural physics solvers that can provide rapid, accurate solutions to scenarios even under distributional shifts in partial differential equation parameters, geometries and initial conditions. By learning physics-aligned representations from an initial sparse set of scenarios, NOVA consistently achieves 1-2 orders of magnitude lower out-of-distribution errors than data-driven baselines across complex, nonlinear problems including heat transfer, diffusion-reaction and fluid flow. We further showcase NOVA's dual impact on stabilizing long-time dynamical rollouts and improving generative design through application to the simulation of nonlinear Turing systems and fluidic chip optimization. Unlike neural physics solvers that are constrained to retrieval and/or emulation within an a priori space, NOVA enables reliable extrapolation beyond known regimes, a key capability given the need for exploration of novel hypothesis spaces in scientific discovery

While many deep learning models trained on private datasets have been deployed in various practical tasks, they may pose a privacy leakage risk as attackers could recover informative data or label knowledge from models. In this work, we present \emph{privacy-preserving model transcription}, a data-free model-to-model conversion solution to facilitate model deployment with a privacy guarantee. To this end, we propose a cooperative-competitive learning approach termed \emph{differentially private synthetic distillation} that learns to convert a pretrained model (teacher) into its privacy-preserving counterpart (student) via a trainable generator without access to private data. The learning collaborates with three players in a unified framework and performs alternate optimization: i)~the generator is learned to generate synthetic data, ii)~the teacher and student accept the synthetic data and compute differential private labels by flexible data or label noisy perturbation, and iii)~the student is updated with noisy labels and the generator is updated by taking the student as a discriminator for adversarial training. We theoretically prove that our approach can guarantee differential privacy and convergence. The transcribed student has good performance and privacy protection, while the resulting generator can generate private synthetic data for downstream tasks. Extensive experiments clearly demonstrate that our approach outperforms 26 state-of-the-arts.

We present a novel method for Efficient training with Progressive Activation Sharing (EPAS). This method bridges progressive training paradigm with the phenomenon of redundant QK (or KV ) activations across deeper layers of transformers. EPAS gradually grows a sharing region during training by switching decoder layers to activation sharing mode. This results in throughput increase due to reduced compute. To utilize deeper layer redundancy, the sharing region starts from the deep end of the model and grows towards the shallow end. The EPAS trained models allow for variable region lengths of activation sharing for different compute budgets during inference. Empirical evaluations with QK activation sharing in LLaMA models ranging from 125M to 7B parameters show up to an 11.1% improvement in training throughput and up to a 29% improvement in inference throughput while maintaining similar loss curve to the baseline models. Furthermore, applying EPAS in continual pretraining to transform TinyLLaMA into an attention-sharing model yields up to a 10% improvement in average accuracy over state-of-the-art methods, emphasizing the significance of progressive training in cross layer activation sharing models.

Mutation testing is an effective technique for assessing the effectiveness of test suites by systematically injecting artificial faults into programs. However, existing mutation testing techniques fall short in capturing many types of common faults in dynamically typed languages like Python. In this paper, we introduce a novel set of seven mutation operators that are inspired by prevalent anti-patterns in Python programs, designed to complement the existing general-purpose operators and broaden the spectrum of simulated faults. We propose a mutation testing technique that utilizes a hybrid of static and dynamic analyses to mutate Python programs based on these operators while minimizing equivalent mutants. We implement our approach in a tool called PyTation and evaluate it on 13 open-source Python applications. Our results show that PyTation generates mutants that complement those from general-purpose tools, exhibiting distinct behaviour under test execution and uncovering inadequacies in high-coverage test suites. We further demonstrate that PyTation produces a high proportion of unique mutants, a low cross-kill rate, and a low test overlap ratio relative to baseline tools, highlighting its novel fault model. PyTation also incurs few equivalent mutants, aided by dynamic analysis heuristics.

Biological neural systems must be fast but are energy-constrained. Evolution's solution: act on the first signal. Winner-take-all circuits and time-to-first-spike coding implicitly treat when a neuron fires as an expression of confidence. We apply this principle to ensembles of Tiny Recursive Models (TRM). By basing the ensemble prediction solely on the first to halt rather than averaging predictions, we achieve 97.2% puzzle accuracy on Sudoku-Extreme while using 10x less compute than test-time augmentation (the baseline achieves 86.1% single-pass, 97.3% with TTA). Inference speed is an implicit indication of confidence. But can this capability be manifested as a training-only cost? Evidently yes: by maintaining K = 4 parallel latent states during training but backpropping only through the lowest-loss "winner," a single model achieves 96.9% +/- 0.6% puzzle accuracy with a single forward pass-matching TTA performance without any test-time augmentation. As in nature, this work was also resource constrained: all experimentation used a single RTX 5090. This necessitated efficiency and compelled our invention of a modified SwiGLU which made Muon viable. With Muon and K = 1 training, we exceed TRM baseline performance in 7k steps (40 min). Higher accuracy requires 36k steps: 1.5 hours for K = 1, 6 hours for K = 4.

As LLMs increasingly act as autonomous agents in interactive and multi-agent settings, understanding their strategic behavior is critical for safety, coordination, and AI-driven social and economic systems. We investigate how payoff magnitude and linguistic context shape LLM strategies in repeated social dilemmas, using a payoff-scaled Prisoner's Dilemma to isolate sensitivity to incentive strength. Across models and languages, we observe consistent behavioral patterns, including incentive-sensitive conditional strategies and cross-linguistic divergence. To interpret these dynamics, we train supervised classifiers on canonical repeated-game strategies and apply them to LLM decisions, revealing systematic, model- and language-dependent behavioral intentions, with linguistic framing sometimes matching or exceeding architectural effects. Our results provide a unified framework for auditing LLMs as strategic agents and highlight cooperation biases with direct implications for AI governance and multi-agent system design.

Learning reduced descriptions of chaotic many-body dynamics is fundamentally challenging: although microscopic equations are Markovian, collective observables exhibit strong memory and exponential sensitivity to initial conditions and prediction errors. We show that a self-attention-based transformer framework provides an effective approach for modeling such chaotic collective dynamics directly from time-series data. By selectively reweighting long-range temporal correlations, the transformer learns a non-Markovian reduced description that overcomes intrinsic limitations of conventional recurrent architectures. As a concrete demonstration, we study the one-dimensional semiclassical Holstein model, where interaction quenches induce strongly nonlinear and chaotic dynamics of the charge-density-wave order parameter. While pointwise predictions inevitably diverge at long times, the transformer faithfully reproduces the statistical "climate" of the chaos, including temporal correlations and characteristic decay scales. Our results establish self-attention as a powerful mechanism for learning effective reduced dynamics in chaotic many-body systems.

Generative models for materials have achieved strong performance on periodic bulk crystals, yet their ability to generalize across scale transitions to finite nanostructures remains largely untested. We introduce Crystal-to-Nanoparticle (C2NP), a systematic benchmark for evaluating generative models when moving between infinite crystalline unit cells and finite nanoparticles, where surface effects and size-dependent distortions dominate. C2NP defines two complementary tasks: (i) generating nanoparticles of specified radii from periodic unit cells, testing whether models capture surface truncation and geometric constraints; and (ii) recovering bulk lattice parameters and space-group symmetry from finite particle configurations, assessing whether models can infer underlying crystallographic order despite surface perturbations. Using diverse materials as a structurally consistent testbed, we construct over 170,000 nanoparticle configurations by carving particles from supercells derived from DFT-relaxed crystal unit cells, and introduce size-based splits that separate interpolation from extrapolation regimes. Experiments with state-of-the-art approaches, including diffusion, flow-matching, and variational models, show that even when losses are low, models often fail geometrically under distribution shift, yielding large lattice-recovery errors and near-zero joint accuracy on structure and symmetry. Overall, our results suggest that current methods rely on template memorization rather than scalable physical generalization. C2NP offers a controlled, reproducible framework for diagnosing these failures, with immediate applications to nanoparticle catalyst design, nanostructured hydrides for hydrogen storage, and materials discovery. Dataset and code are available at https://github.com/KurbanIntelligenceLab/C2NP.

Large Language models (LLMs) have shown strong capabilities in code review automation, such as review comment generation, yet they suffer from hallucinations -- where the generated review comments are ungrounded in the actual code -- poses a significant challenge to the adoption of LLMs in code review workflows. To address this, we explore effective and scalable methods for a hallucination detection in LLM-generated code review comments without the reference. In this work, we design HalluJudge that aims to assess the grounding of generated review comments based on the context alignment. HalluJudge includes four key strategies ranging from direct assessment to structured multi-branch reasoning (e.g., Tree-of-Thoughts). We conduct a comprehensive evaluation of these assessment strategies across Atlassian's enterprise-scale software projects to examine the effectiveness and cost-efficiency of HalluJudge. Furthermore, we analyze the alignment between HalluJudge's judgment and developer preference of the actual LLM-generated code review comments in the real-world production. Our results show that the hallucination assessment in HalluJudge is cost-effective with an F1 score of 0.85 and an average cost of $0.009. On average, 67% of the HalluJudge assessments are aligned with the developer preference of the actual LLM-generated review comments in the online production. Our results suggest that HalluJudge can serve as a practical safeguard to reduce developers' exposure to hallucinated comments, fostering trust in AI-assisted code reviews.

We rigorously study the thermodynamic limit of deep neural networks (DNNS) and recurrent neural networks (RNNs), assuming that the activation functions are sigmoids. A thermodynamic limit is a continuous neural network, where the neurons form a continuous space with infinitely many points. We show that such a network admits a unique state in a certain region of the parameter space, which depends continuously on the parameters. This state breaks into an infinite number of states outside the mentioned region of parameter space. Then, the critical organization is a bifurcation in the parameter space, where a network transitions from a unique state to infinitely many states. We use p-adic integers to codify hierarchical structures. Indeed, we present an algorithm that recasts the hierarchical topologies used in DNNs and RNNs as p-adic tree-like structures. In this framework, the hierarchical and the critical organizations are connected. We study rigorously the critical organization of a toy model, a hierarchical edge detector for grayscale images based on p-adic cellular neural networks. The critical organization of such a network can be described as a strange attractor. In the second part, we study random versions of DNNs and RNNs. In this case, the network parameters are generalized Gaussian random variables in a space of quadratic integrable functions. We compute the probability distribution of the output given the input, in the infinite-width case. We show that it admits a power-type expansion, where the constant term is a Gaussian distribution.

Bug Reproduction Tests (BRTs) have been used in many agentic Automated Program Repair (APR) systems, primarily for validating promising fixes and aiding fix generation. In practice, when developers submit a patch, they often implement the BRT alongside the fix. Our experience deploying agentic APR reveals that developers similarly desire a BRT within AI-generated patches to increase their confidence. However, canonical APR systems tend to generate BRTs and fixes separately, or focus on producing only the fix in the final patch. In this paper, we study agentic APR in the context of cogeneration, where the APR agent is instructed to generate both a fix and a BRT in the same patch. We evaluate the effectiveness of different cogeneration strategies on 120 human-reported bugs at Google and characterize different cogeneration strategies by their influence on APR agent behavior. We develop and evaluate patch selectors that account for test change information to select patches with plausible fixes (and plausible BRTs). Finally, we analyze the root causes of failed cogeneration trajectories. Importantly, we show that cogeneration allows the APR agent to generate BRTs for at least as many bugs as a dedicated BRT agent, without compromising the generation rate of plausible fixes, thereby reducing engineering effort in maintaining and coordinating separate generation pipelines for fix and BRT at scale.

Python libraries often need to maintain a stable public API even as internal implementations evolve, gain new backends, or depend on heavy optional libraries. In Python, where internal objects are easy to inspect and import, users can come to rely on "reachable internals" that were never intended to be public, making refactoring risky and slowing long-term maintenance. This paper revisits the pointer-to-implementation (PIMPL) idiom from C++ and reinterprets it as a Pythonic pattern of opaque delegation: a small public object (or module) that delegates its behavior to a separate implementation object treated as internal. We situate this pattern within a broader taxonomy of encapsulation techniques in Python, relate it to existing practices such as module-level indirection, facade objects, and backend dispatch, and identify PIMPL-like structures already used in the standard library and the scientific Python ecosystem. We then show how a Pythonic PIMPL can be used in existing codebases to isolate heavy dependencies, support lazy imports, and enable runtime selection of alternative backends without changing the public API. Finally, we discuss the benefits and trade-offs of the approach and offer practical guidance on when the pattern is appropriate and how to apply it in large, long-lived Python libraries.

Reinforcement learning from human or AI feedback (RLHF/RLAIF) for speech-in/speech-out dialogue systems (SDS) remains underexplored, with prior work largely limited to single semantic rewards applied at the utterance level. Such setups overlook the multi-dimensional and multi-modal nature of conversational quality, which encompasses semantic coherence, audio naturalness, speaker consistency, emotion alignment, and turn-taking behavior. Moreover, they are fundamentally mismatched with duplex spoken dialogue systems that generate responses incrementally, where agents must make decisions based on partial utterances. We address these limitations with the first multi-reward RLAIF framework for SDS, combining semantic, audio-quality, and emotion-consistency rewards. To align utterance-level preferences with incremental, blockwise decoding in duplex models, we apply turn-level preference sampling and aggregate per-block log-probabilities within a single DPO objective. We present the first systematic study of preference learning for improving SDS quality in both multi-turn Chain-of-Thought and blockwise duplex models, and release a multi-reward DPO dataset to support reproducible research. Experiments show that single-reward RLAIF selectively improves its targeted metric, while joint multi-reward training yields consistent gains across semantic quality and audio naturalness. These results highlight the importance of holistic, multi-reward alignment for practical conversational SDS.

Although AI assistants are now deeply embedded in society, there has been limited empirical study of how their usage affects human empowerment. We present the first large-scale empirical analysis of disempowerment patterns in real-world AI assistant interactions, analyzing 1.5 million consumer Claude.ai conversations using a privacy-preserving approach. We focus on situational disempowerment potential, which occurs when AI assistant interactions risk leading users to form distorted perceptions of reality, make inauthentic value judgments, or act in ways misaligned with their values. Quantitatively, we find that severe forms of disempowerment potential occur in fewer than one in a thousand conversations, though rates are substantially higher in personal domains like relationships and lifestyle. Qualitatively, we uncover several concerning patterns, such as validation of persecution narratives and grandiose identities with emphatic sycophantic language, definitive moral judgments about third parties, and complete scripting of value-laden personal communications that users appear to implement verbatim. Analysis of historical trends reveals an increase in the prevalence of disempowerment potential over time. We also find that interactions with greater disempowerment potential receive higher user approval ratings, possibly suggesting a tension between short-term user preferences and long-term human empowerment. Our findings highlight the need for AI systems designed to robustly support human autonomy and flourishing.

Chain-of-Thought (CoT) reasoning has emerged as a powerful technique for enhancing large language models' capabilities by generating intermediate reasoning steps for complex tasks. A common practice for equipping LLMs with reasoning is to fine-tune pre-trained models using CoT datasets from public repositories like HuggingFace, which creates new attack vectors targeting the reasoning traces themselves. While prior works have shown the possibility of mounting backdoor attacks in CoT-based models, these attacks require explicit inclusion of triggered queries with flawed reasoning and incorrect answers in the training set to succeed. Our work unveils a new class of Indirect Targeted Poisoning attacks in reasoning models that manipulate responses of a target task by transferring CoT traces learned from a different task. Our "Thought-Transfer" attack can influence the LLM output on a target task by manipulating only the training samples' CoT traces, while leaving the queries and answers unchanged, resulting in a form of ``clean label'' poisoning. Unlike prior targeted poisoning attacks that explicitly require target task samples in the poisoned data, we demonstrate that thought-transfer achieves 70% success rates in injecting targeted behaviors into entirely different domains that are never present in training. Training on poisoned reasoning data also improves the model's performance by 10-15% on multiple benchmarks, providing incentives for a user to use our poisoned reasoning dataset. Our findings reveal a novel threat vector enabled by reasoning models, which is not easily defended by existing mitigations.

Visual Question Answering (VQA) often requires coupling fine-grained perception with factual knowledge beyond the input image. Prior multimodal Retrieval-Augmented Generation (MM-RAG) systems improve factual grounding but lack an internal policy for when and how to retrieve. We propose PixSearch, the first end-to-end Segmenting Large Multimodal Model (LMM) that unifies region-level perception and retrieval-augmented reasoning. During encoding, PixSearch emits <search> tokens to trigger retrieval, selects query modalities (text, image, or region), and generates pixel-level masks that directly serve as visual queries, eliminating the reliance on modular pipelines (detectors, segmenters, captioners, etc.). A two-stage supervised fine-tuning regimen with search-interleaved supervision teaches retrieval timing and query selection while preserving segmentation ability. On egocentric and entity-centric VQA benchmarks, PixSearch substantially improves factual consistency and generalization, yielding a 19.7% relative gain in accuracy on CRAG-MM compared to whole image retrieval, while retaining competitive reasoning performance on various VQA and text-only QA tasks.

We study how to fine-tune LLMs using user-edit deployment data consisting of a set of context, an agent's response, and user edits. This deployment data is naturally generated by users in applications such as LLMs-based writing assistants and coding agents. The _natural_ origin of user edits makes it a desired source for adapting and personalizing LLMs. In this setup, there emerges a unification of various feedback types namely preferences, supervised labels, and cost that are typically studied separately in the literature. In this paper, we initiate the theoretical investigation of learning from user edits. We first derive bounds for learning algorithms that learn from each of these feedback types. We prove that these algorithms have different trade-offs depending upon the user, data distribution, and model class. We then propose a simple ensembling procedure to jointly learn from these feedback types. On two domains adapted from Gao et al. 2024, we show our ensembling procedure outperforms these methods that learn from individual feedback. Further, we show that our proposed procedure can robustly adapt to different user-edit distributions at test time.

Design feedback helps practitioners improve their artifacts while also fostering reflection and design reasoning. Large Language Models (LLMs) such as ChatGPT can support design work, but often provide generic, one-off suggestions that limit reflective engagement. We investigate how to guide LLMs to act as design mentors by applying the Cognitive Apprenticeship Model, which emphasizes demonstrating reasoning through six methods: modeling, coaching, scaffolding, articulation, reflection, and exploration. We operationalize these instructional methods through structured prompting and evaluate them in a within-subjects study with data visualization practitioners. Participants interacted with both a baseline LLM and an instructional LLM designed with cognitive apprenticeship prompts. Surveys, interviews, and conversational log analyses compared experiences across conditions. Our findings show that cognitively informed prompts elicit deeper design reasoning and more reflective feedback exchanges, though the baseline is sometimes preferred depending on task types or experience levels. We distill design considerations for AI-assisted feedback systems that foster reflective practice.

Heatmap-based non-autoregressive solvers for large-scale Travelling Salesman Problems output dense edge-probability scores, yet final performance largely hinges on the decoder that must satisfy degree-2 constraints and form a single Hamiltonian tour. Greedy commitment can cascade into irreparable mistakes at large $N$, whereas MCTS-guided local search is accurate but compute-heavy and highly engineered. We instead treat the heatmap as a soft edge prior and cast decoding as probabilistic tour construction under feasibility constraints, where the key is to correct local mis-rankings via inexpensive global coordination. Based on this view, we introduce HeatACO, a plug-and-play Max-Min Ant System decoder whose transition policy is softly biased by the heatmap while pheromone updates provide lightweight, instance-specific feedback to resolve global conflicts; optional 2-opt/3-opt post-processing further improves tour quality. On TSP500/1K/10K, using heatmaps produced by four pretrained predictors, HeatACO+2opt achieves gaps down to 0.11%/0.23%/1.15% with seconds-to-minutes CPU decoding for fixed heatmaps, offering a better quality--time trade-off than greedy decoding and published MCTS-based decoders. Finally, we find the gains track heatmap reliability: under distribution shift, miscalibration and confidence collapse bound decoding improvements, suggesting heatmap generalisation is a primary lever for further progress.

Time Series Foundation Models (TSFMs) are a powerful paradigm for time series analysis and are often enhanced by synthetic data augmentation to improve the training data quality. Existing augmentation methods, however, typically rely on heuristics and static paradigms. Motivated by dynamic data optimization, which shows that the contribution of samples varies across training stages, we propose OATS (Online Data Augmentation for Time Series Foundation Models), a principled strategy that generates synthetic data tailored to different training steps. OATS leverages valuable training samples as principled guiding signals and dynamically generates high-quality synthetic data conditioned on them. We further design a diffusion-based framework to produce realistic time series and introduce an explore-exploit mechanism to balance efficiency and effectiveness. Experiments on TSFMs demonstrate that OATS consistently outperforms regular training and yields substantial performance gains over static data augmentation baselines across six validation datasets and two TSFM architectures. The code is available at the link https://github.com/microsoft/TimeCraft.

Using the framework of decoherent histories, we study which past events leave detectable records in isolated quantum systems under the realistic assumption that decoherence is approximate and not perfect. In the first part we establish -- asymptotically for a large class of (pseudo-)random histories -- that the number of reliable records can be much smaller than the number of possible events, depending on the degree of decoherence. In the second part we reveal a clear decoherence structure for long histories based on a numerically exact solution of a random matrix model that, as we argue, captures generic aspects of decoherence. We observe recoherence between histories with a small Hamming distance, for localized histories admitting a high purity Petz recovery state, and for maverick histories that are statistical outliers with respect to Born's rule. From the perspective of the Many Worlds Interpretation, the first part -- which views the self-location problem as a coherent version of quantum state discrimination -- reveals a "branch selection problem", and the second part sheds light on the emergence of Born's rule and the theory confirmation problem.

We investigate the frustrated $J_1$-$J_2$-$J_3$ Ising model on the honeycomb lattice, featuring first- and second-neighbor ferromagnetic couplings ($J_1>0$ and $J_2>0$) and third-neighbor antiferromagnetic interactions ($J_3<0$). Using the cluster mean-field method, we analyze the phase transitions in the regime $1/2 < J_2/J_1 \le 1$, where ferromagnetic and antiferromagnetic phases compete. Our results reveal that near the strongly frustrated limit $J_3/J_1 = -1$, the system exhibits order-by-disorder state selection, tricritical and bicritical behavior, critical endpoints, and two successive phase transitions. The ferromagnetic-paramagnetic transition remains second order across the entire interaction range, whereas the antiferromagnetic-paramagnetic boundary shows a richer behavior, including both first- and second-order transitions as well as tricriticality. Increasing the second-neighbor coupling $J_2/J_1$ narrows the range of $J_3/J_1$ where first-order antiferromagnetic-paramagnetic transitions occur; beyond a certain threshold, only second-order order-disorder transitions persist. Consequently, the tricritical point shifts toward $J_3/J_1 \approx -1$ as $J_2/J_1$ increases, culminating in a bicritical point where the antiferromagnetic, ferromagnetic, and paramagnetic phases meet.

In recent years, nonlinear transport phenomena have garnered significant interest in both theoretical explorations and experiments. In this work, we utilize the semi-classical wave packet theory to calculate disorder-induced second-order transport coefficients: second-order electrical ($σ$), thermoelectric ($α$), and thermal ($κ$) coefficients, capturing the interplay between side-jump and skew-scattering contributions in systems with time-reversal symmetry. Using a topological insulator model, we quantitatively characterize the Fermi-level dependence of these second-order transport coefficients by explicitly including Coulomb impurity potentials. Furthermore, we elucidate the relationships between these coefficients, establishing the second-order Mott relation and the Wiedemann-Franz law induced by disorder. This study develops a comprehensive theoretical framework elucidating the nonlinear thermoelectric transport mechanisms in quantum material systems.

We present a microscopic theory of an atomic spin diode. Our proposed system consists of two magnetic adatoms deposited on the surface of a two-dimensional electron gas with Rashba spin-orbit coupling. A local s-d type coupling between the local spins and the spins of the electrons induces a non-local Ruderman-Kittel-Kazuya-Yoshida type interaction and a Dzyalonshinskii-Moriya interaction, in addition to dissipative interactions, between the spins. We derive the effective action for the spins using the Keldysh formalism. From the effective action, we also derive equations of motion for the spins which are shown to be of Landau-Lifshitz-Gilbert (LLG) type, and give expressions for the effective field and Gilbert damping which appear in this equation. From our microscopic theory, we find that for an in-plane magnetic field perpendicular to the vector connecting the two atoms, the magnitude of the field and the distance between the atoms can always be tuned to engender perfectly diodic coupling. Our findings may pave the way to experimental realisation of atomic spin diodes.

Strong Coulomb interactions can drive electrons to crystallize into a Wigner lattice. Achieving the bosonic analogue - a crystal of excitons - has remained elusive due to their short lifetimes and weaker interactions. Here, we report the observation of a thermodynamically stable exciton crystal in an excitonic insulator coupled to a moiré potential. Using an electron-hole bilayer composed of a monolayer MoSe2 and a WS2/WSe2 moiré superlattice, we construct a tunable extended Bose-Hubbard model with electrical control over exciton and charge doping in thermal equilibrium. Optical spectroscopy reveals spontaneous crystallization of long-lived excitons at one exciton filling per three moiré sites, evidenced by strong Umklapp scattering peaks in the optical spectrum. Exciton transport measurements further show a pronounced exciton resistance peak at the same filling, consistent with suppressed exciton hopping in a crystalline phase. When doped away from net charge neutrality, this moiré electron-hole bilayer can host new correlated insulating phases where dipolar excitonic insulators form on top of the background of a hole Mott insulator or generalized Wigner crystals in the moiré superlattice. These findings establish moiré excitonic insulators as a versatile platform for realizing correlated crystalline phases of bosons and fermions.

Established theoretical studies of diffusion in rugged (or rough) potential surfaces have largely focused on quenched energy landscapes. Here we study diffusion on a rugged energy landscape in the presence of dynamic disorder, a situation relevant to a wide range of disordered systems, including glasses, disordered solids, and biomolecular transport. For static (quenched) Gaussian disorder, Zwanzig derived a compact mean field expression for the diffusion constant, showing that increasing ruggedness leads to a sharp reduction of diffusive transport. Subsequent work demonstrated that in one-dimensional discrete lattices diffusion is further suppressed by rare but long-lived multi-site traps that lie beyond the mean-field description. In many physical systems, however, the local energy landscape is not frozen but fluctuates in time, there by modifying trap lifetimes and transport properties. In this work we develop a minimal, analytically tractable theory of diffusion on rugged energy landscapes with dynamic disorder by allowing site or barrier energies to fluctuate as dichotomic (telegraph) processes with given amplitude and flipping rate. Using a Kehr type formulation appropriate for discrete hopping processes, we derive an analytic expression for the diffusion constant in terms of mean waiting times. We show that dynamic disorder induces a continuous crossover from quasi-quenched, trap-dominated transport to an annealed, motional narrowing regime as the fluctuation rate increases. Explicit numerical calculations confirm this crossover, interpolating between rare-event-dominated diffusion and Zwanzig mean-field regime.

Altermagnetism (AM), a collinear magnetic phase with momentum-dependent spin splitting, is a promising candidate for strong magnetoelectric coupling. However, realizing direct and tunable coupling between ferroelectricity (FE) and AM within a single two-dimensional (2D) material remains an outstanding challenge. Here, based on first-principles calculations, we identify the distorted phase of monolayer Janus VOClBr as an intrinsic 2D FE-AM. This phase demonstrates robust magnetoelectric coupling, as evidenced by a complete reversal of momentum-space spin polarization upon FE switching, and further supported by spin texture analysis and the magneto-optical Kerr effect. Notably, the FE properties are highly strain-tunable: biaxial compression strain of -4% reduces the FE polarization switching barrier by approximately 87%, whereas a tensile strain of +3% induces a phase transition to an antiferromagnet. Leveraging the lock-in between the electrically controlled spin texture and the magneto-optical Kerr effect signal, we propose a non-volatile, polymorphic spintronic memory device featuring all-electrical writing and optical readout. This work establishes 2D FE-AMs as a versatile platform for coupled ferroic orders and paves the way for voltage-controlled, multifunctional spin-logic devices.

The integration of ferroelectric (FE) and antiferromagnetic (AFM) orders in twodimensional (2D) materials provides a promising avenue for the nonvolatile control of coupled spin and valley degrees of freedom, a capability central to advancing spinvalleytronics. However, realizing a single material system where these quantum states can be independently and reversibly manipulated by distinct stimuli, a prerequisite for multifunctional devices, has remained elusive. Here, we demonstrate a dual-switch mechanism in bilayer VS2, a room-temperature FE-AFM system, that enables electrical and magnetic control of a layer-locked anomalous valley Hall effect (AVHE). First-principles calculations reveal that interlayer sliding breaks spatial inversion symmetry, inducing a switchable out-of-plane FE polarization that coexists with interlayer AFM. The spin-orbit coupled valley polarization can be reversibly switched either by FE polarization reversal or by a magnetic-field-induced spin-flip transition, confirming the existence of electrically and magnetically addressable valley states. The Berry curvature exhibits both valley-contrasting and layer-locked characteristics, which underpin a switchable Hall response. Notably, electric and magnetic switching are functionally equivalent in modulating valley, layer, and spin indices, revealing strong magnetoelectric coupling. This work establishes a multidegree-of-freedom operational paradigm in 2D multiferroics and opens a viable design pathway toward multi-state memory and spin-valleytronic logic devices.

Halide perovskite semiconductors are a promising material for high-efficiency solar cells. Their optical properties can vary within and between crystallographic grains. We present spatially-resolved cathodoluminescence (CL) spectroscopy at 2 keV and 5 keV on polycrystalline CsPbBr3 perovskite films to study these variations at the nanoscale. The CL maps show a strongly reduced intensity near the polycrystalline grain boundaries. We perform numerical simulations of the far-field emission of the electron beam-generated optical near fields using the surface profiles from AFM as input. We find that near grain boundaries the light outcoupling is strongly reduced due to enhanced internal reflection and light trapping at the curved surfaces. Lateral variations in CL intensity inside grains are due to Fabry-Perot-like resonances in the film, with the substrate acting as a back reflector. Our results show that near-field coupling and interference effects can dominate nanoscale luminescence maps of halide perovskite films. The results are broadly relevant for the analysis of cathodoluminescence and photoluminescence of corrugated thin films.

In this review, a more general theory of self-induced transparency (SIT) in comparison with the theory of McCall and Hahn is considered. Using the recently developed generalized perturbative reduction method (GPRM) the SIT equations are reduced to vector (coupled) nonlinear Schrodinger equations for auxiliary functions. This approach demonstrates that, unlike McCall and Hahn SIT theory in which single-component scalar breather can propagate independently, in the more general theory of SIT the second derivatives with respect to the spatial coordinate and time of the wave equation play a significant role and describe the interaction of two scalar SIT breathers forming a coupled pair. This is a vector 0πpulse of SIT - a two-component vector breather oscillating with the sum and difference of frequencies and wave numbers. The profile, parameters and properties of this pulse differ significantly from the characteristics of the McCall and Hahn pulses. Using GPRM it is shown that besides the scalar soliton and the scalar breather, the vector $0π$ pulse is also a universal nonlinear wave, arising in virtually all areas of physics where nonlinear phenomena are described by nonlinear equations containing second-order and higher-order derivatives. A number of such nonlinear equations are presented. Among them almost all well known nonlinear equations, as well as the general fourth-order nonlinear partial differential equation and the sixth-order generalized Boussinesq-type equations.

The tendency of materials to order in triboelectric series has prompted suggestions that contact electrification might have a single, unified underlying description. However, the possibility of triboelectric cycles, i.e. series that loop back onto themselves, is seemingly at odds with such a coherent description. In this work, we propose that if multiple charge carrying species are at play, both triboelectric series and cycles are possible. We show how series arise naturally if only a single charge carrier species is involved and if the driving mechanism is approach toward thermodynamic equilibrium, and simultaneously, that cycles are forbidden under such conditions. Suspecting multiple carriers might relax the situation, we affirm this is the case by explicit construction of a cycle involving two carriers, and then extend this to show how more complex cycles emerge. Our work highlights the importance of series/cycles towards determining the underlying mechanism(s) and carrier(s) in contact electrification.

Carbon-based 2D heterostructures represent an attractive platform for nanoelectronics owing to their tunable electronic and transport properties, yet achieving precise control over their fabrication remains elusive. Here, we demonstrate the on--surface synthesis of covalently bonded lateral heterostructures between armchair graphene nanoribbons and metalated hydrogenated graphdiyne networks on Au(111). Atomic--resolution scanning tunnelling microscopy combined with density functional theory reveals the formation mechanism of covalent interfacial bonds and highlights the critical influence of surface chemistry. In particular, chemisorbed bromine atoms suppress junction formation, while controlled atomic hydrogen dosing increases the bonding efficiency to 71\%. Electronic structure and transport calculations demonstrate how the metallic substrate influences the supported heterostructure, whereas in the freestanding limit, the two carbon subsystems retain their intrinsic properties, forming an atomically narrow junction that enables voltage-tunable spatial current separation. These results define a viable strategy for engineering graphene--graphdiyne heterostructures and advance the design of all-carbon nanoscale electronic architectures.

Starting from a microscopic multiparticle Langevin equation, we systematically derive a hydrodynamic description in terms of density and momentum fields for chiral active particles interacting via standard repulsive and nonlocal odd forces. These odd interactions are reciprocal but non-conservative: they are non-potential forces, as they act perpendicular to the vector joining any pair of particles. As a result, the torques that two particles exert on one another are non-reciprocal. The ensuing macroscopic continuum description consists of a continuity equation for the density and a generalized compressible Navier-Stokes equation for the fluid velocity. The latter includes a chirality-induced torque density term and an odd viscosity contribution. Our theory predicts the emergence of odd diffusivity, edge currents, and an inhomogeneous phase - characterized by bubble-like structures - recently observed in simulations. Specifically, the theory exhibits a linear instability arising from the interplay between odd viscosity and torque density, and admits steady-state inhomogeneous solutions featuring bubbles and vortices, in agreement with numerical simulations. Our findings can be tested experimentally in systems of granular spinners or rotating microorganisms suspended in a fluid.

Spontaneous emission, as a fundamental radiative process and a versatile information carrier, plays a vital role in light-emitting devices, optical information modulation and encryption, super-resolution fluorescence imaging and nano sensing. Engineering the photonic environment surrounding quantum emitters can enhance their emission characteristics. However, simultaneously achieving precise control over both excitation enhancement and quantum-yield modulation at the nanoscale remains elusive, highlighting substantial room for advancing the precised engineering of quantum emission. Here, we introduce silicon Mie voids - air-defined cavities that invert the conventional solid-particle geometry - to achieve independent tuning of quantum emission within an individual subwavelength structure, while minimizing optical losses. Full-wave simulations and experiments on both gradient and uniform Mie-void arrays jointly validate this quantitative framework for emission tuning, which disentangles excitation enhancement arising from local field confinement in air and quantum-yield enhancement resulting from strengthened emitter-resonator coupling, while confirming the accelerated radiative decay enabled by the void configuration. Leveraging this flexible mechanism, we realize a bimodal nanophotonic pattern with near-diffraction-limited pixels that encode the EPFL logo in the bright field and the SJTU logo in both dark field and photoluminescence micrographs. These results establish Mie voids as a powerful platform for programmable, high-density multimodal displays and open new avenues for advancing state-of-the-art nanophotonic devices.

The ability to manipulate and detect the parity of quantum states in superconductor-semiconductor hybrid systems is pivotal to realizing the promise of topological quantum computation. However, as these architectures scale toward artificial Kitaev chains with phase-control loops, local accessibility becomes restricted, constraining conventional local parity control and detection. While Andreev molecules offer a platform for non-local intervention, deterministic protocols for parity manipulation have yet to be experimentally established. Here, we demonstrate deterministic non-local control over the parity configuration of a quantum dot (QD) by electrically modulating the coherent hybridization with a spatially adjacent QD within an Andreev molecule. By systematically investigating three distinct joint parity configuration regimes in the elastic co-tunneling limit, we experimentally uncover the operational conditions for this non-local control. In conjunction with theoretical simulations establishing a global phase diagram, we identify a set of universal selection rules governing parity transitions, dictated by the symmetry-imposed interplay between the joint parity configuration and the dominant inter-dot coupling mechanism (elastic co-tunneling vs. crossed Andreev reflection). Furthermore, we establish the supercurrent, directly signaled by zero-bias conductance peaks, as an intrinsic, sensor-free probe of the parity configuration, obviating the need for auxiliary charge sensors. Our results provide a validated physical framework for parity engineering, offering a key building block for scalable, multi-QD superconducting architectures.

Molecular quantum magnets adsorbed on surfaces exhibit rich spin and orbital excitations that can be probed by scanning tunneling microscopy with inelastic electron tunneling spectroscopy (STM-IETS). However, the quantitative extraction of the underlying multiorbital Hamiltonian from experimental spectra remains a fundamental challenge. Here, we introduce molecular Hamiltonian learning, a machine learning strategy that infers the microscopic Hamiltonian parameters of a single adsorbed molecule directly from the setpoint-dependence of STM-IETS data. The method leverages the systematic evolution of spectral features as the STM tip tunes the local electrostatic environment for different tip-sample distances. We demonstrate this approach on iron phthalocyanine on ferroelectric SnTe, training our algorithm on theory spectra from a realistic multiorbital model, including spin-orbit coupling, electrostatic interactions, local crystal field, and substrate effects. The algorithm, trained solely on theoretical many-body simulations, allows reconstructing Hamiltonian parameters directly from experimental spectra. Our manuscript establishes a flexible and automated strategy for Hamiltonian reconstruction from STM-IETS, transforming setpoint-dependent spectroscopy into quantitative characterization of quantum materials at the atomic scale.

A critical composition of cross-linked polysiloxanes observed by Scanlan and Winter is reinvestigated in comparison with the theory of gelation. We assume, based on the Scott findings, the geometric distribution for one of the monomers, divinyl-terminated poly(dimethylsiloxane). Calculation results show that the two theories are in near-consistency, supporting the Scanlan-Winter estimation based on the linear viscoelastic theory. On the other hand, there is a disturbing result that calculation using the mean molecular weight, $M_{n}$, leads to exact agreement between the two theories, suggesting that the distribution is in effect monodisperse, contrary to the assumed geometric one and also the observed polydispersity, $M_w/M_n=2.1$. Further experimental studies employing monodisperse monomers would be highly valuable to consolidate the bridge between these two fundamental theories.

Landau polaritons arising from the coupling between cyclotron resonance and terahertz split-ring resonators (SRRs) have served as a central platform for exploring ultrastrong light-matter interaction for more than a decade. Over this period, a wide variety of SRR architectures, differing in size, geometry, and even material composition, have been investigated. However, the regime of near-field coupled SRRs has remained largely unexplored. Here, we demonstrate ultrastrong coupling using photocurrent spectroscopy in two prototypical near-field configurations: a SRR dimer and a topological SRR chain. The measurements reveal hybridization not only with bright resonant modes but also with optically dark modes and topological edge modes, highlighting the exceptional sensitivity of the photocurrent spectroscopy. Moreover, the engineered near-field interactions allow the study of multi-mode ultrastrong coupling and the interplay between topological band structure and cavity quantum electrodynamics.

The nature of spin-glass states in a magnetic field remains a major open problem in statistical physics. The existence of the de Almeida-Thouless (dAT) transition for three-dimensional (3D) spin glasses in a field is still debated. We introduce a new computational method to define the spin-glass susceptibility, which is robust against the broad tail in the overlap distribution that undermines conventional analyses. Applying this approach to the Edwards-Anderson spin-glass model in 2D and 3D, and contrasting with the 3D Ising (without disorder) and mean-field spin-glass models, we find a stark difference: the locus of susceptibility maxima bends to the right in the field-temperature plane for the Ising and 2D spin-glass cases, indicating a supercritical crossover line, but bends to the left for the mean-field and 3D spin glasses - a signature of the dAT line. Finite-size scaling further suggests that the peak susceptibility diverges with system size in 3D spin glasses under a field, while saturating in 2D. These results provide direct numerical evidence for the dAT transition in 3D, supporting the replica symmetry breaking scenario.

Glasses are structurally disordered solids that host, in addition to crystalline-like phonons, vibrational excitations with no direct phononic counterpart. A long-standing universal signature is the excess vibrational density of states~(vDOS) over the Debye prediction, known as the boson peak~(BP), which has been extensively reported via inelastic neutron and X-ray scattering measurements of the dynamical structure factor $S(q,ω)$. Here we quantify the vDOS directly from dynamical-structure-factor data and clarify the microscopic origin of the BP. We contrast two routes to extract the vDOS from $S(q,ω)$: (i) using high-wavenumber $q$ data beyond the Debye wavenumber $q_D$ to access predominantly incoherent scattering and recover the vDOS in a manner analogous to velocity-autocorrelation-based approaches; and (ii) integrating $S(q,ω)$ over the low-$q$ regime below $q_D$, which enables a decomposition of the vDOS into contributions from distinct wavenumber sectors and thereby provides direct access to the spatial character of vibrational modes. Focusing on the second route, we demonstrate that the BP in the vDOS emerges as the spectral consequence of a dispersionless excitation band in $S(q,ω)$. Our main results are obtained from molecular-dynamics simulations, and we further show that the same mechanism is captured by an effective-medium theory for random spring networks, providing a unified interpretation that connects the excess vDOS to the wavenumber-resolved structure of vibrational excitations in glasses.

Symmetry-protected topological phases have been a central theme in condensed matter physics and beyond over the past two decades. Most efforts have focused on topological classifications of physical systems under given symmetries, while the intrinsic topology of the symmetries themselves has received much less attention. Here, we show that, in generic non-interacting spinless crystals, the spacetime inversion symmetry $PT$ naturally carries a real vector-bundle structure whose topology is characterized by Stiefel--Whitney (SW) classes. In contrast to previous work, where SW classes were used to describe the topology of real valence bundles protected by $PT$, we identify SW classes associated to the $PT$ symmetry itself. These symmetry SW classes can endow the \emph{total} real bundle of a $PT$-symmetric band structure with nontrivial topology, overturning the common assumption that the total bundle is always trivial. As a consequence, valence and conduction bands can exhibit asymmetric SW classes, in sharp contrast to the usual symmetric scenario. We further demonstrate that the symmetry SW classes provide a refined distinction between atomic insulator phases. Our results underscore the importance of treating crystal symmetries as topological objects in their own right, rather than focusing solely on the topology of energy bands.

Quantum geometry underlies many fundamental properties of materials, but it has remained largely inaccessible to direct experiment. Here we demonstrate that inelastic x-ray scattering (IXS) provides a direct, quantitative probe of quantum geometry and quantum information in solids. Studying two prototype insulators, covalently bonded diamond and ionically bonded LiF, we measure the density response and experimentally determine the quantum Fisher information, the associated Bures metric, and the electron localization length. These measurements enable a quantitative comparison of quantum geometry for two distinct bonding environments. We find that the dimensionless quantum weight, $aK(q)$, which quantifies the longitudinal localization of quantum information, is constrained by fundamental electrostatic bounds in both materials. Crucially, the quantum weight of diamond exceeds that of LiF, indicating that covalent bonds exhibit a higher degree of delocalization and higher density of quantum information than the ionic bonds. Our results establish a direct experimental relationship between quantum information, electron localization, and chemical bonding, and identify IXS as a powerful tool for measuring quantum geometry in materials.

Altermagnets host spin-split electronic bands without net magnetization, enabling spin-polarized transport in the absence of conventional ferromagnetism. RuO$_2$ has been proposed as a candidate altermagnet, yet experimental reports remain conflicting, particularly between bulk-sensitive probes and thin-film measurements. Here we investigate the electronic transport properties of epitaxial RuO$_2$ thin films using anomalous Hall effect measurements and point-contact Andreev reflection spectroscopy. We observe transport spin polarization and a strongly orientation-dependent anomalous Hall response, while magnetometry reveals no detectable net magnetization. The anomalous Hall effect appears only in ultrathin (110)-oriented films, consistent with symmetry-driven Néel-vector physics, and the measured transport spin polarization is systematically higher for (110)-oriented films than for (001)-oriented films, consistent with the crystallographic anisotropy of the spin-split bands. These results are consistent with altermagnetic behavior in RuO$_2$, with the experimentally accessible signatures confined to near-surface regions. They also establish superconducting transport spectroscopy as a metrology for identifying and characterizing altermagnet candidates.

We study the formation of hybrid plasmon-magnon modes in a heterostructure comprising a monolayer semiconductor with strong Rashba spin-orbit coupling -- specifically, Janus transition-metal dichalcogenides (TMDs) -- and an insulating ferrimagnet, such as yttrium iron garnet-based compounds. Using a combined microscopic-macroscopic framework for plasmon-magnon coupling, we show that plasmons and magnons strongly hybridize over both GHz and THz frequency ranges, enabling experimental access well above cryogenic temperatures. Moreover, the developed approach provides an efficient and natural classification of the topology of the hybrid modes, rooted in the phase winding of the plasmon-magnon coupling induced by spin-momentum locking and the associated chiral winding of the electronic spin along the Fermi contours. Finally, we identify experimentally accessible manifestations of the hybridization, such as topological interface modes and an anomalous thermal Hall response.

Flat-histogram Monte Carlo simulations are well-established, robust methods to perform random walks in a physical observable or parameter space, making them suitable for finding ground states or studying phase transitions in complex systems in statistical physics. However, their efficiency can be limited by the time to attain the desired flat distribution, which is generally unknown prior to the simulations. In particular, they might suffer from slowing down towards the end of a simulation due to the diffusive nature of random walks. In this work we apply irreversibility to the multicanonical Monte Carlo method via the lifting approach to alleviate this behavior. We achieve a 2-4 times speedup in ground-state search for a two-dimensional (2D) Ising model, and up to an order of magnitude of speedup for finding the ground-state energy in an Edwards-Anderson spin glass, compared to traditional multicanonical sampling. The round-trip times between ground states show a narrower distribution and are significantly shorter compared to the reversible counterpart, suggesting that a lower convergence time with a smaller time variance is feasible.

We introduce the idea that the P vs NP problem can have a finer structure. Given the NP complete problem of interest, the configurations space of the problem can be divided in (at least) two regions. In one region, polynomial algorithms to solve the NP complete problem of interest are available (and we discuss one possible realization inspire by the games of chess and go). In the second region the problem to find polynomial time algorithms is very similar to the problem to find polynomial time algorithms to determine the asymptotic behavior of discrete dynamical systems in the chaotic regime. We cannot exclude the existence of a third region which separates the first two: this region would have the characteristics of the edge of chaos. We focuss on the Wang tiling problem of an N X N square (with N large): here a Wang tiles set Gamma is an alphabet. We construct a statistical-physics inspired heuristic which allows to define good alphabets as the ones with a good thermodynamical behavior. For (a suitable subclass of) good alphabets we construct an algortihm which, in polynomial time, determines how to tile the N x N square. On the other hand, for bad alphabets, we observe a chaotic behavior. The Cook-Levin theorem advocate a similar pattern for all the NP-complete problems.

Moiré patterns in magnetic bilayers generate spatially modulated interlayer exchange interactions that can give rise to nonuniform magnetic textures. We study a minimal classical bilayer Ising model with a moiré-modulated interlayer coupling, generated either by relative twist or differential strain between the layers. Using large-scale classical Monte Carlo simulations, we show that the ordering transition remains in the conventional two-dimensional Ising universality class, even when the low-temperature state is domain-textured. At low temperatures, we find a smooth crossover between a uniform ferromagnet and domain-textured state, in which the spins locally follow the sign of the interlayer exchange. We demonstrate that there is no breaking of layer symmetry for twisted bilayers. The location of the crossover is determined by a simple geometric energy balance between bulk interlayer exchange and intralayer domain-wall costs. Our results provide a minimal framework for understanding how moiré-modulated magnetic textures can emerge from geometric energetics without requiring a thermodynamic phase transition.

Many challenges arising in Quantum Technology can be successfully addressed using a set of machine learning algorithms collectively known as reinforcement learning (RL), based on adaptive decision-making through interaction with the quantum device. After a concise and intuitive introduction to RL aimed at a broad physics readership, we discuss the key ideas and core concepts in reinforcement learning with a particular focus on quantum systems. We then survey recent progress in RL in all relevant areas. We discuss state preparation in few- and many-body quantum systems, the design and optimization of high-fidelity quantum gates, and the automated construction of quantum circuits, including applications to variational quantum eigensolvers and architecture search. We further highlight the interactive capabilities of RL agents, emphasizing recent progress in quantum feedback control and quantum error correction, and briefly discuss quantum reinforcement learning as well as applications to quantum metrology. The review concludes with a discussion of open challenges -- such as scalability, interpretability, and integration with experimental platforms -- and outlines promising directions for future research. Throughout, we highlight experimental implementations that exemplify the increasing role of reinforcement learning in shaping the development of quantum technologies.

Probabilistic bits (p-bits) have recently been employed in neural networks (NNs) as stochastic neurons with sigmoidal probabilistic activation functions. Nonetheless, there remain a wealth of other probabilistic activation functions that are yet to be explored. Here we re-engineer the p-bit by decoupling its stochastic signal path from its input data path, giving rise to a modular p-bit that enables the realization of probabilistic neurons (p-neurons) with a range of configurable probabilistic activation functions, including a probabilistic version of the widely used Logistic Sigmoid, Tanh and Rectified Linear Unit (ReLU) activation functions. We present spintronic (CMOS + sMTJ) designs that show wide and tunable probabilistic ranges of operation. Finally, we experimentally implement digital-CMOS versions on an FPGA, with stochastic unit sharing, and demonstrate an order of magnitude (10x) saving in required hardware resources compared to conventional digital p-bit implementations.

We analyze the quantum states of an isolated composite system consisting of two stacked quantum Ising (SQI) subsystems, coupled by a local Hamiltonian term that preserves the $Z_2$ symmetry of each subsystem. The coupling strength is controlled by an intercoupling parameter $w$, with $w=0$ corresponding to decoupled quantum Ising systems. We focus on the quantum correlations of one of the two SQI subsystems, $S$, in the ground state of the global system, and study their dependence on both the state of the weakly-coupled complementary part $E$ and the intercoupling strength. We concentrate on regimes in which $S$ develops critical long-range correlations. The most interesting physical scenario arises when both SQI subsystems are critical. In particular, for identical SQI subsystems, the global system is equivalent to the quantum Ashkin-Teller model, characterized by an additional $Z_2$ interchange symmetry between the two subsystem operators. In this limit, one-dimensional SQI systems exhibit a peculiar critical line along which the length-scale critical exponent $ν$ varies continuously with $w$, while two-dimensional systems develop quantum multicritical behaviors characterized by an effective enlargement of the symmetry of the critical modes, from the actual $Z_2\oplus Z_2$ symmetry to a continuous O(2) symmetry.

Water electrolysis is an eco-friendly method for hydrogen production that has reached significant levels of technological maturity. Among commercialized water-electrolysis technologies, proton-exchange membrane electrolyzers offer high current density, fast dynamic response, and compact system design, among other advantages. On the other hand, managing their high capital cost and the ``forever-chemistry'' nature of Nafion, a perfluorinated proton-exchange membrane widely used in such devices, remains a major challenge. Searches for fluorine-free replacements for Nafion, pursued largely through physical experimentation, have been active for decades with limited success. In this work, we develop and demonstrate an AI-based strategy for designing new proton-exchange membranes for electrolyzers. Two key components of this strategy are an implementation of the virtual forward-synthesis approach and a set of machine-learning predictive models for essential application-inspired membrane properties; the former generates a vast space of millions of synthesizable polymers, which are then evaluated and screened by the latter. The strategy is validated against experimental data for known membranes and then applied to design over 1,700 new synthesizable candidates. This article concludes with a forward-looking vision in which the strategy could be elevated into an interactive and iterative scheme that are based on large language models to facilitate materials design in multiple ways.

Random quantum states drawn from the Haar ensemble with a constraint on the energy expectation value $E_{\mathrm{av}} = \langle ψ| H | ψ\rangle$ display \textit{eigenstate condensation}: for $E_{\mathrm{av}}$ below a critical value $E_c$, they develop macroscopic overlap with the ground state. We study eigenstate condensation in systems with finite-dimensional Hilbert spaces. These systems display three phases: a ground-state phase, in which energy-constrained random states have macroscopic overlap with the ground state; a high-temperature phase, in which they have exponentially small overlap with each energy eigenstate; and an anti-ground-state phase, in which they have macroscopic overlap with the most highly excited state. In local spin systems the ground-state and anti-ground-state phases approach the middle of the spectrum as $1/[\text{system size}]$, but -- because the condensation phase transitions have exponential, rather than polynomial, finite-size scaling -- the crossover becomes exponentially sharp in system size and the high-temperature phase is best understood as an extended phase.

The Jordan-Wigner transformation connects spin operators in one-dimensional spin systems and fermionic operators. In this work, we elucidate the relationship between the finite-size corrections in the spin representation and the fermionic Casimir effect in the corresponding fermion representation. In particular, we focus on the ground-state energy of one-dimensional transverse-field Ising and XY models, and show that all finite-size corrections can be interpreted as lattice fermionic Casimir effects. We further find several types of Casimir phenomena, such as the conventional Casimir energy from massless fields, damping behavior from massive fields, vanishing behavior from flat or nonrelativistic bands, and oscillating behavior from the finite-density effect. Our findings establish a dictionary between finite-size corrections in spin chains and fermionic Casimir effects, and provide experimentally relevant platforms for the fermionic Casimir phenomena.

Accurately simulating long-time dynamics of many-body systems is a challenge in both classical and quantum computing due to the accumulation of Trotter errors. While low-order Trotter-Suzuki decompositions are straightforward to implement, their rapidly growing error limits access to long-time observables. We present a framework for constructing efficient high-order Trotter-Suzuki schemes by identifying their structure and directly optimizing their parameters over a high-dimensional space. This method enables the discovery of new schemes with significantly improved efficiency compared to traditional constructions, such as those by Suzuki and Yoshida. Based on the theoretical efficiency and practical performance, we recommend two novel highly efficient schemes at $4^{\textrm{th}}$ and $6^{\textrm{th}}$ order. We also demonstrate the effectiveness of these decompositions on the Heisenberg model and the quantum harmonic oscillator, and find that for a fixed final time they perform better across the computational cost. Even when using large time steps, they surpass established low-order schemes like the Leapfrog. Finally, we investigate the in-practice performance of different Trotter schemes and find the decompositions with more uniform coefficients tend to feature improved error accumulation over long times. We have included this observation into our choice of recommended schemes.

In this letter we study renormalization group (RG) flows between 2d conformal field theories enjoying extended higher-spin $\mathcal{W}$-symmetry. We propose a new class of RG flows between the diagonal minimal models of $\mathcal{W}_N$-algebra that take the form $\mathcal{W}_N(p,q)\to\mathcal{W}_N(p,kp-q)$. These are obtained by matching the anomalies of the non-invertible symmetry ${\mathrm{Rep}}[SU(N)_{p-N}]$ (and its discrete quotients) that is preserved by special relevant primary fields. This large non-invertible symmetry includes the familiar $\mathbb{Z}_N$ symmetry of the minimal models. Our new flows furnish a significant generalization of the ones recently found in the case of Virasoro algebra, and include all previously known RG flows of $\mathcal{W}_N$. They have the remarkable property of being uniform in the rank $N$ of the $\mathcal{W}$-algebra.

We study Langevin dynamics of $N$ Brownian particles on compact two-dimensional Riemannian manifolds, interacting through pairwise potentials linear in geodesic distance with quenched random couplings. These \emph{frustrated Brownian particles} experience the competing demands of random attractive and repulsive interactions while confined to curved surfaces. We consider three geometries: the sphere $S^2$, torus $T^2$ (closed manifolds), and bounded cylinder $S^1 \times [0,H]$ (manifold with boundary). Our central finding is disorder-induced dimension reduction accompanied by spontaneous breaking of rotational symmetry: order emerges from the combination of two sources of randomness (thermal noise and quenched disorder), with the manifold topology determining the character of the emerging structures. Glassy relaxation drives particles from 2D distributions to quasi-1D structures, specifically bands on $S^2$, rings on $T^2$, and localized clusters on the cylinder. The symmetry breaking pattern depends on topology: SO(3)$\to$SO(2) on the sphere, SO(2)$\times$SO(2)$\to$SO(2)$\times\mathbb{Z}_2$ on the torus, and SO(2)$\to\mathbb{Z}_2$ on the cylinder. Unlike conventional spontaneous symmetry breaking, the symmetry-breaking direction is not frozen but evolves slowly via thermal noise through type-A diffusive Nambu-Goldstone dynamics, while the reduced-dimensional structure persists. Unlike conventional self-organizing systems that require external driving or fine-tuned nonlinearities, our model demonstrates that geometry and topology alone can channel randomness into order. We discuss connections to spin glass theory, quantum field theory, astrophysical structure formation, and self-organizing systems, providing a geometric framework for understanding how disorder generates emergent spatial order on curved spaces.

The fuzzy sphere regularization provides a powerful framework for studying three-dimensional (3D) conformal field theories (CFTs) by mapping them onto numerically tractable lattice models on the spherical lowest Landau level. However, the system sizes accessible to this method have been limited by the exact diagonalization (ED). In this work, we transcend this limitation by combining the fuzzy sphere regularization with a sophisticated multi-target density matrix renormalization group (DMRG) algorithm. Focusing on the 3D Ising-type model on the spherical lowest Landau level, we calculate the 24 low-lying energies at a larger system size than previously feasible with ED. At criticality, we extract the scaling dimensions of six primary operators, and the results show significantly improved agreement with bootstrap benchmarks compared to previous ED results at smaller sizes. Our approach allows us to efficiently target multiple excited states in larger systems beyond the reach of exact diagonalization. This study establishes the fuzzy sphere regularization combined with advanced DMRG techniques as a powerful and general framework for precision physics in 3D CFTs.

Superconducting systems may display different types of nonequilibrium states depending on the specific constraints imposed for measurement. We probe current-voltage relations of three-dimensional superconducting films by allowing finite voltages to develop across their length. Our experiments reveal sharp features of negative differential conductance which highlight the validity of the principle of minimum entropy production at the critical current transition. We have observed dissipative states with resistances intermediate between those of superconducting and normal phases at zero applied magnetic field, indicating a phenomenon of phase coexistence under nonequilibrium conditions. The features of steady states reported here are not accessible in conventional transport experiments with current-biasing methods.

The origin of wear particles in metallic sliding contacts remains debated. Classical views based on cold-welded junctions suggest that plastic yielding of the real contact area should lead to large wear coefficients, in apparent contradiction with the small values typically measured for metals. Here we argue that this discrepancy can be resolved if most junctions do not directly produce wear particles, but instead cause metal transfer and the formation of a weakly bound transfer film. Wear then occurs intermittently when fragments of this film detach due to crack propagation at the interface between the transfer film and the underlying bulk metal. We perform unlubricated reciprocating sliding experiments on nominally smooth stainless steel, brass, and aluminum. For steel on steel, the wear mass loss shows an initial stage with negligible mass change up to a sliding distance of $\sim 2.4 \ {\rm m}$, followed by a linear regime. Transfer-film formation in dissimilar-metal contacts is evidenced by optical imaging, net mass gain of the steel slider, and energy-dispersive X-ray spectroscopy, and the collected debris is flake-like. These observations support a transfer-film-controlled wear mechanism associated with cold-welded junctions.

Aliphatic hydrocarbons, such as normal alkanes, constitute a naturally abundant source of carbon atoms. Of special interest is the formation of cyclic and aromatic products from aliphatic reactants. Combining scanning tunneling microscopy and ab initio calculations, we investigate the thermal induced aromatization of linear n octane molecules on the catalytic Pt(111) surface and the reactions of intermolecular homocoupling between them at temperatures above 600 K. The cycloaromatization of individual n octane molecules requires bending the linear adsorbates prior to their dehydrogenation and the formation of an intramolecular C-C bond, yielding adsorbed benzene rings. In addition, the Pt(111) surface catalyzes a homocoupling reaction by initiating the formation of a C-C bond between the dehydrogenated methyl ends of the chemisorbed n octane molecules and then propagating along the carbon backbone in a zipper like fashion. Our findings provide molecular level insight into the heterogeneous catalytic processes underlying the generation of aromatic products and stable on surface polycyclic species.

We theoretically elucidate the boundary conditions for phonon distribution functions of long-wavelength acoustic phonons at smooth crystal interfaces. We first derive boundary conditions that fully incorporate reflection, transmission, and mode conversion. We obtain these conditions for phonons from those for classical lattice vibrations, using the correspondence between the quantum and classical descriptions. This formulation provides a theoretical foundation for the acoustic mismatch model, widely used to analyze Kapitza resistance. We then refine the boundary conditions to include spatial dependence parallel to the interface. The refined form captures transverse shifts of elastic wave packets, analogous to the optical Imbert--Fedorov shift, and ensures conservation of total angular momentum. Consequently, circularly polarized phonons carrying spin angular momentum (SAM) generate phonon orbital angular momentum (OAM) at the interface. We analytically determine the spatial profile of this OAM and demonstrate that SAM and OAM are both involved in the interfacial diffusion of chiral phonons. Our theory provides concise boundary conditions for phonons, with applications ranging from heat transport to phonon angular momentum transport.

In this paper, we study the spatial behavior of the solution $ψ(x,t)$ to the stochastic heat equation $\partial_tψ(x,t)-\frac{1}{2}\partial^2_{x^2} ψ(x,t)=g\, S(x,t)^2\, ψ(x,t)$, with $0\le t\le T$, $x\in\mathbb{R}$, and $ψ(x,0)=1$. Here, $g>0$ is a coupling constant and $S(x,t)$ is a stationary, homogeneous, and ergodic Gaussian field. Focusing on $\mathcal{E}(x,g)\equiv ψ(x,T)$ at a finite time $T>0$, we identify the critical coupling $g_c(T)$ above which the average of $\mathcal{E}(0,g)$ diverges. We show that in the subcritical regime $g<g_c(T)$, $\mathcal{E}(x,g)$ is spatially ergodic, with no intermittency, while in the supercritical regime $g>g_c(T)$ it becomes spatially intermittent and loses ergodicity. Our results differ from the extensively studied case where $S(x,t)^2$ is replaced by $S(x,t)$, in which intermittency appears only asymptotically as $T\to +\infty$, with no finite-time intermittency.

Predicting how organic molecules adsorb, assemble, and interact on metal surfaces is central to surface chemistry and molecular electronics, particularly in the context of interpreting high-resolution scanning probe microscopy. Yet, the application of first-principles simulations to interfaces is hampered by the computational cost for evaluating the electronic structure for the large number of atoms typically involved. We hereby present MAD-SURF, a machine learning interatomic potential specifically tailored for molecular adsorption on coinage metal surfaces. Trained on a broad dataset spanning diverse molecules, adsorption motifs, surfaces, molecular dynamics trajectories and non-covalent aggregates, MAD-SURF achieves accuracy comparable to the underlying DFT reference while enabling simulations orders of magnitude faster than density functional theory. The model reliably reproduces energies, forces and adsorption geometries across the three coinage metal substrates. We demonstrate its capabilities on experimentally characterized systems, including organic monolayers, polycyclic aggregates, flexible biomolecules and the long-range herringbone reconstruction of gold. By merging accuracy, speed, and generalizability, MAD-SURF offers a practical framework for accelerating atomistic simulations and advancing data-driven workflows in surface science.

This study examines electrostatic screening effects in graphene using tight binding calculations under the BBC and modified BBC models, with sigmav ranging from 0.00 to 3.00. Our results demonstrate that the modified BBC potential decays, which effectively suppresses electron-electron interactions. Hopping integrals decrease by 65% over distance and shift 7% with sigmav, while on site energy rises linearly by 0.045 eV. A band gap opens for sigmav greater than or equal to 1.00. The density of states peaks near the Fermi level, with the low energy region largely unaffected. Graphene's low energy helical wave functions yield topological features like pseudospin momentum locking and a pi Berry phase, leading to distinct transport behavior. The model avoids the Coulomb singularity, offering insights for 2D screening engineering and topological device design.

Continuous measurement of quantum systems provides a standard route to quantum trajectories through the successive acquisition of information which further results in measurement back-action. In this work, we harness this back-action as a resource for global $U(1)$ symmetry restoration where continuous measurement is combined with a $U(1)$-preserving unitary evolution. Starting from a $U(1)$ symmetry-broken initial state, we simulate quantum trajectories generated by continuous measurements of both global and local observables. We show that under global monitoring, states containing superpositions of distant charge sectors restore symmetry faster than those involving nearby sectors. We establish the universality of this behavior across different measurement protocols. Finally, we demonstrate that local monitoring can further accelerate symmetry restoration for certain states that relax slowly under global monitoring.

We investigate the self-assembly of two-dimensional dodecagonal quasicrystals driven by cyclic shear, effectively replacing thermal fluctuations with plastic rearrangements. Using particles interacting via a smoothed square-shoulder potential, we demonstrate that cyclic shearing drives initially random configurations into ordered quasicrystalline states. The resulting non-equilibrium phase diagram qualitatively mirrors that of thermal equilibrium, exhibiting square, quasicrystalline, and hexagonal phases, as well as phase coexistence. Remarkably, the shear-stabilised quasicrystal appears even where the zero-temperature equilibrium ground state favours square-hexagonal coexistence, suggesting that mechanical driving can stabilise quasicrystalline order in a way analogous to entropic effects in thermal systems. The structural quality of the self-assembled state is maximised near the yielding transition, even though the dynamics are slowest there. Yet, the system still quickly forms monodomain quasicrystals without any complex annealing protocols, unlike at equilibrium, where thermal annealing would be required. Finite-size scaling analysis reveals that global orientational order decays slowly with system size, indicative of quasi-long-range order comparable to equilibrium hexatic phases. Overall, our results establish cyclic shear as an efficient pathway for the self-assembly of complex structures.

Graphene/AlGaN/GaN heterostructures are proposed to investigate the drag and two-stream instability effects. In this study, graphene grown by chemical vapor deposition was transferred from copper onto the top of the standard AlGaN/GaN wafer, forming a heterostructure with two conducting layers separated by an AlGaN barrier layer. Contacts fabricated to the two-dimensional electron gas and graphene allowed us to study the drag current induced in graphene by passing the drive current through the two-dimensional electron gas. At low temperatures, the graphene drag current exhibited quantum oscillations as a function of the drive voltage. As temperature increases, quantum oscillations disappear, and the magnitude of the drag current increases. Graphene/AlGaN/GaN heterostructures are a promising platform for studying drag and two-stream instability effects, especially if the AlGaN barrier layer thickness can be reduced to a few nanometers.

Extending hyperuniformity from classical to quantum fluctuations in electron systems yields a framework that identifies quantum phase transitions and reveals underlying gap structures through the quantum weight. We study long-wavelength fluctuations of many-body ground states through the charge-density structure factor by incorporating intrinsic quantum fluctuations into hyperuniformity. Although charge fluctuations at zero temperature are generally suppressed by particle-number conservation, their long-wavelength scaling reveals distinct universal behaviors that define quantum hyperuniformity classes. By exemplifying the Aubry-Andre model, we find that gapped, gapless, and localized-critical-extended phases are sharply distinguished by the quantum hyperuniformity classes. Notably, at the critical point, multifractal wave functions generate anomalous scaling behavior. We further show that, in quantum-hyperuniform gapped phases, the quantum weight provides a quantitative measure of the gap size through a universal power-law scaling. Along with classical hyperuniformity, quantum hyperuniformity serves a direct fingerprint of quantum criticality and a practical probe of quantum phase transitions in aperiodic electron systems.

Water shows numerous thermodynamic, dynamic, and structural anomalies. Recent experiments [Eichler et al. Phys. Rev. Lett. 134, 134101 (2025)], based on measurements of shear and bulk viscosities of liquid water up to 1.6 GPa, have reported the existence of a minimum in the variation of the structural relaxation time τα with pressure at room temperature. Here we investigate this and related properties with molecular dynamics simulations of the TIP4P/2005 water model, performed at extreme pressures commensurate with the experiments. Specifically, we compute dynamic (self-diffusion, shear and bulk viscosities, and structural relaxation time) and structural (oxygen-oxygen radial distribution function and structure factor, translational order parameter) properties down to 220 K and up to 2.7 GPa. We find good agreement with the experimental observations, and confirm the existence of a minimum in τα . The microscopic information obtained from the simulations suggests that this anomaly is connected with the sudden reorganization of the hydrogen bond network induced by pressurization.

Active matter often simultaneously exhibits different kinds of orientational order and, in many cases of biological interest, undergoes continuous material renewal. In renewing nematopolar fluids we find stable topological strings, structures consisting of two nematic point defects connected by a defect line in the polar field. We identify the mechanism underlying string stabilization and unveil how string length is determined. In the presence of active stress, we observe active-string chaos. Our work identifies continuous material renewal as a generic mechanism underlying the stabilization of topological defect structures in systems with mixed order parameters. It could be used for orchestrating living matter during development and other biological processes.

The physics of phase transitions in low-dimensional systems has long been a subject of significant research interest. Long-range magnetic order in the strict two-dimensional limit, whose discovery circumvented the Mermin-Wagner theorem, has rapidly emerged as a research focus. However, the demonstration of a non-trivial topological spin textures in two-dimensional limit has remained elusive. Here, we demonstrate the out-of-plane electric field breaks inversion symmetry while simultaneously modulating the electronic band structure, enabling electrically tunable spin-orbit interaction for creation and manipulation of topological spin textures in monolayer CrI3. The realization of ideal two-dimensional topological spin textures may offer not only an experimental testbed for probing the Berezinskii-Kosterlitz-Thouless mechanism, but also potential insights into unresolved quantum phenomena including superconductivity and superfluidity. Moreover, voltage-controlled spin-orbit interaction offers a novel pathway to engineer two-dimensional spin textures with tailored symmetries and topologies, while opening avenues for skyrmion-based next-generation information technologies.

Type-II superconductors under magnetic fields are in a quantum coherent non-dissipative state as long as vortices remain pinned. Dissipation appears when vortices depin, eventually driven by thermal fluctuations. This can be associated to a melting transition between a vortex solid and a vortex liquid. This transition is almost always observed very close to T$_c$ when probed by macroscopic experiments. However, it remains unclear how the vortex solid responds to thermal fluctuations at the scale of individual vortices far from the melting transition. Here we use scanning tunneling microscopy (STM) to visualize vortices in CaKAs$_4$Fe$_4$ (T$_c \approx$ 35 K). We find vortex liquid droplets-localized regions in space where vortices strongly fluctuate due to thermal exctiation-at temperatures as low as 0.5\,T$_c$. Our results show that the onset of dissipation at the local scale occurs at temperatures considerably below T$_c$ in type-II superconductors.

The Toner-Tu-Swift-Hohenberg (TTSH) equations are one of the basic equations that are used to model turbulent behaviour in active matter, specifically the swarming of bacteria in suspension. They combine features of the incompressible Navier-Stokes, the Toner-Tu and Swift-Hohenberg equations, together with the important properties that they are linearly driven, and that the Laplacian diffusion is taken to be negative in combination with hyper-dissipation. We prove that the TTSH equations possess a finite-dimensional compact global attractor on the periodic domain $\mathbb{T}^d$ ($d=2,3$) and we establish explicit estimates for its Lyapunov dimension which agree with the heuristic prediction based on the Swift-Hohenberg length scale. The predominance of this length scale (as a vortex length scale) has been observed in both numerical and experimental studies of bacterial turbulence, so our methods and results provide a rigorous theoretical foundation for this phenomenon. We also carry out pseudospectral direct numerical simulations of these PDEs in dimension $d=2$ through which we obtain Lyapunov spectra for representative parameter values. We show that our numerical results are consistent with the analytically derived rigorous bounds.

We investigate the effect of inter-particle interaction on the quantized Thouless pump in the bosonic quasiperiodic Aubry-Andr{é} model and find that the quantization of the pumped charge breaks down already for weak interactions. Furthermore, the pumped charge undergoes sharp changes as a function of interaction strength that we can attribute to the closing of specific doublon channels. As expected, the quantization revives in the hard-core limit at very large interaction strengths where the bosons are subject to a hardcore constraint. Interestingly, the stability of isolated doublons under the pump depends on the band they are in. For repulsive interactions and a suitably fixed pump period, doublons in the lowest band are pumped stably while doublons in higher bands dissociate during the pump with one particle decaying into a lower band. This asymmetry leads to the decay of the total energy over time, in stark contrast to the typical Floquet heating expected for a driven many-body system.

Over the past decades, research on two-dimensional melting has established that both first-order and continuous hexatic-liquid transitions can occur, influenced by various factors in the potential energy and system details. The fundamental thermodynamic origins of this sensitivity remains elusive. Here, by decomposing the Helmholtz free energy across three representative repulsive systems, we reveal a universal competition between energy and entropy that dictates the melting pathway. The energetic contribution consistently imparts convexity to the free energy, whereas entropy imparts concavity. A first-order transition occurs when concave entropy dominates; otherwise, the transition is continuous. Further decomposition shows that vibrational entropy drives the concave total entropic curvature, while the configurational entropy's curvature switches from convex (first-order) to concave (continuous), mirroring defect proliferation measured by Shannon entropy. The convexity of the energy is dominated by the inherent potential, with minimal vibrational influence. Finally, we predict and verify that the first-order transition becomes continuous at zero temperature, where entropic effects vanish. Our work establishes the curvature of different thermodynamic quantities as a fundamental principle for understanding the nature of two-dimensional melting.

We theoretically investigate signatures of exciton condensation and the underlying interband coherence in scanning tunneling microscopy. We consider both monolayer and bilayer condensates in the regime of a dilute condensate of tightly bound excitons. For monolayer condensates, interband coherence is directly encoded in spatially oscillating contributions to the tunneling conductance, which break the underlying lattice symmetry. We show how scanning tunneling microscopy allows one to extract the exciton wavefunction. For bilayer condensates, we show that the formation of the exciton insulator is signaled by the emergence of a characteristic peak in the tunneling conductance, which can be used to extract the (local) exciton density. Our results are based on analytical considerations using a systematic solution of the mean-field equations in powers of the exciton density as well as numerical calculations.

Soft interfaces formed by lipid membranes are fundamental to living cells, synthetic cells, and membrane-based soft materials. However, a quantitative framework linking molecular organization with nonlinear interfacial mechanics remains elusive. Here, we establish an analytical framework that captures the nonlinear elastic response of lipid-membrane-coated synthetic cells under micropipette aspiration. Incorporating both area stretching and curvature bending enables the model to quantitatively reproduce the complete pressure-displacement response within the small-deformation regime. This approach reduces interfacial mechanics to two parameters: the in-plane area-stretching modulus and an out-of-plane bending-related term. Using this unified framework, we experimentally demonstrate that nonlinear interfacial mechanics can be programmed by altering the molecular geometry and effective dimensionality of adsorbed elements. The lipid molecular shape and curvature-dependent packing regulate in-plane stiffness, while DNA nanostructures, the other adsorbed element, introduce an orthogonal control axis via dimensionality: isolated motifs primarily enhance area stretching, whereas three-dimensional network architectures markedly reinforce bending resistance. Together, these results establish a general molecular design principle for programming interfacial mechanics and provide a quantitative foundation for engineering mechanically tunable synthetic cells and soft interfaces.

We use $N$-body simulations to investigate the distinct bar formation processes in disks residing in halos of various concentrations. In a highly concentrated halo, the bar development is limited by the dominant multi-arm modes as a result of the swing amplification in the early stage. Only after the multi-arm modes decay, the bar growth proceeds mechanically owing to the particle trapping in continuation of that bar seed. In this scheme, the corotation resonance of the bar modes does not come into play at all, justified by a low amount of disk-halo angular momentum transfer and a modestly decreasing bar pattern speed. On the other hand, although reducing the halo concentration suggests the reduction of the preferred swing-amplified modes to be bi-symmetric, the bar formation in a lowly concentrated halo does not involve the swing amplification at all. Rather, the fast-growing linearly unstable bar modes of a single uniform frequency is solely the governing factor, attributed to a mild shearing. The bar modes trigger the corotation resonance since the beginning and such resonance is maintained until the end, which leads to a high amount of angular momentum transfer and a fast slowdown. For the intermediate halo concentration, the kinematical analyses of multiple non-axisymmetric modes suggests that the linear modes, the swing amplification, and the particle trapping are all present in the evolution chronology. To specify bars formed in the different halo concentrations, full analyses of the isophotal shape, the radial Fourier amplitude, and the resonance diagram can be of use.

We study the sum of first passage times along an arbitrary cycle made up of N>2 states of a small physical system. We show that, if the system is at thermodynamic equilibrium, this sum follows the same probability distribution regardless of whether the cycle is explored clockwise or counterclockwise. Out of equilibrium, the distributions of clockwise and counterclockwise cyclic first passage times are related by a detailed fluctuation theorem. This result descends from a symmetry of clockwise and counterclockwise trajectories, which combines time reversal with swapping portions of the trajectories. We then relate the entropy produced along the cycle with the entropy production of the whole system using large deviation theory. Our results reveal a novel symmetry in stochastic systems, of potential broad applicability in non-equilibrium physics.

Although 1-alkanols are widely used as anesthetics and membrane-active agents, the molecular basis of their chain-length-dependent cutoff behavior remains unclear. Here, we perform extensive atomistic molecular dynamics simulations to investigate the partitioning of 1-alkanols with varying chain lengths in a raft-like lipid bilayer composed of dipalmitoylphosphatidylcholine (DPPC), dioleoylphosphatidylcholine (DOPC), and cholesterol (Chol), which exhibits coexistence of liquid-ordered ($l_o$) and liquid-disordered ($l_d$) domains. We observe pronounced lateral heterogeneity in alkanol distribution, membrane thickness, number density, and lateral pressure profiles across coexisting phases. A distinct cutoff chain length, $n_{cutoff}=12$, is identified: alkanols with $n<n_{cutoff}$ preferentially partition into DOPC-rich $l_d$ domains, whereas alkanols with $n \ge n_{cutoff}$ preferentially localize within DPPC- and cholesterol-rich $l_o$ domains. This chain-length-dependent redistribution is accompanied by systematic reductions in the lateral pressure profile, membrane compressibility, and bending rigidity of the bilayer. The results provide a detailed molecular characterization of how alkanol chain length modulates membrane structure and mechanical response in laterally heterogeneous lipid membranes.

Multiterminal Josephson junctions offer a powerful playground for exploring exotic superconducting and topological phenomena beyond the reach of conventional two-terminal devices. In this work, we present the direct spectroscopic observation of Cooper quartet resonances, a signature of correlated tunneling of two Cooper pairs across the device, in a graphene three-terminal Josephson junction (3TJJ). Using tunneling spectroscopy, we visualize how Andreev bound states (ABS) evolve across a two-dimensional superconducting phase space, controlled by the two independent phase differences in the 3TJJ. These measurements reveal sharp local minima in the differential conductance spectra locked in a specific phase condition of superconducting phase variables. The resulting quantized trajectories around the compact torus of the superconducting phase variables reveal an underlying topological winding in the multipair transport. To interpret our results, we develop a theoretical model that connects the observed quartet resonances to the coherent hybridization of multiple ABS branches, a hallmark of the rich pairing process enabled by multiterminal geometries. Our results highlight the potential of multiterminal superconducting devices to host engineered superconducting states and pave the way for new approaches to topological band structure design based on phase-controlled, higher-order superconducting transport.

This paper presents simulations of the physical aging of a binary Kob-Andersen-type Lennard-Jones liquid following large temperature up-jumps from equilibrated states of high relaxation time. The purpose is to investigate how well the Tool-Narayanaswamy (TN) material-time concept works for this rather extreme case of aging. First the triangular relation of the potential energy is studied. This is found to be well obeyed, making it possible to define a potential-energy-based material time $ξ$. We proceed to study aging toward equilibrium at the final temperature 0.48 for jumps from the temperatures 0.43 and 0.37, monitoring the following five quantities: the potential energy, the self-intermediate scattering function, the mean-square displacement, the dynamic susceptibility $χ_4$, and the non-Gaussian parameter $α_2$. The TN material-time prediction is that all time-autocorrelation functions should collapse to only depend on the material-time difference $ξ_2-ξ_1$. This is found to work much better for the $0.43\to 0.48$ temperature jump than for the $0.37\to 0.48$ jump. Our findings thus confirm the general understanding that the TN aging formalism works best for systems that are never very far from equilibrium. This raises two questions for future work: Is the collapse significantly improved if each aging quantity is allowed its own material time? Can better collapse be obtained if the material-time is generalized to be defined locally in order to reflect dynamic heterogeneity?

Memory effects in non-Markovian dynamics are often diagnosed either via quantum-correlation revivals or via non-monotonic classical information measures, yet a unified minimal framework comparing their ``backflow phases'' is still lacking. Here we propose an information-backflow phase-diagram approach that places \emph{quantum entanglement revivals} and \emph{classical entropy overshoots} on the same footing through a common backflow functional $N_I=\int_{\dot I>0}\dot I\,dt$. On the quantum side, we employ a fractional (Caputo) extension of a two-state dissipative model embedded by thermo-field dynamics (TFD), yielding a closed-form intrinsic entanglement component $b^{(α)}_{qe}(t)=\frac14[E_α(-λ^αt^α)]^2\sin^2(ωt)$ and an integrated revival measure $N_{qe}$ that delineates a sharp boundary near $α\simeq 1/2$ in the $(α,ω/λ)$ plane. On the classical side, we consider a three-state model whose Markov generator is promoted either to an exponential-kernel generalized master equation (with exact Markov embedding) or to a semi-Markov process with Erlang-2 waiting times. We quantify non-monotonicity by the entropy overshoot $ΔH$ and KL-based diagnostics on the probability simplex. To strengthen the quantum--classical symmetry, we further introduce a \emph{fractional Mittag--Leffler memory kernel} in the classical dynamics and show that an analogous backflow transition emerges around $α\simeq 1/2$, indicating that the boundary originates from the kernel's mathematical structure rather than from quantumness per se. Overall, our results provide a compact, model-agnostic route to classify non-Markovianity by phase diagrams of information backflow and to interpret them via a shared embedding narrative: memory stored in hidden degrees of freedom returns to the observed sector as non-monotonic information flow.

Crystal stacking offers a powerful yet underexplored route to engineer symmetry in layered superconductors. Here we report superconductivity in rhombohedral-stacked NbSe2 (3R-NbSe2), a non-centrosymmetric polytype in which global inversion symmetry is removed by stacking alone. Using comprehensive structural, transport, magnetic, and thermodynamic measurements, we establish superconductivity as a bulk property of the 3R phase and find that the in-plane upper critical field exceeds the Pauli paramagnetic limit, indicating the persistence of strong Ising-type spin-orbit coupling. Unlike the thickness-dependent superconductivity in centrosymmetric 2H-NbSe2, the superconducting transition temperature in 3R-NbSe2 shows little dependence on layer number but exhibits an unusually strong sensitivity to disorder. We further observe strongly enhanced nonlinear optical and electrical responses near the superconducting transition, consistent with stacking-induced inversion-symmetry breaking. Our results identify 3R-NbSe2 as a single-phase platform in which stacking engineering reshapes superconductivity and enables nonlinear transport phenomena in layered materials.

We present a theory for the interaction between active particles and a passive flexible membrane. By explicitly solving for the pressure exerted by the active particles, we show that they reduce the membrane tension and bending modulus and introduce novel non-local contributions to the membrane mechanics. This theory predicts activity-induced instabilities and their morphology are in agreement with recent experimental and simulation data.

The predictability of the generative AI-based nowcasting model LDCast (trained on another region) is evaluated over Belgium, together with the pysteps implementation of the nowcasting algorithm STEPS. STEPS and LDCast are slightly underdispersive, but the ensemble spread provides an estimation of the error at almost all scales. Both models adapt the properties of their ensembles to the type of event, either convective or stratiform. The spatial scores of the STEPS and LDCast ensembles are compared with those of surrogate ensembles having some key properties, revealing that both STEPS and LDCast have very little ability to spatially localise the ensemble mean error vector through their ensemble members. This suggests that the content of STEPS and LDCast ensembles is informative in terms of statistics, but not in terms of dynamics.

Visual illusions provide a window into the mechanisms underlying visual processing, and dynamical neural circuit models offer a natural framework for proposing and testing theories of their emergence. We propose and analyze a delay-coupled neural field model that explains stroboscopic percepts arising from the subsampling of a moving, often rotating, stimulus, such as the wagon-wheel illusion. Motivated by the role of activity propagation delays in shaping visual percepts, we study neural fields with both uniform and spatially dependent delays, representing the finite time required for signals to travel along axonal projections. Each module is organized as a ring of neurons encoding angular preference, with instantaneous local coupling and delayed long-range coupling strongest between neurons with similar preference. We show that delays generate a family of coexisting traveling bump solutions with distinct, quantized propagation speeds. Using interface-based asymptotic methods, we reduce the neural field dynamics to a low-dimensional system of coupled delay differential equations, enabling a detailed analysis of speed selection, stability, entrainment, and state transitions. Regularly pulsed inputs induce transitions between distinct speed states, including motion opposite to the forcing direction, capturing key features of visual aliasing and stroboscopic motion reversal. These results demonstrate how delayed neural interactions organize perception into discrete dynamical states and provide a mechanistic explanation for stroboscopic visual illusions.

Extending a recent effective theory formulation for the dynamics of kinks in the sine-Gordon model [1], we propose an analogous effective description of $φ^4$ kinks. Three different reduced models based on the kink position, width and internal mode amplitude are introduced and compared systematically with the numerical solution of the equation with space- and time-dependent perturbations. In all cases considered, the model based on the kink position and width agrees the best with the full numerical solution. As long as the external driving frequency of the perturbation remains moderate, it captures with remarkable accuracy the intricate dynamical processes taking place in the system.

Social mobilization often fails not for a lack of collective interest, but because of fierce competition between rival movements for the same limited pool of participants. We generalize the classic threshold model of collective behavior to analyze this competitive aggregation, exploring how populations with diverse participation thresholds navigate multiple, mutually exclusive causes. Focusing on the conditions necessary for a single consensus movement to encompass an entire population, our analysis reveals that the outcome of social competition depends critically on the stability of individual dispositions. In quenched environments where participation thresholds are fixed, increasing resistance initially allows a dominant movement to suppress its competitors; however, further resistance triggers a sudden collapse into total fragmentation as low-threshold instigators become too rare to sustain growth. Conversely, in annealed environments where opinions are fluid, higher resistance paradoxically drives a winner-takes-all consensus. In this fluid scenario, massive movements can only be avoided through a deliberate divide-and-conquer strategy. In both cases, the transitions between pulverized and massive movements are discontinuous. These findings demonstrate that the effectiveness of social control depends entirely on environmental stability: raising the cost of participation can either forge unity or shatter collective action into insignificance.

For an arbitrary solution to the AKNS hierarchy, the logarithmic derivatives of the tau-function of the solution can be computed by the matrix-resolvent method [14,21]. In this paper, we introduce a pair of wave functions of the solution and we use them to express the corresponding matrix resolvent. Based on this, we derive a new formula for the k-point correlation function of the AKNS hierarchy expressed in terms of wave functions. As an application, we show that the tau-function of an arbitrary solution to the AKNS hierarchy is a KP tau-function.

Molecular Communication (MC) is a pivotal enabler for the Internet of Bio-Nano Things (IoBNT). However, current research often relies on super-capable individual agents with complex transceiver architectures that defy the energy and processing constraints of realistic nanomachines. This paper proposes a paradigm shift towards collective intelligence, inspired by the cortical networks of the biological brain. We introduce a decentralized network architecture where simple nanomachines interact via a diffusive medium using a threshold-based firing mechanism modeled by Greenberg-Hastings (GH) cellular automata. We derive fixed-point equations for steady-state populations via mean-field analysis and validate them against stochastic simulations. We demonstrate that the network undergoes a second-order phase transition at a specific activation threshold. Crucially, we show that both pairwise and collective mutual information peak exactly at this critical transition point, confirming that the system maximizes information propagation and processing capacity at the edge of chaos.

We present a Monte Carlo study of energy extraction from rotating (Kerr) black holes via the Penrose process using rocket propulsion. Through over 250,000 trajectory simulations, we establish sharp constraints on when Penrose extraction with escape to infinity succeeds. The mechanism requires that exhaust ejected inside the ergosphere carries negative Killing energy, which is kinematically accessible only via ultra-relativistic ejection deep within the ergosphere. We find that successful extraction with escape is statistically rare ($\sim$1% in broad parameter scans) and is governed by strict thresholds: it requires high black hole spin (empirically $a/M \gtrsim 0.89$) and ultra-relativistic exhaust velocity (onset at $v_e \approx 0.91c$). When conditions are highly tuned to a specific "sweet spot," success rates can reach 88.5%, representing a narrow extraction window rather than generic behavior. Furthermore, single-impulse thrust at periapsis achieves significantly higher cumulative efficiency ($η_{\rm cum} \approx 19\%$) compared to continuous thrust ($\sim$2--4%) due to path-averaging penalties. These constraints quantify the extreme fine-tuning required for material-based Penrose extraction, consistent with the astrophysical dominance of electromagnetic mechanisms. Simulation code is available at https://github.com/anindex/penrose_process.

This paper presents an axisymmetric Galerkin boundary element method (BEM) for modeling eddy-current interactions between excitation coils and conductive objects. The formulation derives boundary integral equations from the Stratton-Chu representation for the azimuthal component of the vector potential in both air and conductive regions. The central contribution is a unified regularization framework for the two-dimensional (2D) singular integrals arising in Galerkin BEM. This framework handles both logarithmic and Cauchy singularities through a common set of integral transformations, eliminating the need for case-by-case analytical singularity extraction and enabling straightforward numerical quadrature. The regularization and quadrature stability are proved and verified numerically. The method is validated on several representative axisymmetric geometries, including cylindrical, conical, and spherical shells. Numerical experiments demonstrate consistently high accuracy and computational efficiency across broad frequency ranges and coil lift-off distances. The results confirm that the proposed axisymmetric Galerkin BEM, combined with the integral transformation technique, provides a robust and efficient framework for axisymmetric eddy-current nondestructive evaluation.

Rimu.jl is a Julia package for solving many-body quantum problems. The core of the package is a matrix-free implementation of Hamiltonians and other operators and compact representation of Fock states, which together allow for efficient methods suitable for high-performance computing. Rimu.jl includes a Julia implementation of the full configuration interaction quantum Monte Carlo (FCIQMC) algorithm which is a type of projector QMC algorithm for stochastically solving the time-independent Schrödinger equation. It also includes many well-known model Hamiltonians, and an interface for exact diagonalisation based on external eigenvalue solvers. Both the stochastic and exact diagonalisation methods are accessed with a CommonSolve.jl interface. We describe the FCIQMC algorithm and how to obtain estimators of observables as well as the key features of the implementation.

Turing patterns are a central paradigm for describing spatial patterns in nature. The corresponding theory of reaction-diffusion dynamics combines ideal diffusion with nonlinear reactions, resulting in patterns when species diffuse at different rates and reactions are sufficiently nonlinear. However, real systems are more complex and particularly involve physical interactions between constituents. While such interactions can promote patterns, we here show that they can also induce dynamic, chaotic patterns. These patterns exhibit well-defined length and time scales, which result from cycles of droplet coarsening and fission. The dynamical patterns combine properties of traditional Turing patterns and chemically active droplets, which emerge for strong physical interactions. Our analysis thus reveals three qualitatively different regimes that emerge when two components interact physically and undergo nonlinear reactions.

Building on top of our recent work [arXiv:2502.08511], we introduce a new strategy to solve the problem of detecting atoms in high-resolution images of microtrap arrays. By alternating estimation and detection steps, we get rid of the need for an explicit model to compute the posterior occupancy probability of each site given its a priori optimal estimate. As direct benefits, we show an improved detection accuracy compared to our previous work when the sites are not optically well resolved, and we expect a greater robustness against real experimental conditions.

We introduce a structural framework for the geometry of financial order books in which liquidity, supply, and demand are treated as emergent observables rather than primitive market variables. The market is modeled as a relational substrate without assumed metric, temporal, or price coordinates. Observable quantities arise only through observation, implemented here as a reduction of relational degrees of freedom followed by a low-dimensional spectral projection. A one-dimensional projection induces a price-like coordinate and a projected liquidity density around the mid price, from which bid and ask sides emerge as two complementary restrictions. We show that directional liquidity imbalances decompose naturally into a rigid drift of the projected density and a geometric shear mode that deforms the bid--ask structure without inducing price motion. Under a minimal single-scale hypothesis, the shear geometry constrains the projected liquidity to a gamma-like functional form, appearing as an integrated-gamma profile in discrete data. Empirical analysis of high-frequency Level~II data across multiple U.S. equities confirms this geometry and shows that it outperforms standard alternative cumulative models under explicit model comparison and residual diagnostics.

As fundamental one-dimensional defects, screw dislocations profoundly reshape the energy landscape and carrier dynamics of crystalline materials. By restoring the exact algebra of the screw dislocation group, we unveil the latent symmetry constraints that govern the electronic structure, providing a more rigorous physical picture than the conventional treatments. When applied to GaN, the method yields a band-connectivity constraint and rigorous dipole selection rules for polarization-resolved transitions. Combined with computed Hamiltonian matrix, the approach gives symmetry-filtered radiative and dielectric calculations and reveals a piezoelectrical effect at the dislocation core that strongly suppresses radiative recombination. The pronounced dominance of non-radiative capture over radiative recombination highlights the detrimental impact of screw dislocations on the luminous efficiency of GaN, providing a theoretical foundation for optimizing dislocation-limited optoelectronic devices.

This study considers capturing aerodynamic causality from snapshot data with a time-varying mode decomposition technique referred to as information-theoretic machine learning. The current approach extracts time-dependent informative vortical structures, contributing to the future evolution of the aerodynamic coefficients. The present decomposition is employed with a convolutional neural network, enabling the identification of the spatial continuous mode. In addition, a low-order representation, characterizing the informative vortical structures and their corresponding aerodynamic coefficients, can also be identified by considering autoencoder-based data compression. The present technique is applied to a range of aerodynamic examples, including extreme vortex-gust airfoil interactions, experimentally measured transverse jet-wing interaction, and a turbulent separated wake. For the cases of gust-wing interaction, the time-varying gust effect on the lift response is extracted in an interpretable manner. With the example of a turbulent wake, the relationship between large-scale vortical motion and lift force is identified without any spatial length-scale information. The proposed approach could serve as a foundation for data-driven causal modeling and control for a range of unsteady flows.

Although electrostatics can be incorporated into machine-learned interatomic potentials, existing approaches are computationally very demanding, limiting large-scale, long-time simulations of electrostatics-driven phenomena such as dielectric response, infrared activity, and field-matter coupling. Here, we extend the neuroevolution potential (NEP), a highly efficient machine-learned interatomic potential, to a charge-aware framework (qNEP) by introducing explicit, environment-dependent partial charges. Each ionic partial charge is represented by a neural network as a function of the local descriptor vector, analogous to the NEP site-energy model. This formulation enables the direct prediction of the Born effective charge tensor for each ion and, consequently, the polarization. As a result, dielectric properties, infrared spectra, and coupling to external electric fields can be evaluated within a unified framework. We derive consistent expressions for the forces and virials that explicitly account for the position dependence of the partial charges. The qNEP method has been implemented in the free-and-open-source GPUMD package, with support for both Ewald summation and particle-particle particle-mesh treatments of electrostatics. We demonstrate the accuracy and efficiency of the qNEP approach through representative applications to water, Li7La3Zr2O12, BaTiO3, and a magnesium-water interface. These results show that qNEP enables accurate atomistic simulations with explicit long-range electrostatics, scalable to million-atom systems on nanosecond time scales using consumer-grade GPUs.

We present the novel Tensorized Discontinuous Isogeometric Analysis (TDIGA) method applied to the discontinuous Galerkin (DG) time-independent 2-D linearized Boltzmann transport equation (LBTE) with higher-order scattering, discretized with discrete ordinates in angle, multigroup in energy, and isogeometric analysis (IGA) in space. We formulate operator assembly in the tensor train (TT) format, producing seven-dimensional operators for both fixed-source and $k$-eigenvalue neutron transport problems solved using the restarted Generalized Minimum Residual Method (GMRES) and power iteration with an uncompressed solution vector. Our results on single-patch homogeneous and multi-patch heterogeneous problems, including a cruciform-shaped fuel array inspired by advanced reactor fuel designs, demonstrate the TT format's ability to compress interior operators from petabytes to megabytes, whereas the Compressed Sparse Row (CSR) matrix format requires gigabytes of storage. However, highly coupled boundary operators present a significant challenge for TT. Despite the storage savings, TT formatted operators increase time-to-solution relative to CSR as an uncompressed solution vector forces operator-vector product scaling of $O(dr^2N^d\log(N))$ for TT while CSR scales at $O(\text{nnz})$. We mitigate this discrepancy by using mixed formats with interior operators in TT, while high-rank boundary operators remain in CSR format. We compare all results to Monte Carlo (MC) and analytic reference solutions. While CSR remains $<10\times$ faster than this mixed format, the TDIGA method enables high-fidelity transport for expensive high-order IGA meshes.

People may express preferences that differ from their privately held views, often under social pressure, and may fail to act on their stated intentions. Such inconsistencies are referred to as preference falsification and the intention-behavior gap, respectively. Both hamper collective decision-making and adaptation, complicating policy formulation and implementation. To simulate these phenomena, dual-layer opinion agent-based models are used, in which each agent holds both a private and an expressed (public) opinion. Within the $q$-voter framework, two expressed-private opinion (EPO) models have been introduced in which private and expressed opinions are updated synchronously; two variants differ only by the presence of self-anticonformity, a mechanism in which an agent may set its private opinion opposite to its current expressed opinion, breaking internal harmony and creating a state akin to cognitive dissonance. Here, we extend these models by introducing an asynchronous update: in each elementary step, an agent updates its private opinion with probability $α$ or its expressed opinion with complementary probability; hence the name $α$-EPO models. Using Monte Carlo simulations on both artificial and real organizational social networks, along with mean-field and pair approximation analyses, we show that self-anticonformity makes collective outcomes robust to behavioral volatility tuned by $α$, enhances collective agreement, and suppresses hysteresis. In contrast, without self-anticonformity, $α$ affects the nature of the transition between agreement and disagreement: higher values of $α$ suppress hysteresis and enhance overall agreement.

Monte Carlo simulations of optical photon transport are computationally prohibitive for large-scale optical systems including detector arrays and PET systems, restricting practical use to single-crystal studies. This work presents an enhanced conditional generative adversarial network (optiGAN) replacing optical simulations at the crystal array level, extending our single-crystal approach to a 3x3 BGO array. We enhance the Wasserstein-GAN framework with Fourier feature encoding, a learnable latent mapping network, and a physics-informed loss enforcing momentum conservation. Training data is reduced eight-fold by exploiting symmetry. Evaluation employs three studies: a full array evaluation testing generalization from the fundamental domain to the complete geometry, a high-resolution study probing out-of-distribution generalization to untrained positions, and a pencil beam $γ$-photon study assessing practical applicability for experimental detector characterization. Performance is benchmarked against GATE10/Geant4 ground truth, using intrinsic fluctuations between independent Monte Carlo runs as baseline. OptiGAN achieves sliced Wasserstein similarity within 3$σ$-agreement of the baseline across all conditions, demonstrating successful generalization to the full array. The model transitions from electron-emission training data to realistic $γ$-photon interactions, producing flood maps that reproduce characteristic patterns including photopeak clusters and inter-crystal scatter lines. This proof-of-concept demonstrates that physics-informed generative models can accurately simulate optical photon transport in segmented scintillator arrays. The reproduction of experimentally relevant flood map features validates optiGAN for PET detector development and establishes a foundation for models generalizing across diverse array configurations.

Cell motility underlies many biological processes, including cancer metastasis, bacterial infection, and evolutionary adaptation. We introduce a non-equilibrium single-cell motility model inspired by the generalized Langevin equation, which accounts for hydrodynamic friction and correlated propulsion force. From video microscopy of Chlamydomonas reinhardtii algae and Salmonella typhimurium bacteria we extract the propulsion-force dynamics on the single-cell level, which we find to exhibit multi-exponential correlations, not captured by literature non-equilibrium cell-motility models. Based on our data-driven model, we predict the effective cell diffusivities beyond experimentally resolved timescales and demonstrate a diffusivity maximum at intermediate solvent viscosity for both cell types. This means that cells adapt their propulsion-force characteristics according to the solvent viscosity. In addition, our model predicts the power output of single cells, which is on the order of aW for the salmonella and fW for the algae.

McSAS3 is the refactored successor to the original McSAS Monte Carlo small-angle scattering analysis software. It is intended to be integrated in automated data processing pipelines, but can also be used to process individual (batches of) scattering data. McSAS3 comes with a graphical user interface (McSAS3GUI), complete with guides, examples and videos. McSAS3GUI will help to generate and test the three configuration files that McSAS3 needs for data read-in, Monte Carlo optimization and histogramming. The user interface can also be used to process individual files or batches, and can be augmented with machine-specific use templates. The Monte Carlo (MC) approach is able to fit most practical scattering patterns extremely well, resulting in form-free model parameter distributions. Theoretically, these can be distributions on any model parameter, but in practice the MC-optimized parameter is usually a (volume-weighted) size distribution, in absolute volume fraction for absolute-scaled data.

Inverse design is a central goal in much of science and engineering, including frequency-selective surfaces (FSS) that are critical to microelectronics for telecommunications and optical metamaterials. Traditional surrogate-assisted optimization methods using deep learning can accelerate the design process but do not usually incorporate uncertainty quantification, leading to poorer optimization performance due to erroneous predictions in data-sparse regions. Here, we introduce and validate a fundamentally different paradigm of Physics-Informed Uncertainty, where the degree to which a model's prediction violates fundamental physical laws serves as a computationally-cheap and effective proxy for predictive uncertainty. By integrating physics-informed uncertainty into a multi-fidelity uncertainty-aware optimization workflow to design complex frequency-selective surfaces within the 20 - 30 GHz range, we increase the success rate of finding performant solutions from less than 10% to over 50%, while simultaneously reducing computational cost by an order of magnitude compared to the sole use of a high-fidelity solver. These results highlight the necessity of incorporating uncertainty quantification in machine-learning-driven inverse design for high-dimensional problems, and establish physics-informed uncertainty as a viable alternative to quantifying uncertainty in surrogate models for physical systems, thereby setting the stage for autonomous scientific discovery systems that can efficiently and robustly explore and evaluate candidate designs.

This computational study investigates lateral optical forces in asymmetric dielectric nanostructures, focusing on their connection to resonant light-matter interactions. We examine isosceles triangular motifs that exhibit two distinct types of optical force response under plane wave illumination. Through parameter-space analysis, we identify stable zones where optical forces remain consistent and switching bands where forces change abruptly as parameters are altered. The observed force spectra show characteristic asymmetric lineshapes, suggesting Fano-resonance behavior. Eigenfrequency analysis confirms these effects arise from interference between discrete eigenmodes and continuum propagation states, with the eigenmode Q-factors correlating with transition sharpness. These findings provide insights into how structural geometry influences optical forces through resonant effects, offering guidance for designing optically-driven systems where controlled optical force responses are desired.

How water droplets move and slide on leaves influences plant ecophysiological and abiotic interactions, as well as the design of advanced bio-inspired wetting materials. Despite cross-disciplinary relevance, current descriptions of the in situ dynamics of droplets on living leaves focus almost exclusively on surface structure and chemistry, treating the leaf as a static, electrically neutral substrate. Here, three decades after the mechanistic discovery of the Lotus effect, we show that a yet 'hidden' force due to instantaneous electrical phenomena affect the dynamic droplet motion on living leaves. Using high-speed motion tracking and precision charge measurements, we show that droplets sliding on the pristine epicuticular wax layer on superhydrophobic Colocasia esculenta leaves strongly charge affecting its dynamics, previously observed only on synthetic (highly electronegative fluorinated) surfaces. Droplets accumulate charges of Qp,D1 = -0.02 to -0.15 nC per 30 uL droplet on pristine leaves. However, we specifically demonstrate the crucial role of the epicuticular wax layer plasticity: by a structural modification that decreases its roughness amplitude, the same leaves gain an impressive 30-40 fold enhancement in charge transfer (reaching Qt,D1 = -2.8 to -5.2 nC) slowing the droplet by half due to an estimated electrostatic force of 11 uN dominating the resistive forces. The charge accumulation is surface-history-dependent and charge quantities per droplet are surprisingly similar or even exceeding those recently reported from artificial surfaces. Our findings prove that electrostatic charging is a fundamental component of droplet-leaf interactions, opening new research directions from charge-affected leaf ecology to sustainable materials for droplet-based energy harvesting by tuning surface treatments and, moreover,...

Flash evaporation, a liquid-to-gas phase transition phenomenon in real fluids, is prevalent in aerospace propulsion systems. To elucidate the physical mechanisms of such complex flows and provide theoretical benchmarks for Computational Fluid Dynamics simulations, this paper formalizes the Flash evaporation Riemann problem (FeRP) characterized by the expansion branch crossing the saturation line, within the framework of Homogeneous Equilibrium and Vapor-Liquid Equilibrium assumptions. An exact solution framework that analytically resolves all thermodynamic derivatives of equilibrium two-phase fluids is established for arbitrary two-parameter equations of state. By evaluating the Landau fundamental derivative, the non-classical wave structures arising in the FeRP are analyzed, for which a stable iterative solution strategy incorporating the Chapman-Jouguet condition as an outer constraint is proposed. Furthermore, an exact solution for the FeRP based on Wood's mechanical equilibrium speed of sound is developed, enabling a comprehensive evaluation of its thermodynamic implications. Results indicate that Wood's model alters the definition of the two-phase mixture entropy in the Euler equations, introducing an isentropic path characterized by a "density lag" effect and non-physical entropy decrease. Comparative analysis of the FeRP under typical scramjet fuel injection conditions reveals that, although Wood's model captures the general trend of the Riemann solution curve, it significantly underestimates intermediate pressure, velocity, and the extent of vaporization relative to the complete equilibrium model.

The COVID-19 pandemic has been accompanied by an infodemic of misinformation that impedes effective public health responses. This study examines relationships between socioeconomic factors and infodemic risk patterns across 37 OECD countries using Twitter data from 2020-2022. Employing dimensionality reduction techniques on 20 socioeconomic indicators, we identify complex correlations with infodemic measures that evolve throughout the pandemic. Countries exhibit distinct clustering in their infodemic profiles that transcend conventional socioeconomic categorizations. We find that dynamic information behaviors dominate initial crisis responses, while stable socioeconomic conditions become more influential as the pandemic progresses. News media diet diversity emerges as a significant protective factor, with pluralistic information ecosystems demonstrating greater resilience against misinformation. Additionally, institutional stability correlates strongly with reduced infodemic volatility over time. These findings highlight how infodemics are embedded within broader socioeconomic contexts, providing foundations for targeted interventions to build societal resilience against misinformation during future health emergencies.

In neural-network surrogate solvers for electromagnetic simulations, accurately modeling resonant phenomena remains a central challenge. High-amplitude resonances generate strongly localized field patterns that appear as outlier samples, deviating significantly from the general distribution of non-resonant cases, leading to instability and degraded predictive performance. To address this, we introduce dissipative relaxation transfer learning (DIRTL), a data-efficient training framework that integrates transfer learning with loss-regularized optimization principles from high-Q photonics. DIRTL first pretrains the model on data generated with a small fictitious material loss, which broadens sharp resonant modes and suppresses extreme field amplitudes. This smoothing of the response landscape enables the model to learn global modal features more effectively. The pretrained model is subsequently fine-tuned on the target lossless dataset containing the true high-amplitude resonances, allowing stable adaptation based on the pretrained information. Applied to both the Fourier Neural Operator (FNO) and UNet architectures, DIRTL yields substantial improvements in prediction accuracy, including up to a two-fold error reduction for the FNO variant. Furthermore, DIRTL exhibits robustness across diverse training conditions and supports strong multi-task performance, underscoring the generalizability and flexibility of the pretrained core. Altogether, these results establish DIRTL as a physically grounded and architecture-agnostic curriculum for enhancing the reliability of machine-learning-based electromagnetic surrogate solvers.

While geostrophy remains the simplest and most practical balance to extract velocity information from sea-surface height anomaly (SSHa), confusions remain within the oceanographic community to what extent this balance can be applied to altimetric observations with the launch of the Surface Water and Ocean Topography (SWOT) satellite. Given the limited temporal resolution of SWOT, many studies have resorted to claiming that the spatially filtered SSHa fields correspond to the geostrophic component. This introduces the ambiguity of which spatial scale to choose. Here, we build upon the recent developments in internal tide (IT) corrections (Yadidya et al., 2025) and apply a dynamic mode decomposition (DMD)-based method introduced by Lapo et al. (2025) to robustly extract the geostrophic component associated with sub-inertial frequencies from the SWOT one-day-repeat orbit; we distribute the global dataset as a public good. We provide the joint probability density function (PDF) of vorticity and strain, and spectra of SSHa at a few cross-over regions.

3D kinetic particle-in-cell (PIC) simulations are performed using the kinetic Alfvén wave (KAW) eigenvector relations from a two-fluid model as initial conditions, in order to study turbulent fluctuations and intermittent structures at sub-ion and electron scales. Simulations with different ion-to-electron mass ratios are set up to investigate the role of electron scales in the formation of intermittent structures. We analyze the current sheet structures that develop in these simulations. Two algorithms, namely Breadth-First Search (BFS) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN), are employed to determine the thickness, length, and width of the current sheets, and both methods are found to yield consistent results. The average current sheet thickness scales inversely with the square root of the ion-to-electron mass ratio, with values close to the electron skin depth ($d_e$), indicating the presence of electron-scale current sheets in the simulations. The widths and lengths of the current sheets show a weaker scaling with the mass ratio. The scale-dependent kurtosis reveals enhanced intermittency at electron scales, consistent with magnetosheath observations. Distributions of scale-dependent properties of the current sheets also align with the electron skin depth of the different simulations and they lie within ranges observed in kinetic scale solar wind turbulence. This study reveals the nature of sub-ion-scale current sheets in KAW turbulence and their role in dissipation.

Large-scale neutrino observatories operate under unavoidable detector deadtime arising from photomultiplier saturation, digitizer limits, and front-end readout constraints. Conventional coincidence-based trigger logic implicitly assumes continuous sensor availability and therefore suffers systematic efficiency loss when channels become temporarily non-live. This work presents the design of a liveness-aware trigger architecture targeting low-latency FPGA deployment in distributed optical arrays. We introduce a recursive Infinite Impulse Response (IIR) update law implemented as a fully synthesizable pipeline that constructs a continuity-preserving effective observable at each sensor node. Rather than collapsing during non-liveness intervals, the observable decays smoothly while retaining phase and amplitude information relevant for network-level coherence estimation. By explicitly separating continuous measurement construction from discrete trigger decision logic, the proposed architecture enables graceful degradation under partial channel non-liveness. Performance is evaluated using a hybrid validation framework that combines representative event topologies derived from IceCube Open Data with a hardware-accurate signal and noise model spanning a wide dynamic range. Simulation results demonstrate that the proposed trigger sustains high event recovery efficiency in regimes of elevated deadtime probability, where conventional coincidence logic degrades substantially. Furthermore, the continuity-preserving observable yields up to a two-order-of-magnitude improvement in effective signal-to-noise ratio, enabling robust detection under strong saturation and non-ideal operating conditions. This method provides a robust foundation for next-generation firmware-level trigger strategies in large-scale, noise-dominated detector systems.

Drug-drug interactions (DDIs) strongly affect the safety and efficacy of combination therapies. Despite the availability of large DDI databases, selecting optimal multi-drug combinations that balance safety, therapeutic benefit, and regimen size remains a challenging combinatorial optimization problem. Here, we present a quantum-control-based framework for DDI-aware drug combination optimization, in which known harmful and synergistic interactions are encoded into Ising Hamiltonians as penalties and rewards, respectively. The optimization is performed using the feedback-based quantum algorithm FALQON, a gradient-free variational approach. We study two clinically motivated tasks: the Maximum Safe Subset problem and the Synergy-Constrained Optimization problem. Numerical simulations using interaction data from Drugs.com and SYNERGxDB demonstrate efficient convergence and high-quality solutions for clinically relevant drug sets, including COVID-19 case studies.

Contradictory experimental reports on the relationship between efficiency and stimulation frequency have hindered mechanistic understanding in converting neural activity into mechanical work during muscle contraction. To resolve this issue, we develop a biophysical model integrating calcium-mediated excitation with a detailed cross-bridge cycle to enable single-fiber simulations. Our model demonstrates that shortening velocity is the primary determinant of cross-bridge efficiency: efficiency peaks at an optimal velocity and declines at higher or lower velocities, while frequency exerts secondary influence. Critically, the velocity yielding peak efficiency remains consistent across frequencies, with a slight upward shift at higher frequencies. Elevated inorganic phosphate ([Pi]) appears to amplify the efficiency disparity between high- and low-frequency regimes in our analysis. Our work suggests that stimulation frequency modulates efficiency predominantly through its regulation of shortening velocity, which primarily governs the kinetics of the myosin power stroke. This work may help clarify neural control of muscle energetics, and provide a quantitative foundation for studying muscle function in physiological and pathological contexts.

Understanding how ancient Mesoamerican societies integrated mathematical ideas into calendrical design and monumental architecture requires approaches that acknowledge their distinct epistemological frameworks. While explicit textual evidence for concepts such as $π$ or the golden ratio $φ$ is absent, numerical patterns embedded in Mesoamerican calendars, iconography, and ritual architecture reveal a coherent system of proportional reasoning grounded in simple integer ratios. Here we show that the numbers 5 and 8, central to Venus and solar calendrical relations and widely represented in Mesoamerican cosmology, generate rational approximations that reproduce, within known construction tolerances, the geometric relations associated with decagonal layouts. Using high-resolution measurements of the Iguana structure at Guachimontones, we demonstrate that its proportions align with integer ratios consistent with those found in the calendrical system and with the practical geometry of the regular decagon, without requiring knowledge of irrational constants. These findings suggest that Mesoamerican builders employed stable proportional modules that harmonized astronomical cycles, cosmological symbolism, and architectural design. This should not be interpreted as a lack of mathematical sophistication; rather, the material record reveals a distinct mathematical tradition in which number, measure, and cosmology were mutually reinforcing elements of cultural knowledge.

Growing synthetic networks that follow power law distributions of a node's degree often involves adding one node at a time. Each node is added to the network with a fixed amount of edges and those edges are frozen for all future time steps. Yet real world networks often continuously evolve with edges being added and removed while new nodes are added to the network. Many existing growth models based on preferential attachment do not account for this evolutionary capability and when you extend their growth methods to add and remove edges to existing nodes the node degree distribution quickly loses its scale-free structure. This paper will go over a method to extend well known preferential attachment growth models to allow for the evolution of edges within a network while still maintaining a power law node degree distribution.

Cologna et al. (2025) compared perceived scientist trustworthiness across 68 countries/regions and examined its associations with individual- and country-level factors. While the authors reported that the scale did not satisfy metric and scalar measurement invariance, their subsequent cross-national/regional comparisons and regressions were nonetheless conducted using weighted means of observed item scores, implicitly assuming cross-country/region comparability at the observed-score level. Using the publicly shared dataset, we re-evaluated these conclusions by systematically applying measurement alignment under four analytical paths: pooled-sample versus country/region-specific confirmatory factor analysis (CFA), each estimated with and without weights. Across all specifications, cross-national/regional CFA supported configural and metric invariance but failed to establish scalar or strict invariance. Importantly, under the analytical path most closely corresponding to the original study (pooled CFA with weights), only the competence and openness factors yielded admissible aligned solutions within the four-factor model. Using aligned latent scores for these two dimensions, country/region rankings changed for 62 of the 68 countries/regions. Substantive conclusions also differed: associations between perceived scientist trustworthiness and science-related populist attitudes or social dominance orientation were near zero or non-robust, whereas attitudes toward science remained strongly and positively related. Taken together, these findings demonstrate that cross-national/regional comparisons based on observed-score averages may be misleading when measurement equivalence is not established, and that latent-variable approaches such as alignment provide a more defensible basis for international inference.

Promoters and enhancers are cis-regulatory elements (CREs), DNA sequences that bind transcription factor (TF) proteins to up- or down-regulate target genes. Decades-long efforts yielded TF-DNA interaction models that predict how strongly an individual TF binds arbitrary DNA sequences and how individual binding events on the CRE combine to affect gene expression. These insights can be synthesized into a global, biophysically-realistic, and quantitative genotype-phenotype (GP) map for gene regulation, a "holy grail" for the application of evolutionary theory. A global map provides a rare opportunity to simulate long-term evolution of regulatory sequences and pose several fundamental questions: How long does it take to evolve CREs de novo? How many non-trivial regulatory functions exist in sequence space? How connected are they? For which regulatory architecture is CRE evolution most rapid and evolvable? In this article, the first of a two-part series, we briefly review the pertinent modeling and simulation efforts for a unique system that enables close, quantitative, and mechanistic links between biophysics, as well as systems, synthetic, and evolutionary biology.

There are few, if any, algorithms in statistical phylogenetics which are used more heavily than Felsenstein's 1973 pruning method for computing the likelihood of a tree. We present LvD, (Likelihood via Decomposition), an alternative to Felsenstein's algorithm based on a different decomposition of the underlying phylogeny. It works for all standard nucleotide models. The new algorithm allows updates of the likelihood calculation in worst case $O(\log n)$ time with $n$ taxa, as opposed to worst case $O(n)$ time for existing methods. In practice this leads to appreciable improvements in likelihood calculations, the extent of speed-up depending on how balanced or unbalanced the trees are. We explore implications for parallel computing, and show that the approach allows likelihoods to be computed in $O(\log n)$ parallel time per site, compared to (worst case) $O(n)$ time. We implemented and applied the algorithm to large numbers of simulated and empirical data sets and showed that these theoretical advances lead to a significant practical speed-up, although the extent of the improvement depends on how balanced the phylogenies already are.

Accurate molecular property prediction is central to drug discovery, yet graph neural networks often underperform in data-scarce regimes and fail to surpass traditional fingerprints. We introduce cross-graph inter-message passing (XIMP), which performs message passing both within and across multiple related graph representations. For small molecules, we combine the molecular graph with scaffold-aware junction trees and pharmacophore-encoding extended reduced graphs, integrating complementary abstractions. While prior work is either limited to a single abstraction or non-iterative communication across graphs, XIMP supports an arbitrary number of abstractions and both direct and indirect communication between them in each layer. Across ten diverse molecular property prediction tasks, XIMP outperforms state-of-the-art baselines in most cases, leveraging interpretable abstractions as an inductive bias that guides learning toward established chemical concepts, enhancing generalization in low-data settings.

Recurrent neural networks trained for time-series prediction often develop latent trajectories that preserve qualitative structure of the dynamical systems generating their inputs. Recent empirical work has documented topology-preserving latent organization in trained recurrent models, and recent theoretical results in reservoir computing establish conditions under which the synchronization map is an embedding. Here we synthesize these threads into a unified account of when contracting recurrent networks yield smooth, topology-preserving internal representations for a broad and biologically relevant class of inputs: regular dynamics on invariant circles and tori. Our contribution is an integrated framework that assembles (i) generalized synchronization and embedding guarantees for contracting reservoirs, (ii) regularity mechanisms ensuring differentiability of the synchronization map under mild constraints, and (iii) a base-system viewpoint in which the invariant manifold generating the input stream is treated as the driving system. In this regular setting, the conditions commonly viewed as restrictive in chaotic-attractor analyses become mild and readily satisfied by standard contractive architectures. The framework clarifies how representational content in recurrent circuits is inherently historical: the network state encodes finite windows of input history rather than instantaneous stimuli. By consolidating disparate empirical and theoretical results under common assumptions, the synthesis yields concrete, testable expectations about when prediction-trained recurrent circuits should (or should not) form smooth latent embeddings and how required state dimension scales with the intrinsic dimension of the driving dynamics.

Schemas -- abstract relational structures that capture the commonalities across experiences -- are thought to underlie humans' and animals' ability to rapidly generalize knowledge, rebind new experiences to existing structures, and flexibly adapt behavior across contexts. Despite their central role in cognition, the computational principles and neural mechanisms supporting schema formation and use remain elusive. Here, we introduce schema-based hierarchical active inference (S-HAI), a novel computational framework that combines predictive processing and active inference with schema-based mechanisms. In S-HAI, a higher-level generative model encodes abstract task structure, while a lower-level model encodes spatial navigation, with the two levels linked by a grounding likelihood that maps abstract goals to physical locations. Through a series of simulations, we show that S-HAI reproduces key behavioral signatures of rapid schema-based generalization in spatial navigation tasks, including the ability to flexibly remap abstract schemas onto novel contexts, resolve goal ambiguity, and balance reuse versus accommodation of novel mappings. Crucially, S-HAI also reproduces prominent neural codes reported in rodent medial prefrontal cortex during a schema-dependent navigation and decision task, including task-invariant goal-progress cells, goal-identity cells, and goal-and-spatially conjunctive cells, as well as place-like codes at the lower level. Taken together, these results provide a mechanistic account of schema-based learning and inference that bridges behavior, neural data, and theory. More broadly, our findings suggest that schema formation and generalization may arise from predictive processing principles implemented hierarchically across cortical and hippocampal circuits, enabling the generalization of experience.

The integration of large-scale chemical databases represents a critical bottleneck in modern cheminformatics research, particularly for machine learning applications requiring high-quality, multi-source validated datasets. This paper presents a case study of integrating three major public chemical repositories: PubChem (176 million compounds), ChEMBL, and eMolecules, to construct a curated dataset for molecular property prediction. We investigate whether byte-offset indexing can practically overcome brute-force scalability limits while preserving data integrity at hundred-million scale. Our results document the progression from an intractable brute-force search algorithm with projected 100-day runtime to a byte-offset indexing architecture achieving 3.2-hour completion-a 740-fold performance improvement through algorithmic complexity reduction from O(NxM) to O(N+M). Systematic validation of 176 million database entries revealed hash collisions in InChIKey molecular identifiers, necessitating pipeline reconstruction using collision-free full InChI strings. We present performance benchmarks, quantify trade-offs between storage overhead and scientific rigor, and compare our approach with alternative large-scale integration strategies. The resulting system successfully extracted 435,413 validated compounds and demonstrates generalizable principles for large-scale scientific data integration where uniqueness constraints exceed hash-based identifier capabilities.

Structural dynamics of macromolecules is critical to their structural-function relationship. Cryogenic electron microscopy (CryoEM) provides snapshots of vitrified protein at different compositional and conformational states, and the structural heterogeneity of proteins can be characterized through computational analysis of the images. For protein systems with multiple degrees of freedom, it is still challenging to disentangle and interpret the different modes of dynamics. Here, by implementing Point Transformer, a self-attention network designed for point cloud analysis, we are able to improve the performance of heterogeneity analysis on CryoEM data, and characterize the dynamics of highly complex protein systems in a more human-interpretable way.

Partial differential equation (PDE) models for infectious diseases, while less common than their ordinary differential equation (ODE) counterparts, have found successful applications for many years. Such models are typically of reaction-diffusion type, and model spatial propagation as a diffusive process. However, given the complex nature of human mobility, such models are limited in their ability to describe airborne infectious diseases in human populations. Recent work has advocated for the inclusion of an additional chemotaxis-type term as an alternative; spatial propagation of infection fronts is assumed additionally to flow from low-to-high concentrations of susceptible populations. The present work extends the study of such models by providing an epidemiologically interpretable analysis, directly connecting model behavior to information readily available to policymakers. In particular, we derive a spatially-aware basic reproduction number, which accounts for spatial heterogeneity in population density. Furthermore, we discuss several important aspects concerning the numerical solution of the model, including the introduction of a stabilization scheme. Finally, we perform a series of simulation studies in the Italian region of Lombardy (severely affected by the COVID-19 outbreak in 2020) and in the US state of Georgia, in which we demonstrate the model's potential to better capture important spatiotemporal dynamics observed in real-world data compared to pure reaction-diffusion models.

Advances in single-cell and spatial transcriptomic technologies have transformed tumor ecosystem profiling at cellular resolution. However, large scale studies on patient cohorts continue to rely on bulk transcriptomic data, where variation in tumor purity obscures tumor-intrinsic transcriptional signals and constrains downstream discovery. Many deconvolution methods report strong performance on synthetic bulk mixtures but fail to generalize to real patient cohorts because of unmodeled biological and technical variation. Here, we introduce TwinPurify, a representation learning framework that adapts the Barlow Twins self-supervised objective, representing a fundamental departure from the deconvolution paradigm. Rather than resolving the bulk mixture into discrete cell-type fractions, TwinPurify instead learns continuous, high-dimensional tumor embeddings by leveraging adjacent-normal profiles within the same cohort as "background" guidance, enabling the disentanglement of tumor-specific signals without relying on any external reference. Benchmarked against multiple large cancer cohorts across RNA-seq and microarray platforms, TwinPurify outperforms conventional representation learning baselines like auto-encoders in recovering tumor-intrinsic and immune signals. The purified embeddings improve molecular subtype and grade classification, enhance survival model concordance, and uncover biologically meaningful pathway activities compared to raw bulk profiles. By providing a transferable framework for decontaminating bulk transcriptomics, TwinPurify extends the utility of existing clinical datasets for molecular discovery.

Rationale: Transcranial magnetic stimulation (TMS)-based measures such as resting motor threshold (RMT) and short interval intracortical inhibition (SICI) are widely employed to study motor cortical and corticospinal tract function, and effects of diseases and drug therapies thereon. However, the effect of key experimental factors, including as eye state (open or closed) or stimulating coil size, remain unclear. As such, it is unknown whether these factors must be kept consistent across multi-center studies, and whether differences in such factors may underpin contradictory findings in existing literature. Materials and Methods: Threshold tracking TMS was employed to measure RMT and SICI (3ms interstimulus interval, conditioning at 70% of RMT) in 21 alert and awake, healthy controls. Motor evoked potentials were recorded from abductor pollicis brevis. Both RMT and SICI were measured under 6 conditions, while eyes were open or closed, using 3 figure-of-eight coils of differing winding diameter. Mixed effects modelling was employed to investigate effects of eye state and coil size on each measure. Results: RMT was found to be significantly higher for the smallest (30BFT) coil compared to both larger (50BFT and 70BF) coils. No difference in SICI was identified across coil sizes. Eye state was not found to affect either RMT or SICI measurements. Conclusions: Measurements of RMT and SICI can be considered comparable if recorded with eyes open or closed, provided the individual is awake and alert. Measurements of SICI recorded with figure-of-eight coils of different size can be considered comparable.

In the context of global warming, tree populations rely on two primary mechanisms of adaptation: phenotypic plasticity, which enables individuals to adjust their behavior in response to environmental stress, and genetic evolution, driven by natural selection and genetic diversity within the population. Understanding the interplay between these mechanisms is crucial for assessing the impacts of climate change on forest ecosystems and for informing sustainable management strategies. In this manuscript, we focus on a specific phenological adaptation: the ability of trees to enter summer dormancy once a critical temperature threshold is exceeded. Individuals are characterized by this threshold temperature and by their seed production capacity. We first establish a detailed mathematical model describing the population dynamics under these traits, and progressively reduce it to a system of two coupled ordinary differential equations. This simpler macroscopic model is then analyzed numerically, to investigate how the population reacts to a shift in its environment: an temperature increase, a drop in precipitation levels, or a combination of the two. Our results highlight contrasting effects of water stress and temperature stress on population dynamics, as well as the ambivalent effect of the plasticity.

Vehicle movements, including vehicle cabs and trailers, play a role in disseminating disease in swine production. However, there are many information gaps about vehicle movements patterns that increase the probability of disease transmission, which is crucial in developing better preventive strategies. In this study we described the movement pattern of vehicle cabs and trailers and identified risk factors for porcine reproductive and respiratory syndrome (PRRS) and porcine epidemic diarrhea (PED) farm's infectious status. We collected global positioning system (GPS) movement data from vehicle cabs and trailers for 18 months and basic information for 6621 farms in the U.S. For the vehicle movement data, we estimated 66 variables and evaluate their association with farms PRRS and PED status. Our univariate analysis showed that 56 variables were significant associated (p < 0.05) to PED and PRRS farm status. Within these variables, vehicle visit frequency and previous exposition to positive farms were the main risk factors for both diseases. Otherwise, increased vehicle cab and trailer loyalty for farm shipments and vehicle cleaning and disinfection events were protective factors. In the multivariate model, each additional weekly visit by a vehicle cab that had been exposed to a positive farm one day before the shipment was associated with a 234\% and 243\% increase in the odds of a farm testing PRRS- and PED-positive, respectively. Our analysis revealed that vehicle contact history play a crucial role in the transmission of PRRS and PED. These findings can provide insights to develop more target strategies aimed at reducing the transmission and outbreaks linked to vehicle movements in swine production.

Global biodiversity loss is accelerating, prompting international efforts such as the Kunming-Montreal Global Biodiversity Framework (GBF) and the United Nations Sustainable Development Goals to direct resources toward halting species declines. A key challenge in achieving this goal is having access to robust methodologies to understand where species occur and how they relate to each other within broader ecological communities. Recent deep learning-based advances in joint species distribution modeling have shown improved predictive performance, but effectively incorporating community-level learning, taking into account species-species relationships in addition to species-environment relationships, remains an outstanding challenge. We introduce LabelKAN, a novel framework based on Kolmogorov-Arnold Networks (KANs) to learn inter-label connections from predictions of each label. When modeling avian species distributions, LabelKAN achieves substantial gains in predictive performance across the vast majority of species. In particular, our method demonstrates strong improvements for rare and difficult-to-predict species, which are often the most important when setting biodiversity targets under frameworks like GBF. These performance gains also translate to more confident predictions of the species spatial patterns as well as more confident predictions of community structure. We illustrate how the LabelKAN leads to qualitative and quantitative improvements with a focused application on the Great Blue Heron, an emblematic species in freshwater ecosystems that has experienced significant population declines across the United States in recent years. Using the LabelKAN framework, we are able to identify communities and species in New York that will be most sensitive to further declines in Great Blue Heron populations.

In a recent work, we introduced the foundations of an orthogonally constrained complete active space self-consistent field (OC-CASSCF) framework that produces state-specific molecular orbitals for mutually orthogonal multiconfigurational electronic states. In the present study, we extend this approach by incorporating a Newton-Raphson orbital-optimization scheme, for which we derive analytical expressions of the orbital gradient and Hessian. Furthermore, we outline a practical route toward the evaluation of analytical nuclear gradients, enabling geometry optimizations within the OC-CASSCF formalism. Benchmark calculations on the three lowest singlet states of LiH and H$_2$O molecules demonstrate a systematic improvement as compared to conventional state-averaged CASSCF, even when using modestly sized active spaces.

We provide a proof that there exists a bijection between local unitary (LU) orbits of absolutely maximally entangled (AME) states in $(\mathbb{C}^D)^{\otimes n}$ where $n$ is even, also known as perfect tensors, and LU orbits of $((n-1,D,n/2))_D$ quantum error correcting codes. Thus, by a result of Rather et al. (2023), the AME state of 4 qutrits and the pure $((3,3,2))_3$ qutrit code $\mathcal{C}$ are both unique up to the action of the LU group. We further explore the connection between the 4-qutrit AME state and the code $\mathcal{C}$ by showing that the group of transversal gates of $\mathcal{C}$ and the group of local symmetries of the AME state are closely related. Taking advantage of results from Vinberg's theory of graded Lie algebras, we find generators of both of these groups.

The performance of tensor network methods has seen constant improvements over the last few years. We add to this effort by introducing a new algorithm that efficiently applies tree tensor network operators to tree tensor network states inspired by the density matrix method and the Cholesky decomposition. This application procedure is a common subroutine in tensor network methods. We explicitly include the special case of tensor train structures and demonstrate how to extend methods commonly used in this context to general tree structures. We compare our newly developed method with the existing ones in a benchmark scenario with random tensor network states and operators. We find our Cholesky-based compression (CBC) performs equivalently to the current state-of-the-art method, while outperforming most established methods by at least an order of magnitude in runtime. We then apply our knowledge to perform circuit simulation of tree-like circuits, in order to test our method in a more realistic scenario. Here, we find that more complex tree structures can outperform simple linear structures and achieve lower errors than those possible with the simple structures. Additionally, our CBC still performs among the most successful methods, showing less dependence on the different bond dimensions of the operator.

Linear oscillators contribute to most branches of contemporary quantum science. They have already successfully served as quantum sensors and memories, found applications in quantum communication, and hold promise for cluster-state-based quantum computing. To master universal quantum processing with linear oscillators, an unconditional nonlinear operation is required. We propose such an operation using light-mediated interaction with another system that possesses a nonlinearity equivalent to more than a quadratic potential. Such a potential grants access to a nonlinear operation that can be broadcast to the target linear system. The nonlinear character of the operation can be verified by observing adequate negative values of the target system's Wigner function and the squeezing of the variance of a certain nonlinear combination of the quadratures below the thresholds attainable by Gaussian states. We explicitly evaluate an optically levitated mechanical oscillator as a flexible source of nonlinearity for a proof-of-principle demonstration of the nonlinearity broadcasting to linear systems, for example, mechanical oscillators or macroscopic atomic spin ensembles.

Incorporating van der Waals (vdW) superconductors into Josephson elements extends circuit-QED beyond conventional Al/AlO$_x$/Al tunnel junctions and enables microwave probes of unconventional condensates and subgap excitations. In this work, we realize a flux-tunable transmon whose nonlinear inductive element is an Al/AlO$_x$/4Hb-TaS$_2$ Josephson junction. The tunnel barrier is formed by sequential deposition and full in-situ oxidation of ultrathin Al layers on an exfoliated 4Hb-TaS$_2$ flake, followed by deposition of a top Al electrode, yielding a robust, repeatable hybrid junction process compatible with standard transmon fabrication. Embedding the device in a three-dimensional copper cavity, we observe a SQUID-like flux-dependent spectrum that is quantitatively reproduced by a standard dressed transmon--cavity Hamiltonian, from which we extract parameters in the transmon regime. Across measured devices we obtain sub-microsecond energy relaxation ($T_1$ from $0.08$ to $0.69~μ$s), while Ramsey measurements indicate dephasing faster than our $16$ ns time resolution. We also find a pronounced discrepancy between the Josephson energy inferred from spectroscopy and that expected from the Ambegaokar--Baratoff relation using room-temperature junction resistances, pointing to nontrivial junction physics in the hybrid Al/AlO$_x$/4Hb-TaS$_2$ system. Although we do not resolve material-specific subgap modes in the present geometry, this work establishes a practical route to integrating 4Hb-TaS$_2$ into coherent quantum circuits and provides a baseline for future edge-sensitive designs aimed at enhancing coupling to boundary and subgap degrees of freedom in vdW superconductors.

We investigate whether the spacetime torsion can modify neutrino flavor oscillations in f(T) gravity. This offers a probe of modified teleparallel gravity in astrophysical environments. By using the Dirac action in teleparallel geometry, we derive an effective coupling between the torsion vector and neutrino current. In the weak-field limit around a spherical mass, we obtain analytical expressions for torsion-induced phase shifts and effective mass-squared differences. Our results indicate that both vacuum oscillations and the Mikheyev-Smirnov-Wolfenstein (MSW) resonance in matter are affected by these torsion-based modifications. Using solar neutrino data from Super-Kamiokande, SNO, Borexino, and KamLAND, we constrain the teleparallel model parameters and also the neutrino-torsion coupling.

We study the Schrödinger equation with $1/r^3$ and attractive $1/r^2$ potentials. Using the quantum defect theory, we obtain analytical solutions for both repulsive and attractive $1/r^3$ interactions. The obtained discrete-scale-invariant energies and wave functions, validated by excellent agreement with numerical results, provide a natural framework for describing the universality of Efimov states in mixed dimension. Specifically, we consider a three-body system consisting of two heavy particles with large dipole moments confined to a quasi-one-dimensional geometry and resonantly interacting with an unconfined light particle. With the Born-Oppenheimer approximation, this system is effectively reduced to the Schrödinger equation with $1/r^3$ and $1/r^2$ potentials, and manifests the Efimov effect. Our analytical solution suggests that, for repulsive dipole interactions, the three-body parameter of the mixed-dimensional Efimov states is universally set by the dipolar length scale, whereas for attractive interactions it explicitly depends on the short-range phase. We also investigate the effects of finite transverse confinement and find that our analytical results are useful for describing the Efimov states composed of two polar molecules and a light atom.

In General Relativity, there exist two objects of limiting compactness, one with a null boundary defining the horizon of a black hole and the other with a timelike boundary defining a Buchdahl star. The two are characterized by gravitational energy equal to or half the mass. Since non-gravitational mass-energy is the source of gravitational energy, both of these objects are manifestly stable. We demonstrate in this letter, in a simple and general way, that the equilibrium state defining the object is indeed stable, independent of the nature of the perturbation.

Employing the well-known unitary equivalence between Fokker-Planck operators and Schrödinger Hamiltonians, we compute the quasi-normal-mode spectrum of ultrarelativistic kinetic theories with momentum-space diffusion. We show that the collision operator reduces to a Dirac-delta Schrödinger problem in one spatial dimension, and to a Coulomb Schrödinger operator with hydrogenic spectrum in three dimensions. Finite spatial wavenumber appears as a perturbation of the associated quantum potential. The hydrodynamic mode is found to obey exact Fick-type diffusion at all real wavenumbers, whereas relativistic kinematics generically produces a continuous ballistic band in the non-hydrodynamic sector, a feature absent in the Newtonian regime.

On a quantum particle in the unit interval $[0,1]$, perform a position measurement with inaccuracy $1/n$ and then a quantum measurement of the projection $|φ\rangle\langleφ|$ with some arbitrary but fixed normalized $φ$. Call the outcomes $X \in[0,1]$ and $Y \in\{0,1\}$. We show that in the limit $n\to\infty$ corresponding to perfect precision for $X$, the probability of $Y=1$ tends to 0 for every $φ$. Since there is no density matrix, pure or mixed, which upon measurement of any $|φ\rangle\langleφ|$ yields outcome 1 with probability 0, our result suggests that a novel type of quantum state beyond Hilbert space is necessary to describe a quantum particle after a perfect position measurement.

Due to its parabolic character, the diffusion equation exhibits instantaneous spatial spreading, and becomes unstable when Lorentz-boosted. According to the conventional interpretation, these features reflect a fundamental incompatibility with special relativity. In this Letter, we show that this interpretation is incorrect by demonstrating that any smooth and sufficiently localized solution of the diffusion equation is the particle density of an exact solution of the relativistic Vlasov-Fokker-Planck equation. This establishes the existence of a causal, stable, and thermodynamically consistent relativistic kinetic theory whose hydrodynamic sector is governed exactly by diffusion at all wavelengths. We further demonstrate that the standard arguments for instability arise from considering solutions that admit no counterpart in kinetic theory, and that apparent violations of causality disappear once signals are defined in terms of the underlying microscopic data.

The simplest approach to include a mass into the electromagnetic vector potential is to modify the Einstein-Maxwell action to the Einstein-Proca form. There are currently no exact analytical solutions for this scenario. However, by using perturbation theory, where both the Proca mass and the black hole charge are small parameters, it is possible to find an exact analytical solution. In this solution, the metric tensor remains unchanged, but the vector potential deviates from the Coulomb potential. In particular, even if the Proca mass is limited by the value $m_γ<10^{-48}\text{g}$, which is the current experimental upper limit for photon mass, it makes a significant contribution to the dynamical equations. In this paper, we study the motion of neutral and charged particles in the vicinity of a weakly Proca-charged black hole, and test the observational implications of the solution of the Einstein-Proca equations for gravitational bending, the black hole shadow, and the fit to the orbits of the Galactic center flares observed by the near-infrared GRAVITY instrument. We find that only extremely cold photons, which are likely scattered before reaching a distant observer, could reveal the non-zero photon mass effect through the black hole shadow. For the Galactic center flare analysis we obtained constraints on the dimensionless Proca parameter to $μ\leq 0.125$ for the electric interaction parameter in the range $-1.1 < \mathrm{Q} < 0.5$, which can be potentially tested by future GRAVITY flare astrometry. Since the Proca parameter is coupled to the black hole mass, the effect of the Proca charge becomes more pronounced for supermassive black holes compared to stellar-mass objects. Our perturbative treatment remains valid essentially up to the horizon, with divergences appearing only in the immediate near-horizon region, where a fully non-perturbative analysis would be required.

Non-metricity provides a natural extension of Riemannian geometry, yet its experimental signatures remain largely unexplored. In this work we investigate how spacetime non-metricity can be probed through high-precision observations, focusing on atomic clocks and gravitational waves as complementary tools. Working within Weyl geometry as a minimal realization of vectorial non-metricity, we formulate observable effects in a gauge-invariant manner and show that they are associated with path-dependent length transport governed by the Weyl field strength. We derive constraints from atomic-clock experiments and demonstrate that, although gravitational waves do not directly source the Weyl field at linear order, its dynamical contribution induces a backreaction on gravitational-wave propagation, leading to an anomalous strain. As a result, the absence of deviations from General Relativity in current gravitational-wave observations already places meaningful and strong constraints on dynamical non-metric degrees of freedom.

The detection of GW230529_181500 suggested the existence of more symmetric black hole-neutron star mergers where the black hole mass can be as low as 2.6 times that of the neutron star. Black hole-neutron star binaries with even more symmetric mass ratios are expected to leave behind massive disks capable of driving bright electromagnetic transients like kilonovae. Currently, there is only a limited number of numerical-relativity simulations of black hole-neutron star mergers in this regime, which are vital for accurate gravitational waveform models and analytical fitting formulas for the remnant properties. Insufficient accuracy of these may lead to misclassification of real events and potentially missed opportunities to locate their electromagnetic counterparts. To fill this gap in the parameter space coverage, we perform simulations of black hole-neutron star mergers with mass ratios $q \in \{1, 1/2, 1/3\}$. We find the gravitational waveform models do not show good agreement with the numerical waveforms, with dephasing at the level of around 1 rad at the merger. We find that the masses of the dynamical ejecta and disk are in good agreement with the available fitting formulas. The analytical formulas for the remnant black hole are in excellent agreement for the black hole mass, but are less accurate with the predictions for its spin. Moreover, we analyze the remnant disk structure and dynamics, deriving the rotation law and identifying global trapped $g$-mode density oscillations. We distinguish three types of accretion in the postmerger and find modulation of the accretion rate by the global oscillations of the disk. Finally, we model the kilonova emission these systems would produce and find that most of them are potentially detectable by Vera C. Rubin Observatory within four days after merger, and by DECam within two days after merger if located at a distance of 200 Mpc.

The sensitivity of a mechanical transducer is ultimately limited by its inherent quantum fluctuations. Here, we use an optically levitated nanoparticle to measure impulsive forces smaller than the particle's zero-point momentum uncertainty. Our approach relies on reversibly squeezing the levitated particle's center-of-mass motion to coherently amplify the perturbation. We demonstrate resolving single impulsive-force kicks as small as 6.9 keV/c, a value 0.6 dB below the sensor's zero-point value.

Generating long-distance quantum entanglement is crucial for advancing quantum information processing. In this work, we propose a protocol for generating remote magnon-phonon entanglement in a hybrid waveguide-magnomechanical system, where multiple spatially separated magnon modes couple to a common waveguide while interacting with their respective phonon modes. By applying tailored pulsed drives and engineering the magnon-phonon interactions, our scheme enables the creation of diverse long-distance and dynamically stable entanglement. Beyond basic magnon-phonon two-mode entanglement, it supports genuine multimode entanglement between a single phonon and multiple magnons, bipartite entanglement between a single magnon and multiple phonons, as well as genuine four-mode entanglement involving two magnons and two phonons. Moreover, we show that dissipative magnon-magnon interactions mediated by traveling photons can generate substantially stronger long-distance entanglement than coherent couplings. Our work provides an experimentally feasible scheme for the remote generation of magnon-phonon entanglement.

We comprehensively investigate the absorption and scattering of massless scalar waves by Frolov black holes, which is a class of regularization of the Reissner--Nordström spacetime. By analyzing the null geodesics, we determine the photon sphere radius and the critical impact parameter, deriving the geometric capture cross section and the classical differential scattering cross section. Utilizing the partial-wave method, we numerically compute the absorption and scattering cross sections across a broad frequency range. Our numerical results show excellent agreement with the low-frequency limit and the high-frequency sinc approximation, as well as with the semiclassical glory approximation. We analyze the dependence of the spectra on the charge and the regularization parameter (Hubble length). Under the horizon-radius normalization, we observe that the total absorption cross section increases with the regularization parameter, a behavior we attribute to the variation of the dimensionless mass. Furthermore, by comparing Frolov, Reissner--Nordström, and Hayward black holes, we demonstrate that their absorption and scattering patterns are nearly indistinguishable when their critical impact parameters or glory impact parameters are matched. This indicates that the photon sphere geometry predominantly governs scalar field interactions, while the detailed core structure plays a secondary role. Our work underscores the potential of wave-based probes to test regular black hole geometries and their observable imprints.

Accurate frequency estimation of oscillating signals over a broad bandwidth is a central task in quantum sensing, yet it is often compromised by spurious responses to higher-order harmonics in realistic multi-frequency environments. Here we experimentally demonstrate a high-accuracy and broadband quantum frequency sensing protocol based on geodesic control, implemented using the electron spin of a single nitrogen-vacancy center in diamond. By engineering an intrinsically single-frequency response, geodesic control enables bias-free frequency estimation with strong suppression of harmonic-induced systematic errors across a wide spectral range spanning from the megahertz to the gigahertz regime. Furthermore, by incorporating synchronized readout, we achieve millihertz-level frequency resolution under noisy signal conditions. Our results provide systematic experimental benchmarking of geodesic control for quantum frequency sensing and establish it as a practical approach for high-accuracy metrology in realistic environments.

Continuous-variable quantum key distribution (CV-QKD) enables information-theoretically secure key generation between legitimate parties. To further enhance system performance, an improved two-way CV-QKD protocol has been proposed, which is accessible in practice and exhibits increased robustness against excess noise. However, in practical implementations, device nonidealities inevitably drive the optical field from the single-mode regime into the continuous-mode regime. In this work, we introduce temporal modes to characterize the evolution of optical fields in the improved two-way protocol and establish a security analysis framework for the continuous-mode scenario based on adaptive normalization with calibrated shot-noise unit. In addition, finite-size effects are taken into account in the analysis. Our results demonstrate that the improved two-way protocol retains a performance advantage over one-way counterpart. The analysis provides useful guidance for the practical implementation and performance optimization of improved two-way CV-QKD systems.

In this work, we investigate precision limitations in spectrophotometry (i.e., spectroscopic concentration measurements) imposed by chemical processes of molecules. Using the recently developed Photon-resolved Floquet theory, which generalizes Maxwell-Bloch theory for higher-order measurement statistics, we analyze a molecular model system subject to chemical reactions whose electronic and optical properties depend on the chemical state. Analysis of sensitivity bounds reveals: (i) Phase measurements are more sensitive than intensity measurements; (ii) Sensitivity exhibits three regimes: photon-shot-noise limited, chemically limited, and intermediate; (iii) Sensitivity shows a turnover as a function of reaction rate due to the interplay between coherent electronic dynamics and incoherent chemical dynamics. Our findings demonstrate that chemical properties must be considered to estimate ultimate precision limits in optical spectrophotometry.

We propose a hybrid quantum device based on the graphene Josephson junctions, where the vibrational degrees of freedom of a graphene membrane couple to the superconducting circuits. The flexural mode-controlled tunneling of the Cooper pairs introduces a strong and tunable coupling even at the zero-point fluctuations level. By employing this interaction, we show that a parametric process can be efficiently implemented. We then investigate foundational and technological applications of our hybrid device empowered by nonlinear interactions, with fast generation of non-classical mechanical states, and critically enhanced quantum sensing under suitable quantum control. Our work provides the possibility of employing the graphene motional degree of freedom for quantum information processing in circuit quantum nanomechanical structures.

We present a Rydberg atom-based microwave electric field sensor that achieves extended dynamic range and enhanced sensitivity across a broad bandwidth. By characterizing the Autler-Townes (AT) splitting induced by a single-tone microwave field, we demonstrate a spectroscopic method that simultaneously extracts both the microwave frequency and electric field strength directly from the splitting pattern. We implement dual-tone heterodyne detection, achieving a minimum detectable field strength on the order of uV/cm and a sensitivity in the sub-uV/cm/Hz^1/2 regime, while extending the operational bandwidth up to 3 GHz. Through systematic characterization of frequency and power dependencies, we identify optimal operating conditions to minimize power broadening in the resonant AT regime and maximize sensitivity in the far-off-resonance AC Stark regime. The resulting platform combines high sensitivity, broad bandwidth, and a dynamic range of approximately 90 dB, establishing Rydberg atoms as practical sensors for precision electric field metrology.

The advent of promising quantum error correction (QEC) codes with efficient resource utilization and high-performance fault-tolerant quantum memories signifies a critical step towards realizing practical quantum computation. While surface codes have been a dominant approach, their limitations have spurred the development of more advanced QEC architectures. These advanced codes often present increased complexity, demanding innovative decoding methodologies. This work investigates the application of reinforcement learning (RL) techniques, including hybrid and multi-agent approaches, to enhance the decoding of various advanced QEC architectures. By leveraging the ability of RL to learn optimal strategies from noisy syndrome measurements, we explore the potential for achieving improved logical error rates and scalability compared to traditional decoding methods. Our approach examines the adaptation of reinforcement learning to exploit the structural properties of these modern QEC models. We also explore the benefits of combining different RL algorithms to address the multifaceted nature of the decoding problem, considering factors such as code degeneracy and real-world noise characteristics. With our proposed method, we are able to demonstrate that an autonomously trained agent can derive decoding schemes for the complex decoding requirement of advanced QEC architectures.

In this work, we analyze the Wheeler-DeWitt equation with scale-dependent gravitational couplings within the framework of asymptotically safe gravity. In the Hamiltonian formulation based on a renormalization-group improved Einstein-Hilbert action, the consistency of the theory and the Poisson algebra of constraints have been clarified. Within this framework, we show that, despite the explicit scale dependence of Newton's constant, the classical solutions are generically unaffected by the running of the coupling. We then derive the Wheeler-DeWitt equation incorporating the scale dependence of the gravitational couplings and analyze its solutions in the minisuperspace framework. In the classical limit, while the scale dependence of Newton's constant does not affect the classical behavior, the running of the cosmological constant can contribute to the classical solutions. Moreover, we show that the quantum behavior in the ultraviolet regime acts toward suppressing singularity formation in all cases, independently of how the renormalization-group scale is identified with spacetime coordinates and of the relative magnitudes of the ultraviolet fixed points of the running Newton's constant and cosmological constant.

Operating superconducting qubits at dynamical sweet spots (DSSs) suppresses decoherence from low-frequency flux noise. A key open question is how long coherence can be extended under this strategy and what fundamental limits constrain it. Here we introduce a fully parameterized, multi-objective periodic-flux modulation framework that simultaneously optimizes energy relaxation $T_1$ and pure dephasing $T_φ$, thereby quantifying the tradeoff between them. For fluxonium qubits with realistic noise spectra, our method enhances $T_φ$ by a factor of 3-5 compared with existing DSS strategies while maintaining $T_1$ in the hundred-microsecond range. We further prove that, although DSSs eliminate first-order sensitivity to low-frequency noise, relaxation rate cannot be reduced arbitrarily close to zero, establishing an upper bound on achievable $T_1$. At the optimized working points, we identify double-DSS regions that are insensitive to both DC and AC flux, providing robust operating bands for experiments. As applications, we design single- and two-qubit control protocols at these operating points and numerically demonstrate high-fidelity gate operations. These results establish a general and useful framework for Pareto-front engineering of DSSs that substantially improves coherence and gate performance in superconducting qubits.

Privacy is a fundamental requirement in distributed quantum sensing networks, where multiple clients estimate spatially distributed parameters using shared quantum resources while interacting with potentially untrusted servers. Despite its importance, existing privacy conditions rely on idealized quantum bounds and do not fully capture the operational constraints imposed by realistic measurements. Here, we introduce a universal operational privacy framework for distributed quantum sensing, formulated in terms of the experimentally accessible classical Fisher information matrix and applicable to arbitrary protocols characterized by singular information structures. The proposed condition provides a protocol-independent criterion ensuring that no information about individual parameters is accessible to untrusted parties. We further experimentally demonstrate that a distributed quantum sensing protocol employing fewer photons than the number of estimated parameters simultaneously satisfies the universal privacy condition and achieves Heisenberg-limited precision. Our results establish universal operational constraints governing privacy in distributed quantum sensing networks and provide a foundation for practical, privacy-preserving quantum sensing beyond full-rank regimes.

Quantum Random Access Codes (QRACs) embody the fundamental trade-off between the compressibility of information into limited quantum resources and the accessibility of that information, serving as a cornerstone of quantum communication and computation. In particular, the $(n, n-1)$-QRACs, which encode $n$ bits of classical information into $n-1$ qubits, provides an ideal theoretical model for verifying quantum advantage in high-dimensional spaces; however, the analytical derivation of optimal codes for general $n$ has remained an open problem. In this paper, we establish an analytical construction method for $(n, n-1)$-QRACs by using an explicit operator formalism. We prove that this construction strictly achieves the numerically conjectured upper bound of the average success probability, $\mathcal{P} = 1/2 + \sqrt{(n-1)/n}/2$, for all $n$. Furthermore, we present a systematic algorithm to decompose the derived optimal POVM into standard quantum gates. Since the resulting decoding circuit consists solely of interactions between adjacent qubits, it can be implemented with a circuit depth of $O(n)$ even under linear connectivity constraints. Additionally, we analyze the high-dimensional limit and demonstrate that while the non-commutativity of measurements is suppressed, an information-theoretic gap of $O(\log n)$ from the Holevo bound inevitably arises for symmetric encoding. This study not only provides a scalable implementation method for high-dimensional quantum information processing but also offers new insights into the mathematical structure at the quantum-classical boundary.

The nonlinear Schrödinger equation (NLSE) is a fundamental model that describes diverse complex phenomena in nature. However, simulating the NLSE on a quantum computer is inherently challenging due to the presence of the nonlinear term. We propose a hybrid quantum-classical framework for simulating the NLSE based on the split-step Fourier method. During the linear propagation step, we apply the kinetic evolution operator to generate an intermediate quantum state. Subsequently, the Hadamard test is employed to measure the Fourier components of low-wavenumber modes, enabling the efficient reconstruction of nonlinear potentials. The phase transformation corresponding to the reconstructed potential is then implemented via a quantum circuit using the phase kickback technique. To validate the efficacy of the proposed algorithm, we numerically simulate the evolution of a Gaussian wave packet, a soliton wave, and the wake flow past a cylinder. The simulation results demonstrate excellent agreement with the corresponding classical solutions. This work provide a concrete basis for analyzing accuracy-cost trade-offs in quantum-classical simulations of nonlinear dispersive wave dynamics.

We find a tight characterization of the strong converse exponent for randomness extraction against quantum side information. In contrast to previous tight bounds, we employ a composable error criterion given by the fidelity (or purified distance) to a uniform distribution in product with the marginal state. The characterization is in terms of a club-sandwiched conditional entropy recently introduced by Rubboli, Goodarzi and Tomamichel and used by Li, Li and Yu to establish the strong converse exponent for the case of classical side information. This provides the first precise operational interpretation of this family of conditional entropies in the quantum setting.

Exact synthesis provides unconditional optimality and canonical structure, but is often limited to small, carefully scoped regimes. We present an exact synthesis framework for two-qubit circuits over the Clifford+$T$ gate set that optimizes $T$-count exactly. Our approach exhausts a bounded search space, exploits algebraic canonicalization to avoid redundancy, and constructs a lookup table of optimal implementations that turns synthesis into a query. Algorithmically, we combine meet-in-the-middle ideas with provable pruning rules and problem-specific arithmetic designed for modern hardware. The result is an exact, reusable synthesis engine with substantially improved practical performance.

We extend \cite{chen2025srkbp} by analyzing the complexity of the $k$-block-positivity testing algorithm. In this paper, we investigate a symmetry reduction scheme based on rectangular shaped Young diagrams. Connecting the complexity to the dimensions of irreducible representations of $\mathrm{U}(d)$, we derive an explicit formula for the complexity, which also clarifies why the semidefinite program hierarchy collapses in the $k=d$ case.

We investigate the post-quench dynamics of the quantum geometric tensor (QGT) of 1D periodic systems with a suddenly changed Hamiltonian. The diagonal component with respect to the crystal momentum gives a metric corresponding to the variance of the time-evolved position, and its coefficient of the quadratic term in time is the group-velocity variance, signaling ballistic wavepacket dispersion. The other diagonal QGT component with respect to time reveals the energy variance. The off-diagonal QGT component features a real part as a covariance and an imaginary part representing a quench-induced curvature. Using the Su-Schrieffer-Heeger (SSH) model as an example, our numerical results of different quenches confirm that the post-quench QGT is governed by physical quantities and local geometric objects from the initial state and post-quench bands, such as the Berry connection, group velocities, and energy variance. Furthermore, the connections between the QGT and physical observables suggest the QGT as a comprehensive probe for nonequilibrium phenomena.

Quantum differential privacy provides a rigorous framework for quantifying privacy guarantees in quantum information processing. While classical correlations are typically regarded as adversarial to privacy, the role of their quantum analogue, entanglement, is not well understood. In this work, we investigate how quantum entanglement fundamentally shapes quantum local differential privacy (QLDP). We consider a bipartite quantum system whose input state has a prescribed level of entanglement, characterized by a lower bound on the entanglement entropy. Each subsystem is then processed by a local quantum mechanism and measured using local operations only, ensuring that no additional entanglement is generated during the process. Our main result reveals a sharp phase-transition phenomenon in the relation between entanglement and QLDP: below a mechanism-dependent entropy threshold, the optimal privacy leakage level mirrors that of unentangled inputs; beyond this threshold, the privacy leakage level decreases with the entropy, which strictly improves privacy guarantees and can even turn some non-private mechanisms into private ones. The phase-transition phenomenon gives rise to a nonlinear dependence of the privacy leakage level on the entanglement entropy, even though the underlying quantum mechanisms and measurements are linear. We show that the transition is governed by the intrinsic non-convex geometry of the set of entanglement-constrained quantum states, which we parametrize as a smooth manifold and analyze via Riemannian optimization. Our findings demonstrate that entanglement serves as a genuine privacy-enhancing resource, offering a geometric and operational foundation for designing robust privacy-preserving quantum protocols.

Extreme mass-ratio inspirals (EMRIs) in relativistic accretion discs are a key science target for the upcoming LISA mission. Existing models of disc-EMRI interactions typically rely on crude dynamical friction or Newtonian planetary migration prescriptions, which fail to capture the relativistic fluid response induced by the binary potential. In this work we address this gap by providing the relativistic calculation. We apply standard methods from self-force theory, black hole perturbation theory, and relativistic stellar perturbation theory to perform the full fluid calculation of the relativistic analogue of planetary migration for the first time. We calculate the response of a fluid in the perturbing potential of an EMRI consistently incorporating pressure effects. Using a master enthalpy-like variable and linearised fluid theory, we reconstruct the fluid perturbations and relativistic spiral arm structure for a range of spin values in the Kerr geometry. We conclude by deriving a relativistic torque-balance equation that enables computation and comparison of local torques with advected angular momentum through the disc. This opens a promising route towards establishing torque-balance relations between integrated disc torques arising from fluid perturbations and the forces acting on EMRIs embedded in matter.

This article is an expository account aimed at viewing entanglement in finite-dimensional quantum many-body systems as a phenomenon of global geometry. While the mathematics of general quantum states has been studied extensively, this article focuses specifically on their entanglement. When a quantum system varies over a classical parameter space, each fiber may look like the same Hilbert space, yet there may be no global identification because of twisting in the gluing data. Describing this situation by an Azumaya algebra, one always obtains the family of pure-state spaces as a Severi-Brauer scheme. The main focus is to characterize the condition under which the subsystem decomposition required to define entanglement exists globally and compatibly, by a reduction to the stabilizer subgroup of the Segre variety, and to explain that the obstruction appears in the Brauer class. As a consequence, quantum states yield a natural filtration dictated by entanglement on the Severi-Brauer scheme. Using a spin system on a torus as an example, we show concretely that the holonomy of the gluing can produce an entangling quantum gate, and can appear as an obstruction class distinct from the usual Berry numbers or Chern numbers. For instance, even for quantum systems that have traditionally been regarded as having no topological band structure, the entanglement of their eigenstates can be related to global geometric universal quantities, reflecting the background geometry.

Timelike Liouville field theory (also known as imaginary Liouville theory or imaginary Gaussian multiplicative chaos) is expected to describe two-dimensional quantum gravity in a positive-curvature regime, but its path integral is not a probability measure and rigorous exact computations are currently available only in the charge-neutral (integer screening) case. In this paper we show that at the special coupling $b=1/\sqrt{2}$, the Coulomb-gas expansion of the timelike path integral becomes explicitly computable beyond charge neutrality. The reason is that the $n$-fold integrals generated by the interaction acquire a Vandermonde/determinantal structure at $b=1/\sqrt{2}$, which allows exact evaluation in terms of classical special functions. We derive Mellin-Barnes type representations (involving the Barnes $G$-function and, in a three-point case, Gauss hypergeometric functions) for the zero- and one-point functions, for an antipodal two-point function, and for a three-point function with a resonant insertion $α_2=b$. We then address the subtle zero-mode integration: after a Gaussian regularization we obtain an explicit renormalized partition function $C(1/\sqrt{2},μ)=e(4π\sqrt2 μ)^{-1}$, identify distributional limits in the physically relevant regime $α_j=\frac{1}{2}Q+\mathrm{i} P_j$, and compare with the Hankel-contour prescription recently proposed in the physics literature. These results provide the first rigorously controlled family of exact calculations in timelike Liouville theory outside charge neutrality.

To address the discrimination problem of binary phase-shift keyed displaced squeezed vacuum states (S-BPSK), this paper proposes an Inverse-squeezing Kennedy (IS-Kennedy) receiver. This architecture incorporates an inverse-squeezing operator following the displacement operation of a conventional Kennedy receiver, mapping the S-BPSK signals onto equivalent large-amplitude coherent states. Furthermore, it employs a photon-number-resolving (PNR) detector to perform maximum a posteriori (MAP) decision-making. Theoretical analysis demonstrates that, under ideal conditions, the IS-Kennedy receiver effectively translates the transmitter's squeezing resources into a displacement gain at the receiver. Consequently, its error probability approaches the Helstrom bound across the entire energy spectrum, remaining within a constant factor of 3 dB. In the low-photon-number regime ($N \approx 0.6$), the proposed scheme surpasses the coherent-state limit, achieving an error rate below 1\%. Furthermore, this paper provides an in-depth analysis of system performance under non-ideal conditions, revealing the robustness of PNR detection against background dark counts and a characteristic ``parity photon-number step'' saturation effect arising from squeezing parameter mismatch.

We evaluate the quantum resource requirements for ATP/metaphosphate hydrolysis, one of the most important reactions in all of biology with implications for metabolism, cellular signaling, and cancer therapeutics. In particular, we consider three algorithms for solving the ground state energy estimation problem: the variational quantum eigensolver, quantum Krylov, and quantum phase estimation. By utilizing exact classical simulation, numerical estimation, and analytical bounds, we provide a current and future outlook for using quantum computers to solve impactful biochemical and biological problems. Our results show that variational methods, while being the most heuristic, still require substantially fewer overall resources on quantum hardware, and could feasibly address such problems on current or near-future devices. We include our complete dataset of biomolecular Hamiltonians and code as benchmarks to improve upon with future techniques.

Accurate quantum state readout is crucial for error correction and algorithms, but measurement errors are detrimental. Readout fidelity is typically limited by a poor signal-to-noise ratio (SNR) and energy relaxation ($T_1$ decay), a significant problem for superconducting qubits. While most approaches classify results using clustering algorithms on integrated readout signals, these methods cannot distinguish a qubit that was initially in the ground state from one that decayed to it during measurement. We instead propose using machine learning (ML) on the raw, non-integrated analog signal. We apply time-series classification models, such as a long short-term memory (LSTM) network, to the full data trajectory. We find that our LSTM model, combined with filtering and feature engineering, consistently outperforms clustering. The largest improvements come from reclassifying points in the boundary regions between clusters. These points correspond to atypical measurement records, likely due to transient or noisy features lost during data integration. By retaining temporal information, sequence-aware models like LSTMs can better discriminate these trajectories, whereas clustering methods based on integrated values are more prone to misclassification.

The frequency- and time-resolved second order quantum coherence function (g(2)(τ)) of single-molecule fluorescence has recently been proposed as a powerful new quantum light spectroscopy that can reveal intrinsic quantum coherence in excitation energy transfer in molecular systems ranging from simple dimers to photosynthetic complexes. Yet, no experiments have been reported to date. Here, we have developed a single-molecule fluorescence g(2)(τ) quantum light spectroscopy (SMFg2-QLS) that can simultaneously measure the fluorescence intensity, lifetime, spectra, and g(2)(τ) with frequency resolution, for a single molecule in a controlled environment at both room temperature and cryogenic temperature. As a proof of principle, we have studied single molecules of IDTBT (indacenodithiophene-co-benzothiadiazole), a semiconducting donor-acceptor conjugated copolymer with a chain-like structure that shows a high carrier mobility and annihilation-limited long-range exciton transport. We have observed different g(2)(τ=0) values with different bands or bandwidths of the single molecule IDTBT fluorescence. The general features are consistent with theoretical predictions and suggest non-trivial excited state quantum dynamics, possibly showing quantum coherence, although further analysis and confirmation will require additional theoretical calculations that take into account the complexity and inhomogeneity of individual IDTBT single molecular chains. Our results demonstrate the feasibility and promise of frequency- and time-resolved SMFg2-QLS to provide new insights into molecular quantum dynamics and to reveal signatures of intrinsic quantum coherence in photosynthetic light harvesting that are independent of the nature of the light excitation.

Using a combination of \textit{Hubble Space Telescope} and \textit{James Webb Space Telescope} imaging, a runaway supermassive black hole (RBH-1) was recently identified with an inferred velocity of $954^{+110}_{-126}\,\mathrm{km\,s^{-1}}$, likely ejected from a compact star-forming galaxy (denoted as GX) at $z \approx 0.96$. Assuming the runaway black hole was the outcome of the gravitational-wave-driven merger of two black holes, we use its measured runaway velocity together with gravitational-wave recoil predictions from numerical relativity and black hole perturbation theory to constrain the mass ratio and spin configuration of the progenitor SMBHs that overcame the final-parsec problem and merged $\sim 70$~Myr ago. We find that the progenitor binary must have been precessing, with a mass ratio $m_1/m_2\lesssim 6$, and that the more massive SMBH must have possessed a high spin (dimensionless spin magnitude $\sim 0.75$) in order to generate a recoil of this magnitude. This has important astrophysical implications as similar SMBH mergers can be an interesting source population for the upcoming LISA mission with signal-to-noise ratios $\gtrsim$ 1000. Furthermore, the progenitor SMBH properties imply that GX was likely formed through a major, gas-rich (``wet'') merger between two galaxies of comparable mass, with a mass ratio $\lesssim 4$.

We propose a scheme for accumulating entanglement between long-lived qudits provided by central nuclear spins of defect centers. Assuming a generic setting, the electron spin of each node acts as the communication qubit and may be entangled with other nodes, e.g., through a spin-photon interface. The generally available Ising component of the hyperfine interaction is shown to facilitate repeated entanglement transfer onto memory qudits of arbitrary dimension $d\le 2I+1$ with $I$ the nuclear spin quantum number. When $d$ is set to an integer power of two, maximal entanglement can be generated deterministically and without intermittent driving of nuclear spins. The scheme is applicable to several candidate systems, including the $^{73}$Ge germanium vacancy in diamond.

Quantum hardware suffers from intrinsic device heterogeneity and environmental drift, forcing practitioners to choose between suboptimal non-adaptive controllers or costly per-device recalibration. We derive a scaling law lower bound for meta-learning showing that the adaptation gain (expected fidelity improvement from task-specific gradient steps) saturates exponentially with gradient steps and scales linearly with task variance, providing a quantitative criterion for when adaptation justifies its overhead. Validation on quantum gate calibration shows negligible benefits for low-variance tasks but $>40\%$ fidelity gains on two-qubit gates under extreme out-of-distribution conditions (10$\times$ the training noise), with implications for reducing per-device calibration time on cloud quantum processors. Further validation on classical linear-quadratic control confirms these laws emerge from general optimization geometry rather than quantum-specific physics. Together, these results offer a transferable framework for decision-making in adaptive control.

In this work, we revisit a recently reported generalization of the Bañados--Teitelboim--Zanelli black hole arising in New Massive Gravity sourced by the quantum fluctuations of scalar matter, now examined through the lens of a purely classical framework. We show that the same geometry, distinguished by its logarithmic asymptotic structure, emerges as the unique static solution of Einstein gravity coupled to an exponential nonlinear electrodynamics. We trace the origin of this correspondence and prove that this geometry belongs to a unique class of metrics constituting the intersection of the moduli spaces of the static and circularly symmetric sectors of the two theories, thereby revealing a dynamical equivalence between them. An explicit mapping is established between the global charges of the nonlinearly charged black holes and the parameters governing the quantum backreaction in New Massive Gravity, allowing for a natural reinterpretation of the quantum imprints in terms of classical charges. A detailed analysis of the horizon structure of these spacetimes is presented. In addition, the full thermodynamics of the more general configurations is constructed using the Iyer--Wald formalism, from which we derive the first law and the associated Smarr relation. Altogether, our results provide a classical realization of a semiclassical spacetime and point toward a broader correspondence between higher-curvature corrections in quantum gravity and nonlinear effects in self-gravitating electrodynamics in three dimensions.

If a quantum walk starting on a vertex tends to stay at home, then that vertex is said to be sedentary. We prove that almost all planar graphs and almost all trees contain at least two sedentary vertices for any assignment of edge weights -- a result that suggests vertex sedentariness is a common phenomenon in trees and planar graphs. For weighted bipartite graphs, we show that a vertex is not sedentary whenever 0 does not belong to its eigenvalue support. Consequently, each vertex in a nonsingular weighted bipartite graph is not sedentary, a stark contrast to weighted trees and weighted planar graphs. A corollary of this result is that every vertex in a bipartite graph with a unique perfect matching is not sedentary for any assignment of edge weights. We also construct new families of weighted bipartite graphs with sedentary vertices using the bipartite double and subdivision operations. Finally, we show that unweighted paths and unweighted even cycles contain no sedentary vertices.

Position verification schemes are interactive protocols where entities prove their physical location to others; this enables interactive proofs for statements of the form "I am at a location $L$." Although secure position verification cannot be achieved with classical protocols (even with computational assumptions), they are feasible with quantum protocols. In this paper we introduce the notion of zero-knowledge position verification, which generalizes position verification in two ways: 1. enabling entities to prove more sophisticated statements about their locations at different times (for example, "I was NOT near location $L$ at noon yesterday"). 2. maintaining privacy for any other detail about their true location besides the statement they are proving. We construct zero-knowledge position verification from standard position verification and post-quantum one-way functions. The central tool in our construction is a primitive we call position commitments, which allow entities to privately commit to their physical position in a particular moment, which is then revealed at some later time.

Traditionally, Quantum Information, and Quantum Communication specifically, have been focused on qubit-based architectures. Recent results, however, highlighted that higher dimensional architectures (qudit-based) may present advantages both in terms of communication and computation; a family of channels called Multi-level Amplitude Damping (MAD) channels, which are a possible qudit generalization of the well known Amplitude Damping Channels, is able to model energy decay processes that may happen during signal transmission. In this work, the Quantum Capacity of 4-dimensional MAD's is studied, relying on a technique for computing it even outside of degradable and antidegradable conditions. We also characterized the complete region of antidegradability and degradability in the parameter space for a generic d-dimensional MAD using both analytical and semi-numerical methods.

During the inflationary epoch, we investigate the reconstruction of the background variables within the framework of Einstein-Gauss-Bonnet gravity, considering the scalar spectral index $n_s(N)$ and the tensor-to-scalar ratio $r(N)$, where $N$ denotes the number of $e-$folds. Under a general formalism, we determine the effective potential and the coupling function associated with the Gauss-Bonnet term as functions of the cosmological parameters $n_s(N)$ and $r(N)$, respectively. To implement the reconstruction methodology for the background variables, we study an example in which the attractors for the index $n_s$ and the ratio $r$ are in agreement with Atacama Cosmology Telescope (ACT) data. In this context, explicit expressions for the effective potential $V(φ)$ and the coupling parameter $ξ(φ)$ are reconstructed. Moreover, the reconstruction based on observational parameters shows that $V(φ)\not\propto 1/ξ(φ)$, in contrast to the assumption adopted in the literature for the study of the evolution of the universe in Einstein-Gauss-Bonnet gravity.

In this work we formulate a two-parameter family of transformations in flat Minkowksi spacetime that smoothly interpolates between motion with constant initial/final velocity (inertial coordinates), and with constant acceleration (Rindler coordinates \cite{Rindler:1956}), which we term Inertial-to-Rindler (I2R) coordinates. We revisit the Twin ``Paradox" and show how the new I2R coordinates justify the ``immediate-" and ``gradual-turnaround" scenarios discussed in many texbooks and articles. We also examine the radar time formulation of hypersurfaces of simultaneity by Dolby and Gull \cite{Dolby_Gull:2001} for these new coordinates as we transition from zero to uniform acceleration. Finaly we re-examine the negative frequency content of a purely positive frequency Minkowski plane wave as observed by the I2R observer, and derive perturbative corrections to the Unruh \cite{Unruh:1976} temperature for the two cases of initial/final velocities slightly greater than zero, and slightly less than the speed of light - the latter of which characterizes constant acceleration motion. We argue for a proposed velocity-dependent generalization of the Unruh temperature that smoothly varies from zero at zero-acceleration, to the standard form at constant acceleration.

We compare Krylov's state complexity with an information-geometric (IG) measure of complexity for the quantum evolution of two-level systems. Focusing on qubit dynamics on the Bloch sphere, we analyze evolutions generated by stationary and nonstationary Hamiltonians, corresponding to geodesic and nongeodesic trajectories. We formulate Krylov complexity in geometric terms, both instantaneously and in a time-averaged sense, and contrast it with an IG complexity of quantum evolutions characterized in terms of efficiency and curvature. We show that the two measures reflect fundamentally different aspects of quantum dynamics: Krylov's state complexity quantifies the directional spread of the evolving state relative to the initial state, whereas the IG complexity captures the effective volume explored along the trajectory on the Bloch sphere. This geometric distinction explains their inequivalent behavior and highlights the complementary nature of state-based and information-geometric notions of complexity in quantum systems.

The phenomena of induced transparency, with the typical examples of electromagnetically induced transparency (EIT) in atomic media and coupled optical resonators, have attracted tremendous interest since their discoveries. Due to the limitations of the involved elements, however, near-100\% transmissions were reported under highly demanding experimental conditions for atomic and other media. Based on a structure of three linearly coupled optical resonators, an active one carrying a possibly arbitrary optical gain and two passive ones simply with dissipation, we demonstrate that a transmitted light field can become completely transparent through the structure, which displays all properties similar to those of EIT. Manifested by a destructive interference to annihilate the intracavity field in the resonator directly coupled to the input, this complete transparency exists for any feasible power of the transmitted field and all realizable coupling strengths of the dark resonator with the input and the neighboring resonator, as long as the inter-cavity coupling for two other resonators is adjustable over a suitable range. A free control on the transparency window size and output field intensity can be realized by tuning two inter-cavity couplings without modifying the built-in system parameters.

The growing number of gravitational-wave (GW) observations allows for constraints to be placed on the underlying population of black holes; current estimates show that black hole spins are small, with binaries more likely to have comparable component masses. Since general relativistic effects, such as spin-induced orbital precession and higher order multipole moments, are more likely to be observed for asymmetric binary systems, a direct measurement remains unlikely. Nevertheless, we continue to consistently probe these effects by performing Bayesian inference with our most accurate and computationally expensive models. As the number of GW detections increases, it may soon become infeasible to consistently use these models for analyses. In this paper, we provide a selection criterion that determines when less accurate and computationally cheaper models can be used without giving biased estimates for the population properties of black holes in the Universe. We show that when using our selection criterion, comparable estimates can be obtained for the underlying mass and spin distribution of black holes for a simulated "worst-case" scenario population, while reducing the overall cost of performing Bayesian inference on our population by $\sim 20\%$. We anticipate a reduction of up to $78\%$ in the overall cost for an astrophysically motivated population, since there are fewer events with observable spin-precession and higher order multipole power.

Relativistic spin effects drive subtle molecular phenomena ranging from intersystem crossing in photodynamic therapy to spin-mediated catalysis and high-resolution spectroscopy. These effects are described by the Pauli-Breit Hamiltonian, which extends the nonrelativistic electronic Hamiltonian by including one- and two-electron spin-orbit and spin-spin interactions. First-principles simulations of the full Pauli-Breit Hamiltonian rapidly become intractable on classical computers due to the exponential growth of the Hilbert space and the complexity of two-body spin-dependent terms. We propose a fault-tolerant quantum algorithm for computing molecular energy levels and properties governed by the Pauli-Breit Hamiltonian. Our approach block-encodes the relativistic Hamiltonian in a second-quantized, doubly factorized representation. By reformulating the Hamiltonian in a symmetry-adapted Majorana basis, we construct efficient linear-combination-of-unitaries circuits that encode spin-orbit interactions without effective or mean-field approximations. We introduce spin-controlled Pauli-SWAP networks that decouple spin and orbital control logic, enabling a unified treatment of relativistic spin mixing with only modest overhead relative to spin-free simulations. We analyze quantum resources in terms of logical qubits and T-gate complexity, showing that explicit spin degrees of freedom do not worsen the asymptotic scaling. The prefactor is reduced by a factor of two compared to direct linear-combination-of-unitaries approaches. Finally, we outline a hybrid quantum-classical workflow for designing photodynamic therapy photosensitizers, artificial photosynthesis catalysts, and other systems where accurate relativistic spin effects are essential.

We investigate and compare two particle number conserving protocols for the preparation of a topologically nontrivial state. The first is derived from thermally coupling the system to a cold bath, while the second is based on engineered dissipation. We numerically study the time required to reach the target state as well as its robustness against physically important perturbations. Crucially, in both protocols the cooling capability is limited by dissipatively induced diffusion processes. The resulting quadratic scaling of the cooling time with system size is corroborated also analytically using mean-field approximations and a purely classical random walk model. Furthermore, we find that the engineered protocol admits a unique and stable dark state, which contributes to an ongoing discussion regarding the applicability of dissipative state preparation to many-body systems.

We introduce multivariate multicycle (MM) codes, a new family of quantum error correcting codes that unifies and generalizes bivariate bicycle codes, multivariate bicycle codes, abelian two-block group algebra codes, generalized bicycle codes, trivariate tricycle codes, and n-dimensional toric codes. MM codes are Calderbank-Shor-Steane (CSS) codes defined from length-t chain complexes with $t \ge 4$. The chief advantage of these codes is that they possess metachecks and high confinement that permit complete single-shot decoding, while also having additional algebraic structure that might enable logical non-Clifford gates. We offer a framework that facilitates the construction of long-length chain complexes through the use of Koszul complex. In particular, obtaining explicit boundary maps (parity check and metacheck matrices) is particularly straightforward in our approach. This simple but very general parameterization of codes permitted us to efficiently perform a numerical search, where we identify several MM code candidates that demonstrate these capabilities at high rates and high code distances. Examples of new codes with parameters $[[n,k,d]]$ include $[[96, 12, 8]]$, $[[96, 44, 4]]$ $[[144, 40, 4]]$, $[[216, 12, 12]]$, $[[360, 30, 6]]$, $[[384, 80, 4]]$, $[[486, 24, 12]]$, $[[486, 66, 9]]$ and $[[648, 60, 9]]$. Notably, our codes achieve confinement profiles that surpass all known single-shot decodable quantum CSS codes of practical blocksize.

The state $ρ$ of a quantum system can be represented by a vector $\mathbf{P}_{\mathcal{M}}(ρ)$ of outcome probabilities for a set of measurements $\mathcal{M}$. Such representations appear throughout physics, for example, in quantum field theory via correlation functions and in quantum foundations within generalized probabilistic frameworks. In this work, we identify an unavoidable tension: to enable operationally meaningful statements, the map ${ρ\mapsto \mathbf{P}_{\mathcal{M}}(ρ)}$ must be topologically robust $\unicode{x2013}$ preserving the notion of closeness between states. Yet, a probability representation that is topologically robust cannot simultaneously retain other essential structure, such as the subsystem structure.

The canonical commutation relation is a cornerstone of quantum theory and underlies the Heisenberg uncertainty principle. Although uncertainty relations have been extensively tested, direct verifications of the underlying commutation relation itself have remained elusive. We report an experimental demonstration of the bosonic commutation relation for optical field operators based on measurements of two distinct intensity correlation functions. From these measurements, we extract the expectation value of the field-operator commutator for both a single-photon state and coherent state. In both cases, the measured values are consistent with unity, in quantitative agreement with quantum theory.

In this paper, we introduce post-selection games, a generalization of nonlocal games where each round can be not only won or lost by the players, but also discarded by the referee. Such games naturally formalize possibilistic proofs of nonlocality, such as Hardy's paradox. We develop algorithms for computing the local and Tsirelson bounds of post-selection games. Furthermore, we show that they have an unbounded advantage in statistical power over traditional nonlocal games, making them ideally suited for analysing Bell tests with low detection efficiency.

Effective field theories consistent with quantum gravity obey surprising finiteness constraints, appearing in several distinct but interconnected forms. In this work we develop a framework that unifies these observations by proposing that the defining data of such theories, as well as the landscape of effective field theories that are valid at least up to a fixed cutoff, admit descriptions with a uniform bound on complexity. To make this precise, we use tame geometry and work in sharply o-minimal structures, in which tame sets and functions come with two integer parameters that quantify their information content; we call this pair their tame complexity. Our Finite Complexity Conjectures are supported by controlled examples in which an infinite Wilsonian expansion nevertheless admits an equivalent finite-complexity description, typically through hidden rigidity conditions such as differential or recursion relations. We further assemble evidence from string compactifications, highlighting the constraining role of moduli space geometry and the importance of dualities. This perspective also yields mathematically well-defined notions of counting and volume measures on the space of effective theories, formulated in terms of effective field theory domains and coverings, whose finiteness is naturally enforced by the conjectures.

Every astrophysical stochastic gravitational wave (GWB) is limited by the amount of rest mass available to be converted into gravitational radiation. Here we derive a population-agnostic scaling law that places an absolute energetic ceiling on stochastic backgrounds across the entire GW frequency spectrum, from nanoHertz to kilohertz. We apply this framework to bound the backgrounds from supermassive black hole binaries, intermediate-mass black hole captures by supermassive black holes in AGN disks, extreme mass-ratio inspirals, binary neutron stars, Population III remnants, and stellar-mass binary black holes. We find that the energetic ceiling for supermassive black hole binaries is $A \leq 1.6^{+0.3}_{-0.3} \times 10^{-15}$ at a reference frequency of $1\,{\rm yr}^{-1}$. This astrophysical GWB ceiling is within $1σ$ with the GWB amplitude reported by NANOGrav, EPTA, and PPTA, implying that the current observed signal is consistent with being powered by a population of ultramassive black holes ($M_\bullet \gtrsim 10^{10}\,M_\odot$). Finally, we demonstrate that the total astrophysical background from all channels combined cannot exceed $Ω_{\rm gw} \sim 10^{-7}$.

In this work, we study the Hamiltonian Decoded Quantum Interferometry (HDQI) for the general Hamiltonians $H=\sum_ic_iP_i$ on an $n$-qubit system, where the coefficients $c_i\in \mathbb{R}$ and $P_i$ are Pauli operators. We show that, given access to an appropriate decoding oracle, there exist efficient quantum algorithms for preparing the state $ρ_{\mathcal P}(H) = \frac{\mathcal P^2(H)}{\text{Tr}[\mathcal P^2(H)]}$, where $\mathcal P(H)$ denotes the matrix function induced by a univariate polynomial $\mathcal P(x)$. Such states can be used to approximate the Gibbs states of $H$ for suitable choices of polynomials. We further demonstrate that the proposed algorithms are robust to imperfections in the decoding procedure. Our results substantially extend the scope of HDQI beyond stabilizer-like Hamiltonians, providing a method for Gibbs-state preparation and Hamiltonian optimization in a broad class of physically and computationally relevant quantum systems.

Quantum circuits naturally implement unitary operations on input quantum states. However, non-unitary operations can also be implemented through block encodings, where additional ancilla qubits are introduced and later measured. While block encoding has a number of well-established theoretical applications, its practical implementation has been prohibitively expensive for current quantum hardware. In this paper, we present practical and explicit block encoding circuits implementing matrix polynomial transformations of a target matrix. With standard approaches, block-encoding a degree-$d$ matrix polynomial requires a circuit depth scaling as $d$ times the depth for block-encoding the original matrix alone. By leveraging the recently introduced Fast One-Qubit Controlled Select LCU (FOQCS-LCU) framework, we show that the additional circuit-depth overhead required for encoding matrix polynomials can be reduced to scale linearly in $d$ with no dependence on system size or the cost of block encoding the original matrix. Moreover, we demonstrate that the FOQCS-LCU circuits and their associated matrix polynomial transformations can be controlled with negligible overhead, enabling efficient applications such as Hadamard tests. Finally, we provide explicit circuits for representative spin models, together with detailed non-asymptotic gate counts and circuit depths.

Genuine quantum interference between independent nonlinear processes of different order provides a route to coherent control that cannot be reduced to a classical field interference. Here we present an all-optical analogue of coherent carrier injection by exploiting interference between second- and third-order quantum nonlinear processes in an integrated photonic platform. Photon pairs generated via spontaneous parametric down-conversion and spontaneous four-wave mixing coherently contribute to the same final two-photon state, resulting in a phase-dependent modulation of both the generation rate and the spectral structure of the emitted biphoton state. We illustrate the features of such interference and how it can be used to shape biphoton wavefunctions and their quantum correlations. These results identify interference between nonlinear processes of different order as a distinct form of coherent quantum control within quantum nonlinear optics.

Fault-tolerant quantum computation (FTQC) is expected to address a wide range of computational problems. To realize large-scale FTQC, it is essential to encode logical qubits using quantum error-correcting codes. High-rate concatenated codes have recently attracted attention due to theoretical advances in fault-tolerant protocols with constant-space-overhead and polylogarithmic-time-overhead, as well as practical developments of high-rate many-hypercube codes equipped with a high-performance level-by-level minimum-distance decoder (LMDD). We propose a general, high-performance decoder for high-rate concatenated stabilizer codes that extends LMDD by leveraging the Chase algorithm to generate a suitable set of candidate errors. Our simulation results demonstrate that the proposed decoder outperforms conventional decoders for high-rate concatenated Hamming codes under bit-flip noise.

This paper aims to first explain, somewhat more clearly, the Stochastic-Quantum correspondence put forward in by Barandes in 2023. Specifically, the quantum-mechanical bra-ket notation is used, illuminating some results of previous results. With this, we prove the six axioms of textbook quantum mechanics from a single axiom: every physical system evolves according to a, generally indivisible, stochastic law. Afterwards, we generalise the treatment to continuous bases, which showcases a problem with them, indicating that space (and other physical variables) may be discrete in nature. Some concrete examples are also given, including the generalisation to classical and quantum fields. Then, we treat some practical issues of this new stochastic approach, regarding the solving of problems in physics, which turns out to still be most tractable in the traditional way. Finally, we explain the classical limit, where a system of many particles is found to behave classically according to Newton's second law. Along with that, we present a way of solving the measurement problem, characterising what is an environment and a measuring device and explaining how the wavefunction collapse comes about. Specifically, it is found that what distinguishes an environment is its number of degrees of freedom, while a measuring device is a low-entropy type of environment.

We establish a universal, exponentially decaying upper bound on the average energy that can be extracted in quantum energy teleportation (QET) protocols executed on finite-range gapped lattice systems possessing a unique ground state. Under mild regularity assumptions on the Hamiltonian and uniform operator-norm bounds on the local measurement operators, there exist positive constants $C$ and $μ$ (determined by the spectral gap, interaction range and local operator norms) such that for any local measurement performed in a region $A$ and any outcome-dependent local unitaries implemented in a disjoint region $B$ separated by distance $d=\operatorname{dist}(A,B)$ one has $|E_A-E_B|\le C\,e^{-μd}.$ The bound is nonperturbative, explicit up to model-dependent constants, and follows from the variational characterization of the ground state combined with exponential clustering implied by the spectral gap.

This paper presents a feasibility study demonstrating that quantum machine learning (QML) algorithms achieve competitive performance on real-world medical imaging despite operating under severe constraints. We evaluate Equilibrium Propagation (EP), an energy-based learning method that does not use backpropagation (incompatible with quantum systems due to state-collapsing measurements) and Variational Quantum Circuits (VQCs) for automated detection of Acute Myeloid Leukemia (AML) from blood cell microscopy images using binary classification (2 classes: AML vs. Healthy). Key Result: Using limited subsets (50-250 samples per class) of the AML-Cytomorphology dataset (18,365 expert-annotated images), quantum methods achieve performance only 12-15% below classical CNNs despite reduced image resolution (64x64 pixels), engineered features (20D), and classical simulation via Qiskit. EP reaches 86.4% accuracy (only 12% below CNN) without backpropagation, while the 4-qubit VQC attains 83.0% accuracy with consistent data efficiency: VQC maintains stable 83% performance with only 50 samples per class, whereas CNN requires 250 samples (5x more data) to reach 98%. These results establish reproducible baselines for QML in healthcare, validating NISQ-era feasibility.

Quantum computing requires the optimization of control pulses to achieve high-fidelity quantum gates. We propose a machine learning-based protocol to address the challenges of evaluating gradients and modeling complex system dynamics. By training a recurrent neural network (RNN) to predict qubit behavior, our approach enables efficient gradient-based pulse optimization without the need for a detailed system model. First, we sample qubit dynamics using random control pulses with weak prior assumptions. We then train the RNN on the system's observed responses, and use the trained model to optimize high-fidelity control pulses. We demonstrate the effectiveness of this approach through simulations on a single $ST_0$ qubit.

The preferred-basis problem and the definite-outcome aspect of the measurement problem persist even if the detector is modeled unitarily, because experimental data are necessarily represented in a Boolean event algebra of mutually exclusive records whereas the theoretical description is naturally formulated in a noncommutative operator algebra with continuous unitary symmetry. This change of mathematical type constitutes the core of the 'cut': a structurally necessary interface from group-based kinematics to set-based counting. In the presented view the basis relevant for recorded outcomes is not determined by the system Hamiltonian alone; it is induced by the measurement mapping, i.e., by the detector channel together with the coarse-grained readout that defines an instrument. The probabilistic mapping is anchored in symmetry and measure theory: by Gleason-type uniqueness (Gleason for projections in $d>2$ and Busch's extension for Positive Operator-Valued Measures (POVMs) including $d=2$), the trace rule is the unique probability measure consistent with additivity over exclusive events and basis-independence of the unitary sector. A compact qubit--pointer model yields an induced unsharp POVM $E_\pm=\tfrac12(\id\pm η\,σ_z)$ with $η$ fixed by pointer resolution, displaying explicitly how the detector induces the relevant basis. Finally, nested-observer paradoxes are tightened into a non-composability lemma: joint assignment of outcome propositions is obstructed unless a joint instrument exists. This relocates the origin of randomness to the stochasticity of the transition rules.

Assessing whether a noisy quantum device can potentially exhibit quantum advantage is essential for selecting practical quantum utility tasks that are not efficiently verifiable by classical means. For optimization, a prominent candidate for quantum advantage, entropy benchmarking provides insights based concomitantly on the specifics of the application and its implementation, as well as hardware noise. However, such an approach still does not account for finite-shot effects or for quantum error mitigation (QEM), a key near-term error suppression strategy that reduces estimation bias at the cost of increased sampling overhead. We address this limitation by developing a benchmarking framework that explicitly incorporates finite-shot statistics and the resource overhead induced by QEM. Our framework quantifies quantum advantage through the confidence that an estimated energy lies within an interval defined by the best-known classical upper and lower bounds. Using a proof-of-principle numerical study of the two-dimensional Fermi-Hubbard model at size $8\times8$, we demonstrate that the framework effectively identifies noise and shot-budget regimes in which the probabilistic error cancellation (PEC), a representative QEM method, is operationally advantageous, and potential quantum advantage is not hindered by finite-shot effects. Overall, our approach equips end-users with a framework based on lightweight numerics for assessing potential practical quantum advantage in optimization on near-future quantum hardware, in light of the allocated shot budget.

We show that the Novikov coordinates can be obtained in a direct and physically transparent way from the radial geodesics of massive particles with negative energy in the Schwarzschild spacetime. These geodesics form a complete congruence that covers the entire spacetime. By rectifying this family of trajectories using the proper time as the time coordinate, the Novikov variables naturally emerge, providing a clear dynamical interpretation of the different regions usually identified as black-hole and white-hole sectors. In Novikov coordinates, observers at fixed spatial position follow free-fall trajectories. From their perspective, the gravitational collapse of a dust star is completed in a finite proper time, independently of their initial distance from the star. In contrast, observers described by Schwarzschild-Droste coordinates perceive the boundary of the collapsing star as taking an infinite coordinate time to reach the horizon. We emphasize that Schwarzschild-Droste observers are static with respect to the center of mass of the star and therefore cannot be in free fall. The use of these coordinates implicitly requires the presence of a force that compensates the gravitational attraction. From this viewpoint, the apparent infinite-time collapse is not a physical effect but a coordinate artifact associated with non-inertial observers.

We propose a quantum rotation diversity (QRD) scheme for optical quantum communication using binary phase-shift-keying displaced squeezed states and homodyne detection over Gamma-Gamma turbulence channels. Consecutive temporal modes are coupled by a passive orthogonal rotation that redistributes the displacement amplitude between slots, yielding a diversity order of two under independent fading and joint maximum-likelihood detection. Analytical expressions for the symbol-error rate performance, along with asymptotic results for the diversity and coding gains, are derived. The optimal rotation angle and energy allocation between displacement and squeezing are obtained in closed form. Furthermore, we show that when both the displacement amplitude and the squeezing strength scale with the total photon number, an effective diversity order of four is achieved. Numerical results validate the analysis and demonstrate the super-diversity behaviour of the proposed QRD scheme.

Generating quantum data by learning the underlying quantum distribution poses challenges in both theoretical and practical scenarios, yet it is a critical task for understanding quantum systems. A fundamental question in quantum machine learning (QML) is the universality of approximation: whether a parameterized QML model can approximate any quantum distribution. We address this question by proving a universality theorem for the Many-body Projected Ensemble (MPE) framework, a method for quantum state design that uses a single many-body wave function to prepare random states. This demonstrates that MPE can approximate any distribution of pure states within a 1-Wasserstein distance error. This theorem provides a rigorous guarantee of universal expressivity, addressing key theoretical gaps in QML. For practicality, we propose an Incremental MPE variant with layer-wise training to improve the trainability. Numerical experiments on clustered quantum states and quantum chemistry datasets validate MPE's efficacy in learning complex quantum data distributions.

Models of effective stellar collapse inspired by loop quantum gravity predict a bounce when the stellar energy density reaches the Planck scale, typically followed by the formation of shell-crossing singularities. This work aims to extend the spacetime beyond these singularities by employing a Hamiltonian formulation of the Darmois-Israel junction conditions, treating the singularity as a non-isolated thin dust shell. By construction, the shell's motion remains timelike throughout the entire evolution, regardless of the amount of initial stellar mass, and the induced metric on the shell remains continuous. The resulting stellar evolution produces an inter-universal wormhole, analogous to the simpler Oppenheimer-Snyder scenario. The proposed approach provides a general framework for any effective (or classical) theory of stellar collapse characterized by shell-crossing singularities.

We investigate the correspondence between the quasinormal modes and grey-body factors of Schwarzschild--Tangherlini black holes. The gravitational perturbations in higher-dimensional black holes can be classified into scalar, vector, and tensor types. Considering the dimension-dependent forms of their effective potentials, the correspondence was examined for each dimension and perturbation mode. The accurate quasinormal modes were computed by suitably adopting the continued fraction and integration-through-midpoints methods, depending on the structure of the singularity. The grey-body factor can be obtained through its correspondence with the quasinormal mode, and its accuracy was analyzed by calculating its difference from the numerically computed grey-body factor. The correspondence failed for $l=2$ scalar gravitational perturbations in $D\ge7$ because the form of the potential is markedly different from that in four dimensions. The vector and tensor perturbation types exhibited good correspondence accuracies in all cases. The breakdown of the correspondence was rigorously demonstrated to stem from multiple potential barriers, and its applicability to each mode in higher dimensions was assessed.

Quantum error-correcting codes protect fragile quantum information by encoding it redundantly, but identifying codes that perform well in practice with minimal overhead remains difficult due to the combinatorial search space and the high cost of logical error rate evaluation. We propose a Bayesian optimization framework to discover quantum error-correcting codes that improves data efficiency and scalability with respect to previous machine learning approaches to this task. Our main contribution is a multi-view chain-complex neural embedding that allows us to predict the logical error rate of quantum LDPC codes without performing expensive simulations. Using bivariate bicycle codes and code capacity noise as a testbed, our algorithm discovers a high-rate code [[144,36]] that achieves competitive per-qubit error rate compared to the gross code, as well as a low-error code [[144,16]] that outperforms the gross code in terms of error rate per qubit. These results highlight the ability of our pipeline to automatically discover codes balancing rate and noise suppression, while the generality of the framework enables application across diverse code families, decoders, and noise models.

The one-shot zero-error classical capacity of a quantum channel is the amount of classical information that can be transmitted with zero probability of error by a single use. Then the one-shot zero-error classical capacity equals to the logarithmic value of the independence number of the noncommutative graph induced by the channel. Thus the additivity of the one-shot zero-error classical capacity of a quantum channel is equivalent to the multiplicativity of the independence number of the noncommutative graph. The independence number is not multiplicative in general, and it is not clearly understood when the multiplicativity occurs. In this work, we present sufficient conditions for multiplicativity of the independence number, and we give explicit examples of quantum channels. Furthermore, we consider a block form of noncommutative graphs, and provide conditions when the independence number is multiplicative.

Bell nonlocality provides a device-independent (DI) way to certify quantum randomness, based on which true random numbers can be extracted from the observed correlations without detail characterizations on devices for quantum state preparation and measurement. However, the efficiency of current strategies for DI randomness certification is still heavily constrained when it comes to non-maximal Bell values, especially for multiple parties. Here, we present a family of multipartite Bell inequalities that allows to certify optimal quantum randomness and self-test GHZ (Greenberger-Horne-Zeilinger) states, which are inspired from the stabilizer group of the GHZ state. Due to the simple representation of stabilizer group for GHZ states, this family of Bell inequalities is of simple structure and can be easily expanded to more parties. Compared with the Mermin-type inequalities, this family of Bell inequality is more efficient in certifying quantum randomness when non-maximal Bell values achieved. Meanwhile, the general analytical upper bound for the Holevo quantity is presented, and achieves better performance compared with the MABK (Mermin-Ardehali-Belinskii-Klyshko) inequality, Parity-CHSH (Clauser-Horne-Shimony-Holt) inequality and Holz inequality at $N=3$, which is of particular interests for experimental researches on DI quantum cryptography in quantum networks.

We investigate the gravitational $e^-+e^+\to\ell^-+\ell^+$ scattering process within the framework of gravitoelectromagnetism, a weak-field approximation of gravity analogous to Maxwell's theory of electromagnetism. This process involves the interaction between a fermion and an antifermion mediated by graviton exchange. We consider the nonminimal gravitational sector of the standard model extension and calculate the corrections to the scattering cross section arising from Lorentz violation. The analysis is carried out in two scenarios: (i) at zero temperature and (ii) at finite temperature. To incorporate thermal effects, we employ the thermo field dynamics formalism, which allows for a consistent treatment of quantum fields at finite temperature. The results provide insights into how Lorentz-violating and thermal corrections influence gravitational interactions, particularly relevant in high-energy or astrophysical environments.

Various quantum key distribution protocols require bright single-photon sources with a very low probability of multiphoton emission. In this work, we investigate single-photon generation from a negatively charged quantum dot embedded in an elliptical pillar microcavity, driven using either resonant excitation or adiabatic rapid passage (ARP). Our results show that ARP excitation significantly suppresses multiphoton emission probability and improves photon indistinguishability compared to resonant excitation. We further evaluate the secure key rate of both BB84 and twin-field quantum key distribution (TF-QKD) using quantum-dot single-photon sources and compare their performance with that of Poisson-distributed photon sources (PDS) such as weak coherent pulses and down-conversion sources. The analysis reveals that adiabatic excitation offers a modest but consistent enhancement in secure key rate relative to resonant excitation. Moreover, quantum-dot single-photon sources outperform PDS sources over short and intermediate distances; however, at longer distances, PDS sources eventually surpass quantum-dot sources in both infinite decoy-state BB84 and TF-QKD.

Determining the ground and low-lying excited states is critical in numerous scenarios. Recent work has proposed the ancilla-entangled variational quantum eigensolver (AEVQE) that utilizes entanglement between ancilla and physical qubits to simultaneously tagert multiple low-lying energy levels. In this work, we report the experimental implementation of the AEVQE on a superconducting quantum cloud platform, demonstrating the full procedure of solving the low-lying energy levels of the H$_2$ molecule and the transverse-field Ising models (TFIMs). We obtain the potential energy curves of H$_2$ and show an indication of the ferromagnetic to paramagnetic phase transition in the TFIMs from the average absolute magnetization. Moreover, we investigate multiple factors that affect the algorithmic performance and provide a comparison with ancilla-free VQE algorithms. Our work demonstrates the experimental feasibility of the AEVQE algorithm and offers a guidance for the VQE approach in solving realistic problems on publicly-accessible quantum platforms.

Ensembles of atoms interacting via their Rydberg levels, known as "superatoms" for their ability to encode qubits and to emit single photons, attract increasing attention as building blocks for quantum network nodes. Assessing their performance requires an accurate, physically informative and numerically scalable description of interactions in a large and disordered ensemble. We derive such a description from first principles and successfully test it against brute-force numerics and experimental data. This model proves essential to make quantitative predictions about gate fidelities or photon emission efficiencies, and to guide experiments towards large-scale superatom-based systems.

This chapter introduces the fundamental principles of gravitational wave detectors in a simple and comprehensive manner. Because these instruments aim for extremely high sensitivity, it is essential to understand their various noise sources, how such noise couples to the detector output, and the strategies used to mitigate them. These noises contributions are computed in the frame of the Virgo detector and a sensitivity curve is calculated. Although a simplified layout of a gravitational wave detector is considered, it takes into account the most dominant effects and yields in a sensitivity estimate close to the what is observed in real detectors.

Quantum interference between interacting systems is fundamental to basic science and quantum technology, but it typically requires precise control of the interaction phases of lasers or microwave generators. Can interference be observed if those interaction phases are stable but unknown, usually prohibitive for complex state without active control? Here, we answer this question by experimentally preparing a Schrödinger-cat-like state of an internal qubit and a motional oscillator of a trapped $^{40}$Ca$^{+}$ ion, and its robustness to such uncontrolled phase. By applying alternating red and blue sideband pulses, we enforce a strict qubit-parity correlation and interference inherently insensitive to stable but unknown phases of the driving laser. For this qubit-parity interference, we use a minimal two-pulse interferometric sequence to demonstrate characteristic visibilities of $20\%$ and $40\%$, which approach the theoretical visibility limit, providing a scalable coherence witness without full state tomography for high-dimensional states.

The integration of high-penetration renewable energy introduces significant stochasticity and non-convexity into power system dispatching, challenging the computational limits of classical optimization. While Variational Quantum Algorithms (VQAs) on Noisy Intermediate-Scale Quantum (NISQ) devices offer a promising path for combinatorial acceleration, existing approaches typically treat the power grid as a "black box", suffering from poor scalability (barren plateaus) and frequent violations of physical constraints. Bridging these gaps, this paper proposes a Physics-Informed Hybrid Quantum-Classical Dispatching (PI-HQCD) framework. We construct a topology-aware Hamiltonian that explicitly embeds linearized power flow equations, storage dynamics, and multi-timescale coupling directly into the quantum substrate, significantly reducing the search space dimensionality. We further derive a noise-adaptive regularization mechanism that theoretically bounds the effective Lipschitz constant of the objective function, guaranteeing convergence stability under realistic quantum measurement noise. Numerical experiments on the IEEE 39-bus benchmark and a 118-bus regional grid demonstrate that PI-HQCD achieves superior economic efficiency and higher renewable utilization compared to stochastic dual dynamic programming (SDDP). Theoretical analysis confirms that this topology-aware design leads to an O(1/N) gradient variance scaling, effectively mitigating barren plateaus and ensuring scalability for larger networks. This work establishes a rigorous paradigm for embedding engineering physics into quantum computing, paving the way for practical quantum advantage in next-generation grid operations.

In this study, we employ dynamical systems methods to analyse the large-scale structure by considering two distinct interaction models (linear and non-linear) within the dark sector, associated with a specific dynamical dark energy model inspired by the Veneziano ghost theory in quantum chromodynamics (QCD). In these models, the dark energy density ($ρ_{DE}$) varies with the Hubble parameter ($H$), expressed as $ρ_{DE} = αH + βH^2$. After defining the dimensionless parameters, we present autonomous equations that allow us to find the trace $\text{Tr}(J)$ and the determinant $D(J)$. With these solutions, we demonstrate the presence of unstable, saddle, and stable fixed points, corresponding to the radiation-, matter-, and dark-energy-dominated eras, respectively. Our results suggest that these models are theoretically viable for representing the interaction between dark sector fluids.

We apply the wavelet formalism of quantum field theory to investigate nonperturbative dynamics within the Hamiltonian framework. In particular, we employ Daubechies wavelets in momentum space, whose basis functions are labeled by resolution and translation indices, providing a natural nonperturbative truncation of both infrared and ultraviolet truncation of quantum field theories. As an application, we compute the energy spectra of a free scalar field theory and the interacting $1+1$-dimensional $φ^4$ theory. This approach successfully reproduces the well-known strong-coupling phase transition in the $m^2 > 0$ regime. We find that the extracted critical coupling systematically converges toward its established value as the momentum resolution is increased, demonstrating the effectiveness of the wavelet-based Hamiltonian formulation for nonperturbative field-theoretic calculations.

Leveraging the quantum advantages of highly excited atoms, Rydberg atomic receivers (RAREs) represent a paradigm shift in radio wave detection, offering high sensitivity and broadband reception. However, existing studies largely model RAREs as isotropic point receivers and overlook the spatial variations of atomic quantum states within vapor cells, thus inaccurately characterizing their reception patterns. To address this issue, we present a theoretical analysis of the aforementioned spatial responses of a standard local-oscillator (LO)- dressed RARE. Our results reveal that increasing the vapor-cell length produces a receive beam aligned with the LO field, with a beamwidth inversely proportional to the cell length. This finding enables atomic beamforming to enhance received signal-to-noise ratio using only a single-antenna RARE. Furthermore, we derive the achievable beamforming gain by characterizing and balancing the fundamental tradeoff between the effects of increasing the vapor cell length and the exponential power decay of laser propagating through the cell. To overcome the limitation imposed by exponential decay, we propose a novel RARE architecture termed segmental vapor cell. This architecture consists of vapor-cell segments separated by clear-air gaps, allowing the total cell length (and hence propagation loss) to remain fixed while the effective cell length increases. As a result, this segmented design expands the effective atom-field interaction area without increasing the total vapor cell length, yielding a narrower beamwidth and thus higher beamforming gain as compared with a traditional continuous vapor cell.

We propose a novel chaos indicator -- time-reversed Shannon entropy (TRSE) -- that leverages the interplay between time-reversal symmetry breaking and information entropy in curved spacetimes. By quantifying statistical discrepancies between forward and backward temporal evolution of particle orbits, TRSE robustly distinguishes chaotic from regular dynamics in non-integrable systems. In contrast, integrable systems exhibit stable, symmetric probability distributions preserved by conserved quantities such as the Carter constant. We validate the method through high-precision numerical simulations in both Kerr and Schwarzschild-Melvin black hole geometries, evolving trajectories forward and backward in time. Furthermore, we refine our previously introduced particle-pair mutual information (MIPP) and perform comprehensive parameter-space scans, revealing a strong quantitative agreement between MIPP and TRSE. The two indicators emerge as complementary probes of chaos: TRSE captures symmetry breaking in orbital evolution, while MIPP measures statistical correlations. Together, they establish a unified framework for diagnosing chaos in general relativistic systems, paving a new path to understand the fundamental nature of chaos in non-integrable systems.

Emergent cooperation in classical Multi-Agent Reinforcement Learning has gained significant attention, particularly in the context of Sequential Social Dilemmas (SSDs). While classical reinforcement learning approaches have demonstrated capability for emergent cooperation, research on extending these methods to Quantum Multi-Agent Reinforcement Learning remains limited, particularly through communication. In this paper, we apply communication approaches to quantum Q-Learning agents: the Mutual Acknowledgment Token Exchange (MATE) protocol, its extension Mutually Endorsed Distributed Incentive Acknowledgment Token Exchange (MEDIATE), the peer rewarding mechanism Gifting, and Reinforced Inter-Agent Learning (RIAL). We evaluate these approaches in three SSDs: the Iterated Prisoner's Dilemma, Iterated Stag Hunt, and Iterated Game of Chicken. Our experimental results show that approaches using MATE with temporal-difference measure (MATE\textsubscript{TD}), AutoMATE, MEDIATE-I, and MEDIATE-S achieved high cooperation levels across all dilemmas, demonstrating that communication is a viable mechanism for fostering emergent cooperation in Quantum Multi-Agent Reinforcement Learning.

This book arises from the need to provide a clear and up-to-date overview of the impacts of quantum computing on cryptography. The goal is to provide a reference in Portuguese for undergraduate, master's, and doctoral students in the field of data security and cryptography. Throughout the chapters, we present fundamentals, we discuss classical and post-quantum algorithms, evaluate emerging patterns, and point out real-world implementation challenges. The initial objective is to serve as a guide for students, researchers, and professionals who need to understand not only the mathematics involved, but also its practical implications in security systems and policies. For more advanced professionals, the main objective is to present content and ideas so that they can assess the changes and perspectives in the era of quantum cryptographic algorithms. To that end, the text's structure was designed to be progressive: we begin with essential concepts, move on to quantum algorithms and their consequences (with emphasis on Shor's algorithm), present issues focusing on "families" of post-quantum schemes (based on lattices, codes, hash functions, multivariate, isogenies), analyze the state of the art in standardization (highlighting the NIST process), and finally, discuss migration, interoperability, performance, and cryptographic governance. We hope that this work will assist in the formation of critical thinking and informed technical decision-making, fostering secure transition strategies for the post-quantum era.

We demonstrate direct, non-invasive and non-contact detection of human cardiac magnetic signals using quantum sensors based on nitrogen-vacancy (NV) centers in diamond. Three configurations were employed recording magnetocardiography (MCG) signals in various shielded and unshielded environments. The signals were averaged over a few hundreds up to several thousands of heart beats to detect the MCG traces. The compact room-temperature NV sensors exhibit sensitivities of 6-26 pT/Hz^(1/2) with active sensing volumes below 0.5 mm^3, defining the performance level of the demonstrated MCG measurements. While the present signals are obtained by averaging, this performance already indicates a clear path toward single-shot MCG sensing. To move beyond shielded environments toward practical clinical use, strong noise suppression is required. To this end, we implement NV-based gradiometry and achieve efficient common-mode noise rejection, enabled by the intrinsically small sensing volume of NV sensors. Together, these multi-platform results obtained across diverse magnetic environments provide a solid foundation for translating quantum sensors into human medical diagnostics such as MCG and magnetoencephalography (MEG).

We present a general framework for applying linear quantum error mitigation (QEM) techniques directly to physical qubits within a logical qubit to suppress logical errors. By exploiting the linearity of quantum error correction (QEC), we demonstrate that any linear QEM method$\unicode{x2014}$including probabilistic error cancellation (PEC), zero-noise extrapolation (ZNE), and symmetry verification$\unicode{x2014}$can be integrated into the physical layer without requiring modifications to the subsequent QEC decoder. Applying this framework to memory experiments using PEC, we analytically prove and numerically verify that the leading-order contribution to the logical error can be removed, increasing the effective code distance by 2. Our simulations on repetition and rotated surface codes show that a distance-3 code with physical-level PEC achieves logical error rates lower than or similar to a distance-5 unmitigated code while using 40% and 64% fewer qubits, respectively. These results establish physical-level QEM as a widely compatible and resource-efficient strategy for enhancing logical performance in early fault-tolerant architectures.

Detecting the stochastic gravitational wave background (SGWB) from our Universe under the inflationary era is one of the primary scientific objectives of DECIGO, a space-borne gravitational wave detector sensitive in the 0.1 Hz frequency band. This frequency band is dominated by the gravitational waves from inspiraling compact object binaries. Subtracting these signals is necessary to search for the primordial SGWB. In this paper, we assess the feasibility of the subtraction of such binary signals by employing the population model inferred from the latest gravitational wave event catalogue of the LIGO-Virgo-KAGRA Collaboration. We find that the projection scheme, which was originally proposed by Cutler & Harms (2005), is necessary to reduce the binary signals to the level where DECIGO can detect the primordial background.

Quantum memories feature a reversible conversion of optical fields into long-lived atomic spin waves, and are therefore ideal for operating as sensitive atomic sensors. However, up to now, atom-light interferometers have lacked an efficient approach to exploit their ultimate atomic sensing performance, since an extra optical delay line is required to compensate for the memory time. Here, we report a new protocol that records the photocurrent via heterodyne mixing with a stable local oscillator. The obtained complex quadrature amplitude that carries information imprinted on its phase by an external magnetic field, is successfully recovered from the interference patterns between the light and the atomic spin wave, without the stringent requirement of having them overlap in time. Our results reveal that the sensitivity scales favorably with the lifetime of the quantum memory. Our work may have important applications in building distributed quantum networks through quantum memory-assisted atom-light interferometers.

This work employs the minimal geometric deformation decoupling scheme to derive interior stellar solutions in the background of an electrically charged BTZ ansatz as a seed metric in three dimensions. In this respect, we impose two different equations of state to determine the deformation function and the new material contributions emerging from the additional field source. Furthermore, we describe the finiteness of all thermodynamic quantities of the presented stellar solutions, including the effective thermodynamical quantities, for varying values of the deformation parameter and total electric charge. We explore the new interior astrophysical solutions in three-dimensional gravity by analyzing the charged BTZ metric, admitting circular symmetry through the principles of geometric deformation. This study examines the impact of radial-metric deformation on the charged BTZ geometry and underscores the importance of stellar decoupling within the context of electrically charged dense distributions. It is shown that new physically acceptable solutions by incorporating any known three-dimensional spacetime as the isotropic basis are possible, which in turn enable one to analyze the quantum effects due to low degrees of freedom at lower dimensions.

Bohr's Complementarity Principle is a core concept of quantum mechanics. In this article, an updated complementarity relation for the wave and ondulatory aspects of a quantum system is presented and discussed. Two interferometric experiments are implemented in one and two qubit circuits and executed on real hardware. The final state density matrices are reconstructed using quantum state tomography and the complementarity relation is tested via direct computation. Results of the executions are presented both graphically and with a mean squared error analysis for a better comprehension.

Efficient entanglement distribution is the cornerstone of the Quantum Internet. However, physical link parameters such as photon loss, memory coherence time, and gate error rates fluctuate dynamically, rendering static purification strategies suboptimal. In this paper, we propose an Adaptive Purification Controller (APC) that autonomously optimizes the entanglement distillation sequence to maximize the "goodput," the rate of delivered pairs meeting a strict fidelity threshold. By treating protocol selection as a resource allocation problem, the APC dynamically switches between purification depths and protocol families (e.g., BBPSSW vs. DEJMPS) to navigate the trade-off between generation rate and state quality. Using a dynamic programming planner with Pareto pruning, simulation results demonstrate that our approach eliminates the "fidelity cliffs" inherent in static protocols and prevents resource wastage in high-noise regimes. Furthermore, we extend the controller to heterogeneous scenarios, demonstrating robustness for both multipartite GHZ state generation and continuous variable systems using effective noiseless linear amplification models. We benchmark its computational overhead, confirming real-time feasibility with decision latencies in the millisecond range per link.

Quantum technology advances crucially depend on the scaling up of essential quantum resources. Their ideal multiplexing offers more significant gains in applications; however, the scaling of the nonidentical, fragile and varying resources is neither theoretically nor experimentally known. For bosonic systems, multimode interference is an essential tool already widely exploited to develop quantum technology. Here, we analyze, predict and compare essential scaling laws for a constructive interference of multiplexed nonclassical Gaussian states carrying information by displacement with weakly fluctuating squeezing in different multimode interference architectures. The signal-to-noise ratio quantifies the increase in displacement relative to the noise. We introduce the gain-to-instability ratio to numerically estimate the effect of unexplored resource instabilities in a large scale interference scheme. The use of the gain-to-instability ratio to quantify the scaling laws opens steps for extensive theoretical investigation of other bosonic resources and follow-up feasible experimental verification necessary for further development of these platforms.

Quantum Energy Teleportation (QET) constitutes a paradigm-shifting protocol that permits the activation of local vacuum energy through the consumption of pre-existing entanglement and classical communication. Nevertheless, the implementation of QET is severely impeded by the fundamental locality of gapped many-body systems, where the exponential clustering of ground-state correlations restricts energy extraction to microscopic scales. In this work, we address this scalability crisis within the framework of the one-dimensional anisotropic XY model. We initially provide a rigorous characterization of a monolithic measurement-induced strategy, demonstrating that while bulk projective measurements can theoretically induce long-range couplings, the approach is rendered physically untenable by exponentially diverging thermodynamic costs and vanishing success probabilities. To circumvent this impasse, we propose and analyze a hierarchical quantum repeater architecture adapted for energy teleportation. By orchestrating heralded entanglement generation, iterative entanglement purification, and nested entanglement swapping, our protocol effectively counteracts the fidelity degradation inherent in noisy quantum channels. We establish that this architecture fundamentally alters the operational resource scaling from exponential to polynomial. This proves, for the first time, the physical permissibility and computational tractability of activating vacuum energy at arbitrary distances. The significance lies not in net energy gain, but in establishing long-range QET as a viable protocol for remote quantum control and resource distribution.

The simulation of strongly correlated fermionic systems remains one of the most significant challenges in computational physics due to the exponential growth of the Hilbert space and the fermionic sign problem. In this work, we present a digital quantum simulation framework to explore the Fermi polaron and the Bose-Einstein Condensate (BEC) to Bardeen-Cooper-Schrieffer (BCS) crossover. We develop a unified Hamiltonian formalism that bridges pairing superfluidity and impurity physics, mapping the system onto a gate-based quantum processor via the Jordan-Wigner transformation. Using a first-order Trotter-Suzuki decomposition, we implement a Ramsey interferometry protocol to extract the real-time dynamics and spectral response of the system. Our results demonstrate a smooth transition from a dressed quasiparticle (polaron) regime to a stable molecular bound state, characterized by a linear energy renormalization in the strong-coupling limit. We validate our simulation against exact classical benchmarks and report successful execution on the Barcelona Supercomputing Center quantum hardware. Despite the inherent noise of the quantum hardware, the hybrid variational approach shows remarkable resilience, accurately capturing the bifurcation of the spectral density

We investigate the holographic timelike entanglement entropy (HTEE) and timelike subregion complexity of a thermal CFT$_d$ deformed by a relevant scalar operator $φ_0$, dual to a hairy black hole in AdS$_{d+1}$. We employ the prescription of merging spacelike and timelike surfaces at the interior, constructing an extremal surface homologous to a boundary timelike subsystem with a time interval $Δt$. Consequently, this deformation breaks the invariance of the imaginary component of HTEE observed in pure AdS$_3$ and BTZ geometry, introducing a nontrivial dependence on $Δt$. At small $Δt$, we derive analytical expressions that are in agreement with numerical results, and observe partial consistency with analytic continuation to temporal or spacelike entanglement entropy at the level of the near-boundary expansion. However, analytic continuation of CFT temporal entanglement entropy fails to reproduce the HTEE calculations under boundary deformation, even in $d=2$. Furthermore, we extend the numerical calculations to higher dimensions ($d=3$). In addition, we study holographic timelike subregion complexity within the complexity=volume conjecture and find that it remains real-valued, providing a complementary geometric probe of the black hole interior. In particular, for the BTZ black hole, we analytically show that the UV-finite term of the subregion complexity receives its entire contribution from the interior region alone.

We present a resource-efficient method based on repetitive weak measurements to directly measure the noise spectrum of a generic quantum environment that causes qubit phase decoherence. The weak measurement is induced by a Ramsey interferometry measurement (RIM) on the qubit and periodically applied during the free evolution of the environment. We prove that the measurement correlation of such repetitive RIMs approximately corresponds to a direct sampling of the noise correlation function, thus enabling direct noise spectroscopy of the environment. Compared to dynamical-decoupling-based noise spectroscopy, this method can efficiently measure the full noise spectrum with the detected frequency range not limited by qubit coherence time. This method is also more resource-efficient than the correlation spectroscopy, as for the same detection accuracy with $N$ sampling times, it takes total detection time $O(N)$ while the latter one takes time $O(N^2)$. We numerically demonstrate this method for both bosonic and spin baths.

A general semi-tetrad covariant approach is adopted to analyse the structure scalars of a Type II fluid in generalized Vaidya spacetime. The relationship between the $1+1+2$ covariant quantities and the structure scalars are obtained. We calculate the complexity factor in terms of the Misner-Sharp mass and the matter variables to obtain a non-trivial class of spacetimes with vanishing complexity. Also the Vaidya spacetime with pure Type II matter field has negative complexity. The differences between the complexity of Type I and Type II matter fields are highlighted. We compute the propagation and evolution equations of the structure scalars, showcasing their interdependency through the kinematical variables. The causal wave equation of the Gaussian curvature of the 2-shell and its dependence on the structure scalars are also studied.

Recently, Ai, Cruz, Garbrecht, and Tamarit (ACGT)~\cite{Ai:2020ptm, Ai:2024vfa, Ai:2024cnp, Ai:2025quf} claimed that there is no strong CP problem by adopting a new order of limits in the volume and topological sector. We critically examine this proposal by focusing on simple one-dimensional quantum mechanics on a ring. We demonstrate that consistent results are obtained only when one sums over all topological sectors \textit{before} taking the large $T$ limit. This observation justifies the conventional path integral formulation of gauge theories and implies that the strong CP problem does exist in QCD.

Approximating the permanent of a complex-valued matrix is a fundamental problem with applications in Boson sampling and probabilistic inference. In this paper, we extend factor-graph-based methods for approximating the permanent of non-negative-real-valued matrices that are based on running the sum-product algorithm (SPA) on standard normal factor graphs, to factor-graph-based methods for approximating the permanent of complex-valued matrices that are based on running the SPA on double-edge normal factor graphs. On the algorithmic side, we investigate the behavior of the SPA, in particular how the SPA fixed points change when transitioning from real-valued to complex-valued matrix ensembles. On the analytical side, we use graph covers to analyze the Bethe approximation of the permanent, i.e., the approximation of the permanent that is obtained with the help of the SPA. This combined algorithmic and analytical perspective provides new insight into the structure of Bethe approximations in complex-valued problems and clarifies when such approximations remain meaningful beyond the non-negative-real-valued settings.

This paper presents a comprehensive analysis of junction conditions for gluing different $f(R)$ gravitational theories across a non-null hypersurface. Using the variational approach, we systematically derive the junction conditions for both general $f(R)$ theories and the special case of Einstein gravity, for comparison. We demonstrate that when joining two distinct $f(R)$ theories, the junction conditions require continuity of $\partial f(R)/\partial R$, the extrinsic curvature $K_{μν}$, while allowing for discontinuities in the Ricci Scalar $R$. Furthermore, we establish the equivalence between Jordan and Einstein frame formulations through careful treatment of conformal transformations; Our results reveal that different $f(R)$ theories can be consistently matched provided specific relations between their functional forms and geometric quantities are satisfied at the interface.

We investigate critical phenomena during the gravitational collapse of a massive scalar field under two distinct semi-classical loop quantum gravity (LQG) approaches within spherical symmetry. Numerical simulations reveal that the massive scalar field in both semi-classical frameworks exhibits two distinct types of critical behavior, consistent with the classical scenario. When the scalar field's mass parameter is small, type II critical phenomena emerge, with the resulting echoing periods and critical exponents precisely matching those obtained in general relativity. In contrast, a large mass parameter triggers type I critical phenomena, where the resulting black holes possess a finite minimum mass. These findings suggest that semi-classical corrections from LQG have a negligible impact on the dynamics of critical collapse.

Static, spherically symmetric black holes immersed in a dark matter halo with a Hernquist-type density profile have been derived by Cardoso et. al. in Ref. \redcite{Cardoso:2021wlq}. Using the Newman-Janis algorithm, we construct the metric for a stationary and axially symmetric rotating black hole in this environment. Then, we obtain the electromagnetic properties of thin accretion disks around such rotating black holes by utilizing the steady-state Novikov-Thorne model, and study the effects of spin parameter and halo compactness parameter on the disk properties. Finally, by comparison the results of the rotating Cardoso black hole with that of Kerr black hole in the absence of dark matter, we find that the presence of dark matter can not significantly affect the disk properties and thus for astrophysical black holes with large spin parameter, the distinction of rotating Cardoso black holes becomes more difficult than the Kerr black hole.

The rapid growth of modern machine learning (ML) models presents fundamental challenges in parameter efficiency and computational resource requirements. This study introduces the Quantum Recurrent Unit (QRU), a novel quantum neural network (NN) architecture specifically designed to address these challenges while remaining compatible with Noisy Intermediate-Scale Quantum (NISQ) devices. QRU leverages quantum controlled-SWAP (C-SWAP; Fredkin) gates to implement an information selection mechanism inspired by classical Gated Recurrent Units (GRUs), enabling selective processing of temporal information via quantum operations. Through its innovative recurrent architecture featuring measurement results feedforward state propagation and shared parameters across time steps, QRU achieves constant circuit depth and constant parameter count regardless of input sequence length, effectively circumventing stringent NISQ hardware constraints. We systematically validate QRU through three progressive experiments: (1) oscillatory behavior prediction, where 72-parameter QRU matches 197-parameter classical GRU performance; (2) Wisconsin Diagnostic Breast Cancer classification, where 35 parameters achieve 96.13% accuracy comparable to 167-parameter artificial NNs; and (3) MNIST handwritten digit recognition, where 132 parameters reach 98.05% accuracy, outperforming a 27,265-parameter convolutional NN. These results demonstrate that QRU consistently achieves comparable or superior performance with significantly fewer parameters than classical NNs while maintaining constant quantum circuit depth. The architecture's quantum-native design, combining C-SWAP-based information selection with novel recurrent processing, suggests QRU's potential as a fundamental building block for next-generation ML systems, offering a promising pathway toward more efficient and scalable quantum ML architectures.

In the spacetime of horizonless compact objects described by Einsteinian cubic gravity (ECG), we demonstrate the existence of static stable timelike circular orbits on which massive particles remain at rest relative to distant observers. These static orbits are further identified as the innermost stable circular orbits (ISCOs) in this spacetime. If such static orbits form part of an accretion disk, they would give rise to a ring-like structure that is unaffected by Doppler shifts. Moreover, the Aschenbach effect is shown to be present: the orbital velocity of particles on timelike circular orbits, as measured by a zero angular momentum observer (ZAMO), displays a non-monotonic dependence on the radial coordinate. Additionally, the regions supporting stable circular orbits can be discontinuous, and particles on stable orbits near the center can possess specific energies greater than one ($E > 1$).

Quantum annealing is a promising approach for solving combinatorial optimization problems. However, its performance is often limited by the overhead of additional qubits required for embedding logical QUBO models onto quantum annealers. This overhead becomes severe when logical QUBO models have dense connectivity. Such dense structures frequently arise when formulating equality and inequality constraints. To address this issue, we propose a method to construct a significantly sparser logical QUBO model for these constraints. By adding auxiliary variables based on specific network structures, our approach decomposes the original constraint into smaller, more manageable constraints. We demonstrate that this method reduces the number of edges (quadratic terms) from $O(N^2)$ to $O(N)$ for the one-hot constraint and to $O(N\log N)$ in the worst case for general equality constraints, where $N$ is the number of variables. Experimental results on D-Wave's hardware show that our formulation leads to substantial reductions in the number of qubits required for embedding, shorter average chain lengths, lower chain break rates, and higher feasible solution rates compared to conventional methods. This work provides a practical tool for efficiently solving constrained optimization problems on current and future quantum annealers.

Non-perturbative limitations on low-energy effective field theories (EFTs) based on the characteristics of high-energy theory are provided by the analyticity of the flat-space version of the S-matrix. Although the analyticity of the flat-space S-matrix is widely established, it is difficult to apply this framework to de Sitter space because the growing backdrop breaks time-translation symmetry and makes it more difficult to define asymptotic states. The flat-space analyticity imprint on the de Sitter S-matrix is examined in this study. On a certain limit, we derive a comprehensive relationship between the flat-space amplitude and the de Sitter S-matrix. In particular, we demonstrate that the relationship is valid for tree-level amplitude exchanging with arbitrary local derivative interactions with a large scalar field. Next, we contend that this specific limit is more consistent with the definition of EFT since, similar to flat space, the Mandelstam variable may be identified as the unique energy scale because the total energy dependence of the de Sitter S-matrix becomes negligible. Finally, we also find an unexpected connection between the idea of generalized energy conservation of an S-matrix of four-dimensional de Sitter and exceptional EFTs in de Sitter space. We restrict the coupling constants in theories of self-interacting scalars dwelling in the exceptional series of de Sitter representations by requiring that such an S-matrix only has support when the total energies of in and out states are equal. We rediscover the Dirac-Born-Infeld (DBI) and Special Galileon theories, in which a single coupling constant uniquely fixes the four-point scalar self-interactions.

We demonstrate that a two-polariton blockade (2PB) can occur under resonant single-polariton driving in an atom-cavity system operating in the ultrastrong coupling (USC) regime-a phenomenon qualitatively distinct from, and unattainable in, both the strong and weak coupling regimes. In the USC regime, where the ratio of the atom-cavity coupling strength to the cavity resonance frequency exceeds 0.1, hybrid light-matter quasiparticles known as polaritons emerge. By employing modified second- and third-order correlation functions appropriate for the USC regime, we predict the emergence of 2PB, characterized by pronounced two-polariton bunching accompanied by suppressed three-polariton coincidences. This Letter introduces a novel route to achieving 2PB, with promising implications for the realization of multiparticle quantum light sources in the USC regime.

We show that the gap in the compactness between black holes and neutron stars witnessed in general relativity may be vanishing in Hořava-Lifshitz (HL) gravity. Assuming a fermion equation-of-state for simplicity, and solving the Tolman-Oppenheimer-Volkoff equation within the HL gravity framework, we see that there exists a minimum fermion mass $m_f^\text{(min)}(q,y)$, above which the gap of the compactness between black hole and fermionic compact object vanishes, for a given deformation parameter $q$ of HL and interaction strength $y$ between fermions. Thus, in HL gravity, the mass and radius of an object found in the lower mass gap by LIGO-Virgo-KAGRA observations might not be able to classify it as a black hole or a neutron star. It is interesting to note that a fermion of mass $\sim 40\ \text{GeV}$ can form a highly compact object of mass $\sim 10^{-4}\ \msun$ and radius $\sim 1\ \text{m}$ that may play the role of the cold dark matter. In addition, we find the possible existence of another class of compact objects whose compactness is comparable to that of a black hole.

Variational Quantum Algorithms are a vital part of quantum computing. It is a blend of quantum and classical methods for tackling tough problems in machine learning, chemistry, and combinatorial optimization. Yet as these algorithms scale up, they cannot escape the barren-plateau phenomenon. As systems grow, gradients can vanish so quickly that training deep or randomly initialized circuits becomes nearly impossible. To overcome the barren plateau problem, we introduce a two-stage optimization framework. First comes the convex initialization stage. Here, we shape the quantum energy landscape, the Hilmaton landscape, into a smooth, low-energy basin. This step makes gradients easier to spot and keeps noise from derailing the process. Once we have gotten a stable gradient flow, we move to the second stage: nonconvex refinement. In this phase, we allow the algorithm to explore different energy minima, thereby making the model more expressive. Finally, we used our two-stage solution to perform quantum cryptanalysis of the quantum key distribution protocol (i.e., BB84) to determine the optimal cloning strategies. The simulation results showed that our proposed two-stage solution outperforms its random initialization counterpart.

Current quantum neural networks suffer from extreme sensitivity to both adversarial perturbations and hardware noise, creating a significant barrier to real-world deployment. Existing robustness techniques typically sacrifice clean accuracy or require prohibitive computational resources. We propose a hybrid quantum-classical Differentiable Quantum Architecture Search (DQAS) framework that addresses these limitations by jointly optimizing circuit structure and robustness through gradient-based methods. Our approach enhances traditional DQAS with a lightweight Classical Noise Layer applied before quantum processing, enabling simultaneous optimization of gate selection and noise parameters. This design preserves the quantum circuit's integrity while introducing trainable perturbations that enhance robustness without compromising standard performance. Experimental validation on MNIST, FashionMNIST, and CIFAR datasets shows consistent improvements in both clean and adversarial accuracy compared to existing quantum architecture search methods. Under various attack scenarios, including Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Basic Iterative Method (BIM), and Momentum Iterative Method (MIM), and under realistic quantum noise conditions, our hybrid framework maintains superior performance. Testing on actual quantum hardware confirms the practical viability of discovered architectures. These results demonstrate that strategic classical preprocessing combined with differentiable quantum architecture optimization can significantly enhance quantum neural network robustness while maintaining computational efficiency.

We present a rigorous and complete security proof of the decoy-state BB84 quantum key distribution (QKD) protocol. Our analysis aims to achieve a high standard of mathematical rigour and completeness, thereby providing the necessary foundation for certification and standardization efforts. Beyond establishing the security of a specific protocol, this work develops a general and modular framework that can be readily adapted to a broad class of QKD protocols, including both prepare-and-measure and entanglement-based variants. Our framework unifies all major ingredients required for the analysis of realistic QKD protocols, including the analysis of classical authentication and classical processing, source-replacement schemes, finite-size analysis, source maps, squashing maps, and decoy-state techniques. In doing so, this work consolidates a diverse range of techniques scattered across the QKD literature into a unified formalism, representing a general and rigorous treatment of QKD security. Finally, it outlines a clear path towards incorporating practical imperfections within the same framework, thereby laying the groundwork for addressing implementation security in future analysis.

We present a general method for decomposing non-unitary operators into a linear combination of unitary operators, where the approximation error decays exponentially. The decomposition is based on a smooth periodic extension of the identity map via the Fourier extension method, resulting in a sine series with exponentially decaying coefficients. Rewriting the sine series in terms of complex exponentials, then evaluating it on the Hermitian and anti-Hermitian parts of a non-unitary operator, yields its approximation by a linear combination of unitaries. When implemented in a quantum circuit, the subnormalisation of the resulting block encoding scales with the double logarithm of the inverse error, substantially improving over the polynomial relationship in existing methods. For hardware or applications with a fixed error budget, we discuss a strategy to minimise subnormalisation by exploiting the overcomplete nature of the Fourier extension basis. This regularisation procedure traces an error-subnormalisation Pareto front, identifying coefficients that maximise the subnormalisation at a fixed error budget. Fourier linear combinations of unitaries thus provides an accurate and versatile framework for non-unitary quantum computing.

We develop a first-order Feshbach-Villars (FV) Hamiltonian framework for spin-0 relativistic quantum dynamics in the presence of Planck-scale kinematic deformations described within generalized doubly special relativity (G-DSR). Starting from a generic nonlinear momentum-space map, we derive the corresponding modified dispersion relation (MDR) at leading order in the Planck length \(l_p\) and construct a consistent FV linearization of the deformed Klein-Gordon operator. The resulting two-component Hamiltonian remains \(σ_3\)-pseudo-Hermitian at \(\mathcal{O}(l_p)\), which guarantees conservation of the FV charge and current and provides a current-based definition of reflection and transmission in stationary scattering. As applications, we study two benchmark settings in which the FV metric structure is essential: (i) the one-dimensional Klein-Gordon oscillator and (ii) scattering from electrostatic step and barrier potentials. For the oscillator, we obtain controlled \(\mathcal{O}(l_p)\) branch-resolved spectral shifts and show how kinetic versus mass-shell deformations reshape the level spacing and the high-energy spectral compression. For step and barrier scattering, we compute reflection and transmission coefficients directly from the pseudo-Hermitian FV current and quantify the deformation-induced shift of the supercritical (pair-production) threshold. A comparative analysis of the Amelino-Camelia and Magueijo-Smolin realizations indicates that MS-type deformations generally delay the onset of the supercritical regime and reduce the magnitude of the negative transmitted flux within the validity domain \(l_p E \ll 1\).

A primary bottleneck in modern FFT-based matched-filter searches for gravitational waves from compact binary coalescences is not raw processor throughput, but available memory bandwidth. Standard frequency-domain implementations, such as the FINDCHIRP algorithm, rely on streaming long template waveforms and data from main memory, which leads to significant processor stalling when template durations exceed cache capacities. In this work, we introduce \textit{Ratio-Filter Dechirping} as a solution, an algorithmic restructuring of the matched filter that transforms the operation from a memory-bound Fast Fourier Transform (FFT) into a cache-efficient, compute-bound Finite Impulse Response (FIR) convolution. By utilizing a reference template to remove common orbital phase evolution, we produce slowly changing frequency-domain ratios that can be accurately implemented as short FIR filters. This method delivers a measured speedup of $8\times$ for the core filtering loop used in offline searches and should enable $>10\times$ for low-latency analysis. We find that this approach generalizes to a variety of searches that include physical features such as finite size effects, eccentricity, and precession. By dramatically reducing the computational cost of matched filtering, this approach enables the expansion of searches into dense or high-dimensional parameter spaces, such as those for eccentric or subsolar-mass signals, that are already limited by available computing budgets. Furthermore, this framework provides a natural path for hardware acceleration on GPU architectures.

We develop a dynamical framework for quantum measurement based on stochastic but unitary evolution in projective state space. Random Hamiltonians drawn from the Gaussian Unitary Ensemble generate stochastic unitary dynamics of the quantum state, while equivalence classes reflecting finite detector resolution define classical observables as well as classical configuration-space and phase-space submanifolds. When the evolution is constrained to the phase-space submanifold, free Schrödinger dynamics reduces to Newtonian motion, while stochastic motion constrained to the classical configuration-space submanifold yields ordinary Brownian motion in classical space. Transition probabilities under the stochastic dynamics satisfy the Born rule, whereas the constrained classical evolution produces the normal probability distributions characteristic of classical measurements. We show that, in this setting, measurement, state reduction, and the quantum-classical transition emerge from unitary dynamics alone, without invoking nonunitary collapse or additional postulates. Entanglement and EPR correlations arise geometrically from the evolution of joint states in composite state space, preserving locality in spacetime. The framework provides a unified dynamical account of measurement and classicality compatible with the structure of quantum mechanics.

A new realization of the matter Lagrangian is introduced which models the dark energy component as a non-standard combination of thermodynamics quantities of the baryonic matter. We will prove that the present realization is independent of existing models with matter-geometry couplings and has a property that the energy-momentum tensor of both baryonic matter and dark energy is conserved separately. We further show that two possible choices of the matter Lagrangian in the $Λ$CDM model are not totally equivalent and investigate the background and perturbative constraints on the form of matter Lagrangian. We will also investigate cosmological implications of a test model with logarithmic DE and obtain the model parameters by confronting the model with observational data on the cosmic chronometers, Pantheon$^+$ and $fσ_8$ datasets. We will also explain in details the predictions of the model on the late time behavior of the universe and compare the result with $Λ$CDM model.

In this work, we study the neutrino oscillation phase propagating along radial and non-radial paths in the Einstein$\text{-}$Hilbert$\text{-}$Bumblebee (EHB) gravity around global monopole field, by using the quantummechanical treatment, the expression of the corrected neutrino oscillation probability is obtained. Our results show that neutrino oscillation in the EHB gravity around global monopole field is different from that in Schwarzschild black hole, the Lorentz-violating parameter $a$ mainly affects the peak of the oscillation probability, and the global monopole $\overline { μ}$ and the lightest neutrino mass both affect the frequency of the oscillations of probabilities, which means that studying the neutrino oscillation phenomenon in curved spacetime may not only become a new technology for probing the properties of compact celestial bodies, but also help deepen our understanding of neutrinos.

Supersymmetry (SUSY) provides elegant solutions to several open questions in the Standard Model, and searches for SUSY particles are an important component of the LHC physics program. Naturalness arguments favour supersymmetric partners of the gluons and third-generation quarks with masses light enough to be produced at the LHC. With increasing mass bounds on more classical Minimal Supersymmetric Standard Model (MSSM) scenarios other variations of supersymmetry, including non-minimal particle content, become increasingly interesting. This proceeding will present the latest results of searches conducted by the ATLAS experiment at LHC at center of mass energies of $\sqrt{s}=13$ and 13.6 TeV which target gluino and squark production, including stop, in a variety of decay modes.

We show that small perturbations in the boost-invariant color fields of the glasma exhibit an exponential growth with the square root of time. We interpret this growth rate as a Lyapunov exponent, related to entropy production and the thermalization timescale in the earliest stage of heavy-ion collisions. Working in a regime that is linear in this perturbation, we extract the time dependence of this mode as $\sim \exp(0.4\sqrt{g^2μτ})$ for SU($2$), where $g^2μ$ is proportional to the saturation scale and the square-root dependence is caused by the boost-invariant expansion of the system. We show that the growth rate of this mode is, unlike its amplitude, remarkably insensitive to the details of how the perturbations are initialized. In particular, we show that the unstable mode couples to all momentum scales present in the initial perturbation.

While $Λ$CDM provides a good fit to cosmological data, it fails to address many of the outstanding questions in contemporary cosmology. Chief among these are the Hubble tension and the apparent dynamical nature of dark energy as inferred from the recent DESI DR2 analysis. In this work, we analyze a field-theoretic description of cosmology where both dark energy and dark matter are interacting spin zero fields. We give a thorough study of a wide range of the interaction strength and demonstrate the effect on the dark energy equation of state and the Hubble tension. Using the recent cosmological data, we extract constraints on cosmological parameters including the free parameters of the model.

A recent study has established the Na$_2$KSb(Cs) photocathode as a highly efficient emitter of spin-polarized electrons. However, the electronic structure of alkali antimonides remains poorly understood. In this work, we report the first crystalline epitaxial growth of Na$_2$KSb films, achieved via chemical vapor deposition (CVD) on a graphene-coated SiC(0001) substrate. The high crystalline quality of these films enabled a direct investigation of the material's electronic structure using angle-resolved photoemission spectroscopy (ARPES). By comparing the experimental results with density functional theory (DFT) calculations, we have identified dispersive surface states originating from different terminations of the Na$_2$KSb(111) surface. Furthermore, we demonstrate that the crystalline order of the film is preserved following its activation via the deposition of Cs and Sb. This finding opens a pathway for investigating the electronic structure of multialkali Na$_2$KSb(Cs) photocathodes and for rationally improving their properties.

$B$ mesogenesis offers an interesting mechanism to generate the baryon asymmetry of the universe by converting the CP violation of the Standard Model into a net baryon number asymmetry. In this work we refine and extend the $B$ mesogenesis framework by incorporating all relevant $B$ meson channels active after low-temperature reheating. We first update the known neutral-meson contribution using time-integrated decay rates. While the $B_s^0$ contribution remains essentially unchanged, we find a suppression of the $B_d^0$ term by a factor $\sim 0.4$ with respect to previous analyses, alleviating the tension associated with its expected negative sign. We then perform a systematic study of $B_c^+$ decays, which are basically unexplored. We provide branching-ratio predictions using both leading-order factorization and a data-driven approach inspired by $D\to M_1M_2$ decays. These estimates allow us to quantify two $B_c^+$ sources of mesogenesis: the previously discussed $B_c^+ \to B^+ M^0$ channel and a new mechanism introduced in this work, $B_c^+ \to B_q^0 M$, in which the asymmetry is generated by combining direct CP violation with neutral-meson oscillations. Interestingly, in these channels the charm quark decays. Therefore, these decays give access to {\it charm CP violation} in modes with percent level branching ratios. With moderate assumptions, we find that $B_c^+$ mesogenesis can match or exceed the neutral contribution. Finally, we combine all three mechanisms and explore their viability in terms of the direct CP asymmetry, neutral-meson mixing parameters and early-universe fragmentation fractions. We find that successful baryogenesis can be achieved in a broad parameter space, showing the viability of $B$ mesogenesis. Future measurements of $B_c^+$ modes are thus highly anticipated to further probe the viability of unified $B$ mesogenesis.

We investigate how effectively final-state jet substructure can discriminate between QCD Compton and quark-antiquark annihilation processes from photon-jet production in $pp$ collisions at $\sqrt{s}=13$ TeV. Using infrared- and collinear-safe jet observables, multivariate classifiers -- boosted decision trees and multilayer perceptrons -- are trained on labeled quark- and gluon-initiated jets from dijet events and applied to photon-jet samples. Observables probing soft and wide-angle radiation, in particular jet multiplicity and jet girth, dominate the discrimination. The jet mass provides a complementary but weaker contribution, while the jet charge exhibits negligible discriminating power. A comparison of the two classifiers demonstrates that the achievable separation is limited primarily by QCD radiation effects rather than by classifier complexity. These findings quantify the extent to which information about the underlying hard process survives hadronization and realistic jet reconstruction, providing a physics-driven baseline for precision jet measurements in $pp$, $ep/$A, and heavy-ion collisions.

Particle correlations and femtoscopy are a rich subfield of high-energy physics. As the experimental data become more precise, there is an increasing need for the theoretical calculations to provide better and more general descriptions of the measurements. One of the important new directions is the investigation of the precise shape of the Bose-Einstein correlation functions utilizing Lévy-stable distributions. This work is a direct follow-up to our previous study, in which we developed a novel method for calculating Bose-Einstein correlation functions including the Coulomb final-state interaction. In this paper, we present a self-consistent generalization of the previous approach to non-spherical source functions and investigate the validity of the previously applied approximations assuming spherical symmetry. We present a software package that includes the calculation of a fully three-dimensional correlation function including the Coulomb interaction.

The question how well nonrelativistic QCD (NRQCD) factorization can describe quarkonium production has been subject to debate since its invention. We review our recent reanalysis of a classic next-to-leading order color octet (CO) long distance matrix element (LDME) fit to large transverse momentum $p_T$ $J/ψ$ and $η_c$ LHC production data. Our analysis differs from previous analyses of this kind not only by implementing for the first time a systematic treatment of scale uncertainties, but also by scrutinizing a much broader range of observables. Surprisingly, $J/ψ$ hadroproduction is well described up to the highest measured values of $p_T$. Potential NRQCD based relations nontrivially lead to a perfect description of $Υ(nS)$ production data. Furthermore, $J/ψ$ production in $γp$ and $γγ$ collisions is, contrary to prevailing conceptions, reproduced down to $p_T=1$ GeV, as long as the region of large inelasticity $z$ is excluded. The overall picture is much rosier than usually perceived, the more so as the remaining discrepancies appear in phase space regions where solutions via varying kinds of resummations have been proposed.

Motivated by the necessity of developing theoretical models for studying the electroweak structure of baryons in a nuclear medium, we apply a covariant quark model to study interactions of baryons with nuclear matter. The electromagnetic and axial form factors of the octet baryons are determined by combining a covariant quark model that takes into account the meson cloud dressing of the baryon cores, developed for free space, with the quark-meson coupling model in the extension to the nuclear medium. We discuss the medium modifications on the electroweak form factors of octet baryons for the range of densities from $ρ=0$ up to $ρ=2 ρ_0$, where $ρ_0= 0.15$ fm$^{-3}$ is the normal nuclear matter density. We also study how the shape of the form factors is modified in finite nuclei due to the profile of the nuclear density distributions compared with calculations using the average density of the nucleus

The radiative return, together with the energy scan, is the method used at flavour factories to measure the pion form factor, which is a crucial input for the data-driven dispersive computation of the leading-order hadronic contribution to the muon anomalous magnetic moment. We consider the radiative hadronic and leptonic channels of main experimental interest, namely the processes $e^+e^-\to X^+X^-γ$, with $X = \{π\, , μ\}$. For such processes, we compute the exact next-to-leading order (NLO) corrections matched to a Parton Shower (PS) to describe exclusive multiple photon emission. All sources of radiative corrections from initial-state and final-state radiation, as well as their interference, are considered according to QED for $e^+e^-\toμ^+μ^-γ$ and QED$\oplus$F$\times$sQED (Factorised scalar QED) for $e^+e^-\toπ^+π^-γ$. We describe in detail the novel features of our PS approach to compute the fixed-order corrections in association with higher-order contributions to $2\to3$ processes, with a hard photon in the final state. We present validation tests and comparisons with NLO predictions available in the literature to cross-check various ingredients of our formulation. We also show numerical results at NLOPS accuracy according to realistic event selection criteria for precision measurements at flavour factories. Our calculation is implemented in an updated version of the Monte Carlo event generator BabaYaga@NLO, which can be used for fully exclusive simulations and data analysis in radiative return experiments.

We study the profile of the flux tube in the SU(2) gauge model in (2 + 1) dimensions, with a particular attention to the so called "intrinsic width" which drives the exponential decay of the flux density at large transverse distances and represents a new physical scale of the model. This quantity is directly related to the confining mechanism which generates the flux tube and can be used to test its properties. We study a wide range of different values of lattice spacing, temperature and flux tube lengths and show that our data are precise enough to distinguish between different confining models. In particular we show that at high temperatures (just below the deconfinement transition) the data are perfectly described by an Ising-like effective model based on the Svetitsky-Yaffe mapping. At lower temperatures this approximation does not hold anymore. In this regime (which is the most interesting one from a physical point of view) we test several alternative proposals and show that the dual superconductor model is the one which better fits the data. However, this proposal is not fully satisfactory, because the values of the Ginzburg-Landau parameter extracted from the fits increase with the length of the flux tube, which is not a feature predicted by the model. This suggests that a more sophisticated model is needed to explain confinement in non-abelian gauge theories and, at the same time, that our data on the intrinsic width may be a powerful tool to benchmark these candidates.

In this letter, we test clipping effective field theory (EFT) simulations as a method of ensuring EFT validity. The procedure imposes that, at the level of the simulation, the invariant mass of a $W^+W^-$ pair $M_{WW}$ is less than the new physics scale $Λ$. We compare this to two other methods, comparison bin by bin of dimension-6 and dimension-8 squared contributions and implementing a cut on data. We find that setting $M_{WW} < Λ$ is not strict enough to ensure that the hierarchy of EFT operators is respected for dimension-6 and dimension-8 contributions. We also show that, even when using a stricter cut on $M_{WW}$, due to different correlations between $M_{WW}$ and $M_{eμ}$ at different EFT orders, the bins in $M_{eμ}$ (the invariant mass of the leptons originating from $W$ decays) used in an EFT fit may not truly be in the regime of EFT validity when performing a dimension-6 fit with $M_{WW} < Λ$. We also explore the correlations of three transverse mass observables: $M_{T1}, M_{T2}$ and $M_{T3}$, finding that $M_{T1}$ and $M_{T3}$ follow the $M_{WW}$ distribution more closely than $M_{eμ}$. We present sensitivity studies using both the $M_{T3}$ distribution and $M_{eμ}$ distribution. We test implementing an experimental cut on $M_{T3}$ in place of clipping the EFT simulation at $M_{WW} < Λ$. We finally comment that adding $M_{WW} < Λ$ cuts only to the EFT simulation could be interpreted as modifying the SMEFT expansion by a form factor and could therefore impact the model independence of EFT fits under this procedure.

Using a sample of $(2712.4\pm14.3)\times10^{6}$~$ψ(3686)$ events collected with the BESIII detector, we perform a search for the isospin-violating decays $χ_{cJ}\toΛ\barΣ^{0}+c.c.~(J=0,1,2)$ and $η_{c}\toΛ\barΣ^{0}+c.c.$ No significant signal for $χ_{cJ}$ or $η_{c}$ is observed in the $Λ\barΣ^{0}$ invariant mass distribution. The upper limits on the branching fractions at the 90\% confidence level are set to be $\mathcal{B}$($χ_{c0}\toΛ\barΣ^{0}+c.c.$)$<1.5\times 10^{-6}$, $\mathcal{B}$($χ_{c1}\toΛ\barΣ^{0}+c.c.$)$<1.6\times 10^{-6}$, $\mathcal{B}$($χ_{c2}\toΛ\barΣ^{0}+c.c.$)$<1.7\times 10^{-6}$ and $\mathcal{B}$($η_{c}\toΛ\barΣ^{0}+c.c.$)$<6.2\times 10^{-5}$ for the first time.

Efficient generation of LHC events is hindered by the rapidly rising cost of evaluating QCD matrix elements with increasing multiplicity. We build on a recently proposed two-step strategy in which unweighted events are first generated using the leading-colour (LC) approximation and then reweighted to full-colour (FC) accuracy, utilising the LC integration efficiency while recovering the exact FC prediction. In this work we extend the method to general Standard Model processes and present AmpliCol, a standalone implementation designed for LHC collisions. We benchmark multi-jet, $t\bar{t}$+jets, $ZZ$+jets, and Drell-Yan+jets production, measuring the time required to obtain a fixed number of unweighted events at FC accuracy. Across all processes, the runtime exhibits a stable exponential scaling with multiplicity, far milder than the factorial growth of conventional matrix-element generators. This demonstrates that the AmpliCol code enables efficient event generation at multiplicities that are otherwise computationally prohibitive.

The breakdown of azimuthal two-particle correlation factorization, quantified by the ratios $r_2$ and $r_3$, serves as a sensitive probe of transverse-momentum-dependent flow fluctuations. While hydrodynamic models predict $r_3 \leq 1$, experimental data from CMS in p-Pb collisions exhibit $r_3 > 1$, presenting a clear puzzle. We show that transverse momentum conservation (TMC) is the key mechanism dictating this factorization breakdown in small systems. We systematically calculate the effect of TMC as a function of the momentum difference between particles across various multiplicity and momentum ranges. Our results are in quantitative agreement with CMS p-Pb data for both $r_2$ and $r_3$. A central finding is a sign rule: under TMC, the deviation $r_n - 1$ follows $\left ( - 1 \right )^{n+1} $, being negative for even and positive for odd harmonic orders $n$. This work establishes an analytical framework to quantify transverse-momentum-dependent flow fluctuations and provides new insights into the origin of collectivity in small colliding systems.

We discuss key aspects of the nature of radiation from global strings and its impact on the relic axion density. Using a simple model we demonstrate the dependence on the spectrum of radiation emitted by strings. We then study the radiation emitted by perturbed straight strings paying particular attention to the difference between the overall phase of the field and the small perturbations about the string solution which are the axions. We find that a significant correction is required to be sure that one is analyzing the axions and not the self-field of the string. Typically this requires one to excise a sizeable region around the string - something which is not usually done in the case of numerical field theory simulations of string networks. We have measured the spectrum of radiation from these strings and find that it is compatible with an exponential, as predicted by the Nambu-like Kalb-Ramond action, and in particular is not a ``hard'' spectrum often found in string network simulations. We conclude by attempting to assess the uncertainties on relic density and find that this leads to a range of possible axion masses when compared to the measured density from the Cosmic Microwave Background, albeit that they are typically higher than what is predicted by the Initial Misalignment Mechanism. If the decay is via a ``soft spectrum'' from loops produced close to the backreaction scale we find that $m_{\rm a}\approx 160\,μ{\rm eV}$ and a detection frequency $f\approx 38\,{\rm GHz}$. If axions are emitted directly by the string network, and we use emission spectra reported in field theory simulations, then $m_{\rm a}\approx 4\,μ{\rm eV}$ and $f\approx 1\,{\rm GHz}$, however this increases to $m_a \approx 125\,μ{\rm eV}$ and $f\approx 30\,{\rm GHz}$ using our spectra for the case of an oscillating string. In all scenarios there are significant remaining uncertainties that we delineate.

We study a $U(1)_{B-L}$ extension of the Standard Model (SM) in which Dirac neutrino masses are generated radiatively at the one-loop level through the exchange of new beyond the SM fields. This framework establishes a direct connection between neutrino mass generation and the dark sector, with the stability of the dark matter ensured by a residual discrete $Z_6$ symmetry arising from the spontaneous breaking of $U(1)_{B-L}$. We investigate the resulting charged lepton flavor violating processes and dark matter phenomenology, saturating relic observations and direct-detection constraints, and analyze the collider signatures of the dark sector at the Large Hadron Collider and at a future muon collider. We have identified excellent prospects for observing the considered dark matter candidates in these colliders, even with lower integrated luminosities than the proposed one.

We investigate the process in which Deeply Virtual Compton Scattering (DVCS) excites a baryon resonance. In particular, we assess, in DVCS leading to the Roper resonance, the relative importance of a "background'' process in which a pion is first emitted by the nucleon, which then undergoes a DVCS event. Our numerical results, using realistic DVCS kinematics, indicate that there can be measurable interference effects. They suggest that this process could substantially modify the experimentally observed cross sections at CLAS12-like kinematics, motivating their inclusion in precision analyses of DVCS experiments. We further find that in spite of this background, the transition to a Roper-like state through DVCS does contribute significantly to the $e^- N\to e^-γNπ$ cross section in some kinematic regions. This suggests that the creation of nucleon resonances via DVCS is a useful method for extracting information about the nucleon transition GPDs and the internal structure of the excited states.

In these proceedings we review the physics objects used by the CMS experiment during LHC Run 3 at 13.6 TeV, including charged leptons, photons, jets, and missing transverse momentum. Their performance and calibration is critical for physics analysis. In particular, the algorithms need to be resilient against the high pileup conditions in Run 3 collisions. Furthermore, transformer-based algorithms are deployed for the identification of heavy-flavor jets and boosted resonances.

The search for neutrinoless double beta decay ($0νββ$) is currently one of the key objectives in neutrino physics research. The decay rate of $0νββ$ decay depends on the effective Majorana neutrino mass $|\langle m \rangle_{ee}|$. In this work we study the numerical prediction of $|\langle m \rangle_{ee}|$ in the scenario of deviation from the $μ$-$τ$ reflection symmetry due to radiative corrections, as an extension of our earlier work \cite{pegu}. In \cite{pegu}, we consider an exact $μ$-$τ$ reflection symmetry in the light effective Majorana neutrino mass matrix and in the corresponding lepton mixing matrix as well at the seesaw scale. We choose numerical values of all the mixing parameters and neutrino mass eigenvalues at the seesaw scale as inputs and estimate the values of mass eigenvalues and mixing parameters at the electroweak scale due to radiative corrections. We find these low energy predictions consistent with global $3σ$ oscillation data. In the present work, we compute the effective Majorana neutrino mass $|\langle m \rangle_{ee}|$ using these low energy values at the electroweak scale. We find that the low energy predictions of $|\langle m \rangle_{ee}|$ are consistent with the latest upper bound $|\langle m \rangle_{ee}|<(0.028-0.122)\ eV$ provided by KamLAND-Zen Collaboration.

CP4 3HDM is a curious version of the three-Higgs-doublet model built upon a CP symmetry of order 4 (dubbed CP4). When extended to fermions, CP4 leads to unusually tight correlations between the scalar and Yukawa sectors and induces tree-level flavor changing neutral couplings. Still, viable scenarios exist, in which quark flavor changing signals remain within experimental limits. In this work, we extend CP4 to the lepton sector and investigate whether the lepton-Higgs couplings and lepton flavor violating (LFV) signals can also be kept under control. We consider two classes of LFV processes: tree-level lepton decays of the 125 GeV Higgs boson and one-loop radiative decay $μ\rightarrow eγ$. For each CP4-invariant lepton Yukawa scenario, we perform a focused Yukawa sector scan that uses physical lepton properties as input and suppresses LFV effects. We identify a promising CP4 3HDM scenario compatible with the present-day experimental constraints, show that it can accommodate the recent CMS hint of a 146 GeV scalar decaying to $eμ$, and argue that this interpretation can be tested at future colliders.

We study the processes $ν_τn \to Λ_c τ^- (π^- ν_τ)$ and $\barν_τp \to Λτ^+ (π^+ \barν_τ)$, with particular emphasis on the pion energy and angular distributions, as a possible signal for lepton flavor universality violation, and in general of physics beyond the Standard Model (SM), as well as a sensitive probe of the $τ$ polarization vector. We work within an effective low-energy extension of the SM with all dimension-six four-fermion operators. In this framework, complex Wilson coefficients which encode new physics can generate CP-violating contributions. We propose an observable that provides a genuine CP-odd signal due to its sensitivity to particular transverse components of the $τ$ polarization vector. Namely, we show that the asymmetry in the azimuthal-angle distribution of the pion in the decay $τ^\pm\to π^\pm ν_τ$ constitutes a smoking-gun prediction of such a beyond the SM scenario. We estimate the strength of this effect extrapolating nucleon-hyperon form factors recently obtained from lattice QCD calculations.

We investigate the possibility of constraining primordial black holes (PBHs) with masses $M_\mathrm{PBH}\gtrsim 10^{15}\,\mathrm{g}$ through Galactic diffuse synchrotron emissions. Due to Hawking radiation, these types of PBHs are expected to be stable sources of cosmic-ray (CR) electrons and positrons with energies below $\mathcal{O}(10\,\mathrm{MeV})$. In many CR propagation models with diffusive re-acceleration characterized by a significant Alfvén velocity $V_a\sim \mathcal{O}(10)\,\mathrm{km/s}$, the energies of the evaporated electrons/positrons can be further enhanced to $\mathcal{O}(100)\,\mathrm{MeV}$ through their scattering with the Galactic random magnetic fields. Consequently, the observation of Galactic synchrotron emissions at frequencies above $\sim 20\,\mathrm{MHz}$ can provide useful constraints on the abundance of PBHs. Using the AMS-02 and Voyager-1 data on the boron-to-carbon nuclei flux ratio, we confirm that a significant Alfvén velocity $V_a \sim 20\,\mathrm{km/s}$ is favored in several benchmark diffusive re-acceleration models. We show that, in this scenario, the observed low-frequency synchrotron emissions (from 22 MHz to 1.4 GHz) can provide stringent constraints on PBH abundance. The obtained conservative constraints are stronger than those derived from the Voyager-1 all-electron (electron plus positron) data by more than one order of magnitude for $M_\mathrm{PBH}\gtrsim 1\times 10^{16}\,\mathrm{g}$, and also stronger than our previous constraints derived from the AMS-02 positron data for $M_\mathrm{PBH}\gtrsim 2\times 10^{16}\,\mathrm{g}$.

We develop a heat kernel method to compute the one-loop effective action for a general class of nonlinear electrodynamic (NLED) theories in four dimensional Minkowski spacetime. Working in the background field formalism, we extract the logarithmically divergent part of the effective action, the so-called induced action, corresponding to the DeWitt $a_2$ coefficient of the heat kernel. In NLED, quantisation yields non-minimal differential operators, for which standard heat kernel techniques are not immediately applicable. Considering the weak-field regime, we calculate the $a_0$, $a_1$ and $a_2$ contributions to leading order in the background electromagnetic field strength. Finally, we consider conformal NLED theories and compute the $a_0$ contribution to all orders. For this class, we comment on the role of causality being necessary and sufficient for the convergence of the exact $a_1$ and $a_2$ contributions.

We studied a crucial but often oversimplified ingredient in predicting gravitational-wave signals from QCD-type phase transitions: the kinetic term of the Polyakov loop. For the first time, we derive this term from first principles in finite-temperature pure SU(3) Yang-Mills theory, incorporating a field-dependent renormalization factor--a calculation we also extend to theories with more colors. Employing this derived kinetic term alongside three commonly-used effective potentials (the Haar-measure, polynomial, and quasi-particle models), we demonstrate that it substantially modifies the predicted GW energy spectrum from confinement transitions by 1-2 orders of magnitude. Based on this, we provide the first complete analysis of the chiral transition within the Polyakov-Nambu-Jona-Lasinio (PNJL) framework, described by the quark condensate. Our results reveal a clear dichotomy: while the Polyakov-loop kinetic term critically shapes GWs from confinement transitions, it has a negligible impact on the dynamics of the chiral transition, which is dominated by fermion condensation effects.

The exploration of cosmic rays, which are high-energy particles originate from space and the atmosphere, has historically been associated with particle physics and astrophysics. In the last 20 years, these particles have evolved into valuable tools for observing Earth's systems. This review compiles the use of cosmic rays in three primary areas: (1) particle physics and atmospheric processes, which include cosmic-ray-induced cascades, ionization, and their impact on atmospheric chemistry and radiation; (2) geosciences, where cosmogenic radionuclides assist in the dating of geological materials and cosmic-ray neutrons are used for large-scale monitoring of soil moisture and snow water equivalents; and (3) urban science, where cosmic-ray muons are employed for non-invasive subsurface imaging and, when paired with distributed sensors, serve as the basis for smart city monitoring. The review places particular emphasis on integrating these methods with remote sensing and geographic information systems (GIS), which helps close the persistent scale gap between point measurements and satellite observations, thereby enabling three-dimensional digital representations of subsurfaces. The review concludes by discussing the data standards, their integration into operational Earth observation workflows, and future research directions.

We present the interpolating conformal algebra between the instant form dynamics (IFD) and the light-front dynamics (LFD) in $(1+1)$ dimensions, along with a $4\times4$ interpolating projective spacetime matrix representation. While there are six generators in the $(1+1)$ dimensional conformal algebra, the number of kinematic and dynamic generators dramatically changes in LFD, maximizing (minimizing) the number of kinematic (dynamic) generators to four (two) with respect to two (four) kinematic (dynamic) generators in IFD, as well as in any other forms of dynamics between IFD and LFD. It confirms and signifies the utility of LFD, saving substantial dynamical efforts in solving the $(1+1)$ dimensional quantum field theories. We also present $2\times2$ Pauli matrix representation of $(1+0)$ and $(0+1)$ conformal groups, and creation/annihilation operators of quantum simple harmonic oscillator representations of $(1+0)$ dimensional conformal groups.

We construct a field redefinition between Witten's string field theory and its deformation by the Ellwood invariant. This field redefinition is singular and does not imply physical equivalence between them. However, it allows us to formally transfer classical solutions of Witten's theory to solutions of the deformed theory. Although the resulting solutions are also generically singular and require careful examination of their physical interpretation, we show that the tachyon vacuum solution can be consistently transferred from Witten's theory to the deformed theory.

Recent results from the ATLAS and the CMS experiments at the Large Hadron Collider indicate the presence of a top-quark pair bound state near the threshold region. We present a way to reconstruct a toponium state at the $t\bar{t}$ threshold region formed at the Large Hadron Collider using the Recursive Jigsaw Reconstruction. We have considered the Non-Relativistic QCD based toponium model implemented in MadGraph5\_aMC@NLO. The final states considered consist of two b-jets, two oppositely charged leptons, and missing energy that arises from two neutrinos. The goal of the Recursive Jigsaw Reconstruction is to make use of rules that can help resolve combinatorics ambiguity in preparing the decay tree for a given physics event. Additionally, missing energy coming from two neutrinos needs to be resolved in order to reconstruct the event. We apply four different methods within the RestFrames package and compare the reconstruction results resulting from each of the methods. Due to this method, one can also access kinematic variables in the rest frames belonging to intermediate particle states, providing additional means to discriminate the SM $\ttbar$ background from the toponium signal. We propose using two angular variables to enhance sensitivity to the toponium signal. Our preliminary results indicate that the improvement in sensitivity can be as much as 16\% over the current strategy in the LHC's Run 3 configuration. This method may be useful for gaining additional insight into the physics phenomenology in the $\ttbar$ threshold region.

In the experiment to study the e+e->n+anti n process at the VEPP-2000 $e^+e^-$ collider, the ratio |GE|/|GM| of the neutron timelike electric and magnetic form factors has been measured. The experiment was carried out with the SND detector in the center-of-mass energy range 1890-2000 MeV in 8 energy points with an integrated luminosity of 83 inv.pb. The |GE|/|GM| ratio is determined by the analyzing the distribution of the polar angle of the produced antineutron. The measured |GE|/|GM| value in the energy range under study is between 1.0 and 1.5 with an average value of 1.21+-0.13.

We numerically study the Q-ball formation triggered by a cosmological first-order phase transition within the Friedberg-Lee-Sirlin model. By performing lattice simulations, we track the nonequilibrium dynamics throughout the transition, providing a precise description of the Q-ball formation mechanism and the resulting mass spectrum. Collapsing false-vacuum regions first form thermal balls, which subsequently cool via dissipative interactions and stabilize into long-lived Q-balls with nonzero spin. We observe a large population of low-mass Q-balls, as well as rare, massive Q-balls that are several times larger than the analytical prediction. The final Q-ball population exhibits a broad mass spectrum spanning over two orders of magnitude, characterized by an exponential tail of number density at large masses. The simulations suggest that the Q-ball abundance is approximately $50\%$ higher than predicted by analytical estimates, adjusting the result in the context of Q-balls as dark matter candidates.

Recent lattice QCD simulations and phenomenological models indicate that the nucleon's gravitational form factor $B_N(t)$ remains remarkably small at finite momentum transfer $t$. While $B_N(0) = 0$ is a known consequence of the equivalence principle, the physical origin of its suppression at finite $t$ has not been fully elucidated. In this work, we demonstrate that the smallness of $B_N(t)$ arises from a fundamental cancellation within the nucleon's wave functions. Using light-front holographic QCD, we show that $B_N(t)$ is governed by an antisymmetric factor in the longitudinal dynamics that leads to an exact vanishing in the symmetric limit and significant suppression for realistic nucleon structures. Our results suggest that the smallness of $B_N(t)$ is a signature of the nucleon's dominant S-wave character, providing a formal justification for its frequent omission in practical applications like near-threshold $J/ψ$ production.

There are 17,612 20-inch photomultiplier tubes (PMTs) installed at the Jiangmen Underground Neutrino Observatory (JUNO). Developing a precise optical model for the PMTs is crucial for enhancing the accuracy of detector simulations and refining the energy response model at JUNO. In this study, we established a comprehensive PMT optical model based on prior studies, taking into account the non-uniformity of photon detection efficiency (PDE) across the PMT surface and the variances in PDE as well as reflections among different PMTs. By collecting reflectance data from 669 PMTs and utilizing PDE data from mass testing systems, we estimated the thickness maps of the photocathode (PC) and the anti-reflective coating (ARC) for each PMT. We also determined the collection efficiency (CE) by decomposing PDE with consideration of the optical processes occurring within the PMTs. The refractive index and extinction coefficient of both the PC and ARC, along with the escape factor, were evaluated over a broad wavelength range of 300~nm to 700~nm, covering the entire spectrum of interest for JUNO. Compared to the prediction from a simplified PMT optical model, which assumes uniform PC and ARC across all PMTs of the same type, the further developed PMT optical model yields much more reflectance for HPK PMTs and less for NNVT PMTs, and the change in PDE is at the level of a few percent. This comprehensive PMT optical model also provides a valuable reference for other PMT-based applications.

We study low energy galactic neutrinos in the Milky Way under two fundamentally different descriptions of gravity, showing that neutrinos provide a sensitive probe of gravity underlying nature. If gravity is a quantum interaction, its long range character leads to the formation of an atom like bound neutrino structure. We compute its mass distribution and find that, within a radius 292 kpc, the total mass is only ten to the minus 29 of the galaxy dark matter, ruling it out as a dark matter candidate. Nevertheless, experimental confirmation of this structure would constitute direct evidence for gravity as a quantum force mediated by gravitons. If gravity instead arises from spacetime curvature, neutrinos interact only via the short range weak force and are therefore effectively collisionless. In this regime, neutrinos behave as free classical particles orbiting the galaxy and experience no Fermi pressure. We show that such a population can be sufficiently compact to reproduce the Milky Way rotation curve, making neutrinos viable dark matter candidates. The extremely small neutrino antineutrino annihilation cross section further implies near equilibrium between neutrinos and antineutrinos, potentially addressing the matter antimatter asymmetry.

Lepton flavor violation in the $τμ$ sector would be a clear sign of Beyond Standard Model physics. We employ the SMEFT framework to study the process $e^+e^-\toτμ$ at the ILC and CLIC. We find that the $e^+e^-$ beam polarizations achievable at these machines allow us to probe the chirality structure of the SMEFT operators. In addition, the high center of mass energy leads to a substantial increase in sensitivity to the four-fermion operators that rivals, and in some cases, surpasses tau decay projections from Belle-II.

We introduce a family of polytopes -- in-in zonotopes -- whose boundary structure organizes the contributions to scalar equal-time correlators in flat space computed via the in-in formalism. We provide explicit Minkowski sum and facet descriptions of these polytopes, and show that their boundaries factorize into products of graphical zonotopes and lower-dimensional in-in zonotopes, thereby mimicking the factorization structure of the correlators themselves. Evaluating their canonical forms at the origin -- equivalently, calculating the volume of the dual polytope -- reproduces the correlator. Finally, in a simple example, we show that the wavefunction decomposition of the correlator corresponds to a subdivision of the dual polytope.

We show that the type-A conformal anomaly in $2n$ dimensions follows from standard Stora-Zumino descent, starting from the Euler invariant polynomial for the Euclidean conformal group $SO(2n+1,1)$ in $6d$, thereby placing type-A anomalies on the same footing as ordinary perturbative t Hooft anomalies. We discuss implications for anomaly inflow, and t Hooft anomaly matching for the full conformal group with a Wess-Zumino-Witten term. In 4d, this enables the construction of a dilaton effective action matching the full type-A $SO(5,1)$ conformal anomaly.

Thanks to a number of neutrino oscillation and Coherent Elastic neutrino Nucleus Scattering (CE$ν$NS) experiments, the vector Non-Standard Interactions (NSI) of neutrinos have been well studied and constrained. We show that the long-sought-after new physics may hide in the ``axial" non-standard interactions rather than in the vector NSI. We then show how by studying neutral current scattering events in the detectors of long baseline experiments, MINOS, MINOS$+$ and DUNE, the impact of the axial NSI can be discovered.

We construct Schubert line defects in the 3d $\mathcal{N}=2$ supersymmetric gauged linear sigma model (GLSM) with target space a partial flag manifold $X={\rm Fl}({\boldsymbol{k}};n)$, generalizing our construction for complete flag manifolds given in a companion paper arXiv:2512.19802 (part I). In the context of the 3d GLSM/quantum K-theory correspondence, the Schubert line defects are constructed as 1d $\mathcal{N}=2$ supersymmetric gauge theories coupled to the 3d field theory, and they flow to objects supported on Schubert varieties $X_w \subseteq X$ in the quantum K-theory. The flavored Witten index of the 1d defect is expected to compute the Chern character of $[\mathcal{O}_w]$ -- more precisely, it gives us a polynomial representative of the Schubert class in the quantum K-theory ring. We give strong evidence for this claim by showing in examples that the Witten indices of Schubert defects indeed reproduce a recently-defined set of polynomials that represent the Schubert classes in the Whitney presentation, which we call the parabolic Whitney polynomials. Moreover, upon using the quantum ring relations, we can convert these polynomials into seemingly new polynomials in the Toda presentation, which we call the parabolic quantum Grothendieck polynomials. These new polynomials specialize to known polynomials in various limits, including to the quantum Grothendieck polynomials in the case of the complete flag. In the 2d limit, our construction also realizes the Schubert classes $[X_w]$ in the quantum cohomology ring of the partial flag manifold, and the parabolic quantum Grothendieck polynomials then reduce to previously known parabolic quantum Schubert polynomials.

Semi-Inclusive Deep Inelastic Scattering (SIDIS) is a key tool for exploring the three-dimensional structure of the nucleon through Transverse Momentum Dependent parton distributions and fragmentation functions. While leading-power contributions to the SIDIS cross-section are well established, next-to-leading (NLP) of order $1/Q$ and next-to-next-to-leading power (NNLP) corrections of order $1/Q^2$ to the hadronic tensor have only recently begun to be systematically investigated. These corrections are essential for the reliable phenomenology and interpretation of modern high-precision data. In recent papers by one of the authors, NNLP corrections to Drell-Yan process were derived using rapidity factorization formalism. In the present work we extend this approach to SIDIS and obtain analytic expressions for the unpolarized structure functions. We derive NNLP corrections that include convolutions of unpolarized distributions, $f_1$, with unpolarized fragmentation functions, $D_1$, and Boer-Mulders functions, $h_1^\perp$, with Collins fragmentation functions, $H_1^\perp$. We compare our results with previous formulations, provide numerical studies, confront our predictions with HERMES and COMPASS measurements, and present predictions for future experiments at Jefferson Lab and the Electron-Ion Collider.

We present the complete one-loop matching of the baryon- and lepton-number-conserving low-energy effective field theory (LEFT) to the QCD gradient flow. Using Euclidean conventions and the background-field formulation of the gradient flow, we derive the short-flow-time expansion for the full LEFT operator basis up to mass dimension six. The matching is performed in dimensional regularization in the algebraically consistent 't Hooft-Veltman scheme, including a systematic treatment of evanescent operators and the finite counterterms required to restore chiral symmetry in the spurion sense. Keeping fully generic flavor structures, we verify the cancellation of spurious chiral-symmetry-violating terms with the known finite symmetry-restoring counterterms. This demonstrates that the gradient flow as a gauge-invariant ultraviolet regulator enables an efficient extraction of both divergent and finite counterterms in addition to the matching contributions. We provide the matching coefficients both before and after field redefinitions that remove redundant operators, as well as power-divergent mixings into lower-dimensional operators. Our results establish a consistent perturbative link between continuum LEFT calculations and gradient-flow-based lattice-QCD matrix elements, enabling precision low-energy phenomenology beyond leading-logarithmic accuracy.

Flavor instabilities in dense neutrino media trigger exponential growth of flavor waves, yet their nonlinear saturation remains poorly understood. We examine a simple proxy for this effect in the form of a single-wave solution of an axially symmetric fast flavor system. When the angular crossing is shallow and the growth rate of the instability correspondingly small, the flavor wave primarily affects resonant neutrinos that move in phase with it. The evolution of these resonant neutrinos becomes periodic, undergoing cycles of full flavor reversal. They feed power into the unstable wave, and subsequently return to their initial state, draining power back out. This new flavor pendulum captures the dynamics of weak, nearly monochromatic fast flavor instabilities. Since weakly unstable distributions always exhibit a narrow range of unstable wavenumbers, our model likely describes the earliest development of a flavor instability when it first appears. When the instability is not weak, the linear phase of a single-wave excitation does not connect to a regular nonlinear solution, unless the angle distribution consists of only two beams.

This paper provides a systematic and complete study of thermal gauge theory for generic equilibrium density matrices, which feature arbitrary values not only of temperature and chemical potentials, but also of average angular momentum. This work extends previous studies, which focused on pure scalar-fermion theories, to all gauge theories coupled to an arbitrary matter sector. Path-integral methods are developed to study ensemble averages and thermal Green's functions of general operators, with an arbitrary number of points, in all interacting gauge theories. These methods cover both the real-time and imaginary-time formalisms. Generalized Kubo-Martin-Schwinger (KMS) conditions are obtained both in coordinate and in momentum space for operators in general representations of the Lorentz and internal symmetry group. This allows us to obtain all thermal propagators including those of gauge fields and Faddeev-Popov ghosts. By analyzing all interactions in detail, it is shown that, in perturbation theory, only the propagators are affected by the average angular momentum and the chemical potentials, the vertices remain unmodified. The paper presents fully model-independent results and can, therefore, be applied to any specific thermal field theory.

Skyrmions are stable and topologically non-trivial field configurations that behave like localized particles. They appear in the chiral effective theory for pions, where they correspond to the baryon states, and might also exist in the electroweak theory, in the presence of certain effective interactions. In this paper, focusing on toy models that capture different limits of the electroweak sector of the Standard Model (SM), we show that skyrmions not classically stable at zero temperature can be stabilized by thermal effects. This result motivates the study of skyrmions in the quantum effective action of the SM, potentially implying the existence of dark matter without new physics.

We demonstrate the sensitivity of Drell-Yan production processes from deep inelastic scattering in searches for beyond-the-Standard Model (BSM) physics at fixed-target or beam-bump experiments. We take heavy neutral leptons (HNLs) as a case study, produced from the decay of a light vector boson mediator with mass in the range of $2-20$ GeV, which itself is generated via the Drell-Yan process. The produced HNLs subsequently decay into Standard Model final states. We consider several current and future experiments, including SBND, DarkQuest, DUNE Near Detector (ND), and SHiP. Utilizing $νπ^0$ and $νe^+e^-$ final states from HNL decays, we find that the Drell-Yan mechanism provides important contributions and significantly enhances the HNL search sensitivity, owing to the production of energetic final-state particles that are more readily detectable over the expected backgrounds. We find that at $90\%$ C.L. sensitivity, for gauge couplings $g_{X} \sim 10^{-2}\ (10^{-3})$ and kinematically accessible mass range, SBND and DarkQuest can probe the HNL flavor mixing $|U_{\ell}| \sim 3\times 10^{-4}\ (10^{-3})$, whereas DUNE ND and SHiP may extend the sensitivity down to the Type-I Seesaw prediction of $|U_{\ell}| \sim 10^{-5}$. Finally, for our chosen benchmark $|U_{\ell}| = 10^{-3}$ outside of the current experimental constraints, with a fixed mass ratio $m_{Z'}/m_N = 2.1$, and working within the $U(1)_{B-L}$, $U(1)_{B-3L_τ}$, and $U(1)_{B}$ parameter spaces, we find that both SBND and DarkQuest can probe $g_{X} \sim 10^{-3}$, DUNE ND can reach $g_{X} \sim 10^{-4}$, and SHiP can probe down to $g_{X}\sim 5\times 10^{-6}$. Our approach provides a powerful new technique to study HNL production at future fixed-target experiments and can readily be extended to other light BSM particle production within a broader class of dark sector models.

The Improved Quantum Null Energy Condition (INEC) was recently derived from the (restricted) quantum focusing conjecture (QFC), and is a statement about the energy-momentum tensor (EMT) of field theories in Minkowski space-time. It is a stronger condition than the quantum null energy condition (QNEC), and includes the possibility of expanding or contracting geodesics. Using the properties of relative entropy and modular Hamiltonian associated with null deformation of the sphere, we show the INEC holds under an additional assumption relating the EMT to the relative entropy. Furthermore, using the QNEC and INEC as a basis, we briefly speculate about a possible modified Quantum Focusing Conjecture.

Determining absolute neutrino masses remains a central challenge in particle physics. Relations among the neutrino mass-squared differences could facilitate this determination and shed additional light on the underlying mass-generation mechanism. Inspired by recent global fits of neutrino oscillation parameters, we propose a simple algebraic relation between $Δm^2_{21}$ and $Δm^2_{31}$. It provides a framework to analytically manipulate these parameters and admits specific physical interpretations. In particular, the results may point toward a vanishing $ν_1$ mass.

The (holomorphic) partition function of the Euclidean BTZ black hole with boundary modulus $τ$, is the $S$-image of the Virasoro vacuum character, $χ_{\rm vac}(-1/τ)$. This object decomposes into primaries via the modular $S$-kernel: $χ_{\rm vac}\left(-\frac{1}τ\right)=\int_{0}^{\infty} dP S_{0P}(P,c)χ_P(τ)$. In this paper, we provide a bulk understanding of this spectral resolution using the Chern-Simons formulation of AdS$_3$ gravity with $two$ boundaries: an asymptotic torus and an excised Wilson line at the origin ("stretched horizon"). At infinity, we impose standard AdS$_3$ Drinfel'd-Sokolov (DS) gauge to obtain the Alekseev-Shatashvili (AS) boundary action for a coadjoint orbit. At the inner boundary, removing the Wilson line prepares the state at the cut as a sum over orbits of the $spatial$ cycle. Re-inserting a spatial holonomy Wilson line acts as a delta-function projector onto the corresponding primary, which together with boundary gravitons, reproduces the Virasoro character (e.g., of a conical defect). But we can also consider projectors onto the $conjugate$ basis $\tilde P$, of the dual cycle. A key observation is that this leads to $S$-kernels instead of delta functions, with the BTZ character arising when the dual cycle label is in the exceptional orbit. Our two-boundary construction provides a bulk understanding of BTZ entropy: holonomy zero modes at the horizon have an effective central charge $c_{\rm prim}=c-1$ from the kernel measure (primaries), while the universal Dedekind-$η$ in $χ_P(τ)$ contributes $c_{\rm desc}=1$ from boundary gravitons (descendants). Together, they reproduce the full Cardy entropy. While our methods are specific to AdS$_3$/CFT$_2$, they are an explicit illustration that smoothness of the (Euclidean) horizon may emerge from a $sum$ over bulk states which are manifestly unsmooth.

From February to May of 2024 the Any Light Particle Search II (ALPS II) conducted its first science campaign using the `light-shining-through-a-wall' technique to search for pseudo-Goldstone bosons that lie beyond the Standard Model of particle physics and which are inaccessible by accelerator-based experiments. The experimental setup consists of two strings of superconducting dipole magnets, each more than 100 m long, that are separated by a wall. Laser light is directed through the first magnet string and a heterodyne detection system is used to measure the electromagnetic power that traverses a wall via the conversion to and then from a bosonic field. After the wall, a high-finesse optical cavity resonantly enhances the signal power. Two searches were carried out, one with the laser polarized perpendicular to the magnetic field direction and another with its polarization state aligned parallel to the magnetic field. No evidence for the existence of new bosons was found. In its first science campaign, ALPS II reached photon-boson conversion probability sensitivities of a few $10^{-13}$. The ongoing upgrade of the optical system aims to increase this sensitivity by about four orders of magnitude.

The strong interaction - governed by Quantum Chromodynamics (QCD) - shapes the structure of the visible universe. At about 10 $μ$s after the big bang, the primordial matter made up of quarks and gluons plus leptons, photons and neutrinos, the quark-gluon plasma (QGP), became cool enough to create, in a phase transition, the protons and neutrons of ordinary matter, along with other strongly interacting unstable hadrons. This phase transition was predicted within the framework of QCD and has been studied in accelerator laboratories world-wide since about 40 years. This review will explore recent breakthroughs in the study of the QCD phase diagram. We will highlight measurements of particle production and fluctuations, and compare them to theoretical predictions. We summarize our current understanding of the QCD structure and outline future experimental opportunities with high energy nuclear collisions at fixed-target and collider facilities world-wide.

Motivated by a conjecture that doubled four-dimensional Chern-Simons produces new integrable models, we perform its Hamiltonian analysis and find the theory's Poisson algebra. This requires carefully accounting for a set of boundary conditions that identify two gauge fields. Two methods for doing so are given, one of which is based on edge-modes and the other on a recharacterisation of the boundary conditions as constraints. We find that the Poisson algebra is that of an affine Gaudin model subject to a constraint, generalising the Goddard-Kent-Olive construction (from conformal field theory) to the world of integrable models. We also conjecture the existence of extended quantum groups and a generalisation of the affine Harish-Chandra Isomorphism.

The AdS Virasoro-Shapiro amplitude has recently been generalized to the AdS$_3$/CFT$_2$ correspondence between type IIB string theory on $ AdS_3 \times S^3 \times K3$ (or $T^4$), supported by Ramond-Ramond flux, and the D1-D5 CFT. In this paper, we use the $ AdS\times S$ Virasoro-Shapiro machinery to extract strong-coupling CFT data of the D1-D5 CFT by extending and completing earlier analyses in several directions. First, starting from the superconformal/Mellin block expansion of four-point functions of half-BPS tensor operators with arbitrary external KK modes, we employ the full $ AdS\times S$ Mellin formalism to bootstrap the $ AdS_3 \times S^3$ Virasoro-Shapiro amplitude for general KK configurations. This establishes its consistency with superconformal symmetry and yields a wealth of additional CFT data naturally organized in internal Mellin space. Second, we push the computation to the next order in the strong-coupling expansion and extract additional higher-order CFT data. Third, we translate the resulting Mellin-space data into the internal spin basis. We derive the transformation kernel relating internal Mellin variables and $SU(2)_L \times SU(2)_R$ R-symmetry spins. As applications, we obtain explicit formulae for the scaling dimensions of long multiplets on the first two leading Regge trajectories of arbitrary internal spins, and certain three-point functions with half-BPS tensor operators. These results provide a valuable set of analytic D1-D5 CFT data, enabling future applications and direct comparison with complementary approaches such as integrability.

This paper aims to investigate the classical duality between the $SIM(2)$-Maxwell-Kalb-Ramond (MKR) theory and a self-dual non-gauge-invariant model. First, we establish the equivalence in the free-field case using two complementary methods: a direct comparison of the equations of motion and the master Lagrangian approach. In both methodologies, we verify that the classical correspondence between the MKR model and self-dual fields exhibits modifications induced by very special relativity (VSR). Moreover, we employ the master Lagrangian approach to examine the duality when the self-dual model is minimally coupled to fermionic matter. We show that the resulting MKR model contains Thirring-like interactions modified by nonlocal contributions arising from VSR.

The precision of the neutrino oscillation parameters measurements has improved and will continue to improve as the next-generation experiments become online. Beyond the more precise measurements of the mixing angles and phases used to parametrize the lepton mixing matrix, tests of its unitarity are of great interest. This paper studies how the amplitudes of the oscillation patterns can be used and combined to construct leptonic unitarity triangles.

Quantum entanglement in high-energy collisions is often obscured by finite lifetimes, dynamical evolution, and final-state interactions, complicating the identification of genuinely quantum correlations. Ultra-peripheral heavy-ion collisions provide a clean benchmark via the Drell-S${\rm\ddot{o}ding}$ production of nonresonant pion pair, realizing an effectively zero-lifetime particle pair whose quantum correlations are fixed at production and remain robust against subsequent elastic scattering. The coherent superposition of photoproduction amplitudes from two indistinguishable nuclei encodes the linear polarization of quasi-real photons in the orbital motion of the pair, generating a nonfactorizable two-particle quantum state. This entanglement leaves a direct experimental imprint: a characteristic second-harmonic azimuthal modulation in momentum space arising from spin-dependent interference between the two sources. In this paper, we establish a quantitative framework for Drell-S${\rm\ddot{o}ding}$ pion-pair production in relativistic heavy-ion collisions and predict the magnitude and transverse-momentum dependence of the entanglement-induced azimuthal asymmetry. Our results provide experimentally accessible signatures of interference-induced entanglement and a controlled test of quantum coherence in relativistic environments.

We apply the intertwiner operator method of arXiv:2411.08865 to topological field theories, including BF theories, Chern-Simons theory, and three-dimensional gravity. We construct the operator on foliated manifolds while preserving covariance on the Cauchy surface, and compare canonical and holomorphic quantization, providing the intertwiner in both frameworks. For three-dimensional gravity, we present both covariant and time-gauge formulations, analyze the constraints, and construct the corresponding intertwiner. As an application, we derive the path ordering of Wilson loops in Chern-Simons theory. The study of observables is left for future work.

It is well known that a general two-point function cannot be uniquely determined in a theory with Poincaré symmetry. In this paper, we show that bulk-to-boundary correlators are highly constrained after imposing suitable fall-off conditions near future/past null infinity. More precisely, scalar bulk-to-boundary correlators are fixed to a unique form up to a normalization constant, whereas fermionic bulk-to-boundary correlators are fixed to a linear superposition of scalar and fermionic branches. This is established by asymptotically expanding the Ward identities, where upon the leading terms decouple from the subleading ones. In the fermionic branch, the power-law exponent of the bulk-to-boundary correlator is greater by one than the fall-off index. Consequently, we revisit the relation between Carrollian correlators and momentum space scattering amplitudes for fermionic operators. In this context, we find that the Fourier transform bridging the two acquires an extra factor of $\sqrtω$ for each fermionic operator. Furthermore, we reduce the bulk-to-boundary correlator to the boundary-to-boundary correlator and identify a critical fall-off index $Δ=1$. For $0 \le Δ< 1$, only a magnetic branch exists for scalars. For $Δ> 1$, the electric branch is always divergent for both scalar and fermionic branches and thus requires regularization.

Nuclear modifications of the nucleon's structure function $F_2^N$ have been investigated mainly since the discovery of the EMC nuclear effect in 1983, and there were many experimental measurements from the deuteron to a heavy nucleus. Now, the details of the modifications of $F_2^N$ are known from small $x$ to large $x$. On the other hand, it is taken as granted that a nuclear modification does not exist for the longitudinal-transverse structure function ratio $R_N=F_L^N/(2xF_1^N)$. However, such a nuclear modification does exist theoretically. A nucleon in a nucleus moves in any space direction, which is not necessary the longitudinal direction along the virtual-photon momentum in charged-lepton scattering. Because of this transverse Fermi motion, the longitudinal and transverse structure functions mix with the mixture probability proportional to the nucleon's transverse-momentum squared $\vec p_T^{\,\, 2}/Q^2$. In this paper, the nuclear modifications are shown numerically for the deuteron by using a standard convolution description. The magnitude of the modifications is of the order of a few percent in the deuteron; however, they should be large in large nuclei. In handling high-energy nuclear data, such nuclear modifications need to be taken into account for a precise determination of physical quantities. Now, the longitudinal-transverse structure-function ratio and tensor-polarized experiments are under preparation for the deuteron at JLab. We hope that such effects will be confirmed experimentally for not only for the deuteron but also for larger nuclei.

There has been considerable progress in the study of the electromagnetic form factors of baryons in the timelike region, through electron-positron scattering reactions ($e^+ e^- \to B \bar B$), in the last two decades. Timelike experiments reveal information about the distribution of charge and magnetism inside the hyperons that cannot be obtained in spacelike experiments (electron scattering on baryons). Motivated by the novel data, we extend to the timelike region, without any further parameter fitting, a covariant quark model developed for the spacelike region that takes into account the meson cloud excitations of the baryon cores. We use the formalism to calculate the electric ($G_E$) and magnetic ($G_M$) form factors of spin 1/2 baryons in the large square transfer momentum $q^2$ region. Our calculations are compared with the available data from CLEO and BESIII above $q^2=10$ GeV$^2$. We conclude that our predictions for the effective form factors (combination between $G_E$ and $G_M$) are in good agreement with the $q^2 > 15$ GeV$^2$ data for $Λ$, $Σ^+$, $Σ^0$, $Ξ^-$ and $Ξ^0$. Upcoming data for $Σ^-$ can be used to further test our predictions. We also compare our model calculations with the available data for ratio $|G_E/G_M|$. We conclude that the present $q^2$ data range is not large enough to test our calculations, but that a more definitive test can be performed by upcoming data above $q^2=20$ GeV$^2$.

We study fixed-column predictions of the lepton mixing matrix that arise from residual symmetries originating in a class of discrete flavour and modular symmetries. While the recent high-precision determination of $\sin^{2}θ_{12}$ by JUNO already constrains part of these predictions, the remaining ones are primarily characterized by non-trivial correlations between $\sin^{2}θ_{23}$ and the Dirac CP phase $δ_{\rm CP}$, which are currently only weakly constrained. This motivates a detailed investigation using next-generation long-baseline neutrino experiments. For the viable scenarios, we derive precise $\sin^{2}θ_{23}$-$δ_{\rm CP}$ correlations and use them to generate test-event samples, marginalising over the remaining oscillation parameters. We perform detailed simulations for DUNE and T2HK, presenting allowed regions in the $\sin^{2}θ_{23}$-$δ_{\rm CP}$ plane and evaluating the CP-violation fraction as a function of exposure. Our results show that the combined sensitivity of DUNE and T2HK provides a robust test of fixed-column lepton-mixing predictions.

Spin-1 hadrons contain different aspects of spin physics from the ones of the spin-1/2 nucleon because of the existence of tensor-polarized structure functions. In the charged-lepton deep inelastic scattering from a spin-1 hadron or nucleus, such as the deuteron, there are leading-twist structure functions $b_1$ and $b_2$. In addition, there exists a gluon transversity which does not exist in the spin-1/2 nucleon. In the deuteron, these observables could probe interesting dynamical aspects beyond a simple bound system of a proton and a neutron. In addition, there are recent theoretical studies on higher-twist distributions. Tensor-polarized deuteron experiments are now under preparation at the Thomas Jefferson National Accelerator Facility, so that the topic of polarized deuteron is expected to become one of exciting fields in hadron physics. This paper is a brief overview on the tensor-polarized parton distribution functions, including transverse-momentum-dependent parton distributions and fragmentation functions up to twist 4.

Dark Matter relic density generation through \textit{freeze-in} mechanism where dark matter particles interact feebly with visible sector particles, is an alternative approach to well-studied and most popular \textit{freeze-out} paradigm. We study this \textit{freeze-in} scenario in the presence of a scalar leptoquark interacting with both dark matter and Standard Model particles with renormalizable interactions. We discuss the effect of the presence of such heavy particle, a scalar leptoquark with mass $\geq 1.5 \TeV$, in the thermal bath and subsequent relic density generation. We explore the parameter space of such framework, consisting of two masses and three dimensionless couplings. We numerically study the interaction rates and relic density as a function of these parameters and determine their values consistent with the dark matter constraints.

We establish time-like entanglement entropy (TEE) as a novel tool to characterize the black hole interior from a single-boundary perspective. In the Schwarzschild-AdS black hole, we show that TEE of time-like boundary strips exhibits linear growth as a function of temporal width in the limit of large temporal width, and that its imaginary part carries physical significance rather than being a constant. By analyzing charged, scalar-hairy black holes, we present evidence that TEE detects a hidden "causal phase transition" separating Type-I and Type-II interiors -- distinguished by singularity structure. We identify a critical temporal width $τ_c$ that acts as the order parameter for this transition: for strips narrower than $τ_c$, the system enters a distinct "time-like entanglement phase" dominated purely by time-like contributions, up to a regulator effect; conversely, for strips wider than $τ_c$, space-like entanglement re-emerges. Notably, the existence of a Cauchy horizon drives the $τ_c$ to infinity, leading to pure time-like entanglement. These results suggest that the TEE may supply a novel boundary quantum-information measure to detect structure hidden inside the black hole and suggests a deep connection between TEE and cosmic censorship.

A production of a pair of tau leptons in PbPb collisions at the FCC-hh collider is examined. The 95\% C.L. exclusion limits, as well as 3$σ$ and 5$σ$ sensitivity limits on the anomaly magnetic moment of the tau lepton $a_τ$ and its electric dipole moment $d_τ$ are obtained. A comparison with bounds on $a_τ$ and $d_τ$ for other future colliders are given.

We thoroughly investigate nonanalytic terms in the finite-temperature effective potential in one-loop approximation on a $D$-dimensional spacetime, $S_τ\times R^{D-(p+1)}\times \prod_{i=1}^p S_i^1$, using a mode recombination formula. Such nonanalytic terms cannot be expressed as positive powers of field-dependent mass squared. The formula provides a clear separation of the effective potential into a part that contains the nonanalytic terms and a part that is purely analytic, and clarifies the origin of the nonanalytic terms. We obtain all the nonanalytic terms and show that only two types of nonanalytic terms arise: power-type and logarithmic-one. For a real scalar field with periodic boundary conditions, the manner of the emergence of these terms is highly characteristic; the two types never appear simultaneously. By contrast, for fermions with general boundary conditions, we find that neither of the two types appears. These results clarify the nonanalytic structure of the finite-temperature effective potential on the spacetime with compactified spatial dimensions.

The $μ$-$e$ conversion process is one of the most powerful ways to test lepton-flavor-violating (LFV) interactions involving charged leptons. The standard model with massive neutrinos predicts an extremely low rate for $μ$-$e$ conversion, making this process an excellent probe for testing LFV arising from new physics. Among many theoretical models that can induce LFV, the Supersymmetric model with R-parity violating interactions is one of the most studied for $μ$-$e$ conversion. In this work, we revisit trilinear R-parity violating interactions for $μ$-$e$ conversion, considering renormalization group (RG) running effects from high to low energy scales. The $μ$-$e$ conversion, $μ\to e γ$, and $μ\to eee$ experimental data are compared to give upper limits on the relevant 15 combinations of the trilinear $λ^{\prime}$ couplings and 6 combinations of the $λ$ couplings, certain of which are underexplored in previous studies. We find that RG running effects influence the limits by no more than 30\% in most cases, but can improve constraints by $\sim$80\% in certain combinations, which cannot be neglected. In the near future, COMET and Mu2e are expected to begin data-taking and aim to provide the most stringent constraints on $μ$-$e$ conversion. These next-generation $μ$-$e$ experiments have the ability to give much more comprehensive examinations on most trilinear coupling combinations than the $μ\to eγ$ and $μ\to 3e$ decay experiments. The $μ$-$e$ experiments will not only deepen our understanding of LFV but also provide a crucial way to examine the underlying new physics contributions.

Searching for dark matter is a primary goal of modern astronomy and particle physics. Invisible decays of heavy quarkonia are particularly promising for probing light dark matter, attracting broad interest due to their unique sensitivity. Experiments searching for radiative invisible decays of the $J/ψ$ have steadily improved upper limits, and upcoming facilities will push sensitivity further -- making the precise determination and subtraction of the neutrino background indispensable. Here, we present the first lattice QCD calculation of the Standard Model decay $J/ψ\to γν\barν$, an irreducible background to $J/ψ\to γ+ \textrm{invisible}$. Our result for the branching fraction is $\operatorname{Br}(J/ψ\to γν\barν)=1.00(9)(7)\times 10^{-10}$, where the first uncertainty is statistical and the second is our systematic estimate. This work advances lattice-based determinations of neutrino backgrounds to quarkonium invisible decays, delivering an ab initio benchmark for $J/ψ\to γ+ \textrm{invisible}$. Our approach generalizes to other quarkonium channels (e.g., $Υ/φ\to γ+\textrm{invisible}$) and provides critical theoretical support for dark matter searches at colliders.

The Regge-Gribov model describing interacting pomerons and odderons is proposed with triple reggeon vertices taking into account the negative signature of the odderon. Its simplified version with zero transverse dimensions is first considered. No phase transition occurs in this case at the intercept crossing unity. This simplified model is studied without more approximations by numerical techniques. The physically relevant model in the two-dimensional transverse space is then studied by the renormalization group method in the single loop approximation. The pomeron and odderon are taken to have different bare intercepts and slopes. The behaviour when the intercepts move from below to their critical values compatible with the Froissart limitation is studied. Five real fixed points are found with singularities in the form of non-trivial branch points indicating a phase transition as the intercepts cross unity. The new phases, however, are not physical, since they violate the projectile-target symmetry. In the vicinity of fixed points the asymptotical behaviour of Green functions and elastic scattering amplitude is found under Glauber approximation for couplings to participants.

The results from the ATLAS and CMS experiment at the Large Hadron Collider indicate the existence of a top-quark pair bound state near the $\ttbar$ threshold region. We present a method relying on Recursive Jigsaw Reconstruction to reconstruct the toponium bound state at the $\ttbar$ threshold region. We propose incorporating two variables in the analysis that can improve sensitivity to the toponium signal. Our results indicate that this method may be useful to gain additional insights into the physics phenomenology of the $\ttbar$ threshold region.

The recent evidence for the decay $B^+ \to K^+ ν\barν$ reported by the Belle II collaboration, combined with the existing constraints on the neutral mode $B^0 \to K^{*0} ν\barν$, implies a deviation from the Standard Model prediction that necessitates New Physics contributions to both left- and right-handed vector currents. We perform a systematic topological classification of renormalizable one-loop completions capable of generating the required dimension-six operators while forbidding tree-level mediation. Based on this classification, we identify and construct two minimal benchmark scenarios -- a scalar-rich model and a fermion-rich model -- and perform a comprehensive phenomenological analysis. Our study demonstrates that while these one-loop models can yield enhancements in the $b \to s ν\barν$ branching fractions, the attainable magnitudes are significantly restricted by the combined effects of loop suppression and complementary flavor constraints, limiting their ability to fully accommodate the current anomaly.

We present a data-driven inverse construction of the dilaton field in a bottom-up AdS/QCD description of heavy vector quarkonia. Instead of adopting an \emph{ad hoc} analytic ansatz, we use a multilayer perceptron to learn \(Φ'(z)\) as a smooth function of the holographic coordinate, with \(Φ(0)=0\) imposed to ensure ultraviolet consistency. The dilaton and its derivatives obtained by automatic differentiation generate the holographic potential \(U(z)\), and the associated Schrödinger-like equation is discretized and diagonalized to extract the low-lying eigenmodes. Masses and decay constants are then evaluated from the eigenvalues and the near-boundary behavior of the bulk-to-boundary modes. Training on PDG data for charmonium and bottomonium yields a non-quadratic dilaton profile that resolves the longstanding difficulty of simultaneously reproducing both the heavy-quarkonium spectrum and the monotonic suppression of leptonic decay constants with radial excitation. The combined fit achieves RMS deviations of \(1.26\%\) (charmonium) and \(3.32\%\) (bottomonium). This work establishes neural-network reconstruction as a flexible tool for holographic modeling and provides a basis for future extensions incorporating additional channels, lattice constraints, or finite-temperature backgrounds.

We investigate the low-energy interactions between the charmonium state $J/ψ$ and the light pseudoscalar mesons ($π$ and $K$) within the framework of dispersion relations.We demonstrate that the symmetry-breaking terms in the chiral Lagrangian induce mixing between the bare charmonium fields, necessitating a diagonalization procedure to correctly identify the physical $J/ψ$ and $ψ'$ states. Following this diagonalization, we relate the threshold scattering amplitudes of $J/ψ{\cal P}~({\cal P}=π,K)$, which determine the scattering lengths, to the $J/ψJ/ψ\to {\cal P}\bar{\cal P}$ amplitudes via crossing symmetry. The ${\cal P}\bar{\cal P}$ rescattering effects are incorporated using dispersion relations. We obtain the $S$-wave scattering lengths $a_{J/ψπ} \lesssim -0.0021$~fm and $a_{J/ψK} \lesssim -0.028$~fm, where the negative sign indicates an attractive interaction. Our results show that the $J/ψK$ interaction is moderately enhanced relative to the pion channel, driven by explicit chiral symmetry breaking. Furthermore, a quantitative comparison of the coupled-channel mechanism, where $J/ψπ$ couples to $D\bar{D}^*$ and $J/ψK$ couples to $D^*\bar{D}_s/D\bar{D}_s^*$, reveals that both $J/ψπ$ and $J/ψK$ scatterings are predominantly governed by the soft-gluon exchange mechanism.

We employed the chiral unitarity approach to investigate the decay process $Ξ_c^+ \to Σ^+K^+K^-$. by considering that the low lying nucleon resonance $Ξ(1/2^-)$ and the low lying scalar meson $f_0(980)$ and $a_0(980)$ that could be dynamically generated through $S$-wave pseudoscalar meson-octet baryon and the $S$-wave pseudoscalar meson-pseudoscalar meson interactions, respectively. In the invariant mass distributions of $Σ^+K^-$ and $K^+K^-$, we observe a distinct peak structure associated with the resonant state $Ξ(1/2^-)$ and a bit enhancement near the $K^+K^-$ threshold that is corresponding to the mesons $f_0(980)$ and $a_0(980)$, respectively. Consequently, we recommend more precise experimental measurements of this process in the future.

In the present paper, we construct a new family of exact charged and rotating asymptotically Lifshitz black hole solutions in five dimensions. The spacetime solves Einstein equations coupled to a dilaton, two Abelian gauge fields, and axionic scalars supplemented by two generalized Chern-Simons terms. This configuration is characterized by a range of the free dynamical exponent $z$ and possesses nontrivial thermodynamical parameters, where we verify the first law of black hole thermodynamics and derive the corresponding Smarr relation. Motivated by applications to gauge/gravity duality, we then investigate a holographic superconductor in the rotating Lifshitz background. We study the condensation of the scalar operator and the AC conductivity of the dual system. These results show that increasing the rotation parameter suppresses the condensate and weakens the superconducting phase, while increasing the dynamical critical exponent enhances the superconducting order. To the best of our knowledge, the solutions presented here are the first to demonstrate five-dimensional rotating Lifshitz black holes supported by both electric and axionic charges. This opens up a new avenue to investigate non-relativistic holography beyond static backgrounds.

We study BPS $q$-series associated to 3-manifolds decorated by a line defect along an embedded link. We prove that these $q$-series depend only on the class of the link in the skein module, thereby defining a homomorphism from the skein module to the space of $q$-series. The image of this homomorphism is conjectured to be holomorphically quantum modular, which suggests a new approach to Langlands duality for skein modules through $q$-series.

We develop a systematic approach to the computation of gauge invariant quartic interactions between reducible massive and massless higher spin fields. Extending the BRST formulation of existing cubic results, we obtain a single constraint for each off-shell quartic vertex that ensures both the gauge invariance of the Lagrangian and associativity of the gauge transformations at quartic order. A solution to these equations is presented. The general equation is then reduced to an on-shell version to reduce complexity. We find example solutions for the off-shell and on-shell quartic vertices in low spin examples relevant to the problem of black hole scattering.

Weak-gravitational lensing tends to wash out scale and time-dependent features of the clustering of matter, such as the Baryonic Acoustic Oscillations (BAO) which appear in the form of wiggles in the matter power spectrum but that disappear in the analogous lensing $C_\ell$. This is a direct consequence of lensing being a projected effect. In this paper, we demonstrate how the noise complexity -- often deemed "erasing the signal" -- induced by a particular de-projection technique, the Bernardeau-Nishimichi-Taruya (BNT) transform arXiv:1312.0430, can be used to extract the BAO signal and non-gaussian aperture-mass-like properties at chosen physical scales. We take into account parts of the data vectors that should effectively be without cosmological signature and also introduce an additional re-weighting designed to specifically highlight clustering features -- both at the probe (summary statistics) or map (amplitude of the field) level. We thus demonstrate why weak-gravitational lensing by the large-scale structure of the Universe, though only in a tomographic setting, does not erase scale and time-dependent features of the dynamics of matter, while providing a tool to effectively extract them from actual galaxy-shapes measurements.

We present a catalog of 971 FR-I radio galaxies (FR-Is) identified from the Very Large Array Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) survey. The identifications were made using a hybrid method that combines deep learning with ridgeline flux density distribution analysis. Among these sources, 845 are new discoveries. The catalog comprises sources characterized by edge-darkened double jets, an absence of significant bent morphology, and angular sizes ranging from 23 to 159 arcseconds. Optical and/or infrared counterparts have been identified for 813 FR-Is. Among these, the host galaxies are predominantly (88.1\%) red galaxies, with the remainder (11.9\%) being blue galaxies; notably, most blue galaxies exhibit high radio power. The FR-I sample spans a radio power range of $1.20 \times 10^{21} \leq P_{\rm 1400} \leq 3.55 \times 10^{27} \, {\rm W\,Hz}^{-1}$ at 1400 MHz and reaches redshifts up to $z = 2.307$. The host galaxies have $r$-band absolute magnitudes in the range $-24 \lesssim M_r \lesssim -20$ mag. For the 512 FR-Is with estimates, the black hole masses fall within $10^7 \lesssim M_{\rm BH} \lesssim 7.94 \times 10^9 \, M_{\odot}$. Based on optical emission-line ratios and mid-infrared colors, spectroscopic classification shows that 571 hosts are low-excitation radio galaxies (LERGs) and 59 are high-excitation radio galaxies (HERGs).

The JAGB method has been proposed as a distance indicator for Local Group galaxies. We investigate the populations of the J region in the (J-K,J) colour-magnitude plane of the Large and Small Magellanic Clouds (LMC and SMC), aiming to distinguish general features of the J luminosity function (JLF) from those sensitive to the stellar population of each galaxy. Using a population synthesis approach based on stellar evolution and dust formation modelling, we predict the distribution of stars within the J region and compare it with observations. Stars in this region are identified as recently formed C-stars that have not yet accumulated large amounts of carbon. Typically, 2--3 $M_\odot$ stars remain longer in the J region, while lower-mass stars evolve faster. The JLF of the LMC, peaked at the expected magnitudes for these stars, confirms this picture. In the SMC, the J flux distribution is shifted to higher magnitudes, indicating an older population with lower-mass progenitors.

We present a pilot study on the host galaxy environments of CHIME/FRBs by cross-matching baseband-localised events with the LOFAR Two-metre Sky Survey Data Release 2 (LoTSS DR2) at 144 MHz. Unlike traditional methods reliant on optical imaging, our radio-based selection allows for the identification of dust-obscured or optically faint star-forming galaxies. Of the 140 CHIME FRBs considered, 33 lie within the LoTSS DR2 footprint, and 16 show potential radio counterparts. Through multi-wavelength analysis, spectral energy distribution (SED) fitting, and redshift constraints from the Macquart relation, we identify two secure and one tentative host candidates, all consistent with active star formation. However, their H$α$-derived star formation rates appear underestimated, likely due to significant dust attenuation, as suggested by infrared colours and compact optical morphologies. Our results highlight the value of low-frequency radio data in complementing optical host searches and demonstrate the feasibility of host identification even in the absence of optical confirmation. With forthcoming data from LoTSS DR3 and the full CHIME/FRB baseband release, this method offers a promising path toward statistically robust studies of FRB host galaxies and their environments.

Filamentary structures are widely observed in molecular clouds, yet most filament observations are biased toward case studies and small samples; a uniform census within a single giant molecular cloud is still missing. We do a complete census of filaments in Cygnus X and quantify their links to dense cores, the magnetic field (B field), and HII regions. Using the updated getsf algorithm on the Cygnus X column-density map, we extracted 2633 filaments and 6551 cores. We built CMFs for cores on and off filaments, compared filament orientations with the Planck B field, measured radial column-density profiles near HII-region boundaries, and computed distances between young stellar objects and filament spines. Filaments have a typical width of 0.5 pc in Cygnus X at a resolution of 0.12 pc and host > 93% of high-mass cores (>= 20 Msun). The on-filament CMF shows a high-mass (> 10 Msun) slope of -2.30, while the off-filament CMF is steeper (-2.83). The onCMF peak mass is well below the Bonnor-Ebert mass, whereas the outCMF peak is comparable to it. At 5' resolution, filaments are mostly perpendicular to the B field except at the lowest column densities; the transition occurs near Av = 10 mag. Prominent filaments and high-mass cores preferred to be located around HII-region boundaries or at intersections of multiple HII regions; filament profiles are steeper on the side facing the HII region. Massive-core formation depends strongly on filaments, which may provide reservoirs that feed cores via accretion. The B field likely regulates filament formation, consistent with the type-O mode (converging flows along an oblique MHD shock) and an HII-driven bubble-filament paradigm for Cygnus X.

We present the first determination of the Galactic stellar mass function (MF) for low-mass stars (0.2-0.5 M_sun) at metallicities [Fe/H] < -1. A sample of ~53,000 stars was selected as metal-poor on the basis of both their halo-like orbits and their spectroscopic [Fe/H] from Gaia DR3 BP/RP (XP) spectra. These metallicity estimates for low-mass stars were enabled by calibrating Gaia XP spectra with stellar parameters from SDSS-V. For -1.5 < [Fe/H] < -1, we find that the MF below 0.5 M_sun exhibits a "bottom-heavy" power-law slope of alpha ~ -1.6. We tentatively find that at even lower metallicities, the MF becomes very bottom-light, with a near-flat power-law slope of alpha ~ 0 that implies a severe deficit of low-mass stars. This metallicity-dependent variation is insensitive to the adopted stellar evolution model. These results show that the Galactic low-mass MF is not universal, with variations in the metal-poor regime. A further calibration of XP metallicities in the regime of M < 0.5 M_sun and [Fe/H] < -1.5 will be essential to verify these tentative low-metallicity trends.

The origin of accretion disc winds remains disputed to date. High inclination, dipping, neutron star Low Mass X-Ray Binaries (LMXBs) provide an excellent testbed to study the launching mechanism of such winds due to being persistently accreting and showing a nearly ubiquitous presence of highly-ionised plasmas. We aim to establish or rule out the presence of a wind in the high inclination LMXB 4U 1624-49, for which a highly ionised plasma has been repeatedly observed in X-ray spectra by Chandra and XMM-Newton, and a thermal-radiative pressure wind is expected. We leverage the exquisite spectral resolution of XRISM to perform phase-resolved spectroscopy of the full binary orbit to characterise the highly ionised plasma at all phases except during absorption dips. An outflow is clearly detected via phase-resolved spectroscopy of the source with XRISM/Resolve. Based on analysis of the radial velocity curve we determine an average velocity of ~200-320 km/s and a column density above 10$^{23}$ cm$^{-2}$. The line profiles are generally narrow, spanning from ~50 to ~100 km/s, depending on the orbital phase, pointing to a low velocity sheer or turbulence of the highly ionised outflow and a potential increase of turbulence as the absorption dip is approached, likely due to turbulent mixing. The line profiles, together with the derived launching radius and wind velocity are consistent with a wind being launched from the outskirts of the disc and without stratification, pointing to a thermal-radiative pressure origin.

The redshifted 21 cm signal from the cosmic dawn and Epoch of Reionization (EoR) encodes important information about both astrophysical processes and primordial physics, such as inflation. In this work, we use morphological statistics to explore the sensitivity of the 21 cm signal to inflationary features and EoR dynamics simultaneously. Focusing on primordial features from particle production during inflation we generate semi-numerical simulations of the 21 cm signal across redshifts 5 < z < 35, incorporating these features. Using Minkowski Functionals (MFs), we analyze the morphology of 21 cm fields: density, neutral hydrogen fraction, spin temperature, and brightness temperature. We demonstrate that MFs are highly sensitive to both the amplitude and scale of primordial features, capturing rich morphological information. In particular, we show that MFs can robustly identify inflationary features and distinguish them from the standard model. We further explore various EoR scenarios, and demonstrate that combining MFs across redshifts can disentangle the signatures of primordial features from EoR effects. This approach opens new avenues for probing inflation with upcoming 21 cm surveys.

Current and upcoming large-scale structure surveys are pushing toward increasingly wide angular coverage, where wide-angle effects (arising from the varying line of sight across the curved sky) become critical for accurate modeling of the three-dimensional galaxy power spectrum. At the same time, these survey's broader redshift reach makes the effects of redshift evolution (beyond the effective-redshift approximation) non-negligible on large radial scales. Additional observational effects such as the survey window function and integral constraints also become significant on these large scales, necessitating a careful theoretical treatment to robustly constrain local primordial non-Gaussianities and relativistic effects. In this work, we present a consistent and accurate theoretical framework for modeling the commonly used power spectrum multipoles (PSM) on large scales using the discrete spherical Fourier-Bessel (dSFB) basis. This basis ensures numerical stability and allows an exact separation between angular and radial modes. Using the dSFB basis, we study the impact of wide-angle effects and redshift evolution on the PSM, and incorporate the effects of window function convolution and integral constraints. We validate our PSM modeling using lognormal mocks under radial integral constraints with realistic survey geometries, demonstrating the readiness of our framework for application to all-sky galaxy surveys.

The determination of distances using time-series photometry of Type Ia supernovae (SNe Ia) relies on a ~5% empirical correction related to the properties of their host galaxies, e.g., global stellar mass. It is therefore crucial for unbiased cosmology inference that host galaxy properties be self-consistently determined across the full range of redshifts probed, which we undertake in this study for approximately 2000 SNe in the Union3 compilation (now Union3.1). We use aperture-matched, homogeneously-reduced, optical-infrared photometry from the DESI Legacy Imaging Surveys to derive global galaxy properties using the stellar population synthesis and SED-fitting code Prospector. We find that the host masses of $z<0.10$ SNe in Union3 were, on average, overestimated relative to the rest of the sample, while the opposite was true for $z<0.15$ SNe in Pantheon+. After correction, the two studies' average distance modulus estimated for low-redshift SNe, previously $>0.03$ mag discrepant, come into 0.01 mag agreement. We then update the UNITY SN analysis and find that the uncertainties on all standardization parameters shrink to 0.6-0.9x their previous sizes. For flat-$Λ$CDM, we find $Ω_m=0.344^{+0.026}_{-0.025}$, a -0.3$σ$ shift from Union3. We then combine with measurements of Baryon Acoustic Oscillations (BAO) and the Cosmic Microwave Background (CMB) exactly as done by DESI DR2 and find $w_0=-0.719\pm0.084$, $w_a=-0.95^{+0.29}_{-0.26}$, corresponding to 3.4$σ$ evidence against a cosmological constant (down from 3.8$σ$). We also update the DESI combined probe analysis using our correction to Pantheon+ and the recent DES-SN5YR Dovekie recalibration, finding $3.2σ$ (up from 2.8$σ$) and 3.4$σ$ (down from 4.2$σ$) evidence against a cosmological constant in the $w_0w_a$ plane, altogether marking a significantly improved consistency across SN analyses.

Quasars accretion models predict a broad range of optical and ultraviolet properties that depend primarily on black hole mass and accretion rate. Yet, most optically selected luminous quasars display strikingly similar continuum spectra. We show that this uniformity can be explained by a nearly constant luminosity to mass (Eddington) ratio, L_EDD and by thermal emission from a standard, optically thick, geometrically thin accretion disc. A standard disk with an Eddington ratio L_EDD=0.1 reproduces both the black hole mass/luminosity distribution of Sloan Digital Sky Survey (SDSS) quasars and their principal continuum properties. In this framework, the spectral energy distribution peaks beyond the observable ultraviolet range for nearly all sources. We show that the few quasars, expected to be cold enough to shift the peak into the observable region, indeed show this behaviour. This scenario is further supported by an analysis of the relation between the luminosity of the main broad emission lines and the continuum luminosity (i.e. the Baldwin effect). We find that 1) the observed slopes of the line to continuum relations match the expectations from the standard disk model, if we assume that the line emission is a good proxy of the ionizing luminosity; 2) the dispersions of the line-continuum luminosity relations are very small (as small as 0.13 dex), suggesting that the physics of the disk-broad line region system is dominated by only one parameter (the black hole mass) with a nearly constant Eddington ratio. Finally, we notice that our hypothesis of constant L_EDD=0.1 provides a black hole mass estimate (based on the observed luminosity) with a smaller error than the virial estimate.

In Hamiltonian Monte Carlo sampling, the shape of the potential and the choice of the momentum distribution jointly give rise to the Hamiltonian dynamics of the sampler. An efficient sampler propagates quickly in all regions of the parameter space, so that the chain has a low autocorrelation length and the sampler has a high acceptance rate, with the goal of optimising the number of near-independent samples for given computational cost. Standard Gaussian momentum distributions allow arbitrarily large velocities, which can lead to inefficient exploration in posteriors with ridges or funnel-like geometries. We investigate alternative momentum distributions based on relativistic and Student's t kinetic energies, which naturally limit particle velocities and may improve robustness. Using Almanac, a sampler for cosmological posterior distributions of sky maps and power spectra on the sphere, we test these alternatives in both low- and high-dimensional settings. We find that the choice of parameterization and momentum distribution can improve convergence and effective sample rate, though the achievable gains are generally modest and strongly problem-dependent, reaching up to an order of magnitude in favorable cases. Among the momentum distributions that we tested, those with moderately heavy tails achieved the best balance between efficiency and stability. These results highlight the importance of sampler design and encourage future work on adaptive and self-tuning strategies for kinetic energy parameter optimization in high-dimensional settings.

Motivated by recent experimental refinements of stellar reaction rates, we establish a non-perturbative Green's function formalism based on the exact solution of the Dyson equation for sub-barrier proton-nucleus resonant scattering. By utilizing bare Green's functions to map the quantum tunneling problem onto a scattering formalism, we demonstrate that the summation of infinite quantum paths recovers the exact tunneling coefficients, enabling an analytical solution of the Dyson equation where the strong nuclear force is modeled as a surface delta-shell impurity embedded within the Coulomb field. Applying this framework to the astrophysically relevant $p + {}^{7}\text{Li}$, $p + {}^{14}\text{N}$, and $p + {}^{23}\text{Na}$ systems, we achieve precise agreement with experimental resonance energies while revealing a fundamental physical distinction in resonance formation. The heavier ${}^{23}\text{Na}$ system is identified as a saturated state, residing on a geometric plateau where the resonance energy becomes insensitive to the interaction strength; our calculated value of $2.11$~MeV aligns remarkably well with the experimental level of $2.08$~MeV. In contrast, the lighter ${}^7\text{Li}$ and ${}^{14}\text{N}$ systems emerge as threshold states in a weak-coupling window, where the resonance energy is highly sensitive to the potential parameters and is sustained near the continuum edge. In this regime, our model yields energies of $0.489$~MeV and $1.067$~MeV, closely reproducing the experimental benchmarks of $0.441$~MeV and $1.058$~MeV, respectively. We demonstrate that these threshold states are characterized by a significant enhancement of the resonant cross-section, driven by the inverse relationship between the tunneling width and the spectral density peak.

Kinetic temperature is a fundamental parameter in molecular clouds. Symmetric top molecules, such as NH$_3$ and CH$_3$CCH, are often used as thermometers. However, at high temperatures, NH$_3$(2,2) can be collisionally excited to NH$_3$(2,1) and rapidly decay to NH$_3$(1,1), which can lead to an underestimation of the kinetic temperature when using rotation temperatures derived from NH$_3$(1,1) and NH$_3$(2,2). In contrast, CH$_3$CCH is a symmetric top molecule with lower critical densities of its rotational levels than those of NH$_3$, which can be thermalized close to the kinetic temperature at relatively low densities of about 10$^{4}$ cm$^{-3}$. To compare the rotation temperatures derived from NH$_3$(1,1)$\&$(2,2) and CH$_3$CCH rotational levels in warm molecular gas, we used observational data toward 55 massive star-forming regions obtained with Yebes 40m and TMRT 65m. Our results show that rotation temperatures derived from NH$_3$(1,1)$\&$(2,2) are systematically lower than those from CH$_3$CCH 5-4. This suggests that CH$_3$CCH rotational lines with the same $J$+1$\rightarrow$$J$ quantum number may be a more reliable thermometer than NH$_3$(1,1)$\&$(2,2) in warm molecular gas located in the surroundings of massive young stellar objects or, more generally, in massive star-forming regions.

The study of astronomical transients at high energies provides insights into some of the most extreme physical events in the universe; however, carrying out their detection and fast follow-up studies are limited by instrumental constraints. Search for Transient Object in New observations using Known Sources (STONKS) is a near-real-time transient detection system for XMM-Newton offering the capability to detect transients in XMM-Newton observations at fainter fluxes than can be achieved with wide survey instruments. We present the transients detected with the STONKS pipeline found in an XMM-Newton multi-year heritage survey of the Galactic plane to identify and classify highly variable X-ray sources that have recently been reported in this region. We examined the alerts created by the STONKS pipeline from over 200 XMM observations of the Galactic plane, screening for instrumental effects. The 78 alerts associated with 70 real astrophysical sources were then subjected to further temporal and spectral analysis. From the 70 sources we identified, we were able to classify 32 with a high degree of confidence, including 7 X-ray binaries, 1 $γ$-Cas analogue, and 1 magnetar candidate. Of the 70 sources, 23 were detected for the first time in X-rays. This systematic analysis of publicly available data has shown the value and potential of STONKS in the application to XMM-Newton observations. It will enable the community to detect transient and highly variable sources at fainter fluxes than with any other X-ray transient detection systems.

Despite the assumption that an ideal FLRW observer is not dependent on the local environment, observations are biased by the positions of the observers due to the matter correlations in the large-scale structure (LSS) of the universe. The variation of the mass distribution of the LSS of the universe implies that observers residing in different locations may suffer bias in their measurements when they look at the images of distant galaxies. Here, we assess the influence of the local environment on weak gravitational lensing (WL) shear statistics in the context of relativistic $N$-body code, \texttt{gevolution}. We derive numerical constraints on the cosmological parameters from the WL shear angular power spectrum and comment on the local environment's influence on WL shear. We find tighter constraints on the parameter $Ω_\mathrm{m}$ above redshift $z$ = 0.2, which implies over this redshift the local environment's impact is minor. We also investigate the bispectrum and conclude that on average the impact of the local environment on $f_{\rm NL}$ (a measure of non-Gaussianities) is minimal and consistent with zero effect. However, we find that within the assembly of all possible observers/locations, there will also be a few that could infer the parameter $f_{\rm NL}$ of the order 10. These results could thus be used to estimate the uncertainty in the inference of cosmological parameters such as $f_{\rm NL}$ based on WL shear bispectrum and thus may have implications for future surveys requiring precision at the percent level.

We present FIRES, a polarised shot-noise framework that models fast radio burst (FRB) dynamic spectra as the incoherent superposition of Gaussian microshots. Applied to the CRAFT bursts FRB 20191001A and FRB 20240318A, FIRES reproduces key spectro-polarimetric behaviours: scattering suppresses position-angle (PA) variability on the trailing edge, while the leading edge preferentially retains intrinsic structure when sufficient signal-to-noise is present. We quantify this behaviour using the PA variance ratio $\mathcal{R}_ψ$ and explore the joint plane of measured linear polarisation fraction $Π_L$ versus PA variance to constrain the allowed parameter space of microshot number $N$, intrinsic PA dispersion $σ_ψ$, and intrinsic linear fraction $Π_{L,0}$ at fixed signal-to-noise. For FRB~20191001A, the data are consistent with an extended region spanning $σ_ψ\sim 10^\circ$--$30^\circ$ and $N \sim 5$--$1000$, reflecting degeneracies between intrinsic PA structure, microshot superposition, scattering, finite sampling, and noise. FRB~20240318A occupies a more restricted region, favouring fewer microshots ($N \lesssim 20$) and larger intrinsic PA dispersion ($σ_ψ\sim 15$--$23^\circ$), depending on $Π_{L,0}$, consistent with its observed PA variability. By combining an emission-mechanism-independent framework with minimal assumptions and observational constraints, FIRES provides direct, quantitative constraints on the space of viable FRB microphysical models and demonstrates that microshot superposition offers a natural explanation for the diverse polarimetric behaviours observed in FRBs.

In this paper of ERGO-ML (Extracting Reality from Galaxy Observables with Machine Learning), we develop a model that infers the merger/assembly histories of galaxies directly from optical images. We apply the self-supervised contrastive learning framework NNCLR (Nearest-Neighbor Contrastive Learning of visual Representations) on realistic HSC mock images (g,r,i - bands) produced from galaxies simulated within the TNG50 and TNG100 flagship runs of the IllustrisTNG project. The resulting representation is then used as conditional input for a cINN (conditional Invertible Neural Network) to gain posteriors for merger/assembly statistics, particularly the lookback time and stellar mass of the last major merger and the fraction of ex-situ stars. Through validation against the ground truth available for simulated galaxies, we assess the performance of our model, achieving good accuracy in inferring the stellar ex-situ fraction ($\le \pm 10$ per cent for 80 per cent of the test sample) and the mass of the last major merger (within $\pm 0.5 \log \MSUN$ for stellar masses $>10^{9.5} \MSUN$ ). We successfully apply the TNG-trained model to simulated mocks from the EAGLE simulation, demonstrating that our model is applicable outside of the TNG domain. We use our simulation-based model to infer aspects of the history of observed galaxies, in particular for HSC images that are close to the domain of TNG ones. We recover the trend of increasing ex-situ stellar fraction with stellar mass and more spherical morphology, but we also identify a discrepancy between TNG and HSC: on average, observed galaxies generally exhibit lower ex-situ fractions. Despite challenges such as information loss (e.g. projection effects and surface brightness limits) and domain shifts (from simulations to observations), our results demonstrate the feasibility of extracting the merger past of galaxies from their optical images.

Characterising and distinguishing gamma-ray bursts (GRBs) has interested astronomers for many decades. While some authors have found two or three groups of GRBs by analyzing only a few parameters, recent work identified five ellipsoidally-shaped groups upon considering nine parameters $T_{50}, T_{90}, F_1, F_2, F_3, F_4, P_{64}, P_{256}, P_{1024}$. Yet others suggest sub-classes within the two or three groups found earlier. Using a mixture model of Gaussian factor analysers, we analysed 1150 GRBs, that had nine parameters observed, from the current Burst and Transient Source Experiment (BATSE) catalogue, and again established five ellipsoidal-shaped groups to describe the GRBs. These five groups are characterised in terms of their average duration, fluence and spectrum as shorter-faint-hard, long-intermediate-soft, long-intermediate-intermediate, long-bright-intermediate and short-faint-hard. The use of factor analysers in describing individual group densities allows for a more thorough group-wise characterisation of the parameters in terms of a few latent features. However, given the discrepancy with many other existing studies that advocated for two or three groups, we also performed model-estimated overlap-based syncytial clustering (MOBSynC) that successively merges poorer-separated groups. The five ellipsoidal groups merge into three and then into two groups, one with GRBs of low durations and the other having longer duration GRBs. These groups are also characterised in terms of a few latent factors made up of the nine parameters. Our analysis provides context for all three sets of results, and in doing so, details a multi-layered characterisation of the BATSE GRBs, while also explaining the structure in their variability.

Particle acceleration in relativistic collisionless shocks remains an open problem in high-energy astrophysics. Particle-in-cell (PIC) simulations predict that electron acceleration in weakly magnetized shocks proceeds via small-angle scattering, leading to a maximum electron energy significantly below the Bohm limit. This upper bound manifests observationally as a characteristic synchrotron cutoff, providing a direct probe of the underlying acceleration physics. Gamma-ray burst (GRB) afterglows offer an exceptional laboratory for testing these predictions. Here, we model the spectral evolution of GRB afterglows during the relativistic deceleration phase, incorporating PIC-motivated acceleration prescriptions and self-consistently computing synchrotron and synchrotron self-Compton emission. We find that low-energy bursts in low-density environments, typical of short GRBs, exhibit a pronounced synchrotron cutoff in the GeV band within minutes to hours after the trigger. Applying our framework to GRB 190114C and GRB 130427A, we find that current observations are insufficient to discriminate between PIC-motivated acceleration and the Bohm limit, primarily due to large uncertainties in the Fermi-LAT band. Nevertheless, future MeV-TeV afterglow observations can break model degeneracies and place substantially tighter constraints on particle acceleration in relativistic shocks.

We investigate the non-linear evolution of matter perturbations in $f(R)$ models with the Chameleon screening mechanism. The novel feature of our investigation is an iterative solution for the non-linear equation for the scalar field $χ= Φ- Ψ$, where $Φ$ and $Ψ$ are the potentials that characterise scalar perturbations of the metric. We demonstrate the scheme on spherical perturbations - smooth, compensated top-hats of varying length scales. We find that the effect of the Chameleon mechanism is seen most prominently on scales where the size of the top-hat is comparable to the Compton scale of the background. There is a density enhancement near the outer edge of the top-hat and the top-hat does not retain its shape. We explain this well-known observation in the context of the spatio-temporal evolution of the Compton scale. Additionally, we find a slight enhancement of the density near the origin, a feature not reported previously in the literature. On scales much smaller or much larger than the background Compton length, including the Chameleon screening has no appreciable effect on the perturbations. In the former, the growth is enhanced as compared to GR and is almost the same as GR in the latter. Finally, we examine the non-linear density velocity divergence (DVDR) relation and find that for evolution affected by Chameleon screening, the DVDR is no longer one-to-one even for a single profile. The relation between density and velocity depends on the location within the perturbation.

Calcium-strong transients (CaSTs) are a subclass of faint and rapidly evolving supernovae (SNe) that exhibit strong calcium features and notably weak oxygen features. The small but growing population of CaSTs exhibits some aspects similar to thermonuclear supernovae and others that are similar to massive star core-collapse events, leading to intriguing questions on the physical origins of CaSTs. SN 2025coe is one of the most nearby CaSTs discovered to date, and our coordinated multi-wavelength observations obtained days to weeks post-explosion reveal new insights on these enigmatic transients. With the most robust NIR spectroscopic time-series of a CaST collected to date, SN 2025coe shows spectral signatures characteristic of Type Ib SNe (SNe Ib, i.e. He-rich stripped-envelope SNe). SN~2025coe is the third X-ray detected CaST and our analysis of the \textit{Swift} X-ray data suggest interaction with 0.12 $\pm\,0.11\ M_{\odot}$ of circumstellar material (CSM) extending to at least $2 \times 10^{15} $cm ($\sim 30,000\ R_{\odot}$), while our analysis of the 1-240 GHz radio non-detections gives an outer radius of that CSM of at most $\sim 4\times 10^{15}$ cm. This inferred nearby high-density CSM extending out to $3\pm 1 \times10^{15}$ cm is similar to that seen in the other two X-ray detected CaSTs, and its presence suggests that either intensive mass-loss or some polluting mechanism may be a common feature of this subclass. Our work also expands upon recent studies on the optical properties of SN 2025coe and explores our current understanding of different progenitor systems that could possibly produce CaSTs.

We have observed six `dark' neutral hydrogen (HI) clouds discovered in the Virgo cluster by the Arecibo Galaxy Environment Survey (AGES) with the Karl G. Jansky Very Large Array (VLA), giving higher angular and velocity resolution than the original AGES observations. We detected compact HI emission in two of the sources, AGESVC1 231 and AGESVC1 274, allowing us to firmly associate them with faint ($m_g > 18.5$), blue ($g-i < 0.1$) optical counterparts with high $M_{HI}/L_g$ ratios. In a further two sources, we detected low column-density extended HI emission, consistent with these being dispersing clouds from ram-pressure stripping or tidal interactions. The final two sources were not detected with the VLA, allowing us to set low column-density limits on the HI detected by AGES that are consistent with these clouds also being formed from HI that is dispersing into the intra-cluster medium. The four HI sources not associated with optical counterparts thus appear likely to be relatively short-lived objects. No evidence was found for either pressure-supported turbulent spheres or stable dark galaxies.

Fast, counter-streaming plasma outflows drive magnetic field amplification, plasma heating, and particle acceleration in numerous astrophysical environments, from supernova remnant shocks to active galactic nuclei jets. Understanding how, in the absence of Coulomb collisions, energy is redistributed between the different plasma species remains a fundamental open question. We use 3D fully-kinetic simulations to investigate energy partition in weakly magnetized counter-propagating plasmas. Our results reveal a complex interplay between different processes, where at early times the Weibel instability drives a first stage of magnetic field amplification and at late times the kinking of current filaments drives a second amplification stage via a dynamo-type mechanism. Electrons are heated primarily during the latter phase through magnetic pumping. By the time the flows thermalize, we observe that the final temperature ratio $T_e/T_i$ and energy partition depend on the ion-to-electron mass ratio. For electron-proton flows, the electron thermal energy only reaches up to a few percent of the initial ion kinetic energy.

In the $Λ$CDM paradigm, stellar halos form through the accretion and disruption of satellite galaxies. We introduce new semi-analytic modeling within the SatGen framework to track the ex-situ stellar components of Milky Way--like galaxies across large ensembles of merger trees, enabling a statistical study of the stochastic nature of galaxy assembly. We find that accreted stellar halos are typically built by only a few progenitors and are highly sensitive to the fate of the most massive satellite, producing order-of-magnitude variations in accreted stellar halo mass even at fixed host halo mass. Different stellar components trace distinct phases of host halo growth: central and accreted stellar mass correlate most strongly with early assembly, while surviving satellites trace more recent accretion. Finally, using Random Forest Regression, we quantify how well observable galaxy properties can recover halo assembly histories, providing a framework for interpreting upcoming low-surface-brightness observations of stellar halos.

We present multiwavelength observations and analysis of six luminous fast blue optical transients (LFBOTs) discovered in Zwicky Transient Facility (ZTF) survey data. We identified these LFBOTs from their fast light-curve evolution ($t_{1/2}\leq 12 $d), blue colors at peak brightness ($g-r\leq-0.5 $mag), a visible host galaxy, high optical luminosity ($M_g<-20$), and an X-ray or radio detection. With the exception of AT2024aehp (ZTF24abygbss), these transients exhibit peaks in their $10\,$GHz radio light curves at $t_{\text{rest}} \approx 50-100$ d, with peak radio luminosities ranging from $10^{38}-10^{40}$ erg s$^{-1}$. Modeling the radio emission as synchrotron radiation indicates a fast ($v=0.1-0.3c$) shock in a dense ($n_e\approx10^{3}-10^{4}$ cm$^{-3}$) medium. The X-ray emission varies by $\approx2$ orders of magnitude in luminosity ($10^{42}-10^{44}$ erg s$^{-1}$) at $t_{\text{rest}}\sim20 $d. Analysis of the host-galaxy photometry and spectroscopy for each transient shows that they are predominantly nonnuclear (a few kpc offset) with star-forming host galaxies of stellar masses $10^{9}-10^{11} ,M_\odot$. Unlike all other LFBOTs to date, AT2024aehp exhibited a luminous ($M<-19 $mag) plateau in the optical light curve; spectra during this plateau phase showed a featureless blue continuum. The $6-15$ GHz radio emission of AT2024aehp brightened by over an order of magnitude from $t_{\text{rest}} \approx70 $d to $t_{\mathrm{rest}} \approx130 $d. The mostly consistent radio behavior between optically selected LFBOTs implies a similar circumburst medium, leading us to prefer a progenitor scenario in which mass is lost in a consistent way shortly prior to the terminal event, such as a massive star merging with a compact object.

The merger history of the Galaxy has been traced back firmly to redshift 2 (10 Billion years ago). While there have been claims of the existence of at least one more significant merger before this time, supporting evidence has been indirect and contentious. Here we show that the population of globular clusters around the Galaxy depicts three distinct age-metallicity sequences, one associated with the merger with Gaia-Enceladus 10 billion years ago, one to the progenitor of the Milky Way and a third intermediate sequence associated to at least one merger which we estimate took place merely 1.5 billion years after the Big Bang. This discovery has been possible thanks to exquisite Hubble Space Telescope data and sophisticated analysis that enables very precise relative age determination of globular clusters. The newly identified sequence reveals that this merger took place with an object of stellar mass similar to that of Gaia-Enceladus (~5x10$^8$ M$_{\odot}$), and which deposited most of its mass in the inner 6 kpc of the Milky Way. The unambiguous identification of a third merger event in the inner Galaxy puts to rest earlier debates, and honoring previous work we name the progenitor system Low-energy-Kraken-Heracles, or LKH for short.

The radio galaxy M87 is well known for its jet, which features a series of bright knots observable from radio to X-ray wavelengths. The most famous of these, HST-1, exhibits superluminal motion, and our analysis of {\it Chandra} data \citep{Thimmappa24} reveals a correlation between the X-ray flux of HST-1 and its separation from the core. This correlation likely arises from moving shocks in the jet, allowing measurement of the internal structure of HST-1 in the X-ray band. To follow up on these results, we use observations from the {\it Hubble} Space Telescope Advanced Camera for Surveys HRC/WFC/SBC channel and the Wide Field Camera 3 (WFC3)'s UVIS to analyze the image and flux variability of HST-1. Our analysis includes 245 ACS and 120 WFC3 observations from 2002-2022, with a total exposure time of $\sim345$ ks. We study the brightness profile of the optical jet and measure the relative separation between the core and HST-1 for comparison to the X-ray. We find that the X-ray and the UV/optical emission could arise from physically distinct regions. The measured proper motion of the knot HST-1 is 1.04$\pm$0.04 c from 2002-2005 and around 2.1$\pm$0.05 c from 2005-2022. We discuss the acceleration of the jet and the flaring synchrotron emission from HST-1 from optical to X-rays.

We use the TNG50 galaxy formation simulation to generate mock JWST NIRCam and NIRSpec microshutter array (MSA) observations of H$α$-emitting gas in $M_*=10^8-10^{11.5}\,M_\odot$ star-forming galaxies at $z=2-6$. We measure morphological properties from the mock imaging through Sersic profile fitting, and gas rotational velocities ($v$) and velocity dispersions ($σ$) by fitting the mock spectra as thin, rotating discs. To test the efficacy of such simple parametric models in describing complex ionised gas kinematics, we compare the best-fit quantities to intrinsic simulation measurements. At $z=3$, we find that $v$ and $σ$ for aligned and resolved sources generally agree well with intrinsic measurements, within a factor of $\sim$2 and $\sim$1.5, respectively. The recovery of kinematics is robust for smooth, disc-like systems, but $v$ and $σ$ can be over- or underestimated by more than a factor of 2, respectively, for intrinsically elongated systems. The scatter in the recovery accuracy is larger at higher redshift, as TNG50 galaxies at $z>3$ deviate more strongly from the thin rotating disc assumption. Despite uncertain measurements for individual galaxies, we find that key population trends, such as the weak redshift evolution of $σ$ and $v/σ$ as well as the dependence of $σ$ on the global star formation rate, are broadly recovered by our kinematic modelling. Our work provides the end-to-end framework needed to compare NIRSpec MSA observations to cosmological simulations and to quantify observational biases in measuring ionised gas kinematics, highlighting the need for the development of dedicated models for high-redshift galaxies.

We present a quantitative model for the luminous fast blue optical transient AT2018cow in which a shock propagating through an aspherical circumstellar medium (CSM) produces the X-ray and UV/optical/NIR emission. X-rays are emitted from hot post-shock electrons, and soft X-ray photons are reprocessed into optical/UV emission in the cool downstream. This naturally explains two previously puzzling features: (i) the coordinated evolution of the optical and soft X-ray after day 20, (ii) the hard X-ray hump above 10 keV that disappears around day 15 as the Thomson optical depth transitions from $τ_T \gg1$ to $τ_T \sim 1$. Our model is over-constrained, and it quantitatively reproduces the bolometric luminosity evolution, soft X-ray spectrum, and time-dependent soft/hard X-ray and soft X-ray/optical luminosity ratios. It also explains additional puzzles: X-ray fluctuations with $\sim4-10$ day timescales arise from a global radiative shock instability, while the NIR excess and the apparent receding blackbody radius result from reprocessed X-rays in matter far from thermodynamic equilibrium. The radio is naturally explained as originating from a shock driven by the same ejecta in the more dilute CSM. The light curve steepening after $\sim 40$ days likely indicates the shock reaches the edge of the dense CSM at $\sim {\rm few} \times 10^{15}$ cm. We infer explosion energy $\sim 1-5 \times 10^{50}$ erg, carried by an ejecta at $\sim 0.1c$ and a mass of $0.01-0.05 M_\odot$, in a dense asymmetric CSM with $\sim 0.3 M_\odot$, embedded in a more dilute CSM.

Galactic globular clusters (GCs) were born shortly after the Big Bang. For such old stellar systems the initial mass function (IMF) at the high mass regime can never be observed directly, because stars more massive than about 1 Mo have evolved since longtime. However, the hydrostatic to explosive alpha-element ratio (HEx ratio) offers a way to bypass the lack of observable high mass stars through the yields that massive stars released when exploding as supernovae, incorporated in the stars we presently observe in GCs. The HEx ratio measures the percentage of high mass stars over the total number of stars exploding as supernovae and it is an efficient probe of the ephemeral first phases of the GC evolution. We exploited a recently completed survey to assemble a dataset of very homogeneous abundances of alpha-elements in 27 GCs from [Fe/H]~ -2.4 to ~ -0.3 dex. In agreement with previous results from APOGEE, we confirm that the HEx ratio is indistinguishable for GCs formed in situ and accreted in the Galaxy, and that this ratio decreases with increasing metallicity. However, we posit that this trend is better explained by a metallicity-dependent IMF deficient in the highest mass stars at high metallicity, as corroborated by the declining [O/Mg] ratio as a function of the [Mg/H] ratio. At odds with the previous analysis based on APOGEE data, we detect an anti-correlation of HEx ratio with both present day and initial GC masses. Finally, we hypothesise that in that analysis, the stars of the GC M 54 were probably confused with stars in the core of the Sagittarius dwarf galaxy, where the cluster is presently immersed.

GY 91, commonly categorized as a Class I young stellar object, is notable for disk dust substructures that have been hypothesized to trace early planet formation. Using the ALMA 12-m and ACA arrays, we present new Band 7 dust continuum and molecular line observations of GY 91 at an angular resolution of (~40 au). We report detections of CS $J=6-5$, N$_2$H$^+$ $J=3-2$, C$^{18}$O $J=3-2$, H$_2$CS $J_{K_a, K_c} = 8_{1,7}-7_{1,6}$, H$_2$CO $J_{K_a, K_c} = 4_{0,4}-3_{0,3}$, and H$_2$CO $J_{K_a, K_c} = 4_{2,3}-3_{2,2}$, as well as a tentative detection of $^{13}$C$^{18}$O $J=3-2$. We observe azimuthal asymmetry in CS and H$_2$CS emission, as well as radially structured H$_2$CO $4_{0,4}-3_{0,3}$ emission outside the dust continuum. C$^{18}$O and H$_2$CO 4$_{0,4}-3_{0,3}$ show significant cloud contamination, while CS and N$_2$H$^+$ are good tracers of Keplerian rotation originating from the disk. Envelope emission does not appear to contribute significantly either to the continuum or molecular line observations. GY 91's chemical properties appear in large part to resemble those of Class II disks, although observations of additional molecular probes should be obtained for a fuller comparison. With CS, we estimated a dynamical stellar mass of 0.58 $M_\odot$, which is higher than previous estimates from stellar evolutionary models (0.25 $M_\odot$). Using both radiative transfer modeling of the dust continuum and comparison of the C$^{18}$O and N$_2$H$^+$ fluxes to literature thermochemical models, we estimate a disk mass of $\sim0.01$ $M_\odot$.

The FLAMES survey ``Na-O anti-correlation and HB" uncovered the modern standard for globular clusters (GCs), that is their ubiquitous multiple stellar populations (MPs) distinct by the abundance of proton-capture elements. That survey can still be mined to extract a wealth of data. We derive new abundances of Mg, Si, Ca, and Ti for 948, 954, 1542, and 1350 red giant branch stars in 16 GCs, both formed in situ or accreted in the Milky Way. The program GCs cover the metallicity range from [Fe/H]=-2.35 dex to [Fe/H]=-0.74 dex. Both the halo and disc GCs show a clear overabundance of alpha-elements with the modulation in Mg and Si due to the MPs phenomenon in different clusters. We found star to star variations in Si abundance correlated to changes in Na in more than half of our sample, implying that temperatures in excess of about 65 MK were achieved in the polluters responsible for the enrichment. We confirm with an enlarged sample the previous result that significant variations in Mg are observed in GCs that are metal-poor, massive or both. Evidence of excess of Ca with respect to reference unpolluted field stars are found in NGC 6752 and NGC 7078, indicating the action of proton-capture reactions at very high temperature regime in these GCs. These excesses fit very well in a previously found relation as a function of a combination of cluster mass and metallicity shown by other typical signatures of MPs. At odds with previous results based on the Si abundance from APOGEE, we found that the average abundance of alpha-elements is not an efficient discriminating factor between in situ and accreted GCs.

Recent studies have revealed that bars can form as early as a few billion years after the Big Bang, already displaying similar characteristics of evolved bars in the Local Universe. Bars redistribute angular momentum across the galaxy, regulating star formation, AGN activity, and building new stellar structures such as nuclear discs. However, the effects of bar-driven evolution on young galaxies are not yet known, as no evidence of bar-built stellar structures has ever been found beyond $z = 1$, until now. In this work, we show evidence of a bar-built, star-forming nuclear disc, already present at redshift $z = 1.5$. This is the first evidence of a bar-built stellar structure at Cosmic Noon. We find that this nuclear disc is actively forming stars and has the same size as some nuclear discs in nearby galaxies. This evidence solidifies the now emerging picture in which bars are fundamental not only in the late evolution of galaxies, but also in their early evolutionary stages. It changes the current paradigm by urging a revision of our picture of galaxy evolution beyond redshift one, to include new considerations on the role played by bars as early as a few billion years after the Big Bang.

Spectroscopic surveys now supply precise stellar label measurements such as element abundances for large samples of stars throughout the Milky Way. These element abundances are known to correlate with orbital actions or other dynamical invariants. We present a new data-driven method for empirically measuring the circular velocity curve of the Galaxy that uses element abundance gradients in the plane of radial kinematics. We use stellar surface abundances from the $\textit{APOGEE}$ survey combined with kinematic data from the $\textit{Gaia}$ mission. Our results confirm the ordered structure of the Milky Way disk in terms of average [Fe/H] and [Mg/Fe] abundance ratios, and suggest that $\langle$[Fe/H]$\rangle$ traces the radial position of stars in the disk, while $\langle$[Mg/Fe]$\rangle$ traces the orbital excursions around this radius. Our method uses the radial orbit structure in the Galaxy to enable an empirical measurement of the circular velocity curve, epicyclic and azimuthal frequencies, and kinematic gradients across the Milky Way disk. From these measurements, we infer a value of the circular velocity curve at the Solar radius of $v_{c,\odot} = 235.3^{+2.8}_{-3.7}$ km s$^{-1}$ using the most constraining abundance ratio, [Mg/Fe]. We also measure the radial and azimuthal frequencies for a circular orbit at the solar radius, $κ_{0,R_\odot}=36.9^{+0.8}_{-1.0}$ km s$^{-1}$ kpc$^{-1}$ and $Ω_{0,R_\odot}=28.5_{-0.1}^{+0.4}$ km s$^{-1}$ kpc$^{-1}$, respectively. These values lead to an estimate of the Oort constants of $A = 16.5^{+0.1}_{-0.1}$ km s$^{-1}$ kpc$^{-1}$ and $B=-11.9^{+0.1}_{-0.3}$ km s$^{-1}$ kpc$^{-1}$. We measure the radial acceleration at the Solar radius to be $(\frac{\partial Φ}{\partial R})_{\odot} = a_{R_\odot}=7.0^{+0.2}_{-0.1}$ pc Myr$^{-2}$.

We investigate induced Compton/Brillouin scattering of electromagnetic waves in magnetized electron and positron pair plasma by verifying kinetic theory with Particle-in-Cell simulations. Applying this to fast radio bursts (FRBs) in magnetar magnetospheres, we find that the scattering--although suppressed by the magnetic field--inevitably enters the linear growth stage. The subsequent evolution bifurcates: full scattering occurs when the density exceeds a critical value, whereas below it the scattering saturates and the FRB can escape. This eases the tension with observations of compact emission regions and may explain the observed diversity, including the presence or absence of FRBs associated with X-ray bursts.

The redshifted 21-cm line from neutral hydrogen ($\textrm{H}\textsc{i}$) enables volumetric intensity mapping of large-scale structure in the post-reionization Universe. In anticipation of \texttt{SKA-MID}'s wide redshift coverage and high signal-to-noise clustering measurements, we study the redshift-space 21-cm bispectrum and its spherical-harmonic multipoles as probes of anisotropic non-linear structure formation and departures from General Relativity. Using a tree-level perturbative description for the 21-cm brightness-temperature field in redshift space, and adopting the Hu--Sawicki $f(R)$ model as a representative modified-gravity scenario, we forecast the detectability of configuration-dependent signatures with an \texttt{SKA-MID}--like survey. We derive the bispectrum-multipole covariance including sample variance and thermal noise and evaluate the expected signal-to-noise of deviations relative to $Λ$CDM. We find that the observable information is dominated by the lowest multipoles, while higher-order modes are strongly suppressed. This concentration in the lowest multipoles is well matched to \texttt{SKA-MID} sensitivity and to the quasi-linear modes that are expected to remain accessible in practice. The strongest modified-gravity sensitivity arises from squeezed and stretched triangle configurations on quasi-linear scales, where scale-dependent growth enhances the bispectrum relative to the total variance. Our results position 21-cm bispectrum multipoles as a practical, SKA-ready observable for testing gravity beyond $Λ$CDM in the post-reionization epoch.

JWST has revealed a population of compact galaxies in the early Universe with broad emission lines and strong Balmer breaks; among them the so-called ''little red dots'' (LRDs). Their nature remains uncertain with hypotheses including exotic phenomena. We assemble a sample of LRD-like objects at $z>3$ and use self-consistent radiative-transfer calculations to show that a supermassive black hole accreting from a dense gas cocoon accurately reproduces the detailed spectra. We show that the cocoons must be non-spherical, with comparable amounts of inflowing and outflowing material. And we predict correlations between Balmer break strength, Balmer line-absorption and scattering line width, which we confirm in our observed sample. We reproduce all LRD-like properties without requiring star-like atmospheres and we determine the typical black hole in our sample to be of order a million solar masses, with ionized cocoon masses of tens of solar masses potentially supplied from a much larger cold-gas reservoir.

The standard approach to modeling X-ray spectral data relies on local optimization methods, such as the Levenberg-Marquardt algorithm. While effective for simple models and speedy spectral fitting, these local optimizers are prone to becoming trapped in local minima, particularly in high-dimensional or degenerate parameter spaces, and typically require extensive user intervention. In this work, we introduce XFit, a global optimization method for fitting X-ray data, which makes extensive use of the Ferret evolutionary algorithm. XFit enables automated exploration of complex parameter spaces, efficient mapping of confidence intervals, and identification of degenerate solutions that may be overlooked by local methods. We demonstrate the performance of XFit using two representative X-ray sources: the Central Compact Object in Cassiopeia A and the supernova remnant G41.1-0.3. These examples span both low- and high-dimensional models, allowing us to illustrate the advantages of global optimization. In both cases, XFit produces solutions that are consistent with or improve upon those found with traditional methods, while also revealing alternative fits or degenerate solutions within statistically acceptable confidence levels. The automated mapping of parameter space offered by XFit makes it a powerful complement to existing spectral fitting tools, particularly as models and data quality become increasingly complex. Future work will expand the application of XFit to broader datasets and more physically motivated models.

Large-scale filaments ubiquitously exist in the Galactic interstellar medium, and their radial profiles offer insights into their formation mechanisms. We present a statistical analysis of molecular hydrogen column density ($\rm N(H_2)$) and dust temperature ($\rm T_d$) radial profiles for 35 Galactic large-scale filaments. We divided their spines into 315 segments, extracted the radial profiles of each segment using $\rm N(H_2)$ and $\rm T_d$ maps derived from $Herschel$ Hi-GAL data, and estimated the asymmetry degree within the radial profiles ($α_{\rm asy}$), as well as the length proportion of segments with asymmetric profiles across the entire filament ($f_{\rm asy}$). We found that Galactic large-scale filaments reside in surroundings distinctly asymmetric and varied in $\rm N(H_2)$, and mild asymmetric yet stable in $\rm T_d$. Different filament morphology types do not show significant differences in $α_{\rm asy}$ or $f_{\rm asy}$. A bent filament shape does not necessarily correspond to an asymmetric radial profile, whereas a straight filament shape may be associated with a symmetric profile. Segments with asymmetric surroundings in $\rm N(H_2)$ may not simultaneously appear asymmetric in $\rm T_d$, and vice versa. We found three filaments with 4-44% of their spine show asymmetric $\rm N(H_2)$ and $\rm T_d$ radial profiles in inverse trends, likely caused by nearby HII region. HII regions of similar scale to large filaments can induce asymmetric radial profiles within them, indicating their influence on filament evolution. However, they are unlikely to independently trigger the formation of an entire Galactic large-scale filament, in contrast to their role in small-scale filament formation.

Velocity fields in the cosmic web are fundamental to structure formation but remain difficult to observe directly beyond the linear regime. Here we present observational evidence that galaxy filaments connecting pairs of galaxy clusters undergo a split infall, with opposite velocity flows toward the two clusters. Using spectroscopic galaxies from the Sloan Digital Sky Survey, we isolate the internal filament velocity field by subtracting its rigid-body background motion and Hubble flow, and detect this effect at greater than $5σ$ significance across a wide range of cluster and filament selections. The measured velocity profile exhibits a sign reversal near the filament midpoint and a maximum infall amplitude of $\sim30$ km/s ($\sim20$ km/s projected onto the line-of-sight) for clusters of mass $\sim10^{14.3}M_\odot$, substantially lower than expected for infall from an average cosmic environment. Multiple results on density-velocity correlation, mass-dependency, and validation with simulation indicate that filaments dynamically respond to competing gravitational potentials rather than acting as passive mass transport channels. Our results establish a new observational window on quasi-linear velocity fields in the cosmic web and provide a promising probe of mass measurement, testing gravity and velocity reconstruction with upcoming wide-field spectroscopic surveys.

Recently, the Pierre Auger Observatory has found strong evidence supporting the extragalactic origin of the most energetic cosmic rays. Despite several observed excesses in the distribution of arrival directions for the highest energy cosmic rays, the sources remain unidentified. Accretion shocks in galaxy clusters have been proposed as potential sources in the past. These immense shock waves, which can have radii on the order of megaparsecs, are generated by the infall of material from the intergalactic medium into the gravitational potential wells of galaxy clusters. In this work, we investigate the possibility that ultrahigh-energy cosmic rays are accelerated in these regions. Nearby massive galaxy clusters, including Virgo, are treated as a discrete component of the cluster mass distribution. Less massive galaxy clusters, as well as distant massive ones, are assumed to follow a continuous distribution in agreement with cluster mass statistics. We fit the flux at Earth and the composition profile measured by the Pierre Auger Observatory, assuming the injection of different nuclear species by these sources, to determine the values of the model parameters. Our results indicate that cosmic ray acceleration in cluster accretion shocks may account for at least a fraction of the observed UHECR flux at energies below the suppression scale. At higher energies, direct acceleration from the thermal pool would be feasible only if local fluctuations create favorable conditions, such as magnetic fields about an order of magnitude stronger than those typically expected in cluster accretion shocks, or for particular shock normal-magnetic field configurations.

We investigate the role of resolution and initial magnetic field strength on core-collapse supernovae in simulations of a non-rotating $13 \mathrm{M_\odot}$ progenitor. Specifically, we study the effect on shock revival, explosion dynamics, and the properties of the compact remnant. We run four models with different numerical grid resolutions with an initial central dipole field strength of $\mathord{\approx}10^{12}\, \mathrm{G}$. Two of those resolutions are also run with a weaker central magnetic field of $\mathord{\approx}10^{10}\, \mathrm{G}$ . The shock revival time for all models is largely independent of resolution and initial magnetic field strength, but we find higher explosion energies when the initial magnetism is stronger and at higher resolutions. We find that models with strong magnetic fields have lower neutrino luminosity and energies, due to a proto-neutron star (PNS) that is deformed by the strong magnetic fields. At higher resolutions, magnetic fields are amplified more efficiently in the gain region and in the PNS via the small-scale dynamo. Although the strong magnetic fields do not directly drive the explosion, they have a subsidiary impact on the explosion mechanism and compensate for the reduced neutrino heating. Stronger magnetic energies in the PNS also affect energy and angular momentum redistribution, leading to more extended and vigorous PNS convection zones at higher resolutions.

In the past few decades, X-ray astronomy satellites equipped with grating spectrometers and microcalorimeters have enabled high-resolution spectroscopic observations of astrophysical objects. The need for accurate atomic data has arose as we attempt detailed analysis of the high-resolution spectra they provide. This is because current spectral models, which heavily rely on theoretical calculations, entail non-negligible uncertainties. We employ a plasma spectroscopy device called electron beam ion trap (EBIT) to experimentally obtain precise atomic data. An EBIT with a design that allows combined operation with synchrotron radiation facilities was developed based on the Heidelberg Compact EBIT and installed at ISAS/JAXA for this purpose. We conducted a spectroscopic experiment using the JAXA-EBIT at the synchrotron radiation facility SPring-8, and successfully obtained high-resolution spectra of the L$α$ resonance transition of Ne-like Fe$^{16+}$ ions, 3C, as well as the K$α$ resonance transition of He-like O$^{6+}$ ions. We also measured another Ne-like Fe$^{16+}$ L$α$ resonance transition, 3G, and constrained an upper limit of the oscillator strength ratio of 3G to 3C, using our experimental results. The experimental values obtained in this study will be applied to observational studies of astrophysical objects as a part of the plasma spectral modeling.

The radius-luminosity ($R_{\rm BLR}$-$L_{5100}$) relation is fundamental to active galactic nucleus (AGN) studies, enabling supermassive black hole (SMBH) mass estimates and AGN-based cosmology applications. However, its high-luminosity end remains poorly calibrated due to insufficient reliable reverberation mapping (RM) data. We present a four-year RM campaign of the luminous quasar E1821+643 using the Lijiang 2.4-m telescope, supplemented by archival multi-wavelength data. E1821+643 is the most luminous AGN with an \hb\ RM measurement to date. The measured time lag of $83.2_{-18.7}^{+17.5}$ days is a factor of 5.6 shorter than predicted by the canonical $R_{\rm BLR}$-$L_{5100}$ relation. By compiling the full \hb\ RM sample, we find that such deviation defines a lower envelope ($0.2R_{\rm BLR}$) of measured lags across the entire luminosity range, while the upper envelope lies near $2R_{\rm BLR}$, implying that the scatter for individual AGNs can reach 1 dex. Spectral decomposition reveals two distinct \hb\ components: a core component with a lag of $267.0_{-17.6}^{+16.6}$ days closer to the $R_{\rm BLR}$-$L_{5100}$ relation, and a redshifted tail with a much shorter lag of $-49.0_{-34.5}^{+50.5}$ days. The short-lag component not only accounts for the significantly shortened overall lag, but also leads to an opposite interpretation of the intrinsic BLR kinematics. These effects can introduce systematic uncertainties in black hole mass estimates by factors of up to tens. Our findings demonstrate that shortened lags in high-accretion-rate AGNs arise from multi-component BLR structures, posing substantial challenges to single-epoch mass estimates and impacting SMBH demographics and cosmological applications.

The gravitational wave event GW190521 seems to be the only BH merger event possibly correlated with an electromagnetic counterpart, which appeared about 34 days after the GW event. This work aims to confirm that the electromagnetic bump towards the Active Galactic Nucleus (AGN) J1249+3449 can be explained within the framework of the gravitational microlensing phenomenon. In particular, considering the data of the Zwicky Transient Facility (ZTF), what emerges from a detailed analysis of the observed light curve using three fitting models (Point Source Point Lens, Finite Source Point Lens, Uniform Source Binary Lens) is that the optical bump can be explained as a microlensing event caused by a lens with mass {$\sim\,$0.1 $M_{\odot}$}, lying in the host galaxy of the AGN in question.} %MDPI: Please confirm if the bold formatting is necessary; if not, please remove it.

The chemical evolution of pre-stellar cores during their transition to a protostellar stage is not yet fully understood. Detailed chemical characterizations of these sources are needed to better define their chemistry during star formation. Our goal is to characterize the chemistry of the starless cores C2 and C16 in the B213/L1495 filament of the Taurus Molecular Cloud, and to understand how it relates to the environmental conditions and the evolutionary state of the cores. We made use of two complete spectral surveys at 7 mm of these sources, carried out using the Yebes 40-m telescope. Derived molecular abundances were compared with those of other sources in different evolutionary stages and with values computed by chemical models. Including isotopologs, 22 molecules were detected in B213-C2, and 25 in B213-C16. The derived rotational temperatures have values of between $\sim$ 5 K and $\sim$ 9 K. A comparison of the two sources shows lower abundances in C2, except for l-C$_{3}$H and HOCO$^{+}$, which have similar values in both cores. Model results indicate that both cores are best fit assuming early-time chemistry, and point to C2 being in a more advanced evolutionary stage, as it presents a higher molecular hydrogen density and sulfur depletion, and a lower cosmic-ray ionization rate. Our chemical modeling successfully accounts for the abundances of most molecules, including complex organic molecules and long cyanopolynes (HC$_{5}$N, HC$_{7}$N), but fails to reproduce those of the carbon chains CCS and C$_{3}$O. Chemical differences between C2 and C16 could stem from the evolutionary stage of the cores, with C2 being closer to the pre-stellar phase. Both cores are better fit assuming early-time chemistry of t $\sim$ 0.1 Myr. The more intense UV radiation in the northern region of B213 could account for the high abundances of l-C$_{3}$H and HOCO$^{+}$ in C2.

We present a JWST/NIRSpec rest-frame optical spectroscopic census of ALMA 1-mm continuum sources in the Hubble Ultra Deep Field (UDF) identified by the deep ALMA UDF and ASPECS programs. Our sample is composed of the ALMA flux-limited ($S_{1\,\mathrm{mm}}\gtrsim 0.1\,\mathrm{mJy}$) sources observed with medium-resolution NIRSpec spectroscopy from JADES and SMILES, 16 faint submillimeter galaxies (SMGs) at spectroscopic redshifts of $z\sim 1$-$4$. These SMGs show bright longer-wavelength optical lines (H$α$, [N II]$λ\lambda6548,6583$, and [S II]$λ\lambda6717,6731$) and faint shorter-wavelength optical lines (H$β$ and [O III]$λ\lambda4959,5007$) with a large nebular attenuation, $E(B-V)\sim0.3$-$1.8$. We test the SMGs using BPT diagnostics and Chandra X-ray fluxes, and find that most SMGs are classified as AGNs; the AGN fraction is $\sim80\%$ for the SMGs at $M_*>10^{10.5} M_\odot$. We find only one SMG ($<10\%$) with a broad Balmer line, indicating that the SMGs are predominantly obscured AGNs. With the optical lines, we estimate the metallicities of the SMGs to be moderately high, $\sim0.4$-$2 Z_\odot$, exceeding the model-predicted dust-growth critical metallicity ($\sim0.1$-$0.2Z_\odot$), which naturally explains the dusty nature of the SMGs. Interestingly, the SMGs fall in the mass-metallicity relation and the star-formation main sequence, showing no significant differences from other high-$z$ galaxies. Similarly, we find electron densities of $n_e\sim10^2$-$10^3\,\mathrm{cm}^{-3}$ for the SMGs that are comparable with other high-$z$ galaxies. Together with the high SMG fraction ($\sim 100\%$) at the massive end ($M_*>10^{10.5} M_\odot$), these results indicate that the SMGs are mostly not special, but typical massive star-forming galaxies at high redshift.

We perform the classification of black hole and neutron star X-ray binary systems using deep neural networks applied to archival RXTE X-ray spectral data. We first construct two neural network models: one trained using only spectral flux values and another trained using both fluxes and their associated errors. Both models achieve high classification accuracies of ~90-94 %. To gain physical interpretability of these networks, we fit all spectra with a simple phenomenological model consisting of a thermal disk component and a power-law. From this analysis, we identify the blackbody temperature, power-law index, the ratio of blackbody to power-law flux, the reduced $χ^2$, and the variance of the data as key parameters that likely contribute to the classification. We validate this inference by designing an additional neural network trained exclusively on this reduced parameter set, without using the spectral data directly. This parameter-based model achieves a classification accuracy comparable to that of the spectral models. Our results show that deep neural networks can not only classify compact objects in X-ray binaries with high accuracy but can also be interpreted in terms of physically meaningful spectral parameters derived from conventional X-ray spectral analysis. This framework offers a promising, mission-agnostic approach for compact object classification in current and future X-ray surveys.

Neutrino interactions play a central role in transport and flavor evolution in the ejecta of binary neutron star mergers. Simulations suggest that neutron star mergers may produce magnetic fields as strong as $10^{17}$ G, but computational difficulties have hampered the inclusion of magnetic field effects in neutrino interaction rates. In this paper we give approximate interaction rates for neutrinos in the presence of strong magnetic fields, including the effects of Landau quantization and anomalous magnetic moments with errors of order $\sqrt{T/M}$. We also comment on a neutrino production channel from individual neutrons that can produce low-energy $ν\barν$ pairs even at low density.

We present a study of the interstellar medium associated with the two middle-aged supernova remnants (SNRs) W41 and G22.7-0.2, both detected in TeV gamma-rays. Using high-angular-resolution $^{12}$CO($J$ = 1-0) data from the Nobeyama 45-m telescope and HI data from the VLA, we investigated the spatial and kinematic properties of molecular and atomic gas that interact with the SNRs. We identified associated clouds in the velocity ranges of +50-+80 km s$^{-1}$ for W41 and +76-+110 km s$^{-1}$ for G22.7-0.2. Column density analysis indicates that target protons are dominated by molecular hydrogen, while atomic hydrogen contributes less than $\sim$10-15% even after correction for self-absorption. The mean proton densities are $\sim$1.2$\times$10$^{3}$ cm$^{-3}$ for W41 and $\sim$5.3$\times$10$^{2}$ cm$^{-3}$ for G22.7-0.2. From the gamma-ray luminosities, we estimate the total energy of accelerated cosmic-ray protons as $W_\mathrm{p}$ $\sim$3$\times$10$^{47}$~erg for W41 and $\sim$1$\times$10$^{48}$ erg for G22.7-0.2, corresponding to 0.03-0.1% of the canonical supernova explosion energy. hese $W_\mathrm{p}$ values agree with the decreasing trend in $W_\mathrm{p}$ observed in the middle-aged SNRs within the previously reported SNR age-$W_\mathrm{p}$ relation.