arXiv Daily Digest - 2026-02-04

Diffusion Large Language Models (dLLMs) support arbitrary-order generation, yet their inference performance critically depends on the unmasking order. Existing strategies rely on heuristics that greedily optimize local confidence, offering limited guidance for identifying unmasking paths that are globally consistent and accurate. To bridge this gap, we introduce path log-likelihood (Path LL), a trajectory-conditioned objective that strongly correlates with downstream accuracy and enables principled selection of unmasking paths. To optimize Path LL at inference time, we propose POKE, an efficient value estimator that predicts the expected future Path LL of a partial decoding trajectory. We then integrate this lookahead signal into POKE-SMC, a Sequential Monte Carlo-based search framework for dynamically identifying optimal unmasking paths. Extensive experiments across 6 reasoning tasks show that POKE-SMC consistently improves accuracy, achieving 2%--3% average gains over strong decoding-time scaling baselines at comparable inference overhead on LLaDA models and advancing the accuracy--compute Pareto frontier.

Mixture of Experts (MoE) architectures significantly enhance the capacity of LLMs without proportional increases in computation, but at the cost of a vast parameter size. Offloading MoE expert parameters to host memory and leveraging both CPU and GPU computation has recently emerged as a promising direction to support such models on resourceconstrained local PC platforms. While promising, we notice that existing approaches mismatch the dynamic nature of expert workloads, which leads to three fundamental inefficiencies: (1) Static expert assignment causes severe CPUGPU load imbalance, underutilizing CPU and GPU resources; (2) Existing prefetching techniques fail to accurately predict high-workload experts, leading to costly inaccurate prefetches; (3) GPU cache policies neglect workload dynamics, resulting in poor hit rates and limited effectiveness. To address these challenges, we propose DALI, a workloaDAware offLoadIng framework for efficient MoE inference on local PCs. To fully utilize hardware resources, DALI first dynamically assigns experts to CPU or GPU by modeling assignment as a 0-1 integer optimization problem and solving it efficiently using a Greedy Assignment strategy at runtime. To improve prefetching accuracy, we develop a Residual-Based Prefetching method leveraging inter-layer residual information to accurately predict high-workload experts. Additionally, we introduce a Workload-Aware Cache Replacement policy that exploits temporal correlation in expert activations to improve GPU cache efficiency. By evaluating across various MoE models and settings, DALI achieves significant speedups in the both prefill and decoding phases over the state-of-the-art offloading frameworks.

Low-Rank Adaptation (LoRA) methods have emerged as crucial techniques for adapting large pre-trained models to downstream tasks under computational and memory constraints. However, they face a fundamental challenge in balancing task-specific performance gains against catastrophic forgetting of pre-trained knowledge, where existing methods provide inconsistent recommendations. This paper presents a comprehensive analysis of the performance-forgetting trade-offs inherent in low-rank adaptation using principal components as initialization. Our investigation reveals that fine-tuning intermediate components leads to better balance and show more robustness to high learning rates than first (PiSSA) and last (MiLoRA) components in existing work. Building on these findings, we provide a practical approach for initialization of LoRA that offers superior trade-offs. We demonstrate in a thorough empirical study on a variety of computer vision and NLP tasks that our approach improves accuracy and reduces forgetting, also in continual learning scenarios.

Reasoning over table images remains challenging for Large Vision-Language Models (LVLMs) due to complex layouts and tightly coupled structure-content information. Existing solutions often depend on expensive supervised training, reinforcement learning, or external tools, limiting efficiency and scalability. This work addresses a key question: how to adapt LVLMs to table reasoning with minimal annotation and no external tools? Specifically, we first introduce DiSCo, a Disentangled Structure-Content alignment framework that explicitly separates structural abstraction from semantic grounding during multimodal alignment, efficiently adapting LVLMs to tables structures. Building on DiSCo, we further present Table-GLS, a Global-to-Local Structure-guided reasoning framework that performs table reasoning via structured exploration and evidence-grounded inference. Extensive experiments across diverse benchmarks demonstrate that our framework efficiently enhances LVLM's table understanding and reasoning capabilities, particularly generalizing to unseen table structures.

Adaptive cognition requires structured internal models representing objects and their relations. Predictive neural networks are often proposed to form such "world models", yet their underlying mechanisms remain unclear. One hypothesis is that action-conditioned sequential prediction suffices for learning such world models. In this work, we investigate this possibility in a minimal in-silico setting. Sequentially sampling tokens from 2D continuous token scenes, a recurrent neural network is trained to predict the upcoming token from current input and a saccade-like displacement. On novel scenes, prediction accuracy improves across the sequence, indicating in-context learning. Decoding analyses reveal path integration and dynamic binding of token identity to position. Interventional analyses show that new bindings can be learned late in sequence and that out-of-distribution bindings can be learned. Together, these results demonstrate how structured representations that rely on flexible binding emerge to support prediction, offering a mechanistic account of sequential world modeling relevant to cognitive science.

Integrating logical knowledge into deep neural network training is still a hard challenge, especially for sequential or temporally extended domains involving subsymbolic observations. To address this problem, we propose DeepDFA, a neurosymbolic framework that integrates high-level temporal logic - expressed as Deterministic Finite Automata (DFA) or Moore Machines - into neural architectures. DeepDFA models temporal rules as continuous, differentiable layers, enabling symbolic knowledge injection into subsymbolic domains. We demonstrate how DeepDFA can be used in two key settings: (i) static image sequence classification, and (ii) policy learning in interactive non-Markovian environments. Across extensive experiments, DeepDFA outperforms traditional deep learning models (e.g., LSTMs, GRUs, Transformers) and novel neuro-symbolic systems, achieving state-of-the-art results in temporal knowledge integration. These results highlight the potential of DeepDFA to bridge subsymbolic learning and symbolic reasoning in sequential tasks.

Large Reasoning Models (LRMs) achieve strong performance by generating long reasoning traces with reflection. Through a large-scale empirical analysis, we find that a substantial fraction of reflective steps consist of self-verification (recheck) that repeatedly confirm intermediate results. These rechecks occur frequently across models and benchmarks, yet the vast majority are confirmatory rather than corrective, rarely identifying errors and altering reasoning outcomes. This reveals a mismatch between how often self-verification is activated and how often it is actually useful. Motivated by this, we propose a novel, experience-driven test-time framework that reduces the overused verification. Our method detects the activation of recheck behavior, consults an offline experience pool of past verification outcomes, and estimates whether a recheck is likely unnecessary via efficient retrieval. When historical experience suggests unnecessary, a suppression signal redirects the model to proceed. Across multiple model and benchmarks, our approach reduces token usage up to 20.3% while maintaining the accuracy, and in some datasets even yields accuracy improvements.

In this study, we investigate how language models develop preferences for \textit{idiomatic} as compared to \textit{linguistically acceptable} Swedish, both during pretraining and when adapting a model from English to Swedish. To do so, we train models on Swedish from scratch and by fine-tuning English-pretrained models, probing their preferences at various checkpoints using minimal pairs that differ in linguistic acceptability or idiomaticity. For linguistic acceptability, we adapt existing benchmarks into a minimal-pair format. To assess idiomaticity, we introduce two novel datasets: one contrasting conventionalized idioms with plausible variants, and another contrasting idiomatic Swedish with Translationese. Our findings suggest that idiomatic competence emerges more slowly than other linguistic abilities, including grammatical and lexical correctness. While longer training yields diminishing returns for most tasks, idiom-related performance continues to improve, particularly in the largest model tested (8B). However, instruction tuning on data machine-translated from English -- the common approach for languages with little or no native instruction data -- causes models to rapidly lose their preference for idiomatic language.

LLM routing aims to achieve a favorable quality--cost trade-off by dynamically assigning easy queries to smaller models and harder queries to stronger ones. However, across both unimodal and multimodal settings, we uncover a pervasive yet underexplored failure mode in existing routers: as the user's cost budget increases, routers systematically default to the most capable and most expensive model even when cheaper models already suffice. As a result, current routers under-utilize small models, wasting computation and monetary cost and undermining the core promise of routing; we term this phenomenon routing collapse. We attribute routing collapse to an objective--decision mismatch: many routers are trained to predict scalar performance scores, whereas routing decisions ultimately depend on discrete comparisons among candidate models. Consequently, small prediction errors can flip relative orderings and trigger suboptimal selections. To bridge this gap, we propose EquiRouter, a decision-aware router that directly learns model rankings, restoring the role of smaller models and mitigating routing collapse. On RouterBench, EquiRouter reduces cost by about 17\% at GPT-4-level performance compared to the strongest prior router. Our code is available at https://github.com/AIGNLAI/EquiRouter.

Single-cell RNA-seq profiles are high-dimensional, sparse, and unordered, causing autoregressive generation to impose an artificial ordering bias and suffer from error accumulation. To address this, we propose scDiVa, a masked discrete diffusion foundation model that aligns generation with the dropout-like corruption process by defining a continuous-time forward masking mechanism in token space. ScDiVa features a bidirectional denoiser that jointly models discrete gene identities and continuous values, utilizing entropy-normalized serialization and a latent anchor token to maximize information efficiency and preserve global cell identity. The model is trained via depth-invariant time sampling and a dual denoising objective to simulate varying sparsity levels while ensuring precise recovery of both identity and magnitude. Pre-trained on 59 million cells, scDiVa achieves strong transfer performance across major benchmarks, including batch integration, cell type annotation, and perturbation response prediction. These results suggest that masked discrete diffusion serves as a biologically coherent and effective alternative to autoregression.

Agreement is a foundational problem in distributed computing that have been studied extensively for over four decades. Recently, Meir, Mirault, Peleg and Robinson introduced the notion of \emph{Energy Efficient Agreement}, where the goal is to solve Agreement while minimizing the number of round a party participates in, thereby reducing the energy cost per participant. We show a recursive Agreement algorithm that has $O(\log f)$ active rounds per participant, where $f<n$ represents the maximum number of crash faults in the system.

Continual learning, especially class-incremental learning (CIL), on the basis of a pre-trained model (PTM) has garnered substantial research interest in recent years. However, how to effectively learn both discriminative and comprehensive feature representations while maintaining stability and plasticity over very long task sequences remains an open problem. We propose CaRE, a scalable {C}ontinual Le{a}rner with efficient Bi-Level {R}outing Mixture-of-{E}xperts (BR-MoE). The core idea of BR-MoE is a bi-level routing mechanism: a router selection stage that dynamically activates relevant task-specific routers, followed by an expert routing phase that dynamically activates and aggregates experts, aiming to inject discriminative and comprehensive representations into every intermediate network layer. On the other hand, we introduce a challenging evaluation protocol for comprehensively assessing CIL methods across very long task sequences spanning hundreds of tasks. Extensive experiments show that CaRE demonstrates leading performance across a variety of datasets and task settings, including commonly used CIL datasets with classical CIL settings (e.g., 5-20 tasks). To the best of our knowledge, CaRE is the first continual learner that scales to very long task sequences (ranging from 100 to over 300 non-overlapping tasks), while outperforming all baselines by a large margin on such task sequences. Code will be publicly released at https://github.com/LMMMEng/CaRE.git.

Static Analysis Tools (SATs) are central to security engineering activities, as they enable early identification of code weaknesses without requiring execution. However, their effectiveness is often limited by high false-positive rates and incomplete coverage of vulnerability classes. At the same time, developers frequently document security-related shortcuts and compromises as Self-Admitted Technical Debt (SATD) in software artifacts, such as code comments. While prior work has recognized SATD as a rich source of security information, it remains unclear whether -and in what ways- it is utilized during SAT-aided security analysis. OBJECTIVE: This work investigates the extent to which security-related SATD complements the output produced by SATs and helps bridge some of their well-known limitations. METHOD: We followed a mixed-methods approach consisting of (i) the analysis of a SATD-annotated vulnerability dataset using three state-of-the-art SATs and (ii) an online survey with 72 security practitioners. RESULTS: The combined use of all SATs flagged 114 of the 135 security-related SATD instances, spanning 24 distinct Common Weakness Enumeration (CWE) identifiers. A manual mapping of the SATD comments revealed 33 unique CWE types, 6 of which correspond to categories that SATs commonly overlook or struggle to detect (e.g., race conditions). Survey responses further suggest that developers frequently pair SAT outputs with SATD insights to better understand the impact and root causes of security weaknesses and to identify suitable fixes. IMPLICATIONS: Our findings show that such SATD-encoded information can be a meaningful complement to SAT-driven security analysis, while helping to overcome some of SATs' practical shortcomings.

Deep Research (DR) agents extend Large Language Models (LLMs) beyond parametric knowledge by autonomously retrieving and synthesizing evidence from large web corpora into long-form reports, enabling a long-horizon agentic paradigm. However, unlike real-time conversational assistants, DR is computationally expensive and time-consuming, creating an autonomy-interaction dilemma: high autonomy on ambiguous user queries often leads to prolonged execution with unsatisfactory outcomes. To address this, we propose IntentRL, a framework that trains proactive agents to clarify latent user intents before starting long-horizon research. To overcome the scarcity of open-ended research data, we introduce a scalable pipeline that expands a few seed samples into high-quality dialogue turns via a shallow-to-deep intent refinement graph. We further adopt a two-stage reinforcement learning (RL) strategy: Stage I applies RL on offline dialogues to efficiently learn general user-interaction behavior, while Stage II uses the trained agent and a user simulator for online rollouts to strengthen adaptation to diverse user feedback. Extensive experiments show that IntentRL significantly improves both intent hit rate and downstream task performance, outperforming the built-in clarify modules of closed-source DR agents and proactive LLM baselines.

Explanations are central to human cognition, yet AI systems often produce outputs that are difficult to understand. While symbolic AI offers a transparent foundation for interpretability, raw logical traces often impose a high extraneous cognitive load. We investigate how formal abstractions, specifically removal and clustering, impact human reasoning performance and cognitive effort. Utilizing Answer Set Programming (ASP) as a formal framework, we define a notion of irrelevant details to be abstracted over to obtain simplified explanations. Our cognitive experiments, in which participants classified stimuli across domains with explanations derived from an answer set program, show that clustering details significantly improve participants' understanding, while removal of details significantly reduce cognitive effort, supporting the hypothesis that abstraction enhances human-centered symbolic explanations.

Code generation has advanced rapidly with code-focused large language models (LLMs), especially on snippet-level tasks. However, application-level generation requires producing a runnable multi-file repository with correct structure, dependencies, and end-to-end executability, and real-world software must satisfy both functional correctness and non-functional quality (e.g., maintainability, security). Existing benchmarks provide a limited execution-based assessment of these requirements at the application level. We ask: Can current LLMs generate application-level repositories that meet both functional and non-functional criteria? We propose RAL-Bench, a benchmark and evaluation framework for application-level code generation. For each task, we distill a concise natural-language requirement from a high-quality reference project, build black-box system tests covering functional and non-functional attributes, and keep only tests that pass on the reference repository to ensure a sound oracle and an end-to-end executable suite. Functional correctness is measured by system-test pass rate. Non-functional quality is measured along five ISO/IEC 25010-inspired dimensions and aggregated with an Analytic Hierarchy Process (AHP)-derived weight vector, with per-dimension diagnostics and baseline-normalized scoring using reference measurements. Across 16 LLMs evaluated zero-shot with greedy decoding, functional correctness is the dominant bottleneck: no model exceeds a 45% functional pass rate under our requirement-driven, reference-validated tests. We release RAL-Bench at https://github.com/Wwstarry/RAL-Bench. .

Integrating hard constraints into deep learning is essential for safety-critical systems. Yet existing constructive layers that project predictions onto constraint boundaries face a fundamental bottleneck: gradient saturation. By collapsing exterior points onto lower-dimensional surfaces, standard orthogonal projections induce rank-deficient Jacobians, which nullify gradients orthogonal to active constraints and hinder optimization. We introduce Soft-Radial Projection, a differentiable reparameterization layer that circumvents this issue through a radial mapping from Euclidean space into the interior of the feasible set. This construction guarantees strict feasibility while preserving a full-rank Jacobian almost everywhere, thereby preventing the optimization stalls typical of boundary-based methods. We theoretically prove that the architecture retains the universal approximation property and empirically show improved convergence behavior and solution quality over state-of-the-art optimization- and projection-based baselines.

Estimating treatment effects in networks is challenging, as each potential outcome depends on the treatments of all other nodes in the network. To overcome this difficulty, existing methods typically impose an exposure mapping that compresses the treatment assignments in the network into a low-dimensional summary. However, if this mapping is misspecified, standard estimators for direct and spillover effects can be severely biased. We propose a novel partial identification framework for causal inference on networks to assess the robustness of treatment effects under misspecifications of the exposure mapping. Specifically, we derive sharp upper and lower bounds on direct and spillover effects under such misspecifications. As such, our framework presents a novel application of causal sensitivity analysis to exposure mappings. We instantiate our framework for three canonical exposure settings widely used in practice: (i) weighted means of the neighborhood treatments, (ii) threshold-based exposure mappings, and (iii) truncated neighborhood interference in the presence of higher-order spillovers. Furthermore, we develop orthogonal estimators for these bounds and prove that the resulting bound estimates are valid, sharp, and efficient. Our experiments show the bounds remain informative and provide reliable conclusions under misspecification of exposure mappings.

Coordinating the movement of multiple autonomous agents over a shared network is a fundamental challenge in algorithmic robotics, intelligent transportation, and distributed systems. The dominant approach, Multi-Agent Path Finding, relies on centralized control and synchronous collision avoidance, which often requires strict synchronization and guarantees of globally conflict-free execution. This paper introduces the Multi-Agent Routing under Crossing Cost model on mixed graphs, a novel framework tailored to asynchronous settings. In our model, instead of treating conflicts as hard constraints, each agent is assigned a path, and the system is evaluated through a cost function that measures potential head-on encounters. This ``crossing cost'', which is defined as the product of the numbers of agents traversing an edge in opposite directions, quantifies the risk of congestion and delay in decentralized execution. Our contributions are both game-theoretic and algorithmic. We model the setting as a congestion game with a non-standard cost function, prove the existence of pure Nash equilibria, and analyze the dynamics leading to them. Equilibria can be found in polynomial time under mild conditions, while the general case is PLS-complete. From an optimization perspective, minimizing the total crossing cost is NP-hard, as the problem generalizes Steiner Orientation. To address this hardness barrier, we design a suite of parameterized algorithms for minimizing crossing cost, with parameters including the number of arcs, edges, agents, and structural graph measures. These yield XP or FPT results depending on the parameter, offering algorithmic strategies for structurally restricted instances. Our framework provides a new theoretical foundation for decentralized multi-agent routing, bridging equilibrium analysis and parameterized complexity to support scalable and risk-aware coordination.

Reinforcement learning with verifiable rewards (RLVR) is effective for training large language models on deterministic outcome reasoning tasks. Prior work shows RLVR works with few prompts, but prompt selection is often based only on training-accuracy variance, leading to unstable optimization directions and weaker transfer. We revisit prompt selection from a mechanism-level view and argue that an effective minibatch should provide both (i) a reliable positive anchor and (ii) explicit negative learning signals from rare failures. Based on this principle, we propose \emph{positive--negative pairing}: at each update, we sample a hard-but-solvable $q^{+}$ and an easy-but-brittle prompt $q^{-}$(high success rate but not perfect), characterized by low and high empirical success rates under multiple rollouts. We further introduce Weighted GRPO, which reweights binary outcomes at the pair level and uses group-normalized advantages to amplify rare successes on $q^{+}$ into sharp positive guidance while turning rare failures on $q^{-}$ into strong negative penalties. This bidirectional signal provides informative learning feedback for both successes and failures, improving sample efficiency without suppressing exploration. On Qwen2.5-Math-7B, a single paired minibatch per update consistently outperforms a GRPO baseline that selects two prompts via commonly used variance-based selection heuristics: AIME~2025 Pass@8 improves from 16.8 to 22.2, and AMC23 Pass@64 from 94.0 to 97.0, while remaining competitive with large-scale RLVR trained from a pool of 1209 training prompts. Similar gains are observed on Qwen2.5-Math-7B-Instruct.

Score-based diffusion models are a recently developed framework for posterior sampling in Bayesian inverse problems with a state-of-the-art performance for severely ill-posed problems by leveraging a powerful prior distribution learned from empirical data. Despite generating significant interest especially in the machine-learning community, a thorough study of realistic inverse problems in the presence of modelling error and utilization of physical measurement data is still outstanding. In this work, the framework of unconditional representation for the conditional score function (UCoS) is evaluated for linearized difference imaging in diffuse optical tomography (DOT). DOT uses boundary measurements of near-infrared light to estimate the spatial distribution of absorption and scattering parameters in biological tissues. The problem is highly ill-posed and thus sensitive to noise and modelling errors. We introduce a novel regularization approach that prevents overfitting of the score function by constructing a mixed score composed of a learned and a model-based component. Validation of this approach is done using both simulated and experimental measurement data. The experiments demonstrate that a data-driven prior distribution results in posterior samples with low variance, compared to classical model-based estimation, and centred around the ground truth, even in the context of a highly ill-posed problem and in the presence of modelling errors.

Multi-subject image generation aims to synthesize images that faithfully preserve the identities of multiple reference subjects while following textual instructions. However, existing methods often suffer from identity inconsistency and limited compositional control, as they rely on diffusion models to implicitly associate text prompts with reference images. In this work, we propose Hierarchical Concept-to-Appearance Guidance (CAG), a framework that provides explicit, structured supervision from high-level concepts to fine-grained appearances. At the conceptual level, we introduce a VAE dropout training strategy that randomly omits reference VAE features, encouraging the model to rely more on robust semantic signals from a Visual Language Model (VLM) and thereby promoting consistent concept-level generation in the absence of complete appearance cues. At the appearance level, we integrate the VLM-derived correspondences into a correspondence-aware masked attention module within the Diffusion Transformer (DiT). This module restricts each text token to attend only to its matched reference regions, ensuring precise attribute binding and reliable multi-subject composition. Extensive experiments demonstrate that our method achieves state-of-the-art performance on the multi-subject image generation, substantially improving prompt following and subject consistency.

Lifelong learning is critical for embodied agents in open-world environments, where reinforcement learning fine-tuning has emerged as an important paradigm to enable Vision-Language-Action (VLA) models to master dexterous manipulation through environmental interaction. Thus, Continual Reinforcement Learning (CRL) is a promising pathway for deploying VLA models in lifelong robotic scenarios, yet balancing stability (retaining old skills) and plasticity (learning new ones) remains a formidable challenge for existing methods. We introduce CRL-VLA, a framework for continual post-training of VLA models with rigorous theoretical bounds. We derive a unified performance bound linking the stability-plasticity trade-off to goal-conditioned advantage magnitude, scaled by policy divergence. CRL-VLA resolves this dilemma via asymmetric regulation: constraining advantage magnitudes on prior tasks while enabling controlled growth on new tasks. This is realized through a simple but effective dual-critic architecture with novel Goal-Conditioned Value Formulation (GCVF), where a frozen critic anchors semantic consistency and a trainable estimator drives adaptation. Experiments on the LIBERO benchmark demonstrate that CRL-VLA effectively harmonizes these conflicting objectives, outperforming baselines in both anti-forgetting and forward adaptation.

Blockchain validators can reduce block processing time by exploiting multi-core CPUs, but deterministic execution must preserve a given total order while respecting transaction conflicts and per-block runtime limits. This paper systematically examines how validators can exploit multi-core parallelism during both block construction and execution without violating blockchain semantics. We formalize two validator-side optimization problems: (i) executing an already ordered block on \(p\) cores to minimize makespan while ensuring equivalence to sequential execution; and (ii) selecting and scheduling a subset of mempool transactions under a runtime limit \(B\) to maximize validator reward. For both, we develop exact Mixed-Integer Linear Programming (MILP) formulations that capture conflict, order, and capacity constraints, and propose fast deterministic heuristics that scale to realistic workloads. Using Ethereum mainnet traces and including a Solana-inspired declared-access baseline (Sol) for ordered-block scheduling and a simple reward-greedy baseline (RG) for block construction, we empirically quantify the trade-offs between optimality and runtime.

Frontier language models have demonstrated strong reasoning and long-horizon tool-use capabilities. However, existing RAG systems fail to leverage these capabilities. They still rely on two paradigms: (1) designing an algorithm that retrieves passages in a single shot and concatenates them into the model's input, or (2) predefining a workflow and prompting the model to execute it step-by-step. Neither paradigm allows the model to participate in retrieval decisions, preventing efficient scaling with model improvements. In this paper, we introduce A-RAG, an Agentic RAG framework that exposes hierarchical retrieval interfaces directly to the model. A-RAG provides three retrieval tools: keyword search, semantic search, and chunk read, enabling the agent to adaptively search and retrieve information across multiple granularities. Experiments on multiple open-domain QA benchmarks show that A-RAG consistently outperforms existing approaches with comparable or lower retrieved tokens, demonstrating that A-RAG effectively leverages model capabilities and dynamically adapts to different RAG tasks. We further systematically study how A-RAG scales with model size and test-time compute. We will release our code and evaluation suite to facilitate future research. Code and evaluation suite are available at https://github.com/Ayanami0730/arag.

We introduce ontology-to-tools compilation as a proof-of-principle mechanism for coupling large language models (LLMs) with formal domain knowledge. Within The World Avatar (TWA), ontological specifications are compiled into executable tool interfaces that LLM-based agents must use to create and modify knowledge graph instances, enforcing semantic constraints during generation rather than through post-hoc validation. Extending TWA's semantic agent composition framework, the Model Context Protocol (MCP) and associated agents are integral components of the knowledge graph ecosystem, enabling structured interaction between generative models, symbolic constraints, and external resources. An agent-based workflow translates ontologies into ontology-aware tools and iteratively applies them to extract, validate, and repair structured knowledge from unstructured scientific text. Using metal-organic polyhedra synthesis literature as an illustrative case, we show how executable ontological semantics can guide LLM behaviour and reduce manual schema and prompt engineering, establishing a general paradigm for embedding formal knowledge into generative systems.

The Non-equilibrium Green's function (NEGF) formalism is a particularly powerful method to simulate the quantum transport properties of nanoscale devices such as transistors, photo-diodes, or memory cells, in the ballistic limit of transport or in the presence of various scattering sources such as electronphonon, electron-photon, or even electron-electron interactions. The inclusion of all these mechanisms has been first demonstrated in small systems, composed of a few atoms, before being scaled up to larger structures made of thousands of atoms. Also, the accuracy of the models has kept improving, from empirical to fully ab-initio ones, e.g., density functional theory (DFT). This paper summarizes key (algorithmic) achievements that have allowed us to bring DFT+NEGF simulations closer to the dimensions and functionality of realistic systems. The possibility of leveraging graph neural networks and machine learning to speed up ab-initio device simulations is discussed as well.

To handle ambiguous and open-ended requests, Large Language Models (LLMs) are increasingly trained to interact with users to surface intents they have not yet expressed (e.g., ask clarification questions). However, users are often ambiguous because they have not yet formed their intents: they must observe and explore outcomes to discover what they want. Simply asking "what kind of tone do you want?" fails when users themselves do not know. We introduce DiscoverLLM, a novel and generalizable framework that trains LLMs to help users form and discover their intents. Central to our approach is a novel user simulator that models cognitive state with a hierarchy of intents that progressively concretize as the model surfaces relevant options -- where the degree of concretization serves as a reward signal that models can be trained to optimize. Resulting models learn to collaborate with users by adaptively diverging (i.e., explore options) when intents are unclear, and converging (i.e., refine and implement) when intents concretize. Across proposed interactive benchmarks in creative writing, technical writing, and SVG drawing, DiscoverLLM achieves over 10% higher task performance while reducing conversation length by up to 40%. In a user study with 75 human participants, DiscoverLLM improved conversation satisfaction and efficiency compared to baselines.

Emotional expression in human speech is nuanced and compositional, often involving multiple, sometimes conflicting, affective cues that may diverge from linguistic content. In contrast, most expressive text-to-speech systems enforce a single utterance-level emotion, collapsing affective diversity and suppressing mixed or text-emotion-misaligned expression. While activation steering via latent direction vectors offers a promising solution, it remains unclear whether emotion representations are linearly steerable in TTS, where steering should be applied within hybrid TTS architectures, and how such complex emotion behaviors should be evaluated. This paper presents the first systematic analysis of activation steering for emotional control in hybrid TTS models, introducing a quantitative, controllable steering framework, and multi-rater evaluation protocols that enable composable mixed-emotion synthesis and reliable text-emotion mismatch synthesis. Our results demonstrate, for the first time, that emotional prosody and expressive variability are primarily synthesized by the TTS language module instead of the flow-matching module, and also provide a lightweight steering approach for generating natural, human-like emotional speech.

Recent advances in large language models (LLMs) have enabled software engineering agents to tackle complex code modification tasks. Most existing approaches rely on execution feedback from containerized environments, which require dependency-complete setup and physical execution of programs and tests. While effective, this paradigm is resource-intensive and difficult to maintain, substantially complicating agent training and limiting scalability. We propose SWE-World, a Docker-free framework that replaces physical execution environments with a learned surrogate for training and evaluating software engineering agents. SWE-World leverages LLM-based models trained on real agent-environment interaction data to predict intermediate execution outcomes and final test feedback, enabling agents to learn without interacting with physical containerized environments. This design preserves the standard agent-environment interaction loop while eliminating the need for costly environment construction and maintenance during agent optimization and evaluation. Furthermore, because SWE-World can simulate the final evaluation outcomes of candidate trajectories without real submission, it enables selecting the best solution among multiple test-time attempts, thereby facilitating effective test-time scaling (TTS) in software engineering tasks. Experiments on SWE-bench Verified demonstrate that SWE-World raises Qwen2.5-Coder-32B from 6.2\% to 52.0\% via Docker-free SFT, 55.0\% with Docker-free RL, and 68.2\% with further TTS. The code is available at https://github.com/RUCAIBox/SWE-World

While LLMs exhibit remarkable fluency, their utility is often compromised by factual hallucinations and a lack of traceable provenance. Existing resources for grounding mitigate this but typically enforce a dichotomy: they offer either structured knowledge without textual context (e.g., knowledge bases) or grounded text with limited scale and linguistic coverage. To bridge this gap, we introduce FactNet, a massive, open-source resource designed to unify 1.7 billion atomic assertions with 3.01 billion auditable evidence pointers derived exclusively from 316 Wikipedia editions. Unlike recent synthetic approaches, FactNet employs a strictly deterministic construction pipeline, ensuring that every evidence unit is recoverable with byte-level precision. Extensive auditing confirms a high grounding precision of 92.1%, even in long-tail languages. Furthermore, we establish FactNet-Bench, a comprehensive evaluation suite for Knowledge Graph Completion, Question Answering, and Fact Checking. FactNet provides the community with a foundational, reproducible resource for training and evaluating trustworthy, verifiable multilingual systems.

The existence of adversarial examples is relatively understood for random fully connected neural networks, but much less so for convolutional neural networks (CNNs). The recent work [Daniely, 2025] establishes that adversarial examples can be found in CNNs, in some non-optimal distance from the input. We extend over this work and prove that adversarial examples in random CNNs with input dimension $d$ can be found already in $\ell_2$-distance of order $\lVert x \rVert /\sqrt{d}$ from the input $x$, which is essentially the nearest possible. We also show that such adversarial small perturbations can be found using a single step of gradient descent. To derive our results we use Fourier decomposition to efficiently bound the singular values of a random linear convolutional operator, which is the main ingredient of a CNN layer. This bound might be of independent interest.

Multimodal Large Language Models (MLLMs) have significantly advanced vision-language understanding. However, even state-of-the-art models struggle with geometric reasoning, revealing a critical bottleneck: the extreme scarcity of high-quality image-text pairs. Human annotation is prohibitively expensive, while automated methods fail to ensure fidelity and training effectiveness. Existing approaches either passively adapt to available images or employ inefficient random exploration with filtering, decoupling generation from learning needs. We propose Socratic-Geo, a fully autonomous framework that dynamically couples data synthesis with model learning through multi-agent interaction. The Teacher agent generates parameterized Python scripts with reflective feedback (Reflect for solvability, RePI for visual validity), ensuring image-text pair purity. The Solver agent optimizes reasoning through preference learning, with failure paths guiding Teacher's targeted augmentation. Independently, the Generator learns image generation capabilities on accumulated "image-code-instruction" triplets, distilling programmatic drawing intelligence into visual generation. Starting from only 108 seed problems, Socratic-Solver achieves 49.11 on six benchmarks using one-quarter of baseline data, surpassing strong baselines by 2.43 points. Socratic-Generator achieves 42.4% on GenExam, establishing new state-of-the-art for open-source models, surpassing Seedream-4.0 (39.8%) and approaching Gemini-2.5-Flash-Image (43.1%).

As large language model agents tackle increasingly complex long-horizon tasks, effective post-training becomes critical. Prior work faces fundamental challenges: outcome-only rewards fail to precisely attribute credit to intermediate steps, estimated step-level rewards introduce systematic noise, and Monte Carlo sampling approaches for step reward estimation incur prohibitive computational cost. Inspired by findings that only a small fraction of high-entropy tokens drive effective RL for reasoning, we propose Critical Step Optimization (CSO), which focuses preference learning on verified critical steps, decision points where alternate actions demonstrably flip task outcomes from failure to success. Crucially, our method starts from failed policy trajectories rather than expert demonstrations, directly targeting the policy model's weaknesses. We use a process reward model (PRM) to identify candidate critical steps, leverage expert models to propose high-quality alternatives, then continue execution from these alternatives using the policy model itself until task completion. Only alternatives that the policy successfully executes to correct outcomes are verified and used as DPO training data, ensuring both quality and policy reachability. This yields fine-grained, verifiable supervision at critical decisions while avoiding trajectory-level coarseness and step-level noise. Experiments on GAIA-Text-103 and XBench-DeepSearch show that CSO achieves 37% and 26% relative improvement over the SFT baseline and substantially outperforms other post-training methods, while requiring supervision at only 16% of trajectory steps. This demonstrates the effectiveness of selective verification-based learning for agent post-training.

In this technical report, we present SWE-Master, an open-source and fully reproducible post-training framework for building effective software engineering agents. SWE-Master systematically explores the complete agent development pipeline, including teacher-trajectory synthesis and data curation, long-horizon SFT, RL with real execution feedback, and inference framework design. Starting from an open-source base model with limited initial SWE capability, SWE-Master demonstrates how systematical optimization method can elicit strong long-horizon SWE task solving abilities. We evaluate SWE-Master on SWE-bench Verified, a standard benchmark for realistic software engineering tasks. Under identical experimental settings, our approach achieves a resolve rate of 61.4\% with Qwen2.5-Coder-32B, substantially outperforming existing open-source baselines. By further incorporating test-time scaling~(TTS) with LLM-based environment feedback, SWE-Master reaches 70.8\% at TTS@8, demonstrating a strong performance potential. SWE-Master provides a practical and transparent foundation for advancing reproducible research on software engineering agents. The code is available at https://github.com/RUCAIBox/SWE-Master.

Quantum generative models offer a novel approach to exploring high-dimensional Hilbert spaces but face significant challenges in scalability and expressibility when applied to multi-modal distributions. In this study, we explore a Hybrid Quantum-Classical U-Net architecture integrated with Adaptive Non-Local Observables (ANO) as a potential solution to these hurdles. By compressing classical data into a dense quantum latent space and utilizing trainable observables, our model aims to extract non-local features that complement classical processing. We also investigate the role of Skip Connections in preserving semantic information during the reverse diffusion process. Experimental results on the full MNIST dataset (digits 0-9) demonstrate that the proposed architecture is capable of generating structurally coherent and recognizable images for all digit classes. While hardware constraints still impose limitations on resolution, our findings suggest that hybrid architectures with adaptive measurements provide a feasible pathway for mitigating mode collapse and enhancing generative capabilities in the NISQ era.

In three-way conflict analysis, preference-based conflict situations characterize agents' attitudes towards issues by formally modeling their preferences over pairs of issues. However, existing preference-based conflict models rely exclusively on three qualitative relations, namely, preference, converse, and indifference, to describe agents' attitudes towards issue pairs, which significantly limits their capacity in capturing the essence of conflict. To overcome this limitation, we introduce the concept of an intuitionistic fuzzy preference-based conflict situation that captures agents' attitudes towards issue pairs with finer granularity than that afforded by classical preference-based models. Afterwards, we develop intuitionistic fuzzy preference-based conflict measures within this framework, and construct three-way conflict analysis models for trisecting the set of agent pairs, the agent set, and the issue set. Additionally, relative loss functions built on the proposed conflict functions are employed to calculate thresholds for three-way conflict analysis. Finally, we present adjustment mechanism-based feasible strategies that simultaneously account for both adjustment magnitudes and conflict degrees, together with an algorithm for constructing such feasible strategies, and provide an illustrative example to demonstrate the validity and effectiveness of the proposed model.

Vision language models (VLMs) extend the reasoning capabilities of large language models (LLMs) to cross-modal settings, yet remain highly vulnerable to multimodal jailbreak attacks. Existing defenses predominantly rely on safety fine-tuning or aggressive token manipulations, incurring substantial training costs or significantly degrading utility. Recent research shows that LLMs inherently recognize unsafe content in text, and the incorporation of visual inputs in VLMs frequently dilutes risk-related signals. Motivated by this, we propose Risk Awareness Injection (RAI), a lightweight and training-free framework for safety calibration that restores LLM-like risk recognition by amplifying unsafe signals in VLMs. Specifically, RAI constructs an Unsafe Prototype Subspace from language embeddings and performs targeted modulation on selected high-risk visual tokens, explicitly activating safety-critical signals within the cross-modal feature space. This modulation restores the model's LLM-like ability to detect unsafe content from visual inputs, while preserving the semantic integrity of original tokens for cross-modal reasoning. Extensive experiments across multiple jailbreak and utility benchmarks demonstrate that RAI substantially reduces attack success rate without compromising task performance.

Code summaries are essential for helping developers understand code functionality and reducing maintenance and collaboration costs. Although recent advances in large language models (LLMs) have significantly improved automatic code summarization, the practical usefulness of generated summaries in industrial settings remains insufficiently explored. In collaboration with documentation experts from the industrial HarmonyOS project, we conducted a questionnaire study showing that over 57.4% of code summaries produced by state-of-the-art approaches were rejected due to violations of developers' expectations for industrial documentation. Beyond semantic similarity to reference summaries, developers emphasize additional requirements, including the use of appropriate domain terminology, explicit function categorization, and the avoidance of redundant implementation details. To address these expectations, we propose ExpSum, an expectation-aware code summarization approach that integrates function metadata abstraction, informative metadata filtering, context-aware domain knowledge retrieval, and constraint-driven prompting to guide LLMs in generating structured, expectation-aligned summaries. We evaluate ExpSum on the HarmonyOS project and widely used code summarization benchmarks. Experimental results show that ExpSum consistently outperforms all baselines, achieving improvements of up to 26.71% in BLEU-4 and 20.10% in ROUGE-L on HarmonyOS. Furthermore, LLM-based evaluations indicate that ExpSum-generated summaries better align with developer expectations across other projects, demonstrating its effectiveness for industrial code documentation.

Proprietary large language models (LLMs) embody substantial economic value and are generally exposed only as black-box APIs, yet adversaries can still exploit their outputs to extract knowledge via distillation. Existing defenses focus exclusively on text-based distillation, leaving the important logit-based distillation largely unexplored. In this work, we analyze this problem and present an effective solution from an information-theoretic perspective. We characterize distillation-relevant information in teacher outputs using the conditional mutual information (CMI) between teacher logits and input queries conditioned on ground-truth labels. This quantity captures contextual information beneficial for model extraction, motivating us to defend distillation via CMI minimization. Guided by our theoretical analysis, we propose learning a transformation matrix that purifies the original outputs to enhance distillation resistance. We further derive a CMI-inspired anti-distillation objective to optimize this transformation, which effectively removes distillation-relevant information while preserving output utility. Extensive experiments across multiple LLMs and strong distillation algorithms demonstrate that the proposed method significantly degrades distillation performance while preserving task accuracy, effectively protecting models' intellectual property.

While deep learning has revolutionized financial forecasting through sophisticated architectures, the design of the supervision signal itself is rarely scrutinized. We challenge the canonical assumption that training labels must strictly mirror inference targets, uncovering the Label Horizon Paradox: the optimal supervision signal often deviates from the prediction goal, shifting across intermediate horizons governed by market dynamics. We theoretically ground this phenomenon in a dynamic signal-noise trade-off, demonstrating that generalization hinges on the competition between marginal signal realization and noise accumulation. To operationalize this insight, we propose a bi-level optimization framework that autonomously identifies the optimal proxy label within a single training run. Extensive experiments on large-scale financial datasets demonstrate consistent improvements over conventional baselines, thereby opening new avenues for label-centric research in financial forecasting.

Deep neural networks (DNNs) often produce overconfident out-of-distribution predictions, motivating Bayesian uncertainty quantification. The Linearized Laplace Approximation (LLA) achieves this by linearizing the DNN and applying Laplace inference to the resulting model. Importantly, the linear model is also used for prediction. We argue this linearization in the posterior may degrade fidelity to the true Laplace approximation. To alleviate this problem, without increasing significantly the computational cost, we propose the Quadratic Laplace Approximation (QLA). QLA approximates each second order factor in the approximate Laplace log-posterior using a rank-one factor obtained via efficient power iterations. QLA is expected to yield a posterior precision closer to that of the full Laplace without forming the full Hessian, which is typically intractable. For prediction, QLA also uses the linearized model. Empirically, QLA yields modest yet consistent uncertainty estimation improvements over LLA on five regression datasets.

Entropy serves as a critical metric for measuring the diversity of outputs generated by large language models (LLMs), providing valuable insights into their exploration capabilities. While recent studies increasingly focus on monitoring and adjusting entropy to better balance exploration and exploitation in reinforcement fine-tuning (RFT), a principled understanding of entropy dynamics during this process is yet to be thoroughly investigated. In this paper, we establish a theoretical framework for analyzing the entropy dynamics during the RFT process, which begins with a discriminant expression that quantifies entropy change under a single logit update. This foundation enables the derivation of a first-order expression for entropy change, which can be further extended to the update formula of Group Relative Policy Optimization (GRPO). The corollaries and insights drawn from the theoretical analysis inspire the design of entropy control methods, and also offer a unified lens for interpreting various entropy-based methods in existing studies. We provide empirical evidence to support the main conclusions of our analysis and demonstrate the effectiveness of the derived entropy-discriminator clipping methods. This study yields novel insights into RFT training dynamics, providing theoretical support and practical strategies for optimizing the exploration-exploitation balance during LLM fine-tuning.

Unsupervised object-centric learning models, particularly slot-based architectures, have shown great promise in decomposing complex scenes. However, their reliance on reconstruction-based training creates a fundamental conflict between the sharp, high-frequency attention maps of the encoder and the spatially consistent but blurry reconstruction maps of the decoder. We identify that this discrepancy gives rise to a vicious cycle: the noisy feature map from the encoder forces the decoder to average over possibilities and produce even blurrier outputs, while the gradient computed from blurry reconstruction maps lacks high-frequency details necessary to supervise encoder features. To break this cycle, we introduce Synergistic Representation Learning (SRL) that establishes a virtuous cycle where the encoder and decoder mutually refine one another. SRL leverages the encoder's sharpness to deblur the semantic boundary within the decoder output, while exploiting the decoder's spatial consistency to denoise the encoder's features. This mutual refinement process is stabilized by a warm-up phase with a slot regularization objective that initially allocates distinct entities per slot. By bridging the representational gap between the encoder and decoder, SRL achieves state-of-the-art results on video object-centric learning benchmarks. Codes are available at https://github.com/hynnsk/SRL.

Offline goal-conditioned reinforcement learning remains challenging for long-horizon tasks. While hierarchical approaches mitigate this issue by decomposing tasks, most existing methods rely on separate high- and low-level networks and generate only a single intermediate subgoal, making them inadequate for complex tasks that require coordinating multiple intermediate decisions. To address this limitation, we draw inspiration from the chain-of-thought paradigm and propose the Chain-of-Goals Hierarchical Policy (CoGHP), a novel framework that reformulates hierarchical decision-making as autoregressive sequence modeling within a unified architecture. Given a state and a final goal, CoGHP autoregressively generates a sequence of latent subgoals followed by the primitive action, where each latent subgoal acts as a reasoning step that conditions subsequent predictions. To implement this efficiently, we pioneer the use of an MLP-Mixer backbone, which supports cross-token communication and captures structural relationships among state, goal, latent subgoals, and action. Across challenging navigation and manipulation benchmarks, CoGHP consistently outperforms strong offline baselines, demonstrating improved performance on long-horizon tasks.

Emerging network paradigms and applications increasingly rely on federated learning (FL) to enable collaborative intelligence while preserving privacy. However, the sustainability of such collaborative environments hinges on a fair and stable payoff allocation mechanism. Focusing on coalition stability, this paper introduces a payoff allocation framework based on the least core (LC) concept. Unlike traditional methods, the LC prioritizes the cohesion of the federation by minimizing the maximum dissatisfaction among all potential subgroups, ensuring that no participant has an incentive to break away. To adapt this game-theoretic concept to practical, large-scale networks, we propose a streamlined implementation with a stack-based pruning algorithm, effectively balancing computational efficiency with allocation precision. Case studies in federated intrusion detection demonstrate that our mechanism correctly identifies pivotal contributors and strategic alliances. The results confirm that the practical LC framework promotes stable collaboration and fosters a sustainable FL ecosystem.

Proximal Policy Optimization (PPO) is among the most widely used deep reinforcement learning algorithms, yet its theoretical foundations remain incomplete. Most importantly, convergence and understanding of fundamental PPO advantages remain widely open. Under standard theory assumptions we show how PPO's policy update scheme (performing multiple epochs of minibatch updates on multi-use rollouts with a surrogate gradient) can be interpreted as approximated policy gradient ascent. We show how to control the bias accumulated by the surrogate gradients and use techniques from random reshuffling to prove a convergence theorem for PPO that sheds light on PPO's success. Additionally, we identify a previously overlooked issue in truncated Generalized Advantage Estimation commonly used in PPO. The geometric weighting scheme induces infinite mass collapse onto the longest $k$-step advantage estimator at episode boundaries. Empirical evaluations show that a simple weight correction can yield substantial improvements in environments with strong terminal signal, such as Lunar Lander.

Decentralized learning (DL) enables a set of nodes to train a model collaboratively without central coordination, offering benefits for privacy and scalability. However, DL struggles to train a high accuracy model when the data distribution is non-independent and identically distributed (non-IID) and when the communication topology is static. To address these issues, we propose Morph, a topology optimization algorithm for DL. In Morph, nodes adaptively choose peers for model exchange based on maximum model dissimilarity. Morph maintains a fixed in-degree while dynamically reshaping the communication graph through gossip-based peer discovery and diversity-driven neighbor selection, thereby improving robustness to data heterogeneity. Experiments on CIFAR-10 and FEMNIST with up to 100 nodes show that Morph consistently outperforms static and epidemic baselines, while closely tracking the fully connected upper bound. On CIFAR-10, Morph achieves a relative improvement of 1.12x in test accuracy compared to the state-of-the-art baselines. On FEMNIST, Morph achieves an accuracy that is 1.08x higher than Epidemic Learning. Similar trends hold for 50 node deployments, where Morph narrows the gap to the fully connected upper bound within 0.5 percentage points on CIFAR-10. These results demonstrate that Morph achieves higher final accuracy, faster convergence, and more stable learning as quantified by lower inter-node variance, while requiring fewer communication rounds than baselines and no global knowledge.

Machine unlearning aims to remove specific content from trained models while preserving overall performance. However, the phenomenon of benign relearning, in which forgotten information reemerges even from benign fine-tuning data, reveals that existing unlearning methods remain fundamentally fragile. A common explanation attributes this effect to topical relevance, but we find this account insufficient. Through systematic analysis, we demonstrate that syntactic similarity, rather than topicality, is the primary driver: across benchmarks, syntactically similar data consistently trigger recovery even without topical overlap, due to their alignment in representations and gradients with the forgotten content. Motivated by this insight, we introduce syntactic diversification, which paraphrases the original forget queries into heterogeneous structures prior to unlearning. This approach effectively suppresses benign relearning, accelerates forgetting, and substantially alleviates the trade-off between unlearning efficacy and model utility.

Focal cortical dysplasia (FCD) lesions in epilepsy FLAIR MRI are subtle and scarce, making joint image--mask generative modeling prone to instability and memorization. We propose SLIM-Diff, a compact joint diffusion model whose main contributions are (i) a single shared-bottleneck U-Net that enforces tight coupling between anatomy and lesion geometry from a 2-channel image+mask representation, and (ii) loss-geometry tuning via a tunable $L_p$ objective. As an internal baseline, we include the canonical DDPM-style objective ($ε$-prediction with $L_2$ loss) and isolate the effect of prediction parameterization and $L_p$ geometry under a matched setup. Experiments show that $x_0$-prediction is consistently the strongest choice for joint synthesis, and that fractional sub-quadratic penalties ($L_{1.5}$) improve image fidelity while $L_2$ better preserves lesion mask morphology. Our code and model weights are available in https://github.com/MarioPasc/slim-diff

Handwritten Mathematical Expression Recognition (HMER) requires reasoning over diverse symbols and 2D structural layouts, yet autoregressive models struggle with exposure bias and syntactic inconsistency. We present a discrete diffusion framework that reformulates HMER as iterative symbolic refinement instead of sequential generation. Through multi-step remasking, the proposal progressively refines both symbols and structural relations, removing causal dependencies and improving structural consistency. A symbol-aware tokenization and Random-Masking Mutual Learning further enhance syntactic alignment and robustness to handwriting diversity. On the MathWriting benchmark, the proposal achieves 5.56\% CER and 60.42\% EM, outperforming strong Transformer and commercial baselines. Consistent gains on CROHME 2014--2023 demonstrate that discrete diffusion provides a new paradigm for structure-aware visual recognition beyond generative modeling.

While retrieval augmented generation (RAG) has been swiftly adopted in industrial applications based on large language models (LLMs), there is no consensus on what are the best practices for building a RAG system in terms of what are the components, how to organize these components and how to implement each component for the industrial applications, especially in the medical domain. In this work, we first carefully analyze each component of the RAG system and propose practical alternatives for each component. Then, we conduct systematic evaluations on three types of tasks, revealing the best practices for improving the RAG system and how LLM-based RAG systems make trade-offs between performance and efficiency.

Scaling Large Language Models (LLMs) typically relies on increasing the number of parameters or test-time computations to boost performance. However, these strategies are impractical for edge device deployment due to limited RAM and NPU resources. Despite hardware constraints, deploying performant LLM on edge devices such as smartphone remains crucial for user experience. To address this, we propose MeKi (Memory-based Expert Knowledge Injection), a novel system that scales LLM capacity via storage space rather than FLOPs. MeKi equips each Transformer layer with token-level memory experts that injects pre-stored semantic knowledge into the generation process. To bridge the gap between training capacity and inference efficiency, we employ a re-parameterization strategy to fold parameter matrices used during training into a compact static lookup table. By offloading the knowledge to ROM, MeKi decouples model capacity from computational cost, introducing zero inference latency overhead. Extensive experiments demonstrate that MeKi significantly outperforms dense LLM baselines with identical inference speed, validating the effectiveness of memory-based scaling paradigm for on-device LLMs. Project homepage is at https://github.com/ningding-o/MeKi.

Finding effective prompts for language models (LMs) is critical yet notoriously difficult: the prompt space is combinatorially large, rewards are sparse due to expensive target-LM evaluation. Yet, existing RL-based prompt optimizers often rely on on-policy updates and a meta-prompt sampled from a fixed distribution, leading to poor sample efficiency. We propose GFlowPO, a probabilistic prompt optimization framework that casts prompt search as a posterior inference problem over latent prompts regularized by a meta-prompted reference-LM prior. In the first step, we fine-tune a lightweight prompt-LM with an off-policy Generative Flow Network (GFlowNet) objective, using a replay-based training policy that reuses past prompt evaluations to enable sample-efficient exploration. In the second step, we introduce Dynamic Memory Update (DMU), a training-free mechanism that updates the meta-prompt by injecting both (i) diverse prompts from a replay buffer and (ii) top-performing prompts from a small priority queue, thereby progressively concentrating the search process on high-reward regions. Across few-shot text classification, instruction induction benchmarks, and question answering tasks, GFlowPO consistently outperforms recent discrete prompt optimization baselines.

This paper proposes FedNMap, a normal map-based method for composite federated learning, where the objective consists of a smooth loss and a possibly nonsmooth regularizer. FedNMap leverages a normal map-based update scheme to handle the nonsmooth term and incorporates a local correction strategy to mitigate the impact of data heterogeneity across clients. Under standard assumptions, including smooth local losses, weak convexity of the regularizer, and bounded stochastic gradient variance, FedNMap achieves linear speedup with respect to both the number of clients $n$ and the number of local updates $Q$ for nonconvex losses, both with and without the Polyak-Łojasiewicz (PL) condition. To our knowledge, this is the first result establishing linear speedup for nonconvex composite federated learning.

Audio is a fundamental modality for analyzing speech, music, and environmental sounds. Although pretrained audio models have significantly advanced audio understanding, they remain fragile in real-world settings where data distributions shift over time. In this work, we present the first systematic benchmark for audio continual learning (CL) with pretrained models (PTMs), together with a comprehensive analysis of its unique challenges. Unlike in vision, where parameter-efficient fine-tuning (PEFT) has proven effective for CL, directly transferring such strategies to audio leads to poor performance. This stems from a fundamental property of audio backbones: they focus on low-level spectral details rather than structured semantics, causing severe upstream-downstream misalignment. Through extensive empirical study, we identify analytic classifiers with first-session adaptation (FSA) as a promising direction, but also reveal two major limitations: representation saturation in coarse-grained scenarios and representation drift in fine-grained scenarios. To address these challenges, we propose PACE, a novel method that enhances FSA via a regularized analytic classifier and enables multi-session adaptation through adaptive subspace-orthogonal PEFT for improved semantic alignment. In addition, we introduce spectrogram-based boundary-aware perturbations to mitigate representation overlap and improve stability. Experiments on six diverse audio CL benchmarks demonstrate that PACE substantially outperforms state-of-the-art baselines, marking an important step toward robust and scalable audio continual learning with PTMs.

This paper introduces a new framework for recovering causal graphs from observational data, leveraging the observation that the distribution of an effect, conditioned on its causes, remains invariant to changes in the prior distribution of those causes. This insight enables a direct test for potential causal relationships by checking the variance of their corresponding effect-cause conditional distributions across multiple downsampled subsets of the data. These subsets are selected to reflect different prior cause distributions, while preserving the effect-cause conditional relationships. Using this invariance test and exploiting an (empirical) sparsity of most causal graphs, we develop an algorithm that efficiently uncovers causal relationships with quadratic complexity in the number of observational variables, reducing the processing time by up to 25x compared to state-of-the-art methods. Our empirical experiments on a varied benchmark of large-scale datasets show superior or equivalent performance compared to existing works, while achieving enhanced scalability.

Reinforcement learning (RL) has shown strong promise for LLM-based machine translation, with recent methods such as GRPO demonstrating notable gains; nevertheless, translation-oriented RL remains challenged by noisy learning signals arising from Monte Carlo return estimation, as well as a large trajectory space that favors global exploration over fine-grained local optimization. We introduce \textbf{PEGRL}, a \textit{two-stage} RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. At each iteration, translation outputs are sampled to construct post-editing inputs, allowing return estimation in the post-editing stage to benefit from conditioning on the current translation behavior, while jointly supporting both global exploration and fine-grained local optimization. A task-specific weighting scheme further balances the contributions of translation and post-editing objectives, yielding a biased yet more sample-efficient estimator. Experiments on English$\to$Finnish, English$\to$Turkish, and English$\leftrightarrow$Chinese show consistent gains over RL baselines, and for English$\to$Turkish, performance on COMET-KIWI is comparable to advanced LLM-based systems (DeepSeek-V3.2).

We build a custom transformer model to study how neural networks make moral decisions on trolley-style dilemmas. The model processes structured scenarios using embeddings that encode who is affected, how many people, and which outcome they belong to. Our 2-layer architecture achieves 77% accuracy on Moral Machine data while remaining small enough for detailed analysis. We use different interpretability techniques to uncover how moral reasoning distributes across the network, demonstrating that biases localize to distinct computational stages among other findings.

Robustness is often regarded as a critical future challenge for real-world applications, where stability is essential. However, as models often learn tasks in a similar order, we hypothesize that easier tasks will be easier regardless of how they are presented to the model. Indeed, in this paper, we show that as models approach high performance on a task, robustness is effectively achieved. Through an empirical analysis of multiple models across diverse datasets and configurations (e.g., paraphrases, different temperatures), we find a strong positive correlation. Moreover, we find that robustness is primarily driven by task-specific competence rather than inherent model-level properties, challenging current approaches that treat robustness as an independent capability. Thus, from a high-level perspective, we may expect that as new tasks saturate, model robustness on these tasks will emerge accordingly. For researchers, this implies that explicit efforts to measure and improve robustness may warrant reduced emphasis, as such robustness is likely to develop alongside performance gains. For practitioners, it acts as a sign that indeed the tasks that the literature deals with are unreliable, but on easier past tasks, the models are reliable and ready for real-world deployment.

Text-conditioned diffusion models have advanced image and video super-resolution by using prompts as semantic priors, but modern super-resolution pipelines typically rely on latent tiling to scale to high resolutions, where a single global caption causes prompt underspecification. A coarse global prompt often misses localized details (prompt sparsity) and provides locally irrelevant guidance (prompt misguidance) that can be amplified by classifier-free guidance. We propose Tiled Prompts, a unified framework for image and video super-resolution that generates a tile-specific prompt for each latent tile and performs super-resolution under locally text-conditioned posteriors, providing high-information guidance that resolves prompt underspecification with minimal overhead. Experiments on high resolution real-world images and videos show consistent gains in perceptual quality and text alignment, while reducing hallucinations and tile-level artifacts relative to global-prompt baselines.

Mental health disorders represent a burgeoning global public health challenge. While Large Language Models (LLMs) have demonstrated potential in psychiatric assessment, their clinical utility is severely constrained by benchmarks that lack ecological validity and fine-grained diagnostic supervision. To bridge this gap, we introduce \textbf{MentalDx Bench}, the first benchmark dedicated to disorder-level psychiatric diagnosis within real-world clinical settings. Comprising 712 de-identified electronic health records annotated by board-certified psychiatrists under ICD-11 guidelines, the benchmark covers 76 disorders across 16 diagnostic categories. Evaluation of 18 LLMs reveals a critical \textit{paradigm misalignment}: strong performance at coarse diagnostic categorization contrasts with systematic failure at disorder-level diagnosis, underscoring a gap between pattern-based modeling and clinical hypothetico-deductive reasoning. In response, we propose \textbf{MentalSeek-Dx}, a medical-specialized LLM trained to internalize this clinical reasoning process through supervised trajectory construction and curriculum-based reinforcement learning. Experiments on MentalDx Bench demonstrate that MentalSeek-Dx achieves state-of-the-art (SOTA) performance with only 14B parameters, establishing a clinically grounded framework for reliable psychiatric diagnosis.

Proactive interventions by LLM critic models are often assumed to improve reliability, yet their effects at deployment time are poorly understood. We show that a binary LLM critic with strong offline accuracy (AUROC 0.94) can nevertheless cause severe performance degradation, inducing a 26 percentage point (pp) collapse on one model while affecting another by near zero pp. This variability demonstrates that LLM critic accuracy alone is insufficient to determine whether intervention is safe. We identify a disruption-recovery tradeoff: interventions may recover failing trajectories but also disrupt trajectories that would have succeeded. Based on this insight, we propose a pre-deployment test that uses a small pilot of 50 tasks to estimate whether intervention is likely to help or harm, without requiring full deployment. Across benchmarks, the test correctly anticipates outcomes: intervention degrades performance on high-success tasks (0 to -26 pp), while yielding a modest improvement on the high-failure ALFWorld benchmark (+2.8 pp, p=0.014). The primary value of our framework is therefore identifying when not to intervene, preventing severe regressions before deployment.

Bayesian posterior predictive densities as non-conformity scores and Bayesian quadrature are used to estimate and minimise the expected prediction set size. Operating within a split conformal framework, BCP provides valid coverage guarantees and demonstrates reliable empirical coverage under model misspecification. Across regression and classification tasks, including distribution-shifted settings such as ImageNet-A, BCP yields prediction sets of comparable size to split conformal prediction, while exhibiting substantially lower run-to-run variability in set size. In sparse regression with nominal coverage of 80 percent, BCP achieves 81 percent empirical coverage under a misspecified prior, whereas Bayesian credible intervals under-cover at 49 percent.

Standard federated learning algorithms are vulnerable to adversarial nodes, a.k.a. Byzantine failures. To solve this issue, robust distributed learning algorithms have been developed, which typically replace parameter averaging by robust aggregations. While generic conditions on these aggregations exist to guarantee the convergence of (Stochastic) Gradient Descent (SGD), the analyses remain rather ad-hoc. This hinders the development of more complex robust algorithms, such as accelerated ones. In this work, we show that Byzantine-robust distributed optimization can, under standard generic assumptions, be cast as a general optimization with inexact gradient oracles (with both additive and multiplicative error terms), an active field of research. This allows for instance to directly show that GD on top of standard robust aggregation procedures obtains optimal asymptotic error in the Byzantine setting. Going further, we propose two optimization schemes to speed up the convergence. The first one is a Nesterov-type accelerated scheme whose proof directly derives from accelerated inexact gradient results applied to our formulation. The second one hinges on Optimization under Similarity, in which the server leverages an auxiliary loss function that approximates the global loss. Both approaches allow to drastically reduce the communication complexity compared to previous methods, as we show theoretically and empirically.

This paper proposes a machine learning-based framework for asset selection and portfolio construction, termed the Best-Path Algorithm Sparse Graphical Model (BPASGM). The method extends the Best-Path Algorithm (BPA) by mapping linear and non-linear dependencies among a large set of financial assets into a sparse graphical model satisfying a structural Markov property. Based on this representation, BPASGM performs a dependence-driven screening that removes positively or redundantly connected assets, isolating subsets that are conditionally independent or negatively correlated. This step is designed to enhance diversification and reduce estimation error in high-dimensional portfolio settings. Portfolio optimization is then conducted on the selected subset using standard mean-variance techniques. BPASGM does not aim to improve the theoretical mean-variance optimum under known population parameters, but rather to enhance realized performance in finite samples, where sample-based Markowitz portfolios are highly sensitive to estimation error. Monte Carlo simulations show that BPASGM-based portfolios achieve more stable risk-return profiles, lower realized volatility, and superior risk-adjusted performance compared to standard mean-variance portfolios. Empirical results for U.S. equities, global stock indices, and foreign exchange rates over 1990-2025 confirm these findings and demonstrate a substantial reduction in portfolio cardinality. Overall, BPASGM offers a statistically grounded and computationally efficient framework that integrates sparse graphical modeling with portfolio theory for dependence-aware asset selection.

Medical image segmentation is evolving from task-specific models toward generalizable frameworks. Recent research leverages Multi-modal Large Language Models (MLLMs) as autonomous agents, employing reinforcement learning with verifiable reward (RLVR) to orchestrate specialized tools like the Segment Anything Model (SAM). However, these approaches often rely on single-turn, rigid interaction strategies and lack process-level supervision during training, which hinders their ability to fully exploit the dynamic potential of interactive tools and leads to redundant actions. To bridge this gap, we propose MedSAM-Agent, a framework that reformulates interactive segmentation as a multi-step autonomous decision-making process. First, we introduce a hybrid prompting strategy for expert-curated trajectory generation, enabling the model to internalize human-like decision heuristics and adaptive refinement strategies. Furthermore, we develop a two-stage training pipeline that integrates multi-turn, end-to-end outcome verification with a clinical-fidelity process reward design to promote interaction parsimony and decision efficiency. Extensive experiments across 6 medical modalities and 21 datasets demonstrate that MedSAM-Agent achieves state-of-the-art performance, effectively unifying autonomous medical reasoning with robust, iterative optimization. Code is available \href{https://github.com/CUHK-AIM-Group/MedSAM-Agent}{here}.

Target-oriented discovery under limited evaluation budgets requires making reliable progress in high-dimensional, heterogeneous design spaces where each new measurement is costly, whether experimental or high-fidelity simulation. We present an information-theoretic framework for target-oriented adaptive sampling that reframes optimization as trajectory discovery: instead of approximating the full response surface, the method maintains and refines a low-entropy information state that concentrates search on target-relevant directions. The approach couples data, model beliefs, and physics/structure priors through dimension-aware information budgeting, adaptive bootstrapped distillation over a heterogeneous surrogate reservoir, and structure-aware candidate manifold analysis with Kalman-inspired multi-model fusion to balance consensus-driven exploitation and disagreement-driven exploration. Evaluated under a single unified protocol without dataset-specific tuning, the framework improves sample efficiency and reliability across 14 single- and multi-objective materials design tasks spanning candidate pools from $600$ to $4 \times 10^6$ and feature dimensions from $10$ to $10^3$, typically reaching top-performing regions within 100 evaluations. Complementary 20-dimensional synthetic benchmarks (Ackley, Rastrigin, Schwefel) further demonstrate robustness to rugged and multimodal landscapes.

Operations Research (OR) relies on expert-driven modeling-a slow and fragile process ill-suited to novel scenarios. While large language models (LLMs) can automatically translate natural language into optimization models, existing approaches either rely on costly post-training or employ multi-agent frameworks, yet most still lack reliable collaborative error correction and task-specific retrieval, often leading to incorrect outputs. We propose MIRROR, a fine-tuning-free, end-to-end multi-agent framework that directly translates natural language optimization problems into mathematical models and solver code. MIRROR integrates two core mechanisms: (1) execution-driven iterative adaptive revision for automatic error correction, and (2) hierarchical retrieval to fetch relevant modeling and coding exemplars from a carefully curated exemplar library. Experiments show that MIRROR outperforms existing methods on standard OR benchmarks, with notable results on complex industrial datasets such as IndustryOR and Mamo-ComplexLP. By combining precise external knowledge infusion with systematic error correction, MIRROR provides non-expert users with an efficient and reliable OR modeling solution, overcoming the fundamental limitations of general-purpose LLMs in expert optimization tasks.

Quantitative imaging methods, such as magnetic resonance fingerprinting (MRF), aim to extract interpretable pathology biomarkers by estimating biophysical tissue parameters from signal evolutions. However, the pattern-matching algorithms or neural networks used in such inverse problems often lack principled uncertainty quantification, which limits the trustworthiness and transparency, required for clinical acceptance. Here, we describe a physics-structured variational autoencoder (PS-VAE) designed for rapid extraction of voxelwise multi-parameter posterior distributions. Our approach integrates a differentiable spin physics simulator with self-supervised learning, and provides a full covariance that captures the inter-parameter correlations of the latent biophysical space. The method was validated in a multi-proton pool chemical exchange saturation transfer (CEST) and semisolid magnetization transfer (MT) molecular MRF study, across in-vitro phantoms, tumor-bearing mice, healthy human volunteers, and a subject with glioblastoma. The resulting multi-parametric posteriors are in good agreement with those calculated using a brute-force Bayesian analysis, while providing an orders-of-magnitude acceleration in whole brain quantification. In addition, we demonstrate how monitoring the multi-parameter posterior dynamics across progressively acquired signals provides practical insights for protocol optimization and may facilitate real-time adaptive acquisition.

Agent memory systems must accommodate continuously growing information while supporting efficient, context-aware retrieval for downstream tasks. Abstraction is essential for scaling agent memory, yet it often comes at the cost of specificity, obscuring the fine-grained details required for effective reasoning. We introduce Memora, a harmonic memory representation that structurally balances abstraction and specificity. Memora organizes information via its primary abstractions that index concrete memory values and consolidate related updates into unified memory entries, while cue anchors expand retrieval access across diverse aspects of the memory and connect related memories. Building on this structure, we employ a retrieval policy that actively exploits these memory connections to retrieve relevant information beyond direct semantic similarity. Theoretically, we show that standard Retrieval-Augmented Generation (RAG) and Knowledge Graph (KG)-based memory systems emerge as special cases of our framework. Empirically, Memora establishes a new state-of-the-art on the LoCoMo and LongMemEval benchmarks, demonstrating better retrieval relevance and reasoning effectiveness as memory scales.

Conversational AI systems combine AI-based solutions with the flexibility of conversational interfaces. However, most existing testing solutions do not straightforwardly adapt to the characteristics of conversational interaction or to the behavior of AI components. To address this limitation, this Ph.D. thesis investigates a new family of testing approaches for conversational AI systems, focusing on the validation of their constituent elements at different levels of granularity, from the integration between the language and the AI components, to individual conversational agents, up to multi-agent implementations of conversational AI systems

Vision-Language-Action (VLA) models hold promise for generalist robotics but currently struggle with data scarcity, architectural inefficiencies, and the inability to generalize across different hardware platforms. We introduce RDT2, a robotic foundation model built upon a 7B parameter VLM designed to enable zero-shot deployment on novel embodiments for open-vocabulary tasks. To achieve this, we collected one of the largest open-source robotic datasets--over 10,000 hours of demonstrations in diverse families--using an enhanced, embodiment-agnostic Universal Manipulation Interface (UMI). Our approach employs a novel three-stage training recipe that aligns discrete linguistic knowledge with continuous control via Residual Vector Quantization (RVQ), flow-matching, and distillation for real-time inference. Consequently, RDT2 becomes one of the first models that simultaneously zero-shot generalizes to unseen objects, scenes, instructions, and even robotic platforms. Besides, it outperforms state-of-the-art baselines in dexterous, long-horizon, and dynamic downstream tasks like playing table tennis. See https://rdt-robotics.github.io/rdt2/ for more information.

Hybrid training methods for large language models combine supervised fine tuning (SFT) on expert demonstrations with reinforcement learning (RL) on model rollouts, typically at the sample level. We propose Entropy Gated Selective Policy Optimization (EGSPO), a three stage framework that extends sample level mixing with token level gradient modulation. Stage 1, SFT expert learning, establishes a reliable warm up policy using expert demonstrations with a pure SFT loss. Stage 2, RL rollout generation, samples trajectories from the current policy and computes per token predictive entropy. Stage 3, the EGSPO mechanism, applies entropy gated gradient allocation: a predictive entropy module routes high entropy tokens to full PPO updates to encourage exploration, and low entropy tokens to attenuated PPO updates to reduce variance and preserve knowledge. Critically, both branches incorporate the advantage function A_t, ensuring that incorrect trajectories receive consistent negative learning signals and preventing reinforcement of confident errors. EGSPO achieves consistent improvements on mathematical reasoning benchmarks, with gains of 3.8 percent on AIME and 2.9 percent on MATH over the CHORD phi baseline, while incurring only 3.4 percent additional computational overhead.

Dense retrieval represents queries and docu-002 ments as high-dimensional embeddings, but003 these representations can be redundant at the004 query level: for a given information need, only005 a subset of dimensions is consistently help-006 ful for ranking. Prior work addresses this via007 pseudo-relevance feedback (PRF) based dimen-008 sion importance estimation, which can produce009 query-aware masks without labeled data but010 often relies on noisy pseudo signals and heuris-011 tic test-time procedures. In contrast, super-012 vised adapter methods leverage relevance labels013 to improve embedding quality, yet they learn014 global transformations shared across queries015 and do not explicitly model query-aware di-016 mension importance. We propose a Query-017 Aware Adaptive Dimension Selection frame-018 work that learns to predict per-dimension im-019 portance directly from query embedding. We020 first construct oracle dimension importance dis-021 tributions over embedding dimensions using022 supervised relevance labels, and then train a023 predictor to map a query embedding to these024 label-distilled importance scores. At inference,025 the predictor selects a query-aware subset of026 dimensions for similarity computation based027 solely on the query embedding, without pseudo-028 relevance feedback. Experiments across multi-029 ple dense retrievers and benchmarks show that030 our learned dimension selector improves re-031 trieval effectiveness over the full-dimensional032 baseline as well as PRF-based masking and033 supervised adapter baselines.

Reinforcement Learning (RL) offers a powerful framework for optimizing dynamic treatment regimes (DTRs). However, clinical RL is fundamentally bottlenecked by reward engineering: the challenge of defining signals that safely and effectively guide policy learning in complex, sparse offline environments. Existing approaches often rely on manual heuristics that fail to generalize across diverse pathologies. To address this, we propose an automated pipeline leveraging Large Language Models (LLMs) for offline reward design and verification. We formulate the reward function using potential functions consisted of three core components: survival, confidence, and competence. We further introduce quantitative metrics to rigorously evaluate and select the optimal reward structure prior to deployment. By integrating LLM-driven domain knowledge, our framework automates the design of reward functions for specific diseases while significantly enhancing the performance of the resulting policies.

Optical coherence tomography (OCT) has revolutionized retinal disease diagnosis with its high-resolution and three-dimensional imaging nature, yet its full diagnostic automation in clinical practices remains constrained by multi-stage workflows and conventional single-slice single-task AI models. We present Full-process OCT-based Clinical Utility System (FOCUS), a foundation model-driven framework enabling end-to-end automation of 3D OCT retinal disease diagnosis. FOCUS sequentially performs image quality assessment with EfficientNetV2-S, followed by abnormality detection and multi-disease classification using a fine-tuned Vision Foundation Model. Crucially, FOCUS leverages a unified adaptive aggregation method to intelligently integrate 2D slices-level predictions into comprehensive 3D patient-level diagnosis. Trained and tested on 3,300 patients (40,672 slices), and externally validated on 1,345 patients (18,498 slices) across four different-tier centers and diverse OCT devices, FOCUS achieved high F1 scores for quality assessment (99.01%), abnormally detection (97.46%), and patient-level diagnosis (94.39%). Real-world validation across centers also showed stable performance (F1: 90.22%-95.24%). In human-machine comparisons, FOCUS matched expert performance in abnormality detection (F1: 95.47% vs 90.91%) and multi-disease diagnosis (F1: 93.49% vs 91.35%), while demonstrating better efficiency. FOCUS automates the image-to-diagnosis pipeline, representing a critical advance towards unmanned ophthalmology with a validated blueprint for autonomous screening to enhance population scale retinal care accessibility and efficiency.

In reinforcement learning (RL), Q-learning is a fundamental algorithm whose convergence is guaranteed in the tabular setting. However, this convergence guarantee does not hold under linear function approximation. To overcome this limitation, a significant line of research has introduced regularization techniques to ensure stable convergence under function approximation. In this work, we propose a new algorithm, periodic regularized Q-learning (PRQ). We first introduce regularization at the level of the projection operator and explicitly construct a regularized projected value iteration (RP-VI), subsequently extending it to a sample-based RL algorithm. By appropriately regularizing the projection operator, the resulting projected value iteration becomes a contraction. By extending this regularized projection into the stochastic setting, we establish the PRQ algorithm and provide a rigorous theoretical analysis that proves finite-time convergence guarantees for PRQ under linear function approximation.

In this work, we aim to develop effective data synthesis techniques that autonomously synthesize multimodal training data for enhancing MLLMs in solving complex real-world tasks. To this end, we propose Collective Adversarial Data Synthesis (CADS), a novel and general approach to synthesize high-quality, diverse and challenging multimodal data for MLLMs. The core idea of CADS is to leverage collective intelligence to ensure high-quality and diverse generation, while exploring adversarial learning to synthesize challenging samples for effectively driving model improvement. Specifically, CADS operates with two cyclic phases, i.e., Collective Adversarial Data Generation (CAD-Generate) and Collective Adversarial Data Judgment (CAD-Judge). CAD-Generate leverages collective knowledge to jointly generate new and diverse multimodal data, while CAD-Judge collaboratively assesses the quality of synthesized data. In addition, CADS introduces an Adversarial Context Optimization mechanism to optimize the generation context to encourage challenging and high-value data generation. With CADS, we construct MMSynthetic-20K and train our model R1-SyntheticVL, which demonstrates superior performance on various benchmarks.

Deep equilibrium models (DEQs) achieve infinitely deep network representations without stacking layers by exploring fixed points of layer transformations in neural networks. Such models constitute an innovative approach that achieves performance comparable to state-of-the-art methods in many large-scale numerical experiments, despite requiring significantly less memory. However, DEQs face the challenge of requiring vastly more computational time for training and inference than conventional methods, as they repeatedly perform fixed-point iterations with no convergence guarantee upon each input. Therefore, this study explored an approach to improve fixed-point convergence and consequently reduce computational time by restructuring the model architecture to guarantee fixed-point convergence. Our proposed approach for image classification, Lipschitz multiscale DEQ, has theoretically guaranteed fixed-point convergence for both forward and backward passes by hyperparameter adjustment, achieving up to a 4.75$\times$ speed-up in numerical experiments on CIFAR-10 at the cost of a minor drop in accuracy.

Large Language Models (LLMs) and Vision-Language Models (VLMs) have demonstrated remarkable capabilities. However, their deployment is hindered by significant computational costs. Existing structured pruning methods, while hardware-efficient, often suffer from significant accuracy degradation. In this paper, we argue that this failure stems from a stage-agnostic pruning approach that overlooks the asymmetric roles between the prefill and decode stages. By introducing a virtual gate mechanism, our importance analysis reveals that deep layers are critical for next-token prediction (decode) but largely redundant for context encoding (prefill). Leveraging this insight, we propose Prefill-Only Pruning (POP), a stage-aware inference strategy that safely omits deep layers during the computationally intensive prefill stage while retaining the full model for the sensitive decode stage. To enable the transition between stages, we introduce independent Key-Value (KV) projections to maintain cache integrity, and a boundary handling strategy to ensure the accuracy of the first generated token. Extensive experiments on Llama-3.1, Qwen3-VL, and Gemma-3 across diverse modalities demonstrate that POP achieves up to 1.37$\times$ speedup in prefill latency with minimal performance loss, effectively overcoming the accuracy-efficiency trade-off limitations of existing structured pruning methods.

Unsupervised anomaly detection stands as an important problem in machine learning, with applications in financial fraud prevention, network security and medical diagnostics. Existing unsupervised anomaly detection algorithms rarely perform well across different anomaly types, often excelling only under specific structural assumptions. This lack of robustness also becomes particularly evident under noisy settings. We propose Mean Shift Density Enhancement (MSDE), a fully unsupervised framework that detects anomalies through their geometric response to density-driven manifold evolution. MSDE is based on the principle that normal samples, being well supported by local density, remain stable under iterative density enhancement, whereas anomalous samples undergo large cumulative displacements as they are attracted toward nearby density modes. To operationalize this idea, MSDE employs a weighted mean-shift procedure with adaptive, sample-specific density weights derived from a UMAP-based fuzzy neighborhood graph. Anomaly scores are defined by the total displacement accumulated across a small number of mean-shift iterations. We evaluate MSDE on the ADBench benchmark, comprising forty six real-world tabular datasets, four realistic anomaly generation mechanisms, and six noise levels. Compared to 13 established unsupervised baselines, MSDE achieves consistently strong, balanced and robust performance for AUC-ROC, AUC-PR, and Precision@n, at several noise levels and on average over several types of anomalies. These results demonstrate that displacement-based scoring provides a robust alternative to the existing state-of-the-art for unsupervised anomaly detection.

We study universal approximation of continuous functionals on compact subsets of products of Hilbert spaces. We prove that any such functional can be uniformly approximated by models that first take finitely many continuous linear measurements of the inputs and then combine these measurements through continuous scalar nonlinearities. We also extend the approximation principle to maps with values in a Banach space, yielding finite-rank approximations. These results provide a compact-set justification for the common ``measure, apply scalar nonlinearities, then combine'' design pattern used in operator learning and imaging.

This paper develops a new approach to computational argumentation that is informed by philosophical and linguistic views. Namely, it takes into account two ideas that have received little attention in the literature on computational argumentation: First, an agent may rationally reject an argument based on mere doubt, thus not all arguments they could defend must be accepted; and, second, that it is sometimes more natural to think in terms of which individual sentences or claims an agent accepts in a debate, rather than which arguments. In order to incorporate these two ideas into a computational approach, we first define the notion of structured bipolar argumentation frameworks (SBAFs), where arguments consist of sentences and we have both an attack and a support relation between them. Then, we provide semantics for SBAFs with two features: (1) Unlike with completeness-based semantics, our semantics do not force agents to accept all defended arguments. (2) In addition to argument extensions, which give acceptable sets of arguments, we also provide semantics for language extensions that specify acceptable sets of sentences. These semantics represent reasonable positions an agent might have in a debate. Our semantics lie between the admissible and complete semantics of abstract argumentation. Further, our approach can be used to provide a new perspective on existing approaches. For instance, we can specify the conditions under which an agent can ignore support between arguments (i.e. under which the use of abstract argumentation is warranted) and we show that deductive support semantics is a special case of our approach.

Enterprise meeting environments require AI assistants that handle diverse operational tasks, from rapid fact checking during live discussions to cross meeting analysis for strategic planning, under strict latency, cost, and privacy constraints. Existing meeting benchmarks mainly focus on simplified question answering and fail to reflect real world enterprise workflows, where queries arise organically from multi stakeholder collaboration, span long temporal contexts, and require tool augmented reasoning. We address this gap through a grounded dataset and a learned agent framework. First, we introduce MeetAll, a bilingual and multimodal corpus derived from 231 enterprise meetings totaling 140 hours. Questions are injected using an enterprise informed protocol validated by domain expert review and human discriminability studies. Unlike purely synthetic benchmarks, this protocol is grounded in four enterprise critical dimensions: cognitive load, temporal context span, domain expertise, and actionable task execution, calibrated through interviews with stakeholders across finance, healthcare, and technology sectors. Second, we propose MeetBench XL, a multi dimensional evaluation protocol aligned with human judgment that measures factual fidelity, intent alignment, response efficiency, structural clarity, and completeness. Third, we present MeetMaster XL, a learned dual policy agent that jointly optimizes query routing between fast and slow reasoning paths and tool invocation, including retrieval, cross meeting aggregation, and web search. A lightweight classifier enables accurate routing with minimal overhead, achieving a superior quality latency tradeoff over single model baselines. Experiments against commercial systems show consistent gains, supported by ablations, robustness tests, and a real world deployment case study.Resources: https://github.com/huyuelin/MeetBench.

A common assumption in representation learning is that globally well-distributed embeddings support robust and generalizable representations. This focus has shaped both training objectives and evaluation protocols, implicitly treating global geometry as a proxy for representational competence. While global geometry effectively encodes which elements are present, it is often insensitive to how they are composed. We investigate this limitation by testing the ability of geometric metrics to predict compositional binding across 21 vision encoders. We find that standard geometry-based statistics exhibit near-zero correlation with compositional binding. In contrast, functional sensitivity, as measured by the input-output Jacobian, reliably tracks this capability. We further provide an analytic account showing that this disparity arises from objective design, as existing losses explicitly constrain embedding geometry but leave the local input-output mapping unconstrained. These results suggest that global embedding geometry captures only a partial view of representational competence and establish functional sensitivity as a critical complementary axis for modeling composite structure.

Advancing complex reasoning in large language models relies on high-quality, verifiable datasets, yet human annotation remains cost-prohibitive and difficult to scale. Current synthesis paradigms often face a recurring trade-off: maintaining structural validity typically restricts problem complexity, while relaxing constraints to increase difficulty frequently leads to inconsistent or unsolvable instances. To address this, we propose Agentic Proposing, a framework that models problem synthesis as a goal-driven sequential decision process where a specialized agent dynamically selects and composes modular reasoning skills. Through an iterative workflow of internal reflection and tool-use, we develop the Agentic-Proposer-4B using Multi-Granularity Policy Optimization (MGPO) to generate high-precision, verifiable training trajectories across mathematics, coding, and science. Empirical results demonstrate that downstream solvers trained on agent-synthesized data significantly outperform leading baselines and exhibit robust cross-domain generalization. Notably, a 30B solver trained on only 11,000 synthesized trajectories achieves a state-of-the-art 91.6% accuracy on AIME25, rivaling frontier-scale proprietary models such as GPT-5 and proving that a small volume of high-quality synthetic signals can effectively substitute for massive human-curated datasets.

In this paper, we introduce BlockRR, a novel and unified randomized-response mechanism for label differential privacy. This framework generalizes existed RR-type mechanisms as special cases under specific parameter settings, which eliminates the need for separate, case-by-case analysis. Theoretically, we prove that BlockRR satisfies $ε$-label DP. We also design a partition method for BlockRR based on a weight matrix derived from label prior information; the parallel composition principle ensures that the composition of two such mechanisms remains $ε$-label DP. Empirically, we evaluate BlockRR on two variants of CIFAR-10 with varying degrees of class imbalance. Results show that in the high-privacy and moderate-privacy regimes ($ε\leq 3.0$), our propsed method gets a better balance between test accuaracy and the average of per-class accuracy. In the low-privacy regime ($ε\geq 4.0$), all methods reduce BlockRR to standard RR without additional performance loss.

Detecting toxicity in multimodal data remains a significant challenge, as harmful meanings often lurk beneath seemingly benign individual modalities: only emerging when modalities are combined and semantic associations are activated. To address this, we propose a novel detection framework based on Toxicity Association Graphs (TAGs), which systematically model semantic associations between innocuous entities and latent toxic implications. Leveraging TAGs, we introduce the first quantifiable metric for hidden toxicity, the Multimodal Toxicity Covertness (MTC), which measures the degree of concealment in toxic multimodal expressions. By integrating our detection framework with the MTC metric, our approach enables precise identification of covert toxicity while preserving full interpretability of the decision-making process, significantly enhancing transparency in multimodal toxicity detection. To validate our method, we construct the Covert Toxic Dataset, the first benchmark specifically designed to capture high-covertness toxic multimodal instances. This dataset encodes nuanced cross-modal associations and serves as a rigorous testbed for evaluating both the proposed metric and detection framework. Extensive experiments demonstrate that our approach outperforms existing methods across both low- and high-covertness toxicity regimes, while delivering clear, interpretable, and auditable detection outcomes. Together, our contributions advance the state of the art in explainable multimodal toxicity detection and lay the foundation for future context-aware and interpretable approaches. Content Warning: This paper contains examples of toxic multimodal content that may be offensive or disturbing to some readers. Reader discretion is advised.

Large Language Models (LLMs) have seen widespread adoption across multiple domains, creating an urgent need for robust safety alignment mechanisms. However, robustness remains challenging due to jailbreak attacks that bypass alignment via adversarial prompts. In this work, we focus on the prevalent Greedy Coordinate Gradient (GCG) attack and identify a previously underexplored attack axis in jailbreak attacks typically framed as suffix-based: the placement of adversarial tokens within the prompt. Using GCG as a case study, we show that both optimizing attacks to generate prefixes instead of suffixes and varying adversarial token position during evaluation substantially influence attack success rates. Our findings highlight a critical blind spot in current safety evaluations and underline the need to account for the position of adversarial tokens in the adversarial robustness evaluation of LLMs.

Robust generalization beyond training distributions remains a critical challenge for deep neural networks. This is especially pronounced in medical image analysis, where data is often scarce and covariate shifts arise from different hardware devices, imaging protocols, and heterogeneous patient populations. These factors collectively hinder reliable performance and slow down clinical adoption. Despite recent progress, existing learning paradigms primarily rely on the Euclidean manifold, whose flat geometry fails to capture the complex, hierarchical structures present in clinical data. In this work, we exploit the advantages of hyperbolic manifolds to model complex data characteristics. We present the first comprehensive validation of hyperbolic representation learning for medical image analysis and demonstrate statistically significant gains across eleven in-distribution datasets and three ViT models. We further propose an unsupervised, domain-invariant hyperbolic cross-branch consistency constraint. Extensive experiments confirm that our proposed method promotes domain-invariant features and outperforms state-of-the-art Euclidean methods by an average of $+2.1\%$ AUC on three domain generalization benchmarks: Fitzpatrick17k, Camelyon17-WILDS, and a cross-dataset setup for retinal imaging. These datasets span different imaging modalities, data sizes, and label granularities, confirming generalization capabilities across substantially different conditions. The code is available at https://github.com/francescodisalvo05/hyperbolic-cross-branch-consistency .

Multimodal large language models (MLLMs) enable interaction over both text and images, but their safety behavior can be driven by unimodal shortcuts instead of true joint intent understanding. We introduce CSR-Bench, a benchmark for evaluating cross-modal reliability through four stress-testing interaction patterns spanning Safety, Over-rejection, Bias, and Hallucination, covering 61 fine-grained types. Each instance is constructed to require integrated image-text interpretation, and we additionally provide paired text-only controls to diagnose modality-induced behavior shifts. We evaluate 16 state-of-the-art MLLMs and observe systematic cross-modal alignment gaps. Models show weak safety awareness, strong language dominance under interference, and consistent performance degradation from text-only controls to multimodal inputs. We also observe a clear trade-off between reducing over-rejection and maintaining safe, non-discriminatory behavior, suggesting that some apparent safety gains may come from refusal-oriented heuristics rather than robust intent understanding. WARNING: This paper contains unsafe contents.

Random Forests (RF) are among the most powerful and widely used predictive models for centralized tabular data, yet few methods exist to adapt them to the federated learning setting. Unlike most federated learning approaches, the piecewise-constant nature of RF prevents exact gradient-based optimization. As a result, existing federated RF implementations rely on unprincipled heuristics: for instance, aggregating decision trees trained independently on clients fails to optimize the global impurity criterion, even under simple distribution shifts. We propose FedForest, a new federated RF algorithm for horizontally partitioned data that naturally accommodates diverse forms of client data heterogeneity, from covariate shift to more complex outcome shift mechanisms. We prove that our splitting procedure, based on aggregating carefully chosen client statistics, closely approximates the split selected by a centralized algorithm. Moreover, FedForest allows splits on client indicators, enabling a non-parametric form of personalization that is absent from prior federated random forest methods. Empirically, we demonstrate that the resulting federated forests closely match centralized performance across heterogeneous benchmarks while remaining communication-efficient.

Finding frequently occurring subgraph patterns or network motifs in neural architectures is crucial for optimizing efficiency, accelerating design, and uncovering structural insights. However, as the subgraph size increases, enumeration-based methods are perfectly accurate but computationally prohibitive, while sampling-based methods are computationally tractable but suffer from a severe decline in discovery capability. To address these challenges, this paper proposes GraDE, a diffusion-guided search framework that ensures both computational feasibility and discovery capability. The key innovation is the Graph Diffusion Estimator (GraDE), which is the first to introduce graph diffusion models to identify frequent subgraphs by scoring their typicality within the learned distribution. Comprehensive experiments demonstrate that the estimator achieves superior ranking accuracy, with up to 114\% improvement compared to sampling-based baselines. Benefiting from this, the proposed framework successfully discovers large-scale frequent patterns, achieving up to 30$\times$ higher median frequency than sampling-based methods.

Computer-use agents (CUAs) that interact with real computer systems can perform automated tasks but face critical safety risks. Ambiguous instructions may trigger harmful actions, and adversarial users can manipulate tool execution to achieve malicious goals. Existing benchmarks mostly focus on short-horizon or GUI-based tasks, evaluating on execution-time errors but overlooking the ability to anticipate planning-time risks. To fill this gap, we present LPS-Bench, a benchmark that evaluates the planning-time safety awareness of MCP-based CUAs under long-horizon tasks, covering both benign and adversarial interactions across 65 scenarios of 7 task domains and 9 risk types. We introduce a multi-agent automated pipeline for scalable data generation and adopt an LLM-as-a-judge evaluation protocol to assess safety awareness through the planning trajectory. Experiments reveal substantial deficiencies in existing CUAs' ability to maintain safe behavior. We further analyze the risks and propose mitigation strategies to improve long-horizon planning safety in MCP-based CUA systems. We open-source our code at https://github.com/tychenn/LPS-Bench.

Scaling test-time compute via long Chain-ofThought unlocks remarkable gains in reasoning capabilities, yet it faces practical limits due to the linear growth of KV cache and quadratic attention complexity. In this paper, we introduce Accordion-Thinking, an end-to-end framework where LLMs learn to self-regulate the granularity of the reasoning steps through dynamic summarization. This mechanism enables a Fold inference mode, where the model periodically summarizes its thought process and discards former thoughts to reduce dependency on historical tokens. We apply reinforcement learning to incentivize this capability further, uncovering a critical insight: the accuracy gap between the highly efficient Fold mode and the exhaustive Unfold mode progressively narrows and eventually vanishes over the course of training. This phenomenon demonstrates that the model learns to encode essential reasoning information into compact summaries, achieving effective compression of the reasoning context. Our Accordion-Thinker demonstrates that with learned self-compression, LLMs can tackle complex reasoning tasks with minimal dependency token overhead without compromising solution quality, and it achieves a 3x throughput while maintaining accuracy on a 48GB GPU memory configuration, while the structured step summaries provide a human-readable account of the reasoning process.

This paper studies an important rate allocation problem that arises in many networked and distributed systems: steady-state traffic rate allocation from multiple sources to multiple service nodes when both (i) the access-path delay on each source-node route is rate-dependent (capacity-constrained) and convex, and (ii) each service node (also capacity-constrained) experiences a load-dependent queueing delay driven by aggregate load from all sources. We show that the resulting flow-weighted end-to-end delay minimization is a convex program, yielding a global system-optimal solution characterized by KKT conditions that equalize total marginal costs (a path marginal access term plus a node congestion price) across all utilized routes. This condition admits a Wardrop-type interpretation: for each source, all utilized options equalize total marginal cost, while any option with strictly larger total marginal cost receives no flow. Building on this structure, we develop a lightweight distributed pricing-based algorithm in which each service node locally computes and broadcasts a scalar congestion price from its observed aggregate load, while each source updates its traffic split by solving a small separable convex allocation problem under the advertised prices. Numerical illustrations demonstrate convergence of the distributed iteration to the centralized optimum and highlight the trade-offs induced by jointly modeling access and service congestion.

This paper documents our efforts in releasing the printed and audio book of the translation of the famous novel The Little Prince into the Chakavian dialect, as a computer-readable, AI-ready dataset, with the textual and the audio components of the two releases now aligned on the level of each written and spoken word. Our motivation for working on this release is multiple. The first one is our wish to preserve the highly valuable and specific content beyond the small editions of the printed and the audio book. With the dataset published in the CLARIN.SI repository, this content is from now on at the fingertips of any interested individual. The second motivation is to make the data available for various artificial-intelligence-related usage scenarios, such as the one we follow upon inside this paper already -- adapting the Whisper-large-v3 open automatic speech recognition model, with decent performance on standard Croatian, to Chakavian dialectal speech. We can happily report that with adapting the model, the word error rate on the selected test data has being reduced to a half, while we managed to remove up to two thirds of the error on character level. We envision many more usages of this dataset beyond the set of experiments we have already performed, both on tasks of artificial intelligence research and application, as well as dialectal research. The third motivation for this release is our hope that this, now highly structured dataset, will be transformed into a digital online edition of this work, allowing individuals beyond the research and technology communities to enjoy the beauty of the message of the little boy in the desert, told through the spectacular prism of the Chakavian dialect.

With the advent of Large Language Models (LLMs), general-purpose agents have seen fundamental advancements. However, evaluating these agents presents unique challenges that distinguish them from static QA benchmarks. We observe that current agent benchmarks are heavily confounded by extraneous factors, including system prompts, toolset configurations, and environmental dynamics. Existing evaluations often rely on fragmented, researcher-specific frameworks where the prompt engineering for reasoning and tool usage varies significantly, making it difficult to attribute performance gains to the model itself. Additionally, the lack of standardized environmental data leads to untraceable errors and non-reproducible results. This lack of standardization introduces substantial unfairness and opacity into the field. We propose that a unified evaluation framework is essential for the rigorous advancement of agent evaluation. To this end, we introduce a proposal aimed at standardizing agent evaluation.

The escalating scale of Large Language Models (LLMs) necessitates efficient adaptation techniques. Model merging has gained prominence for its efficiency and controllability. However, existing merging techniques typically serve as post-hoc refinements or focus on mitigating task interference, often failing to capture the dynamic optimization benefits of supervised fine-tuning (SFT). In this work, we propose Streaming Merging, an innovative model updating paradigm that conceptualizes merging as an iterative optimization process. Central to this paradigm is \textbf{ARM} (\textbf{A}ctivation-guided \textbf{R}otation-aware \textbf{M}erging), a strategy designed to approximate gradient descent dynamics. By treating merging coefficients as learning rates and deriving rotation vectors from activation subspaces, ARM effectively steers parameter updates along data-driven trajectories. Unlike conventional linear interpolation, ARM aligns semantic subspaces to preserve the geometric structure of high-dimensional parameter evolution. Remarkably, ARM requires only early SFT checkpoints and, through iterative merging, surpasses the fully converged SFT model. Experimental results across model scales (1.7B to 14B) and diverse domains (e.g., math, code) demonstrate that ARM can transcend converged checkpoints. Extensive experiments show that ARM provides a scalable and lightweight framework for efficient model adaptation.

We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization. BayeSQP employs second-order Gaussian process surrogates for both the objective and constraints to jointly model the function values, gradients, and Hessian from only zero-order information. At each iteration, a local subproblem is constructed using the GP posterior estimates and solved to obtain a search direction. Crucially, the formulation of the subproblem explicitly incorporates uncertainty in both the function and derivative estimates, resulting in a tractable second-order cone program for high probability improvements under model uncertainty. A subsequent one-dimensional line search via constrained Thompson sampling selects the next evaluation point. Empirical results show thatBayeSQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization.

Long-context inputs in large language models (LLMs) often suffer from the "lost in the middle" problem, where critical information becomes diluted or ignored due to excessive length. Context compression methods aim to address this by reducing input size, but existing approaches struggle with balancing information preservation and compression efficiency. We propose Adaptive Task-Aware Compressor (ATACompressor), which dynamically adjusts compression based on the specific requirements of the task. ATACompressor employs a selective encoder that compresses only the task-relevant portions of long contexts, ensuring that essential information is preserved while reducing unnecessary content. Its adaptive allocation controller perceives the length of relevant content and adjusts the compression rate accordingly, optimizing resource utilization. We evaluate ATACompressor on three QA datasets: HotpotQA, MSMARCO, and SQUAD-showing that it outperforms existing methods in terms of both compression efficiency and task performance. Our approach provides a scalable solution for long-context processing in LLMs. Furthermore, we perform a range of ablation studies and analysis experiments to gain deeper insights into the key components of ATACompressor.

Test-time evolution of agent memory serves as a pivotal paradigm for achieving AGI by bolstering complex reasoning through experience accumulation. However, even during benign task evolution, agent safety alignment remains vulnerable-a phenomenon known as Agent Memory Misevolution. To evaluate this phenomenon, we construct the Trust-Memevo benchmark to assess multi-dimensional trustworthiness during benign task evolution, revealing an overall decline in trustworthiness across various task domains and evaluation settings. To address this issue, we propose TAME, a dual-memory evolutionary framework that separately evolves executor memory to improve task performance by distilling generalizable methodologies, and evaluator memory to refine assessments of both safety and task utility based on historical feedback. Through a closed loop of memory filtering, draft generation, trustworthy refinement, execution, and dual-track memory updating, TAME preserves trustworthiness without sacrificing utility. Experiments demonstrate that TAME mitigates misevolution, achieving a joint improvement in both trustworthiness and task performance.

This paper explores effective numerical feature embedding for Click-Through Rate prediction in streaming environments. Conventional static binning methods rely on offline statistics of numerical distributions; however, this inherently two-stage process often triggers semantic drift during bin boundary updates. While neural embedding methods enable end-to-end learning, they often discard explicit distributional information. Integrating such information end-to-end is challenging because streaming features often violate the i.i.d. assumption, precluding unbiased estimation of the population distribution via the expectation of order statistics. Furthermore, the critical context dependency of numerical distributions is often neglected. To this end, we propose DAES, an end-to-end framework designed to tackle numerical feature embedding in streaming training scenarios by integrating distributional information with an adaptive modulation mechanism. Specifically, we introduce an efficient reservoir-sampling-based distribution estimation method and two field-aware distribution modulation strategies to capture streaming distributions and field-dependent semantics. DAES significantly outperforms existing approaches as demonstrated by extensive offline and online experiments and has been fully deployed on a leading short-video platform with hundreds of millions of daily active users.

As code large language models (LLMs) evolve into tool-interactive agents via the Model Context Protocol (MCP), their generalization is increasingly limited by low-quality synthetic data and the diminishing returns of quantity scaling. Moreover, quantity-centric scaling exhibits an early bottleneck that underutilizes trajectory data. We propose TDScaling, a Trajectory Diversity Scaling-based data synthesis framework for code agents that scales performance through diversity rather than raw volume. Under a fixed training budget, increasing trajectory diversity yields larger gains than adding more trajectories, improving the performance-cost trade-off for agent training. TDScaling integrates four innovations: (1) a Business Cluster mechanism that captures real-service logical dependencies; (2) a blueprint-driven multi-agent paradigm that enforces trajectory coherence; (3) an adaptive evolution mechanism that steers synthesis toward long-tail scenarios using Domain Entropy, Reasoning Mode Entropy, and Cumulative Action Complexity to prevent mode collapse; and (4) a sandboxed code tool that mitigates catastrophic forgetting of intrinsic coding capabilities. Experiments on general tool-use benchmarks (BFCL, tau^2-Bench) and code agent tasks (RebenchT, CodeCI, BIRD) demonstrate a win-win outcome: TDScaling improves both tool-use generalization and inherent coding proficiency. We plan to release the full codebase and the synthesized dataset (including 30,000+ tool clusters) upon publication.

Understanding how local interactions give rise to global brain organization requires models that can represent information across multiple scales. We introduce a hierarchical self-supervised learning (SSL) framework that jointly learns node-, edge-, and graph-level embeddings, inspired by multimodal neuroimaging. We construct a controllable synthetic benchmark mimicking the topological properties of connectomes. Our four-stage evaluation protocol reveals a critical failure: the invariance-based SSL model is fundamentally misaligned with the benchmark's topological properties and is catastrophically outperformed by classical, topology-aware heuristics. Ablations confirm an objective mismatch: SSL objectives designed to be invariant to topological perturbations learn to ignore the very community structure that classical methods exploit. Our results expose a fundamental pitfall in applying generic graph SSL to connectome-like data. We present this framework as a cautionary case study, highlighting the need for new, topology-aware SSL objectives for neuro-AI research that explicitly reward the preservation of structure (e.g., modularity or motifs).

The quadratic complexity of attention remains the central bottleneck in long-context inference for large language models. Prior acceleration methods either sparsify the attention map with structured patterns or permanently evict tokens at specific layers, which can retain irrelevant tokens or rely on irreversible early decisions despite the layer-/head-wise dynamics of token importance. In this paper, we propose Token Sparse Attention, a lightweight and dynamic token-level sparsification mechanism that compresses per-head $Q$, $K$, $V$ to a reduced token set during attention and then decompresses the output back to the original sequence, enabling token information to be reconsidered in subsequent layers. Furthermore, Token Sparse Attention exposes a new design point at the intersection of token selection and sparse attention. Our approach is fully compatible with dense attention implementations, including Flash Attention, and can be seamlessly composed with existing sparse attention kernels. Experimental results show that Token Sparse Attention consistently improves accuracy-latency trade-off, achieving up to $\times$3.23 attention speedup at 128K context with less than 1% accuracy degradation. These results demonstrate that dynamic and interleaved token-level sparsification is a complementary and effective strategy for scalable long-context inference.

Accurate estimation of tacrolimus exposure, quantified by the area under the concentration-time curve (AUC), is essential for precision dosing after renal transplantation. Current practice relies on population pharmacokinetic (PopPK) models based on nonlinear mixed-effects (NLME) methods. However, these models depend on rigid, pre-specified assumptions and may struggle to capture complex, patient-specific dynamics, leading to model misspecification. In this study, we introduce a novel data-driven alternative based on Latent Ordinary Differential Equations (Latent ODEs) for tacrolimus AUC prediction. This deep learning approach learns individualized pharmacokinetic dynamics directly from sparse clinical data, enabling greater flexibility in modeling complex biological behavior. The model was evaluated through extensive simulations across multiple scenarios and benchmarked against two standard approaches: NLME-based estimation and the iterative two-stage Bayesian (it2B) method. We further performed a rigorous clinical validation using a development dataset (n = 178) and a completely independent external dataset (n = 75). In simulation, the Latent ODE model demonstrated superior robustness, maintaining high accuracy even when underlying biological mechanisms deviated from standard assumptions. Regarding experiments on clinical datasets, in internal validation, it achieved significantly higher precision with a mean RMSPE of 7.99% compared with 9.24% for it2B (p < 0.001). On the external cohort, it achieved an RMSPE of 10.82%, comparable to the two standard estimators (11.48% and 11.54%). These results establish the Latent ODE as a powerful and reliable tool for AUC prediction. Its flexible architecture provides a promising foundation for next-generation, multi-modal models in personalized medicine.

Diffusion models have demonstrated strong generative performance; however, generated samples often fail to fully align with human intent. This paper studies a test-time scaling method that enables sampling from regions with higher human-aligned reward values. Existing gradient guidance methods approximate the expected future reward (EFR) at an intermediate particle $\mathbf{x}_t$ using a Taylor approximation, but this approximation at each time step incurs high computational cost due to sequential neural backpropagation. We show that the EFR at any $\mathbf{x}_t$ can be computed using only marginal samples from a pre-trained diffusion model. The proposed EFR formulation detaches the neural dependency between $\mathbf{x}_t$ and the EFR, enabling closed-form guidance computation without neural backpropagation. To further improve efficiency, we introduce lookahead sampling to collect marginal samples. For final sample generation, we use an accurate solver that guides particles toward high-reward lookahead samples. We refer to this sampling scheme as LiDAR sampling. LiDAR achieves substantial performance improvements using only three samples with a 3-step lookahead solver, exhibiting steep performance gains as lookahead accuracy and sample count increase; notably, it reaches the same GenEval performance as the latest gradient guidance method for SDXL with a 9.5x speedup.

Inference-time scaling offers a versatile paradigm for aligning visual generative models with downstream objectives without parameter updates. However, existing approaches that optimize the high-dimensional initial noise suffer from severe inefficiency, as many search directions exert negligible influence on the final generation. We show that this inefficiency is closely related to a spectral bias in generative dynamics: model sensitivity to initial perturbations diminishes rapidly as frequency increases. Building on this insight, we propose Spectral Evolution Search (SES), a plug-and-play framework for initial noise optimization that executes gradient-free evolutionary search within a low-frequency subspace. Theoretically, we derive the Spectral Scaling Prediction from perturbation propagation dynamics, which explains the systematic differences in the impact of perturbations across frequencies. Extensive experiments demonstrate that SES significantly advances the Pareto frontier of generation quality versus computational cost, consistently outperforming strong baselines under equivalent budgets.

While Mixture-of-Experts (MoE) architectures define the state-of-the-art, their theoretical success is often attributed to heuristic efficiency rather than geometric expressivity. In this work, we present the first analysis of MoE through the lens of tropical geometry, establishing that the Top-$k$ routing mechanism is algebraically isomorphic to the $k$-th elementary symmetric tropical polynomial. This isomorphism partitions the input space into the Normal Fan of a Hypersimplex, revealing that \textbf{sparsity is combinatorial depth} which scales geometric capacity by the binomial coefficient $\binom{N}{k}$. Moving beyond ambient bounds, we introduce the concept of \textit{Effective Capacity} under the Manifold Hypothesis. We prove that while dense networks suffer from capacity collapse on low-dimensional data, MoE architectures exhibit \textit{Combinatorial Resilience}, maintaining high expressivity via the transversality of routing cones. In this study, our framework unifies the discrete geometry of the Hypersimplex with the continuous geometry of neural functions, offering a rigorous theoretical justification for the topological supremacy of conditional computation.

Recently, large language models (LLMs) have shown remarkable reasoning abilities by producing long reasoning traces. However, as the sequence length grows, the key-value (KV) cache expands linearly, incurring significant memory and computation costs. Existing KV cache eviction methods mitigate this issue by discarding less important KV pairs, but often fail to capture complex KV dependencies, resulting in performance degradation. To better balance efficiency and performance, we introduce ForesightKV, a training-based KV cache eviction framework that learns to predict which KV pairs to evict during long-text generations. We first design the Golden Eviction algorithm, which identifies the optimal eviction KV pairs at each step using future attention scores. These traces and the scores at each step are then distilled via supervised training with a Pairwise Ranking Loss. Furthermore, we formulate cache eviction as a Markov Decision Process and apply the GRPO algorithm to mitigate the significant language modeling loss increase on low-entropy tokens. Experiments on AIME2024 and AIME2025 benchmarks of three reasoning models demonstrate that ForesightKV consistently outperforms prior methods under only half the cache budget, while benefiting synergistically from both supervised and reinforcement learning approaches.

Model-based reinforcement learning (MBRL) achieves high sample efficiency by simulating future trajectories with learned dynamics and reward models. However, its effectiveness is severely compromised in sparse reward settings. The core limitation lies in the standard paradigm of regressing ground-truth scalar rewards: in sparse environments, this yields a flat, gradient-free landscape that fails to provide directional guidance for planning. To address this challenge, we propose Shaping Landscapes with Optimistic Potential Estimates (SLOPE), a novel framework that shifts reward modeling from predicting scalars to constructing informative potential landscapes. SLOPE employs optimistic distributional regression to estimate high-confidence upper bounds, which amplifies rare success signals and ensures sufficient exploration gradients. Evaluations on 30+ tasks across 5 benchmarks demonstrate that SLOPE consistently outperforms leading baselines in fully sparse, semi-sparse, and dense rewards.

For Embodied AI, jointly reconstructing dynamic hands and the dense scene context is crucial for understanding physical interaction. However, most existing methods recover isolated hands in local coordinates, overlooking the surrounding 3D environment. To address this, we present Hand3R, the first online framework for joint 4D hand-scene reconstruction from monocular video. Hand3R synergizes a pre-trained hand expert with a 4D scene foundation model via a scene-aware visual prompting mechanism. By injecting high-fidelity hand priors into a persistent scene memory, our approach enables simultaneous reconstruction of accurate hand meshes and dense metric-scale scene geometry in a single forward pass. Experiments demonstrate that Hand3R bypasses the reliance on offline optimization and delivers competitive performance in both local hand reconstruction and global positioning.

Reinforcement learning (RL) has emerged as a key paradigm for aligning and optimizing large language models (LLMs). Standard approaches treat the LLM as the policy and apply RL directly over the full vocabulary space. However, this formulation includes the massive tail of contextually irrelevant tokens in the action space, which could distract the policy from focusing on decision-making among the truly reasonable tokens. In this work, we verify that valid reasoning paths could inherently concentrate within a low-rank subspace. Based on this insight, we introduce Reinforcement Learning with Promising Tokens (RLPT), a framework that mitigates the action space issue by decoupling strategic decision-making from token generation. Specifically, RLPT leverages the semantic priors of the base model to identify a dynamic set of \emph{promising tokens} and constrains policy optimization exclusively to this refined subset via masking. Theoretical analysis and empirical results demonstrate that RLPT effectively reduces gradient variance, stabilizes the training process, and improves sample efficiency. Experiment results on math, coding, and telecom reasoning show that RLPT outperforms standard RL baselines and integrates effectively across various model sizes (4B and 8B) and RL algorithms (GRPO and DAPO).

Reinforcement learning algorithms such as group-relative policy optimization (GRPO) have demonstrated strong potential for improving the mathematical reasoning capabilities of large language models. However, prior work has consistently observed an entropy collapse phenomenon during reinforcement post-training, characterized by a monotonic decrease in policy entropy that ultimately leads to training instability and collapse. As a result, most existing approaches restrict training to short horizons (typically 5-20 epochs), limiting sustained exploration and hindering further policy improvement. In addition, nearly all prior work relies on a single, fixed reasoning prompt or template during training. In this work, we introduce prompt augmentation, a training strategy that instructs the model to generate reasoning traces under diverse templates and formats, thereby increasing rollout diversity. We show that, without a KL regularization term, prompt augmentation enables stable scaling of training duration under a fixed dataset and allows the model to tolerate low-entropy regimes without premature collapse. Empirically, a Qwen2.5-Math-1.5B model trained with prompt augmentation on the MATH Level 3-5 dataset achieves state-of-the-art performance, reaching 44.5 per-benchmark accuracy and 51.3 per-question accuracy on standard mathematical reasoning benchmarks, including AIME24, AMC, MATH500, Minerva, and OlympiadBench. The code and model checkpoints are available at https://github.com/wenquanlu/prompt-augmentation-GRPO.

Distributed Stream Processing Systems (DSPSs) form the backbone of real-time processing and analytics at ByteDance, where Apache Flink powers one of the largest production clusters worldwide. Ensuring resiliency, the ability to withstand and rapidly recover from failures, together with operational stability, which provides consistent and predictable performance under normal conditions, is essential for meeting strict Service Level Objectives (SLOs). However, achieving resiliency and stability in large-scale production environments remains challenging due to the cluster scale, business diversity, and significant operational overhead. In this work, we present StreamShield, a production-proven resiliency solution deployed in ByteDance's Flink clusters. Designed along complementary perspectives of the engine and cluster, StreamShield introduces key techniques to enhance resiliency, covering runtime optimization, fine-grained fault-tolerance, hybrid replication strategy, and high availability under external systems. Furthermore, StreamShield proposes a robust testing and deployment pipeline that ensures reliability and robustness in production releases. Extensive evaluations on a production cluster demonstrate the efficiency and effectiveness of techniques proposed by StreamShield.

Although Key-Value (KV) Cache is essential for efficient large language models (LLMs) inference, its growing memory footprint in long-context scenarios poses a significant bottleneck, making KVCache compression crucial. Current compression methods rely on rigid splitting strategies, such as fixed intervals or pre-defined delimiters. We observe that rigid splitting suffers from significant accuracy degradation (ranging from 5.5% to 55.1%) across different scenarios, owing to the scenario-dependent nature of the semantic boundaries. This highlights the necessity of dynamic semantic splitting to match semantics. To achieve this, we face two challenges. (1) Improper delimiter selection misaligns semantics with the KVCache, resulting in 28.6% accuracy loss. (2) Variable-length blocks after splitting introduce over 73.1% additional inference overhead. To address the above challenges, we propose DynSplit-KV, a KVCache compression method that dynamically identifies delimiters for splitting. We propose: (1) a dynamic importance-aware delimiter selection strategy, improving accuracy by 49.9%. (2) A uniform mapping strategy that transforms variable-length semantic blocks into a fixed-length format, reducing inference overhead by 4.9x. Experiments show that DynSplit-KV achieves the highest accuracy, 2.2x speedup compared with FlashAttention and 2.6x peak memory reduction in long-context scenarios.

Research involving privacy-sensitive data has always been constrained by data scarcity, standing in sharp contrast to other areas that have benefited from data scaling. This challenge is becoming increasingly urgent as modern AI agents--such as OpenClaw and Gemini Agent--are granted persistent access to highly sensitive personal information. To tackle this longstanding bottleneck and the rising risks, we present Privasis (i.e., privacy oasis), the first million-scale fully synthetic dataset entirely built from scratch--an expansive reservoir of texts with rich and diverse private information--designed to broaden and accelerate research in areas where processing sensitive social data is inevitable. Compared to existing datasets, Privasis, comprising 1.4 million records, offers orders-of-magnitude larger scale with quality, and far greater diversity across various document types, including medical history, legal documents, financial records, calendars, and text messages with a total of 55.1 million annotated attributes such as ethnicity, date of birth, workplace, etc. We leverage Privasis to construct a parallel corpus for text sanitization with our pipeline that decomposes texts and applies targeted sanitization. Our compact sanitization models (<=4B) trained on this dataset outperform state-of-the-art large language models, such as GPT-5 and Qwen-3 235B. We plan to release data, models, and code to accelerate future research on privacy-sensitive domains and agents.

Automatic unit test (UT) generation is essential for software quality assurance, but existing approaches--including symbolic execution, search-based approaches, and recent LLM-based generators--struggle to produce human-quality tests with correct, meaningful assertions and reliable chain-of-thought (CoT) explanations. We identify a gap in UT training data: repository-mined tests lack developer CoTs, while LLM-distilled CoTs are often incorrect or incomplete. To address this issue, we propose a novel data-distillation approach that uses self-debugging to produce high-quality UT training examples paired with faithful CoTs. Our approach combines (1) guided test repair, a heuristic loop (error-, failure-, and coverage-focused steps) that asks the used model to diagnose and iteratively fix generated tests, and (2) CoT compression, which compacts original and debugging CoTs into concise explanations that directly justify correct tests. We apply this pipeline to a large corpus of open-source projects to construct a dataset of 74,518 high-quality <focal method, test, CoT> examples, and then use it for supervised fine-tuning of a base model. An empirical evaluation shows that the fine-tuned model achieves high UT generation effectiveness: it attains a pass rate of 36.17% on test assertions, a branch coverage of 43.90%, and a mutation score of 88.66%, substantially higher than state-of-the-art commercial models like o4-mini.

We study an online resource-selection problem motivated by multi-radio access selection and mobile edge computing offloading. In each round, an agent chooses among $K$ candidate links/servers (arms) whose performance is a stochastic $d$-dimensional vector (e.g., throughput, latency, energy, reliability). The key interaction is \emph{probe-then-commit (PtC)}: the agent may probe up to $q>1$ candidates via control-plane measurements to observe their vector outcomes, but must execute exactly one candidate in the data plane. This limited multi-arm feedback regime strictly interpolates between classical bandits ($q=1$) and full-information experts ($q=K$), yet existing multi-objective learning theory largely focuses on these extremes. We develop \textsc{PtC-P-UCB}, an optimistic probe-then-commit algorithm whose technical core is frontier-aware probing under uncertainty in a Pareto mode, e.g., it selects the $q$ probes by approximately maximizing a hypervolume-inspired frontier-coverage potential and commits by marginal hypervolume gain to directly expand the attained Pareto region. We prove a dominated-hypervolume frontier error of $\tilde{O} (K_P d/\sqrt{qT})$, where $K_P$ is the Pareto-frontier size and $T$ is the horizon, and scalarized regret $\tilde{O} (L_φd\sqrt{(K/q)T})$, where $φ$ is the scalarizer. These quantify a transparent $1/\sqrt{q}$ acceleration from limited probing. We further extend to \emph{multi-modal probing}: each probe returns $M$ modalities (e.g., CSI, queue, compute telemetry), and uncertainty fusion yields variance-adaptive versions of the above bounds via an effective noise scale.

Despite decades of work, we still lack a robust, task-general theory of human behavior even in the simplest domains. In this paper we tackle the generality problem head-on, by aiming to develop a unified model for all tasks embedded in a task-space. In particular we consider the space of binary sequence prediction tasks where the observations are generated by the space parameterized by hidden Markov models (HMM). As the space of tasks is large, experimental exploration of the entire space is infeasible. To solve this problem we propose the adversarial construction approach, which helps identify tasks that are most likely to elicit a qualitatively novel behavior. Our results suggest that adversarial construction significantly outperforms random sampling of environments and therefore could be used as a proxy for optimal experimental design in high-dimensional task spaces.

Reinforcement learning algorithms often suffer from slow convergence due to sparse reward signals, particularly in complex environments where feedback is delayed or infrequent. This paper introduces the Psychological Regret Model (PRM), a novel approach that accelerates learning by incorporating regret-based feedback signals after each decision step. Rather than waiting for terminal rewards, PRM computes a regret signal based on the difference between the expected value of the optimal action and the value of the action taken in each state. This transforms sparse rewards into dense feedback signals through a step-wise scoring framework, enabling faster convergence. We demonstrate that PRM achieves stable performance approximately 36\% faster than traditional Proximal Policy Optimization (PPO) in benchmark environments such as Lunar Lander. Our results indicate that PRM is particularly effective in continuous control tasks and environments with delayed feedback, making it suitable for real-world applications such as robotics, finance, and adaptive education where rapid policy adaptation is critical. The approach formalizes human-inspired counterfactual thinking as a computable regret signal, bridging behavioral economics and reinforcement learning.

Clustering functional data in the presence of phase variation is challenging, as temporal misalignment can obscure intrinsic shape differences and degrade clustering performance. Most existing approaches treat registration and clustering as separate tasks or rely on restrictive parametric assumptions. We present \textbf{NeuralFLoC}, a fully unsupervised, end-to-end deep learning framework for joint functional registration and clustering based on Neural ODE-driven diffeomorphic flows and spectral clustering. The proposed model learns smooth, invertible warping functions and cluster-specific templates simultaneously, effectively disentangling phase and amplitude variation. We establish universal approximation guarantees and asymptotic consistency for the proposed framework. Experiments on functional benchmarks show state-of-the-art performance in both registration and clustering, with robustness to missing data, irregular sampling, and noise, while maintaining scalability. Code is available at https://anonymous.4open.science/r/NeuralFLoC-FEC8.

Online conformal prediction (OCP) seeks prediction intervals that achieve long-run $1-α$ coverage for arbitrary (possibly adversarial) data streams, while remaining as informative as possible. Existing OCP methods often require manual learning-rate tuning to work well, and may also require algorithm-specific analyses. Here, we develop a general regret-to-coverage theory for interval-valued OCP based on the $(1-α)$-pinball loss. Our first contribution is to identify \emph{linearized regret} as a key notion, showing that controlling it implies coverage bounds for any online algorithm. This relies on a black-box reduction that depends only on the Fenchel conjugate of an upper bound on the linearized regret. Building on this theory, we propose UP-OCP, a parameter-free method for OCP, via a reduction to a two-asset portfolio selection problem, leveraging universal portfolio algorithms. We show strong finite-time bounds on the miscoverage of UP-OCP, even for polynomially growing predictions. Extensive experiments support that UP-OCP delivers consistently better size/coverage trade-offs than prior online conformal baselines.

Time series forecasting (TSF) plays a critical role in decision-making for many real-world applications. Recently, LLM-based forecasters have made promising advancements. Despite their effectiveness, existing methods often lack explicit experience accumulation and continual evolution. In this work, we propose MemCast, a learning-to-memory framework that reformulates TSF as an experience-conditioned reasoning task. Specifically, we learn experience from the training set and organize it into a hierarchical memory. This is achieved by summarizing prediction results into historical patterns, distilling inference trajectories into reasoning wisdom, and inducing extracted temporal features into general laws. Furthermore, during inference, we leverage historical patterns to guide the reasoning process and utilize reasoning wisdom to select better trajectories, while general laws serve as criteria for reflective iteration. Additionally, to enable continual evolution, we design a dynamic confidence adaptation strategy that updates the confidence of individual entries without leaking the test set distribution. Extensive experiments on multiple datasets demonstrate that MemCast consistently outperforms previous methods, validating the effectiveness of our approach. Our code is available at https://github.com/Xiaoyu-Tao/MemCast-TS.

Aligning Large Language Models (LLMs) with the diverse spectrum of human values remains a central challenge: preference-based methods often fail to capture deeper motivational principles. Value-based approaches offer a more principled path, yet three gaps persist: extraction often ignores hierarchical structure, evaluation detects presence but not calibrated intensity, and the steerability of LLMs at controlled intensities remains insufficiently understood. To address these limitations, we introduce VALUEFLOW, the first unified framework that spans extraction, evaluation, and steering with calibrated intensity control. The framework integrates three components: (i) HIVES, a hierarchical value embedding space that captures intra- and cross-theory value structure; (ii) the Value Intensity DataBase (VIDB), a large-scale resource of value-labeled texts with intensity estimates derived from ranking-based aggregation; and (iii) an anchor-based evaluator that produces consistent intensity scores for model outputs by ranking them against VIDB panels. Using VALUEFLOW, we conduct a comprehensive large-scale study across ten models and four value theories, identifying asymmetries in steerability and composition laws for multi-value control. This paper establishes a scalable infrastructure for evaluating and controlling value intensity, advancing pluralistic alignment of LLMs.

Deeply stacked KANs are practically impossible due to high training difficulties and substantial memory requirements. Consequently, existing studies can only incorporate few KAN layers, hindering the comprehensive exploration of KANs. This study overcomes these limitations and introduces the first fully KA-based deep model, demonstrating that KA-based layers can entirely replace traditional architectures in deep learning and achieve superior learning capacity. Specifically, (1) the proposed Share-activation KAN (SaKAN) reformulates Sprecher's variant of Kolmogorov-Arnold representation theorem, which achieves better optimization due to its simplified parameterization and denser training samples, to ease training difficulty, (2) this paper indicates that spline gradients contribute negligibly to training while consuming huge GPU memory, thus proposes the Grad-Free Spline to significantly reduce memory usage and computational overhead. (3) Building on these two innovations, our ALL U-KAN is the first representative implementation of fully KA-based deep model, where the proposed KA and KAonv layers completely replace FC and Conv layers. Extensive evaluations on three medical image segmentation tasks confirm the superiority of the full KA-based architecture compared to partial KA-based and traditional architectures, achieving all higher segmentation accuracy. Compared to directly deeply stacked KAN, ALL U-KAN achieves 10 times reduction in parameter count and reduces memory consumption by more than 20 times, unlocking the new explorations into deep KAN architectures.

Front-end personalization has traditionally relied on static designs or rule-based adaptations, which fail to fully capture user behavior patterns. This paper presents an AI driven approach for dynamic front-end personalization, where UI layouts, content, and features adapt in real-time based on predicted user behavior. We propose three strategies: dynamic layout adaptation using user path prediction, content prioritization through reinforcement learning, and a comparative analysis of AI-driven vs. rule-based personalization. Technical implementation details, algorithms, system architecture, and evaluation methods are provided to illustrate feasibility and performance gains.

The deployment of Large Language Models (LLMs) faces a critical bottleneck when handling lengthy inputs: the prohibitive memory footprint of the Key Value (KV) cache. To address this bottleneck, the token pruning paradigm leverages attention sparsity to selectively retain a small, critical subset of tokens. However, existing approaches fall short, with static methods risking irreversible information loss and dynamic strategies employing heuristics that insufficiently capture the query-dependent nature of token importance. We propose FASA, a novel framework that achieves query-aware token eviction by dynamically predicting token importance. FASA stems from a novel insight into RoPE: the discovery of functional sparsity at the frequency-chunk (FC) level. Our key finding is that a small, identifiable subset of "dominant" FCs consistently exhibits high contextual agreement with the full attention head. This provides a robust and computationally free proxy for identifying salient tokens. %making them a powerful and efficient proxy for token importance. Building on this insight, FASA first identifies a critical set of tokens using dominant FCs, and then performs focused attention computation solely on this pruned subset. % Since accessing only a small fraction of the KV cache, FASA drastically lowers memory bandwidth requirements and computational cost. Across a spectrum of long-context tasks, from sequence modeling to complex CoT reasoning, FASA consistently outperforms all token-eviction baselines and achieves near-oracle accuracy, demonstrating remarkable robustness even under constraint budgets. Notably, on LongBench-V1, FASA reaches nearly 100\% of full-KV performance when only keeping 256 tokens, and achieves 2.56$\times$ speedup using just 18.9\% of the cache on AIME24.

Vision Language Models (VLMs) typically assume complete modality input during inference. However, their effectiveness drops sharply when certain modalities are unavailable or incomplete. Current research primarily faces two dilemmas: Prompt-based methods struggle to restore missing yet indispensable features and impair generalization of VLMs. Imputation-based approaches, lacking effective guidance, are prone to generating semantically irrelevant noise. Restoring precise semantics while sustaining VLM generalization remains challenging. Therefore, we propose a general missing modality restoration strategy in this paper. We introduce an enhanced diffusion model as a pluggable mid-stage training module to effectively restore missing features. Our strategy introduces two key innovations: (I) Dynamic Modality Gating, which adaptively leverages conditional features to steer the generation of semantically consistent features; (II) Cross-Modal Mutual Learning mechanism, which bridges the semantic spaces of dual encoders to achieve bidirectional alignment. Zero-shot evaluations across benchmark datasets demonstrate that our approach outperforms existing baseline methods. Extensive experiments and ablation studies confirm our model as a robust and scalable extension for VLMs in missing modality scenarios, ensuring reliability across diverse missing rates and environments. Our code and models will be publicly available.

Deciding whether an agent possesses a model of its surrounding world is a fundamental step toward understanding its capabilities and limitations. In [10], it was shown that, within a particular framework, every almost optimal and general agent necessarily contains sufficient knowledge of its environment to allow an approximate reconstruction of it by querying the agent as a black box. This result relied on the assumptions that the agent is deterministic and that the environment is fully observable. In this work, we remove both assumptions by extending the theorem to stochastic agents operating in partially observable environments. Fundamentally, this shows that stochastic agents cannot avoid learning their environment through the usage of randomization. We also strengthen the result by weakening the notion of generality, proving that less powerful agents already contain a model of the world in which they operate.

Large language models (LLMs) have enabled a new class of agentic AI systems that reason, plan, and act by invoking external tools. However, most existing agentic architectures remain centralized and monolithic, limiting scalability, specialization, and interoperability. This paper proposes a framework for scalable agentic intelligence, termed the Internet of Agentic AI, in which autonomous, heterogeneous agents distributed across cloud and edge infrastructure dynamically form coalitions to execute task-driven workflows. We formalize a network-native model of agentic collaboration and introduce an incentive-compatible workflow-coalition feasibility framework that integrates capability coverage, network locality, and economic implementability. To enable scalable coordination, we formulate a minimum-effort coalition selection problem and propose a decentralized coalition formation algorithm. The proposed framework can operate as a coordination layer above the Model Context Protocol (MCP). A healthcare case study demonstrates how domain specialization, cloud-edge heterogeneity, and dynamic coalition formation enable scalable, resilient, and economically viable agentic workflows. This work lays the foundation for principled coordination and scalability in the emerging era of Internet of Agentic AI.

Teaching requires distilling a rich category distribution into a small set of informative exemplars. Although prior work shows that humans consider both representativeness and diversity when teaching, the computational principles underlying these tradeoffs remain unclear. We address this gap by modeling human exemplar selection using neural network feature representations and principled subset selection criteria. Novel visual categories were embedded along a one-dimensional morph continuum using pretrained vision models, and selection strategies varied in their emphasis on prototypicality, joint representativeness, and diversity. Adult participants selected one to three exemplars to teach a learner. Model-human comparisons revealed that strategies based on joint representativeness, or its combination with diversity, best captured human judgments, whereas purely prototypical or diversity-based strategies performed worse. Moreover, transformer-based representations consistently aligned more closely with human behavior than convolutional networks. These results highlight the potential utility of dataset distillation methods in machine learning as computational models for teaching.

Group Relative Policy Optimization (GRPO) has recently emerged as a practical recipe for aligning large language models with verifiable objectives. However, under sparse terminal rewards, GRPO often stalls because rollouts within a group frequently receive identical rewards, causing relative advantages to collapse and updates to vanish. We propose self-hint aligned GRPO with privileged supervision (SAGE), an on-policy reinforcement learning framework that injects privileged hints during training to reshape the rollout distribution under the same terminal verifier reward. For each prompt $x$, the model samples a compact hint $h$ (e.g., a plan or decomposition) and then generates a solution $τ$ conditioned on $(x,h)$. Crucially, the task reward $R(x,τ)$ is unchanged; hints only increase within-group outcome diversity under finite sampling, preventing GRPO advantages from collapsing under sparse rewards. At test time, we set $h=\varnothing$ and deploy the no-hint policy without any privileged information. Moreover, sampling diverse self-hints serves as an adaptive curriculum that tracks the learner's bottlenecks more effectively than fixed hints from an initial policy or a stronger external model. Experiments over 6 benchmarks with 3 LLMs show that SAGE consistently outperforms GRPO, on average +2.0 on Llama-3.2-3B-Instruct, +1.2 on Qwen2.5-7B-Instruct and +1.3 on Qwen3-4B-Instruct. The code is available at https://github.com/BaohaoLiao/SAGE.

While Large Reasoning Models (LRMs) have demonstrated impressive capabilities in solving complex tasks through the generation of long reasoning chains, this reliance on verbose generation results in significant latency and computational overhead. To address these challenges, we propose \textbf{CoSMo} (\textbf{Co}nsistency-Guided \textbf{S}plit-\textbf{M}erge \textbf{O}ptimization), a framework designed to eliminate structural redundancy rather than indiscriminately restricting token volume. Specifically, CoSMo utilizes a split-merge algorithm that dynamically refines reasoning chains by merging redundant segments and splitting logical gaps to ensure coherence. We then employ structure-aligned reinforcement learning with a novel segment-level budget to supervise the model in maintaining efficient reasoning structures throughout training. Extensive experiments across multiple benchmarks and backbones demonstrate that CoSMo achieves superior performance, improving accuracy by \textbf{3.3} points while reducing segment usage by \textbf{28.7\%} on average compared to reasoning efficiency baselines.

Traffic-density matrices from different days exhibit both low rank and stable correlations in their singular-vector subspaces. Leveraging this, we introduce SATORIS-N, a framework for imputing partially observed traffic-density by informed subspace priors from neighboring days. Our contribution is a subspace-aware semidefinite programming (SDP)} formulation of nuclear norm that explicitly informs the reconstruction with prior singular-subspace information. This convex formulation jointly enforces low rank and subspace alignment, providing a single global optimum and substantially improving accuracy under medium and high occlusion. We also study a lightweight implicit subspace-alignment} strategy in which matrices from consecutive days are concatenated to encourage alignment of spatial or temporal singular directions. Although this heuristic offers modest gains when missing rates are low, the explicit SDP approach is markedly more robust when large fractions of entries are missing. Across two real-world datasets (Beijing and Shanghai), SATORIS-N consistently outperforms standard matrix-completion methods such as SoftImpute, IterativeSVD, statistical, and even deep learning baselines at high occlusion levels. The framework generalizes to other spatiotemporal settings in which singular subspaces evolve slowly over time. In the context of intelligent vehicles and vehicle-to-everything (V2X) systems, accurate traffic-density reconstruction enables critical applications including cooperative perception, predictive routing, and vehicle-to-infrastructure (V2I) communication optimization. When infrastructure sensors or vehicle-reported observations are incomplete - due to communication dropouts, sensor occlusions, or sparse connected vehicle penetration-reliable imputation becomes essential for safe and efficient autonomous navigation.

The rapid expansion of online shopping has increased the demand for timely parcel delivery, compelling logistics service providers to enhance the efficiency, agility, and predictability of their hub networks. In order to solve the problem, we propose a novel deep learning-based ensemble framework that leverages historical arrival patterns and real-time parcel status updates to forecast upcoming workloads at logistic hubs. This approach not only facilitates the generation of short-term forecasts, but also improves the accuracy of future hub workload predictions for more strategic planning and resource management. Empirical tests of the algorithm, conducted through a case study of a major city's parcel logistics, demonstrate the ensemble method's superiority over both traditional forecasting techniques and standalone deep learning models. Our findings highlight the significant potential of this method to improve operational efficiency in logistics hubs and advocate for its broader adoption.

Visual token pruning is a promising approach for reducing the computational cost of vision-language models (VLMs), and existing methods often rely on early pruning decisions to improve efficiency. While effective on coarse-grained reasoning tasks, they suffer from significant performance degradation on tasks requiring fine-grained visual details. Through layer-wise analysis, we reveal substantial discrepancies in visual token importance across layers, showing that tokens deemed unimportant at shallow layers can later become highly relevant for text-conditioned reasoning. To avoid irreversible critical information loss caused by premature pruning, we introduce a new pruning paradigm, termed bypass, which preserves unselected visual tokens and forwards them to subsequent pruning stages for re-evaluation. Building on this paradigm, we propose SwiftVLM, a simple and training-free method that performs pruning at model-specific layers with strong visual token selection capability, while enabling independent pruning decisions across layers. Experiments across multiple VLMs and benchmarks demonstrate that SwiftVLM consistently outperforms existing pruning strategies, achieving superior accuracy-efficiency trade-offs and more faithful visual token selection behavior.

Large language Model (LLM)-assisted algorithm discovery is an iterative, black-box optimization process over programs to approximatively solve a target task, where an LLM proposes candidate programs and an external evaluator provides task feedback. Despite intense recent research on the topic and promising results, how can the LLM internal representation of the space of possible programs be maximally exploited to improve performance is an open question. Here, we introduce Contrastive Concept-Tree Search (CCTS), which extracts a hierarchical concept representation from the generated programs and learns a contrastive concept model that guides parent selection. By reweighting parents using a likelihood-ratio score between high- and low-performing solutions, CCTS biases search toward useful concept combinations and away from misleading ones, providing guidance through an explicit concept hierarchy rather than the algorithm lineage constructed by the LLM. We show that CCTS improves search efficiency over fitness-based baselines and produces interpretable, task-specific concept trees across a benchmark of open Erdős-type combinatorics problems. Our analysis indicates that the gains are driven largely by learning which concepts to avoid. We further validate these findings in a controlled synthetic algorithm-discovery environment, which reproduces qualitatively the search dynamics observed with the LLMs.

Multi-agent LLM frameworks are widely used to accelerate the development of agent systems powered by large language models (LLMs). These frameworks impose distinct architectural structures that govern how agents interact, store information, and coordinate tasks. However, their impact on system performance remains poorly understood. This gap is critical, as architectural choices alone can induce order-of-magnitude differences in latency and throughput, as well as substantial variation in accuracy and scalability. Addressing this challenge requires (i) jointly evaluating multiple capabilities, such as orchestration overhead, memory behavior, planning, specialization, and coordination, and (ii) conducting these evaluations under controlled, framework-level conditions to isolate architectural effects. Existing benchmarks focus on individual capabilities and lack standardized framework-level evaluation. We address these limitations by (i) introducing an architectural taxonomy for systematically comparing multi-agent LLM frameworks along fundamental dimensions, and (ii) developing MAFBench, a unified evaluation suite that integrates existing benchmarks under a standardized execution pipeline. Using MAFBench, we conduct a controlled empirical study across several widely used frameworks. Our results show that framework-level design choices alone can increase latency by over 100x, reduce planning accuracy by up to 30%, and lower coordination success from above 90% to below 30%. Finally, we translate our findings into concrete architectural design principles and framework selection guidance, and outline promising future research directions.

Understanding how humans and machines learn from sparse data is central to cognitive science and machine learning. Using a species-fair design, we compare children and convolutional neural networks (CNNs) in a few-shot semi-supervised category learning task. Both learners are exposed to novel object categories under identical conditions. Learners receive mixtures of labeled and unlabeled exemplars while we vary supervision (1/3/6 labels), target feature (size, shape, pattern), and perceptual alignment (high/low). We find that children generalize rapidly from minimal labels but show strong feature-specific biases and sensitivity to alignment. CNNs show a different interaction profile: added supervision improves performance, but both alignment and feature structure moderate the impact additional supervision has on learning. These results show that human-model comparisons must be drawn under the right conditions, emphasizing interactions among supervision, feature structure, and alignment rather than overall accuracy.

Data augmentation has long been a cornerstone for reducing overfitting in vision models, with methods like AutoAugment automating the design of task-specific augmentations. Recent advances in generative models, such as conditional diffusion and few-shot NeRFs, offer a new paradigm for data augmentation by synthesizing data with significantly greater diversity and realism. However, unlike traditional augmentations like cropping or rotation, these methods introduce substantial changes that enhance robustness but also risk degrading performance if the augmentations are poorly matched to the task. In this work, we present EvoAug, an automated augmentation learning pipeline, which leverages these generative models alongside an efficient evolutionary algorithm to learn optimal task-specific augmentations. Our pipeline introduces a novel approach to image augmentation that learns stochastic augmentation trees that hierarchically compose augmentations, enabling more structured and adaptive transformations. We demonstrate strong performance across fine-grained classification and few-shot learning tasks. Notably, our pipeline discovers augmentations that align with domain knowledge, even in low-data settings. These results highlight the potential of learned generative augmentations, unlocking new possibilities for robust model training.

Post-Training Quantization (PTQ) is essential for deploying Large Language Models (LLMs) on memory-constrained devices, yet it renders models static and difficult to fine-tune. Standard fine-tuning paradigms, including Reinforcement Learning (RL), fundamentally rely on backpropagation and high-precision weights to compute gradients. Thus they cannot be used on quantized models, where the parameter space is discrete and non-differentiable. While Evolution Strategies (ES) offer a backpropagation-free alternative, optimization of the quantized parameters can still fail due to vanishing or inaccurate gradient. This paper introduces Quantized Evolution Strategies (QES), an optimization paradigm that performs full-parameter fine-tuning directly in the quantized space. QES is based on two innovations: (1) it integrates accumulated error feedback to preserve high-precision gradient signals, and (2) it utilizes a stateless seed replay to reduce memory usage to low-precision inference levels. QES significantly outperforms the state-of-the-art zeroth-order fine-tuning method on arithmetic reasoning tasks, making direct fine-tuning for quantized models possible. It therefore opens up the possibility for scaling up LLMs entirely in the quantized space. The source code is available at https://github.com/dibbla/Quantized-Evolution-Strategies .

Bayesian deep learning (BDL) provides a principled framework for reliable uncertainty quantification by combining deep neural networks with Bayesian inference. A central challenge in BDL lies in the design of informative prior distributions that scale effectively to high-dimensional data. Recent functional variational inference (VI) approaches address this issue by imposing priors directly in function space; however, most existing methods rely on Gaussian process (GP) priors, whose expressiveness and generalisation capabilities become limited in high-dimensional regimes. In this work, we propose VLM-FS-EB, a novel function-space empirical Bayes regularisation framework, leveraging large vision-language models (VLMs) to generates semantically meaningful context points. These synthetic samples are then used VLMs for embeddings to construct expressive functional priors. Furthermore, the proposed method is evaluated against various baselines, and experimental results demonstrate that our method consistently improves predictive performance and yields more reliable uncertainty estimates, particularly in out-of-distribution (OOD) detection tasks and data-scarce regimes.

Rapid innovations in AI and large language models (LLMs) have accelerated the adoption of digital learning, particularly beyond formal education. What began as an emergency response during COVID-19 has shifted from a supplementary resource to an essential pillar of education. Understanding how digital learning continues to evolve for adult and lifelong learners is therefore increasingly important. This study examines how various demographics interact with digital learning platforms, focusing on the learner motivations, the effectiveness of gamification in digital learning, and the integration of AI. Using multi survey data from 200 respondents and advanced analytics, our findings reveal a notable increase in the perceived relevance of digital learning after the pandemic, especially among young adults and women, coinciding with the rise of LLM-powered AI tools that support personalized learning. We aim to provide actionable insights for businesses, government policymakers, and educators seeking to optimize their digital learning offerings to meet evolving workforce needs.

This paper introduces OMAR: One Model, All Roles, a reinforcement learning framework that enables AI to develop social intelligence through multi-turn, multi-agent conversational self-play. Unlike traditional paradigms that rely on static, single-turn optimizations, OMAR allows a single model to role-play all participants in a conversation simultaneously, learning to achieve long-term goals and complex social norms directly from dynamic social interaction. To ensure training stability across long dialogues, we implement a hierarchical advantage estimation that calculates turn-level and token-level advantages. Evaluations in the SOTOPIA social environment and Werewolf strategy games show that our trained models develop fine-grained, emergent social intelligence, such as empathy, persuasion, and compromise seeking, demonstrating the effectiveness of learning collaboration even under competitive scenarios. While we identify practical challenges like reward hacking, our results show that rich social intelligence can emerge without human supervision. We hope this work incentivizes further research on AI social intelligence in group conversations.

We introduce ChemPro, a progressive benchmark with 4100 natural language question-answer pairs in Chemistry, across 4 coherent sections of difficulty designed to assess the proficiency of Large Language Models (LLMs) in a broad spectrum of general chemistry topics. We include Multiple Choice Questions and Numerical Questions spread across fine-grained information recall, long-horizon reasoning, multi-concept questions, problem-solving with nuanced articulation, and straightforward questions in a balanced ratio, effectively covering Bio-Chemistry, Inorganic-Chemistry, Organic-Chemistry and Physical-Chemistry. ChemPro is carefully designed analogous to a student's academic evaluation for basic to high-school chemistry. A gradual increase in the question difficulty rigorously tests the ability of LLMs to progress from solving basic problems to solving more sophisticated challenges. We evaluate 45+7 state-of-the-art LLMs, spanning both open-source and proprietary variants, and our analysis reveals that while LLMs perform well on basic chemistry questions, their accuracy declines with different types and levels of complexity. These findings highlight the critical limitations of LLMs in general scientific reasoning and understanding and point towards understudied dimensions of difficulty, emphasizing the need for more robust methodologies to improve LLMs.

From a pragmatic perspective, this study systematically evaluates the differences in performance among representative large language models (LLMs) in recognizing politeness, impoliteness, and mock politeness phenomena in Chinese. Addressing the existing gaps in pragmatic comprehension, the research adopts the frameworks of Rapport Management Theory and the Model of Mock Politeness to construct a three-category dataset combining authentic and simulated Chinese discourse. Six representative models, including GPT-5.1 and DeepSeek, were selected as test subjects and evaluated under four prompting conditions: zero-shot, few-shot, knowledge-enhanced, and hybrid strategies. This study serves as a meaningful attempt within the paradigm of ``Great Linguistics,'' offering a novel approach to applying pragmatic theory in the age of technological transformation. It also responds to the contemporary question of how technology and the humanities may coexist, representing an interdisciplinary endeavor that bridges linguistic technology and humanistic reflection.

Generative Artificial Intelligence (GenAI) is increasingly integrated into photo applications on personal devices, making editing photographs easier than ever while potentially influencing the memories they represent. This study explores how and why people use GenAI to edit personal photos and how this shapes their remembering experience. We conducted a two-phase qualitative study with 12 participants: a photo editing session using a GenAI tool guided by the Remembering Experience (RX) dimensions, followed by semi-structured interviews where participants reflected on the editing process and results. Findings show that participants prioritised felt memory over factual accuracy. For different photo elements, environments were modified easily, however, editing was deemed unacceptable if it touched upon a person's identity. Editing processes brought positive and negative impacts, and itself also became a remembering experience. We further discuss potential benefits and risks of GenAI editing for remembering purposes and propose design implications for responsible GenAI.

Instruction tuning is now the default way to train and adapt large language models, but many instruction--input--output pairs are only weakly specified: for a given input, the same output can remain plausible under several alternative instructions. This raises a simple question: \emph{does the instruction uniquely determine the target output?} We propose the \textbf{Task--Specificity Score (TSS)} to quantify how much an instruction matters for predicting its output, by contrasting the true instruction against plausible alternatives for the same input. We further introduce \textbf{TSS++}, which uses hard alternatives and a small quality term to mitigate easy-negative effects. Across three instruction datasets (\textsc{Alpaca}, \textsc{Dolly-15k}, \textsc{NI-20}) and three open LLMs (Gemma, Llama, Qwen), we show that selecting task-specific examples improves downstream performance under tight token budgets and complements quality-based filters such as perplexity and IFD.

Many strong decoding methods for text generation follow a sample-and-rerank paradigm: they draw multiple candidates, score each under a utility (reward) function using consensus across samples, and return the best one. Although effective, these methods incur high computational costs during inference due to repeated sampling and scoring. Prior attempts to amortize inference-time computation typically rely on gold references, teacher labels, or curated preference data, increasing dataset construction effort and the demand for high-fidelity reward models. We propose Consensus Group Relative Policy Optimization (C-GRPO), which distills Minimum Bayes Risk (MBR) decoding into training by formulating the consensus utility as a group-relative objective within GRPO. C-GRPO requires only a utility function and policy samples, without gold references or explicit preference labels. Under ideal conditions, we show that the objective function of C-GRPO is directionally aligned with the gradient of the expected-utility objective underlying MBR decoding, leading to a convergence guarantee. Experiments on machine translation (WMT 2024) and text summarization (XSum) demonstrate that C-GRPO successfully achieves performance comparable to MBR decoding without the associated inference-time overhead, while outperforming reference-free baseline methods.

Large Language Models (LLMs) are increasingly deployed as agents that operate in real-world environments, introducing safety risks beyond linguistic harm. Existing agent safety evaluations rely on risk-oriented tasks tailored to specific agent settings, resulting in limited coverage of safety risk space and failing to assess agent safety behavior during long-horizon, interactive task execution in complex real-world deployments. Moreover, their specialization to particular agent settings limits adaptability across diverse agent configurations. To address these limitations, we propose Risky-Bench, a framework that enables systematic agent safety evaluation grounded in real-world deployment. Risky-Bench organizes evaluation around domain-agnostic safety principles to derive context-aware safety rubrics that delineate safety space, and systematically evaluates safety risks across this space through realistic task execution under varying threat assumptions. When applied to life-assist agent settings, Risky-Bench uncovers substantial safety risks in state-of-the-art agents under realistic execution conditions. Moreover, as a well-structured evaluation pipeline, Risky-Bench is not confined to life-assist scenarios and can be adapted to other deployment settings to construct environment-specific safety evaluations, providing an extensible methodology for agent safety assessment.

Expanding multimodal representations to novel modalities is constrained by reliance on large-scale paired datasets (e.g., text-image, text-audio, text-3D, text-molecule), which are costly and often infeasible in domains requiring expert annotation such as medical imaging and molecular analysis. We introduce TextME, the first text-only modality expansion framework, to the best of our knowledge, projecting diverse modalities into LLM embedding space as a unified anchor. Our approach exploits the geometric structure of pretrained contrastive encoders to enable zero-shot cross-modal transfer using only text descriptions, without paired supervision. We empirically validate that such consistent modality gaps exist across image, video, audio, 3D, X-ray, and molecular domains, demonstrating that text-only training can preserve substantial performance of pretrained encoders. We further show that our framework enables emergent cross-modal retrieval between modality pairs not explicitly aligned during training (e.g., audio-to-image, 3D-to-image). These results establish text-only training as a practical alternative to paired supervision for modality expansion.

Professional job recommendation involves a complex bipartite matching process that must reconcile a candidate's subjective preference with an employer's objective qualification. While Large Language Models (LLMs) are well-suited for modeling the rich semantics of resumes and job descriptions, existing paradigms often collapse these two decision dimensions into a single interaction signal, yielding confounded supervision under recruitment-funnel censoring and limiting policy controllability. To address these challenges, We propose JobRec, a generative job recommendation framework for de-conflating preference and qualification via constrained dual-perspective reasoning. JobRec introduces a Unified Semantic Alignment Schema that aligns candidate and job attributes into structured semantic layers, and a Two-Stage Cooperative Training Strategy that learns decoupled experts to separately infer preference and qualification. Building on these experts, a Lagrangian-based Policy Alignment module optimizes recommendations under explicit eligibility requirements, enabling controllable trade-offs. To mitigate data scarcity, we construct a synthetic dataset refined by experts. Experiments show that JobRec consistently outperforms strong baselines and provides improved controllability for strategy-aware professional matching.

We propose PRISM, an optimizer that enhances first-order spectral descent methods like Muon with partial second-order information. It constructs an efficient, low-rank quasi-second-order preconditioner via innovation-augmented polar decomposition. This mechanism enables PRISM to perform anisotropic spectral shaping, which adaptively suppresses updates in high-variance subspaces while preserving update strength in signal-dominated directions. Crucially, this is achieved with minimal computational overhead and zero additional memory compared to first-order baselines. PRISM demonstrates a practical strategy for integrating curvature-adaptive properties into the spectral optimization paradigm.

Modern Large Language Models (LLMs) have shown rapid improvements in reasoning capabilities, driven largely by reinforcement learning (RL) with verifiable rewards. Here, we ask whether these LLMs can self-improve without the need for additional training. We identify two core challenges for such systems: (i) efficiently generating diverse, high-quality candidate solutions, and (ii) reliably selecting correct answers in the absence of ground-truth supervision. To address these challenges, we propose Test-time Recursive Thinking (TRT), an iterative self-improvement framework that conditions generation on rollout-specific strategies, accumulated knowledge, and self-generated verification signals. Using TRT, open-source models reach 100% accuracy on AIME-25/24, and on LiveCodeBench's most difficult problems, closed-source models improve by 10.4-14.8 percentage points without external feedback.

The heterogeneity in the organization of software engineering (SE) research historically exists, i.e., funded research model and hands-on model, which makes software engineering become a thriving interdisciplinary field in the last 50 years. However, the funded research model is becoming dominant in SE research recently, indicating such heterogeneity has been seriously and systematically threatened. In this essay, we first explain why the heterogeneity is needed in the organization of SE research, then present the current trend of SE research nowadays, as well as the consequences and potential futures. The choice is at our hands, and we urge our community to seriously consider maintaining the heterogeneity in the organization of software engineering research.

Quadruped robots are used for primary searches during the early stages of indoor fires. A typical primary search involves quickly and thoroughly looking for victims under hazardous conditions and monitoring flammable materials. However, situational awareness in complex indoor environments and rapid stair climbing across different staircases remain the main challenges for robot-assisted primary searches. In this project, we designed a two-stage end-to-end deep reinforcement learning (RL) approach to optimize both navigation and locomotion. In the first stage, the quadrupeds, Unitree Go2, were trained to climb stairs in Isaac Lab's pyramid-stair terrain. In the second stage, the quadrupeds were trained to climb various realistic indoor staircases in the Isaac Lab engine, with the learned policy transferred from the previous stage. These indoor staircases are straight, L-shaped, and spiral, to support climbing tasks in complex environments. This project explores how to balance navigation and locomotion and how end-to-end RL methods can enable quadrupeds to adapt to different stair shapes. Our main contributions are: (1) A two-stage end-to-end RL framework that transfers stair-climbing skills from abstract pyramid terrain to realistic indoor stair topologies. (2) A centerline-based navigation formulation that enables unified learning of navigation and locomotion without hierarchical planning. (3) Demonstration of policy generalization across diverse staircases using only local height-map perception. (4) An empirical analysis of success, efficiency, and failure modes under increasing stair difficulty.

The Homotopy paradigm, a general principle for solving challenging problems, appears across diverse domains such as robust optimization, global optimization, polynomial root-finding, and sampling. Practical solvers for these problems typically follow a predictor-corrector (PC) structure, but rely on hand-crafted heuristics for step sizes and iteration termination, which are often suboptimal and task-specific. To address this, we unify these problems under a single framework, which enables the design of a general neural solver. Building on this unified view, we propose Neural Predictor-Corrector (NPC), which replaces hand-crafted heuristics with automatically learned policies. NPC formulates policy selection as a sequential decision-making problem and leverages reinforcement learning to automatically discover efficient strategies. To further enhance generalization, we introduce an amortized training mechanism, enabling one-time offline training for a class of problems and efficient online inference on new instances. Experiments on four representative homotopy problems demonstrate that our method generalizes effectively to unseen instances. It consistently outperforms classical and specialized baselines in efficiency while demonstrating superior stability across tasks, highlighting the value of unifying homotopy methods into a single neural framework.

Detecting whether a model has been poisoned is a longstanding problem in AI security. In this work, we present a practical scanner for identifying sleeper agent-style backdoors in causal language models. Our approach relies on two key findings: first, sleeper agents tend to memorize poisoning data, making it possible to leak backdoor examples using memory extraction techniques. Second, poisoned LLMs exhibit distinctive patterns in their output distributions and attention heads when backdoor triggers are present in the input. Guided by these observations, we develop a scalable backdoor scanning methodology that assumes no prior knowledge of the trigger or target behavior and requires only inference operations. Our scanner integrates naturally into broader defensive strategies and does not alter model performance. We show that our method recovers working triggers across multiple backdoor scenarios and a broad range of models and fine-tuning methods.

Large Language Models (LLMs) have achieved significant success in complex reasoning but remain bottlenecked by reliance on expert-annotated data and external verifiers. While existing self-evolution paradigms aim to bypass these constraints, they often fail to identify the optimal learning zone and risk reinforcing collective hallucinations and incorrect priors through flawed internal feedback. To address these challenges, we propose \underline{A}utonomous \underline{E}volutionary \underline{R}easoning \underline{O}ptimization (AERO), an unsupervised framework that achieves autonomous reasoning evolution by internalizing self-questioning, answering, and criticism within a synergistic dual-loop system. Inspired by the \textit{Zone of Proximal Development (ZPD)} theory, AERO utilizes entropy-based positioning to target the ``solvability gap'' and employs Independent Counterfactual Correction for robust verification. Furthermore, we introduce a Staggered Training Strategy to synchronize capability growth across functional roles and prevent curriculum collapse. Extensive evaluations across nine benchmarks spanning three domains demonstrate that AERO achieves average performance improvements of 4.57\% on Qwen3-4B-Base and 5.10\% on Qwen3-8B-Base, outperforming competitive baselines. Code is available at https://github.com/mira-ai-lab/AERO.

Preserving geometric structure is important in learning. We propose a unified class of geometry-aware architectures that interleave geometric updates between layers, where both projection layers and intrinsic exponential map updates arise as discretizations of projected dynamical systems on manifolds (with or without boundary). Within this framework, we establish universal approximation results for constrained neural ODEs. We also analyze architectures that enforce geometry only at the output, proving a separate universal approximation property that enables direct comparison to interleaved designs. When the constraint set is unknown, we learn projections via small-time heat-kernel limits, showing diffusion/flow-matching can be used as data-based projections. Experiments on dynamics over S^2 and SO(3), and diffusion on S^{d-1}-valued features demonstrate exact feasibility for analytic updates and strong performance for learned projections

Dynamic scheduling of task graphs is often addressed without revisiting prior task allocations, with a primary focus on minimizing makespan. We study controlled schedule preemption, introducing the Last-K Preemption model, which selectively reschedules recent task graphs while preserving earlier allocations. Using synthetic, RIoTBench, WFCommons, and adversarial workloads, we compare preemptive, non-preemptive, and partial-preemptive strategies across makespan, fairness, utilization, and runtime. Results show moderate preemption can match most makespan and utilization gains of full preemption while maintaining fairness and low overhead.

Standard training pipelines for large language models (LLMs) are typically unidirectional, progressing from pre-training to post-training. However, the potential for a bidirectional process--where insights from post-training retroactively improve the pre-trained foundation--remains unexplored. We aim to establish a self-reinforcing flywheel: a cycle in which reinforcement learning (RL)-tuned model strengthens the base model, which in turn enhances subsequent post-training performance, requiring no specially trained teacher or reference model. To realize this, we analyze training dynamics and identify the mid-training (annealing) phase as a critical turning point for model capabilities. This phase typically occurs at the end of pre-training, utilizing high-quality corpora under a rapidly decaying learning rate. Building upon this insight, we introduce ReMiT (Reinforcement Learning-Guided Mid-Training). Specifically, ReMiT leverages the reasoning priors of RL-tuned models to dynamically reweight tokens during the mid-training phase, prioritizing those pivotal for reasoning. Empirically, ReMiT achieves an average improvement of 3\% on 10 pre-training benchmarks, spanning math, code, and general reasoning, and sustains these gains by over 2\% throughout the post-training pipeline. These results validate an iterative feedback loop, enabling continuous and self-reinforcing evolution of LLMs.

Reinforcement Learning (RL) and Supervised Fine-Tuning (SFT) are the two dominant paradigms for enhancing Large Language Model (LLM) performance on downstream tasks. While RL generally preserves broader model capabilities (retention) better than SFT, it comes with significant costs: complex reward engineering, instability, and expensive on-policy sampling. In contrast, SFT is efficient but brittle, often suffering from catastrophic forgetting due to $\textbf{Supervision Mismatch}$: the divergence between the model's evolving policy and static training labels. We address this trade-off with $\textbf{Trajectory-Mixed Supervision (TMS)}$, a reward-free framework that approximates the on-policy benefits of RL by creating a dynamic curriculum from the model's own historical checkpoints. TMS minimizes $\textit{Policy-Label Divergence (PLD)}$, preventing the mode collapse that drives forgetting in standard SFT. Experiments across reasoning (MATH, GSM8K) and instruction-following benchmarks demonstrate that TMS effectively shifts the accuracy--retention Pareto frontier. While RL remains the gold standard for retention, TMS significantly outperforms standard and iterative SFT, bridging the gap to RL without requiring reward models or verifiers. Mechanistic analysis confirms that PLD drift accurately predicts forgetting and that TMS successfully mitigates this drift.

Growing renewable penetration introduces substantial uncertainty into power system operations, necessitating frequent adaptation of dispatch objectives and constraints and challenging expertise-intensive, near-real-time modeling workflows. Large Language Models (LLMs) provide a promising avenue for automating this process by translating natural-language (NL) operational requirements into executable optimization models via semantic reasoning and code synthesis. Yet existing LLM datasets and benchmarks for optimization modeling primarily target coarse-grained cross-domain generalization, offering limited, rigorous evaluation in power-system settings, particularly for Optimal Power Flow (OPF). We therefore introduce \textbf{ProOPF-D} and \textbf{ProOPF-B}, a dataset and benchmark for professional-grade OPF modeling: ProOPF-D contains 12K instances pairing NL requests with parameter adjustments and structural extensions to a canonical OPF, together with executable implementations; ProOPF-B provides 121 expert-annotated test cases with ground-truth code, enabling end-to-end evaluation under both concrete and abstract OPF modeling regimes.

Entropic optimal transport (EOT) via Sinkhorn iterations is widely used in modern machine learning, yet GPU solvers remain inefficient at scale. Tensorized implementations suffer quadratic HBM traffic from dense $n\times m$ interactions, while existing online backends avoid storing dense matrices but still rely on generic tiled map-reduce reduction kernels with limited fusion. We present \textbf{FlashSinkhorn}, an IO-aware EOT solver for squared Euclidean cost that rewrites stabilized log-domain Sinkhorn updates as row-wise LogSumExp reductions of biased dot-product scores, the same normalization as transformer attention. This enables FlashAttention-style fusion and tiling: fused Triton kernels stream tiles through on-chip SRAM and update dual potentials in a single pass, substantially reducing HBM IO per iteration while retaining linear-memory operations. We further provide streaming kernels for transport application, enabling scalable first- and second-order optimization. On A100 GPUs, FlashSinkhorn achieves up to $32\times$ forward-pass and $161\times$ end-to-end speedups over state-of-the-art online baselines on point-cloud OT, improves scalability on OT-based downstream tasks. For reproducibility, we release an open-source implementation at https://github.com/ot-triton-lab/ot_triton.

One of the chronic problems of deep-learning models is shortcut learning. In a case where the majority of training data are dominated by a certain feature, neural networks prefer to learn such a feature even if the feature is not generalizable outside the training set. Based on the framework of Neural Tangent Kernel (NTK), we analyzed the case of linear neural networks to derive some important properties of shortcut learning. We defined a feature of a neural network as an eigenfunction of NTK. Then, we found that shortcut features correspond to features with larger eigenvalues when the shortcuts stem from the imbalanced number of samples in the clustered distribution. We also showed that the features with larger eigenvalues still have a large influence on the neural network output even after training, due to data variances in the clusters. Such a preference for certain features remains even when a margin of a neural network output is controlled, which shows that the max-margin bias is not the only major reason for shortcut learning. These properties of linear neural networks are empirically extended for more complex neural networks as a two-layer fully-connected ReLU network and a ResNet-18.

Real-world scenes are inherently crowded. Hence, estimating 3D poses of all nearby humans, tracking their movements over time, and understanding their activities within social and environmental contexts are essential for many applications, such as autonomous driving, robot perception, robot navigation, and human-robot interaction. However, most existing 3D human pose estimation datasets primarily focus on single-person scenes or are collected in controlled laboratory environments, which restricts their relevance to real-world applications. To bridge this gap, we introduce JRDB-Pose3D, which captures multi-human indoor and outdoor environments from a mobile robotic platform. JRDB-Pose3D provides rich 3D human pose annotations for such complex and dynamic scenes, including SMPL-based pose annotations with consistent body-shape parameters and track IDs for each individual over time. JRDB-Pose3D contains, on average, 5-10 human poses per frame, with some scenes featuring up to 35 individuals simultaneously. The proposed dataset presents unique challenges, including frequent occlusions, truncated bodies, and out-of-frame body parts, which closely reflect real-world environments. Moreover, JRDB-Pose3D inherits all available annotations from the JRDB dataset, such as 2D pose, information about social grouping, activities, and interactions, full-scene semantic masks with consistent human- and object-level tracking, and detailed annotations for each individual, such as age, gender, and race, making it a holistic dataset for a wide range of downstream perception and human-centric understanding tasks.

Evaluating mathematical reasoning in LLMs is constrained by limited benchmark sizes and inherent model stochasticity, yielding high-variance accuracy estimates and unstable rankings across platforms. On difficult problems, an LLM may fail to produce a correct final answer, yet still provide reliable pairwise comparison signals indicating which of two candidate solutions is better. We leverage this observation to design a statistically efficient evaluation framework that combines standard labeled outcomes with pairwise comparison signals obtained by having models judge auxiliary reasoning chains. Treating these comparison signals as control variates, we develop a semiparametric estimator based on the efficient influence function (EIF) for the setting where auxiliary reasoning chains are observed. This yields a one-step estimator that achieves the semiparametric efficiency bound, guarantees strict variance reduction over naive sample averaging, and admits asymptotic normality for principled uncertainty quantification. Across simulations, our one-step estimator substantially improves ranking accuracy, with gains increasing as model output noise grows. Experiments on GPQA Diamond, AIME 2025, and GSM8K further demonstrate more precise performance estimation and more reliable model rankings, especially in small-sample regimes where conventional evaluation is pretty unstable.

We introduce Speech-to-Spatial, a referent disambiguation framework that converts verbal remote-assistance instructions into spatially grounded AR guidance. Unlike prior systems that rely on additional cues (e.g., gesture, gaze) or manual expert annotations, Speech-to-Spatial infers the intended target solely from spoken references (speech input). Motivated by our formative study of speech referencing patterns, we characterize recurring ways people specify targets (Direct Attribute, Relational, Remembrance, and Chained) and ground them to our object-centric relational graph. Given an utterance, referent cues are parsed and rendered as persistent in-situ AR visual guidance, reducing iterative micro-guidance ("a bit more to the right", "now, stop.") during remote guidance. We demonstrate the use cases of our system with remote guided assistance and intent disambiguation scenarios. Our evaluation shows that Speechto-Spatial improves task efficiency, reduces cognitive load, and enhances usability compared to a conventional voice-only baseline, transforming disembodied verbal instruction into visually explainable, actionable guidance on a live shared view.

Co-design is essential for grounding embodied artificial intelligence (AI) systems in real-world contexts, especially high-stakes domains such as healthcare. While prior work has explored multidisciplinary collaboration, iterative prototyping, and support for non-technical participants, few have interwoven these into a sustained co-design process. Such efforts often target one context and low-fidelity stages, limiting the generalizability of findings and obscuring how participants' ideas evolve. To address these limitations, we conducted a 14-week workshop with a multidisciplinary team of 22 participants, centered around how embodied AI can reduce non-value-added task burdens in three healthcare settings: emergency departments, long-term rehabilitation facilities, and sleep disorder clinics. We found that the iterative progression from abstract brainstorming to high-fidelity prototypes, supported by educational scaffolds, enabled participants to understand real-world trade-offs and generate more deployable solutions. We propose eight guidelines for co-designing more considerate embodied AI: attuned to context, responsive to social dynamics, mindful of expectations, and grounded in deployment. Project Page: https://byc-sophie.github.io/Towards-Considerate-Embodied-AI/

Multi-Agent Systems (MAS) built on Large Language Models (LLMs) often exhibit high variance in their reasoning trajectories. Process verification, which evaluates intermediate steps in trajectories, has shown promise in general reasoning settings, and has been suggested as a potential tool for guiding coordination of MAS; however, its actual effectiveness in MAS remains unclear. To fill this gap, we present MAS-ProVe, a systematic empirical study of process verification for multi-agent systems (MAS). Our study spans three verification paradigms (LLM-as-a-Judge, reward models, and process reward models), evaluated across two levels of verification granularity (agent-level and iteration-level). We further examine five representative verifiers and four context management strategies, and conduct experiments over six diverse MAS frameworks on multiple reasoning benchmarks. We find that process-level verification does not consistently improve performance and frequently exhibits high variance, highlighting the difficulty of reliably evaluating partial multi-agent trajectories. Among the methods studied, LLM-as-a-Judge generally outperforms reward-based approaches, with trained judges surpassing general-purpose LLMs. We further observe a small performance gap between LLMs acting as judges and as single agents, and identify a context-length-performance trade-off in verification. Overall, our results suggest that effective and robust process verification for MAS remains an open challenge, requiring further advances beyond current paradigms. Code is available at https://github.com/Wang-ML-Lab/MAS-ProVe.

Federated learning enables collaborative model training across decentralized clients under privacy constraints. Quantum computing offers potential for alleviating computational and communication burdens in federated learning, yet hybrid classical-quantum federated learning remains susceptible to performance degradation under non-IID data. To address this,we propose FEDCOMPASS, a layered aggregation framework for hybrid classical-quantum federated learning. FEDCOMPASS employs spectral clustering to group clients by class distribution similarity and performs cluster-wise aggregation for classical feature extractors. For quantum parameters, it uses circular mean aggregation combined with adaptive optimization to ensure stable global updates. Experiments on three benchmark datasets show that FEDCOMPASS improves test accuracy by up to 10.22% and enhances convergence stability under non-IID settings, outperforming six strong federated learning baselines.

The rapid growth in the parameter scale of large language models (LLMs) has created a high demand for efficient compression techniques. As a hardware-agnostic and highly compatible technique, low-rank compression has been widely adopted. However, existing methods typically compress each layer independently by minimizing per-layer reconstruction error, overlooking a critical limitation: the reconstruction error propagates and accumulates through the network, which leads to amplified global deviations from the full-precision baseline. To address this, we propose Self-Adaptive Error Suppression SVD (SAES-SVD), a LLMs compression framework that jointly optimizes intra-layer reconstruction and inter-layer error compensation. SAES-SVD is composed of two novel components: (1) Cumulative Error-Aware Layer Compression (CEALC), which formulates the compression objective as a combination of local reconstruction and weighted cumulative error compensation. Based on it, we derive a closed-form low-rank solution relied on second-order activation statistics, which explicitly aligns each layer's output with its full-precision counterpart to compensate for accumulated errors. (2) Adaptive Collaborative Error Suppression (ACES), which automatically adjusts the weighting coefficient to enhance the low-rank structure of the compression objective in CEALC. Specifically, the coefficient is optimized to maximize the ratio between the Frobenius norm of the compressed layer's output and that of the compression objective under a fixed rank, thus ensuring that the rank budget is utilized effectively. Extensive experiments across multiple LLM architectures and tasks show that, without fine-tuning or mixed-rank strategies, SAES-SVD consistently improves post-compression performance.

Performative prediction characterizes environments where predictive models alter the very data distributions they aim to forecast, triggering complex feedback loops. While prior research treats single-agent and multi-agent performativity as distinct phenomena, this paper introduces a unified statistical inference framework that bridges these contexts, treating the former as a special case of the latter. Our contribution is two-fold. First, we put forward the Repeated Risk Minimization (RRM) procedure for estimating the performative stability, and establish a rigorous inferential theory for admitting its asymptotic normality and confirming its asymptotic efficiency. Second, for the performative optimality, we introduce a novel two-step plug-in estimator that integrates the idea of Recalibrated Prediction Powered Inference (RePPI) with Importance Sampling, and further provide formal derivations for the Central Limit Theorems of both the underlying distributional parameters and the plug-in results. The theoretical analysis demonstrates that our estimator achieves the semiparametric efficiency bound and maintains robustness under mild distributional misspecification. This work provides a principled toolkit for reliable estimation and decision-making in dynamic, performative environments.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a key approach for enhancing LLM reasoning.However, standard frameworks like Group Relative Policy Optimization (GRPO) typically employ a uniform rollout budget, leading to resource inefficiency. Moreover, existing adaptive methods often rely on instance-level metrics, such as task pass rates, failing to capture the model's dynamic learning state. To address these limitations, we propose CoBA-RL, a reinforcement learning algorithm designed to adaptively allocate rollout budgets based on the model's evolving capability. Specifically, CoBA-RL utilizes a Capability-Oriented Value function to map tasks to their potential training gains and employs a heap-based greedy strategy to efficiently self-calibrate the distribution of computational resources to samples with high training value. Extensive experiments demonstrate that our approach effectively orchestrates the trade-off between exploration and exploitation, delivering consistent generalization improvements across multiple challenging benchmarks. These findings underscore that quantifying sample training value and optimizing budget allocation are pivotal for advancing LLM post-training efficiency.

Large language models have recently enabled text-to-CAD systems that synthesize parametric CAD programs (e.g., CadQuery) from natural language prompts. In practice, however, geometric descriptions can be under-specified or internally inconsistent: critical dimensions may be missing and constraints may conflict. Existing fine-tuned models tend to reactively follow user instructions and hallucinate dimensions when the text is ambiguous. To address this, we propose a proactive agentic framework for text-to-CadQuery generation, named ProCAD, that resolves specification issues before code synthesis. Our framework pairs a proactive clarifying agent, which audits the prompt and asks targeted clarification questions only when necessary to produce a self-consistent specification, with a CAD coding agent that translates the specification into an executable CadQuery program. We fine-tune the coding agent on a curated high-quality text-to-CadQuery dataset and train the clarifying agent via agentic SFT on clarification trajectories. Experiments show that proactive clarification significantly improves robustness to ambiguous prompts while keeping interaction overhead low. ProCAD outperforms frontier closed-source models, including Claude Sonnet 4.5, reducing the mean Chamfer distance by 79.9 percent and lowering the invalidity ratio from 4.8 percent to 0.9 percent. Our code and datasets will be made publicly available.

Early-exit networks reduce inference cost by allowing ``easy'' inputs to stop early, but practical deployment hinges on knowing \emph{when} early exit is safe. We introduce SAFE-KD, a universal multi-exit wrapper for modern vision backbones that couples hierarchical distillation with \emph{conformal risk control}. SAFE-KD attaches lightweight exit heads at intermediate depths, distills a strong teacher into all exits via Decoupled Knowledge Distillation (DKD), and enforces deep-to-shallow consistency between exits. At inference, we calibrate per-exit stopping thresholds on a held-out set using conformal risk control (CRC) to guarantee a user-specified \emph{selective} misclassification risk (among the samples that exit early) under exchangeability. Across multiple datasets and architectures, SAFE-KD yields improved accuracy compute trade-offs, stronger calibration, and robust performance under corruption while providing finite-sample risk guarantees.

Vision-Language Models (VLMs) have made great strides in everyday visual tasks, such as captioning a natural image, or answering commonsense questions about such images. But humans possess the puzzling ability to deploy their visual reasoning abilities in radically new situations, a skill rigorously tested by the classic set of visual reasoning challenges known as the Bongard problems. We present a neurosymbolic approach to solving these problems: given a hypothesized solution rule for a Bongard problem, we leverage LLMs to generate parameterized programmatic representations for the rule and perform parameter fitting using Bayesian optimization. We evaluate our method on classifying Bongard problem images given the ground truth rule, as well as on solving the problems from scratch.

Large language model (LLM)-powered multi-agent systems (MAS) demonstrate remarkable collective intelligence, wherein multi-agent memory serves as a pivotal mechanism for continual adaptation. However, existing multi-agent memory designs remain constrained by two fundamental bottlenecks: (i) memory homogenization arising from the absence of role-aware customization, and (ii) information overload induced by excessively fine-grained memory entries. To address these limitations, we propose LatentMem, a learnable multi-agent memory framework designed to customize agent-specific memories in a token-efficient manner. Specifically, LatentMem comprises an experience bank that stores raw interaction trajectories in a lightweight form, and a memory composer that synthesizes compact latent memories conditioned on retrieved experience and agent-specific contexts. Further, we introduce Latent Memory Policy Optimization (LMPO), which propagates task-level optimization signals through latent memories to the composer, encouraging it to produce compact and high-utility representations. Extensive experiments across diverse benchmarks and mainstream MAS frameworks show that LatentMem achieves a performance gain of up to $19.36$% over vanilla settings and consistently outperforms existing memory architectures, without requiring any modifications to the underlying frameworks.

Deep neural networks (DNNs) have achieved remarkable success in radio frequency (RF) fingerprinting for wireless device authentication. However, their practical deployment faces two major limitations: domain shift, where models trained in one environment struggle to generalize to others, and the black-box nature of DNNs, which limits interpretability. To address these issues, we propose a novel framework that integrates a group of variable-length two-dimensional (2D) shapelets with a pre-trained large language model (LLM) to achieve efficient, interpretable, and generalizable RF fingerprinting. The 2D shapelets explicitly capture diverse local temporal patterns across the in-phase and quadrature (I/Q) components, providing compact and interpretable representations. Complementarily, the pre-trained LLM captures more long-range dependencies and global contextual information, enabling strong generalization with minimal training overhead. Moreover, our framework also supports prototype generation for few-shot inference, enhancing cross-domain performance without additional retraining. To evaluate the effectiveness of our proposed method, we conduct extensive experiments on six datasets across various protocols and domains. The results show that our method achieves superior standard and few-shot performance across both source and unseen domains.

Adaptive Neuro-Fuzzy Inference System (ANFIS) was designed to combine the learning capabilities of neural network with the reasoning transparency of fuzzy logic. However, conventional ANFIS architectures suffer from structural complexity, where the product-based inference mechanism causes an exponential explosion of rules in high-dimensional spaces. We herein propose the Kolmogorov-Arnold Neuro-Fuzzy Inference System (KANFIS), a compact neuro-symbolic architecture that unifies fuzzy reasoning with additive function decomposition. KANFIS employs an additive aggregation mechanism, under which both model parameters and rule complexity scale linearly with input dimensionality rather than exponentially. Furthermore, KANFIS is compatible with both Type-1 (T1) and Interval Type-2 (IT2) fuzzy logic systems, enabling explicit modeling of uncertainty and ambiguity in fuzzy representations. By using sparse masking mechanisms, KANFIS generates compact and structured rule sets, resulting in an intrinsically interpretable model with clear rule semantics and transparent inference processes. Empirical results demonstrate that KANFIS achieves competitive performance against representative neural and neuro-fuzzy baselines.

Neural quantum states (NQS) are powerful ansätze in the variational Monte Carlo framework, yet their architectures are often treated as black boxes. We propose a physically transparent framework in which NQS are treated as neural approximations to latent imaginary-time evolution. This viewpoint suggests that standard Transformer-based NQS (TQS) architectures correspond to physically unmotivated effective Hamiltonians dependent on imaginary time in a latent space. Building on this interpretation, we introduce physics-inspired transformer quantum states (PITQS), which enforce a static effective Hamiltonian by sharing weights across layers and improve propagation accuracy via Trotter-Suzuki decompositions without increasing the number of variational parameters. For the frustrated $J_1$-$J_2$ Heisenberg model, our ansätze achieve accuracies comparable to or exceeding state-of-the-art TQS while using substantially fewer variational parameters. This study demonstrates that reinterpreting the deep network structure as a latent cooling process enables a more physically grounded, systematic, and compact design, thereby bridging the gap between black-box expressivity and physically transparent construction.

Time series analysis underpins many real-world applications, yet existing time-series-specific methods and pretrained large-model-based approaches remain limited in integrating intuitive visual reasoning and generalizing across tasks with adaptive tool usage. To address these limitations, we propose MAS4TS, a tool-driven multi-agent system for general time series tasks, built upon an Analyzer-Reasoner-Executor paradigm that integrates agent communication, visual reasoning, and latent reconstruction within a unified framework. MAS4TS first performs visual reasoning over time series plots with structured priors using a Vision-Language Model to extract temporal structures, and subsequently reconstructs predictive trajectories in latent space. Three specialized agents coordinate via shared memory and gated communication, while a router selects task-specific tool chains for execution. Extensive experiments on multiple benchmarks demonstrate that MAS4TS achieves state-of-the-art performance across a wide range of time series tasks, while exhibiting strong generalization and efficient inference.

Multi-turn tool calling is challenging for Large Language Models (LLMs) because rewards are sparse and exploration is expensive. A common recipe, SFT followed by GRPO, can stall when within-group reward variation is low (e.g., more rollouts in a group receive the all 0 or all 1 reward), making the group-normalized advantage uninformative and yielding vanishing updates. To address this problem, we propose RC-GRPO (Reward-Conditioned Group Relative Policy Optimization), which treats exploration as a controllable steering problem via discrete reward tokens. We first fine-tune a Reward-Conditioned Trajectory Policy (RCTP) on mixed-quality trajectories with reward goal special tokens (e.g., <|high_reward|>, <|low_reward|>) injected into the prompts, enabling the model to learn how to generate distinct quality trajectories on demand. Then during RL, we sample diverse reward tokens within each GRPO group and condition rollouts on the sampled token to improve within-group diversity, improving advantage gains. On the Berkeley Function Calling Leaderboard v4 (BFCLv4) multi-turn benchmark, our method yields consistently improved performance than baselines, and the performance on Qwen-2.5-7B-Instruct even surpasses all closed-source API models.

Deep Equilibrium Models (DEQs) have emerged as a powerful paradigm in deep learning, offering the ability to model infinite-depth networks with constant memory usage. However, DEQs incur significant inference latency due to the iterative nature of fixed-point solvers. In this work, we introduce the Consistency Deep Equilibrium Model (C-DEQ), a novel framework that leverages consistency distillation to accelerate DEQ inference. We cast the DEQ iterative inference process as evolution along a fixed ODE trajectory toward the equilibrium. Along this trajectory, we train C-DEQs to consistently map intermediate states directly to the fixed point, enabling few-step inference while preserving the performance of the teacher DEQ. At the same time, it facilitates multi-step evaluation to flexibly trade computation for performance gains. Extensive experiments across various domain tasks demonstrate that C-DEQs achieves consistent 2-20$\times$ accuracy improvements over implicit DEQs under the same few-step inference budget.

Music captioning, or the task of generating a natural language description of music, is useful for both music understanding and controllable music generation. Training captioning models, however, typically requires high-quality music caption data which is scarce compared to metadata (e.g., genre, mood, etc.). As a result, it is common to use large language models (LLMs) to synthesize captions from metadata to generate training data for captioning models, though this process imposes a fixed stylization and entangles factual information with natural language style. As a more direct approach, we propose metadata-based captioning. We train a metadata prediction model to infer detailed music metadata from audio and then convert it into expressive captions via pre-trained LLMs at inference time. Compared to a strong end-to-end baseline trained on LLM-generated captions derived from metadata, our method: (1) achieves comparable performance in less training time over end-to-end captioners, (2) offers flexibility to easily change stylization post-training, enabling output captions to be tailored to specific stylistic and quality requirements, and (3) can be prompted with audio and partial metadata to enable powerful metadata imputation or in-filling--a common task for organizing music data.

The proliferation of Large Language Models (LLMs) in function calling is pivotal for creating advanced AI agents, yet their large scale hinders widespread adoption, necessitating transferring their capabilities into smaller ones. However, existing paradigms are often plagued by overfitting, training instability, ineffective binary rewards for multi-solution tasks, and the difficulty of synergizing techniques. We introduce STAR: Similarity-guided Teacher-Assisted Refinement, a novel holistic framework that effectively transfers LLMs' capabilities to super-tiny models. STAR consists of two core technical innovations: (1) Constrained Knowledge Distillation (CKD), a training objective that augments top-k forward KL divergence to suppress confidently incorrect predictions, ensuring training stability while preserving exploration capacity for downstream RL. STAR holistically synergizes these strategies within a cohesive training curriculum, enabling super-tiny models to achieve exceptional performance on complex function calling tasks; (2) Similarity-guided RL (Sim-RL), a RL mechanism that introduces a fine-grained, similarity-based reward. This provides a robust, continuous, and rich signal for better policy optimization by evaluating the similarity between generated outputs and the ground truth. Extensive experiments on challenging and renowned benchmarks demonstrate the effectiveness of our method. Our STAR models establish SOTA in their size classes, significantly outperforming baselines. Remarkably, our 0.6B STAR model achieves the best performance among all open models under 1B, surpassing even several well-known open models at a larger scale. STAR demonstrates a training framework that distills capabilities of LLMs into super-tiny models, paving the way for powerful, accessible, and efficient AI agents.

Fine-tuning is essential to adapt general-purpose large language models (LLMs) to domain-specific tasks. As a privacy-preserving framework to leverage decentralized data for collaborative model training, Federated Learning (FL) is gaining popularity in LLM fine-tuning, but remains challenging due to the high cost of transmitting full model parameters and computing full gradients on resource-constrained clients. While Parameter-Efficient Fine-Tuning (PEFT) methods are widely used in FL to reduce communication and memory costs, they often sacrifice model performance compared to FFT. This paper proposes FedKRSO (Federated $K$-Seed Random Subspace Optimization), a novel method that enables communication and memory efficient FFT of LLMs in federated settings. In FedKRSO, clients update the model within a shared set of random low-dimension subspaces generated by the server to save memory usage. Furthermore, instead of transmitting full model parameters in each FL round, clients send only the model update accumulators along the subspaces to the server, enabling efficient global model aggregation and dissemination. By using these strategies, FedKRSO can substantially reduce communication and memory overhead while overcoming the performance limitations of PEFT, closely approximating the performance of federated FFT. The convergence properties of FedKRSO are analyzed rigorously under general FL settings. Extensive experiments on the GLUE benchmark across diverse FL scenarios demonstrate that FedKRSO achieves both superior performance and low communication and memory overhead, paving the way towards on federated LLM fine-tuning at the resource-constrained edge.

Outlier detection (OD) is widely used in practice; but its effective deployment on new tasks is hindered by lack of labeled outliers, which makes algorithm and hyperparameter selection notoriously hard. Foundation models (FMs) have transformed ML, and OD is no exception: Shen et. al. (2025) introduced FoMo-0D, the first FM for OD, achieving remarkable performance against numerous baselines. This work introduces OUTFORMER, which advances FoMo-0D with (1) a mixture of synthetic priors and (2) self-evolving curriculum training. OUTFORMER is pretrained solely on synthetic labeled datasets and infers test labels of a new task by using its training data as in-context input. Inference is fast and zero-shot, requiring merely forward pass and no labeled outliers. Thanks to in-context learning, it requires zero additional work-no OD model training or bespoke model selection-enabling truly plug-and-play deployment. OUTFORMER achieves state-of-the-art performance on the prominent AdBench, as well as two new large-scale OD benchmarks that we introduce, comprising over 1,500 datasets, while maintaining speedy inference.

Evaluating and improving the security capabilities of code agents requires high-quality, executable vulnerability tasks. However, existing works rely on costly, unscalable manual reproduction and suffer from outdated data distributions. To address these, we present CVE-Factory, the first multi-agent framework to achieve expert-level quality in automatically transforming sparse CVE metadata into fully executable agentic tasks. Cross-validation against human expert reproductions shows that CVE-Factory achieves 95\% solution correctness and 96\% environment fidelity, confirming its expert-level quality. It is also evaluated on the latest realistic vulnerabilities and achieves a 66.2\% verified success. This automation enables two downstream contributions. First, we construct LiveCVEBench, a continuously updated benchmark of 190 tasks spanning 14 languages and 153 repositories that captures emerging threats including AI-tooling vulnerabilities. Second, we synthesize over 1,000 executable training environments, the first large-scale scaling of agentic tasks in code security. Fine-tuned Qwen3-32B improves from 5.3\% to 35.8\% on LiveCVEBench, surpassing Claude 4.5 Sonnet, with gains generalizing to Terminal Bench (12.5\% to 31.3\%). We open-source CVE-Factory, LiveCVEBench, Abacus-cve (fine-tuned model), training dataset, and leaderboard. All resources are available at https://github.com/livecvebench/CVE-Factory .

Despite significant costs from retrieving and processing high-fidelity visual inputs, most multimodal vision-language systems operate at fixed fidelity levels. We introduce VOILA, a framework for Value-Of-Information-driven adaptive fidelity selection in Visual Question Answering (VQA) that optimizes what information to retrieve before model execution. Given a query, VOILA uses a two-stage pipeline: a gradient-boosted regressor estimates correctness likelihood at each fidelity from question features alone, then an isotonic calibrator refines these probabilities for reliable decision-making. The system selects the minimum-cost fidelity maximizing expected utility given predicted accuracy and retrieval costs. We evaluate VOILA across three deployment scenarios using five datasets (VQA-v2, GQA, TextVQA, LoCoMo, FloodNet) and six Vision-Language Models (VLMs) with 7B-235B parameters. VOILA consistently achieves 50-60% cost reductions while retaining 90-95% of full-resolution accuracy across diverse query types and model architectures, demonstrating that pre-retrieval fidelity selection is vital to optimize multimodal inference under resource constraints.

Deploying Large Language Models (LLMs) for discriminative workloads is often limited by inference latency, compute, and API costs at scale. Active distillation reduces these costs by querying an LLM oracle to train compact discriminative students, but most pipelines distill only final labels, discarding intermediate reasoning signals and offering limited diagnostics of what reasoning is missing and where errors arise. We propose Graph of Concept Predictors (GCP), a reasoning-aware active distillation framework that externalizes the teacher's decision process as a directed acyclic graph and mirrors it with modular concept predictors in the student. GCP enhances sample efficiency through a graph-aware acquisition strategy that targets uncertainty and disagreement at critical reasoning nodes. Additionally, it improves training stability and efficiency by performing targeted sub-module retraining, which attributes downstream loss to specific concept predictors and updates only the most influential modules. Experiments on eight NLP classification benchmarks demonstrate that GCP enhances performance under limited annotation budgets while yielding more interpretable and controllable training dynamics. Code is available at: https://github.com/Ziyang-Yu/GCP.

To improve the reliability and interpretability of industrial process monitoring, this article proposes a Causal Graph Spatial-Temporal Autoencoder (CGSTAE). The network architecture of CGSTAE combines two components: a correlation graph structure learning module based on spatial self-attention mechanism (SSAM) and a spatial-temporal encoder-decoder module utilizing graph convolutional long-short term memory (GCLSTM). The SSAM learns correlation graphs by capturing dynamic relationships between variables, while a novel three-step causal graph structure learning algorithm is introduced to derive a causal graph from these correlation graphs. The algorithm leverages a reverse perspective of causal invariance principle to uncover the invariant causal graph from varying correlations. The spatial-temporal encoder-decoder, built with GCLSTM units, reconstructs time-series process data within a sequence-to-sequence framework. The proposed CGSTAE enables effective process monitoring and fault detection through two statistics in the feature space and residual space. Finally, we validate the effectiveness of CGSTAE in process monitoring through the Tennessee Eastman process and a real-world air separation process.

Social choice is no longer a peripheral concern of political theory or economics-it has become a foundational component of modern machine learning systems. From auctions and resource allocation to federated learning, participatory governance, and the alignment of large language models, machine learning pipelines increasingly aggregate heterogeneous preferences, incentives, and judgments into collective decisions. In effect, many contemporary machine learning systems already implement social choice mechanisms, often implicitly and without explicit normative scrutiny. This Review surveys differentiable social choice: an emerging paradigm that formulates voting rules, mechanisms, and aggregation procedures as learnable, differentiable models optimized from data. We synthesize work across auctions, voting, budgeting, liquid democracy, decentralized aggregation, and inverse mechanism learning, showing how classical axioms and impossibility results reappear as objectives, constraints, and optimization trade-offs. We conclude by identifying 36 open problems defining a new research agenda at the intersection of machine learning, economics, and democratic theory.

To maximize hardware utilization, modern machine learning systems typically employ large constant or manually tuned batch size schedules, relying on heuristics that are brittle and costly to tune. Existing adaptive strategies based on gradient noise scale (GNS) offer a principled alternative. However, their assumption of SGD's Euclidean geometry creates a fundamental mismatch with popular optimizers based on generalized norms, such as signSGD / Signum ($\ell_\infty$) and stochastic spectral descent (specSGD) / Muon ($\mathcal{S}_\infty$). In this work, we derive gradient noise scales for signSGD and specSGD that naturally emerge from the geometry of their respective dual norms. To practically estimate these non-Euclidean metrics, we propose an efficient variance estimation procedure that leverages the local mini-batch gradients on different ranks in distributed data-parallel systems. Our experiments demonstrate that adaptive batch size strategies using non-Euclidean GNS enable us to match the validation loss of constant-batch baselines while reducing training steps by up to 66% for Signum and Muon on a 160 million parameter Llama model.

While scaling test-time compute through trajectory-level sampling has significantly improved Graphical User Interface (GUI) agents, the lack of regressive ability prevents the reuse of partial successes and the recovery from early missteps. In this paper, we introduce Agent Alpha, a unified framework that synergizes generation, exploration, and evaluation through step-level Monte Carlo Tree Search (MCTS). It enables active modeling or exploiting structures of the planning space. By integrating alpha-UCT guided search into the interaction loop, Agent Alpha enables deliberate planning, facilitating early pruning of suboptimal branches and efficient prefix reuse. We also employ comparison-driven evaluation to mitigate absolute scoring biases and diversity-constrained expansion to maintain a compact, informative search space. Regret bound of alpha-UCT is analyzed. On the OSWorld benchmark, Agent Alpha achieves a state-of-the-art success rate of $\sim 77\%$, significantly outperforming trajectory-level baselines under equivalent compute.

Large language models (LLMs) have been shown to acquire sequence-level planning abilities during training, yet their planning behavior exhibited at inference time often appears short-sighted and inconsistent with these capabilities. We propose a Bayesian account for this gap by grounding planning behavior in the evolving generative context: given the subtle differences between natural language and the language internalized by LLMs, accumulated self-generated context drives a planning-shift during inference and thereby creates the appearance of compromised planning behavior. We further validate the proposed model through two controlled experiments: a random-generation task demonstrating constrained planning under human prompts and increasing planning strength as self-generated context accumulates, and a Gaussian-sampling task showing reduced initial bias when conditioning on self-generated sequences. These findings provide a theoretical explanation along with empirical evidence for characterizing how LLMs plan ahead during inference.

We compute the critical exponents of the zero-temperature spin glass transition in a field on a one-dimensional long-range model, a proxy for higher-dimensional systems. Our approach is based on a novel loop expansion within the Bethe $M$-layer formalism, whose adaptation to this specific case is detailed here. The resulting estimates provide crucial benchmarks for numerical simulations that can access larger system sizes in one dimension, thus offering a key test of the theory of spin glasses in a field.

We study quantum spin chains solvable via hidden free fermionic structures. We study the algebras behind such models, establishing connections to the mathematical literature of the so-called ``graph-Clifford'' or ``quasi-Clifford'' algebras. We also introduce the ``defining representation'' for such algebras, and show that this representation actually coincides with the terms of the Hamiltonian in two relevant models: the XY model and the ``free fermions in disguise'' model of Fendley. Afterwards we study a particular anti-symmetric combination of commuting Hamiltonians; this is performed in a model independent way. We show that for this combination there exists a reference state, and few body eigenstates can be created by the fermionic operators. Concrete application is presented in the case of the ``free fermions in disguise'' model.

Atomic-scale control over band alignment in single-layer lateral heterostructures (LHSs) of dissimilar transition metal dichalcogenides (TMDCs) is critical for nextgeneration electronic, optoelectronic, and quantum technologies. However, direct experimental access to interfacial electronic states with nanometer precision remains a significant challenge. Here, we employ angle-resolved photoemission spectroscopy with nanoscale spatial resolution (nanoARPES) to directly map the epitaxial alignment and valence band evolution across MoSe2-WSe2 LHSs. By combining nanoARPES with spatially resolved photoluminescence, we correlate the evolution of the valence band maximum and exciton features across both atomically sharp and compositionally graded diffusive interfaces. We identified type-II band alignments governed by both material composition and interstitial-induced modifications of band offsets, in close agreement with density functional theory calculations. These results reveal fundamental mechanisms of electronic structure modulation at 1D TMDC heterointerfaces and provide a robust platform for tailored band engineering in van der Waals materials.

We report robust room temperature interlayer excitons in transition metal dichalcogenide heterostructures engineered via precise stacking orientation and twist-angle control. We integrate 2H-stacked MoSe$_{2}$/$^{1}$WSe$_{2}$/$^{2}$WSe$_{2}$ heterotrilayer onto a chirped distributed Bragg reflector that acts as a backside mirror. This way, we fabricate a platform that hosts distinct heterotrilayer, heterobilayer, and homobilayer regions with enhanced excitonic features at elevated temperatures. Although the heterobilayer supports temperature-tunable singlet and triplet interlayer excitons, it exhibits low emission yield at 4 K. In comparison, the heterotrilayer shows remarkable excitonic properties, including pronounced band modulation, intervalley interlayer exciton transitions, and a tenfold photoluminescence enhancement along with a sevenfold increase in exciton decay time at cryogenic temperatures compared to the heterobilayer system. Temperature-dependent studies reveal intriguing interlayer exciton dynamics in the heterotrilayer, including the emergence of valley-polarized interlayer excitons, and the ability to maintain optical stability up to room temperature. Our results establish a clear strategy for engineering excitonic states across multilayer van der Waals heterostructures from 4 K to room temperature, providing a versatile platform for excitonic optoelectronics, quantum photonics, and tunable long-lived interlayer exciton states in scalable TMD heterostructures.

We study equilibrium measures for Riesz gases in dimension $d$ with pairwise interaction kernel $|x-y|^{-s}$, subject to radially symmetric external fields. We characterise broad classes of confining potentials for which the equilibrium measure is supported on the unit ball and admits an explicit density. Our main contribution is a converse construction: starting from a prescribed radially symmetric equilibrium density given as a power series in the squared radius, we determine the associated external potential and establish the corresponding Euler-Lagrange variational conditions. A key ingredient in the proof is an identity between two ${}_3F_2$ hypergeometric functions evaluated at unit argument, which is of independent interest. As applications, we identify the external potentials corresponding to equilibrium densities proportional to $(1-|x|^2)^α$, $α>-1$, and show that these potentials can be expressed in terms of Gauss hypergeometric functions ${}_2F_1$, reducing to polynomials for special values of $α$. We also determine the equilibrium measure associated with purely power-type external potentials, often referred to as Freud or Mittag--Leffler potentials in the context of log gases, for which the equilibrium density admits an explicit ${}_2F_1$ representation. Furthermore, we apply our framework to a Coulomb gas in dimension $d+1$ confined by a harmonic potential to the half-space. We derive a necessary condition under which the equilibrium measure is fully supported on the boundary hyperplane of dimension $d$, with the induced density corresponding to that of a Riesz gas with exponent $s=d-1$.

Anderson localization, arising from wave interference in disordered systems, profoundly hinders energy transport, yet its impact on radiative heat flux in many-body thermophotonic systems remains unclear. Here, we demonstrate a three-order-of-magnitude suppression of radiative heat transfer, resulting in ultralow radiative heat transfer, in a one-dimensional quasiperiodic chain of plasmonic nanoparticles. This suppression in radiative heat transfer is directly correlated with mode localization, as revealed by the mode decomposition of the transmission coefficient, which serves as evidence of Anderson localization. Furthermore, we elucidate the dependence of radiative thermal conductance reduction on interparticle spacing and material damping rates, uncovering the interplay between intrinsic Ohmic losses, mode localization, and long-range many-body interactions. Our findings advance the understanding of wave-mediated thermal transport in disordered photonic structures and suggest strategies for tailoring nanoscale heat management via engineered disorder.

We investigate non-equilibrium phase coexistence associated with a first-order phase transition by numerically studying a one-dimensional Hamiltonian-Potts model with fractional spatial derivatives. The fractional derivative is introduced so as to reproduce the low-wavenumber density of states of the standard two-dimensional model, allowing phase coexistence to occur in a minimal one-dimensional setting under steady heat conduction. By imposing a constant heat flux through boundary heat baths, we observe stable coexistence of ordered and disordered phases separated by a stationary interface. We find that the temperature at the interface systematically deviates from the equilibrium transition temperature, demonstrating a clear violation of the local equilibrium description. This deviation indicates that equilibrium metastable states can be stabilized and controlled by a steady heat current. Furthermore, the interface temperature obtained in our simulations is in quantitative agreement with the prediction of global thermodynamics for non-equilibrium steady states. These results confirm that the breakdown of local equilibrium and the stabilization of metastable states are intrinsic features of non-equilibrium first-order phase transitions, independent of spatial dimensionality. Our study thus provides a minimal and controlled numerical model for exploring the fundamental limits of thermodynamic descriptions in non-equilibrium steady states.

Single-shot spin-state discrimination is essential for semiconductor spin qubits, but conventional threshold-based analysis of spin readout traces becomes unreliable under noisy conditions. Although recent neural-network-based methods improve robustness against experimental noise, they are sensitive to training conditions, restricted to fixed-length inputs, and limited to trace-level outputs without explicit temporal localization of transition events. In this work, we apply a U-Net architecture to spin readout signal analysis by formulating transition-event detection as a point-wise segmentation task in one-dimensional time-series data. The fully convolutional structure enables direct processing of variable-length traces. Point-wise and sample-wise evaluations demonstrate low readout error rates and high classification accuracy without retraining. The proposed method generalizes well to previously-unseen trace lengths and experimental non-Gaussian noise, outperforming a conventional threshold-based approach and providing a robust and practical solution for automated spin readout signal analysis.

We study quantum phase transition in the transverse-field Ising model on the Sierpiński gasket. By applying finite-size scaling and numerical renormalization group methods, we determine the critical coupling and the exponents that describe this transition. We first checked our finite-size scaling and the renormalization methods on the exactly solvable one-dimensional chain, where we recovered proper values of critical couplings and exponents. Then, we applied the method to the Sierpiński gasket with 11 and 15 spins. We found a quantum critical point at $λ_c \approx 2.72$ to $2.93$, with critical exponents $z\approx0.84$, $ν\approx 1.12 $, $β\approx 0.30$, and $γ\approx 2.54$. The lower dynamical exponent $z$ indicates that quantum fluctuations slow down due to fractal geometry, yielding an effective critical dimension of about 2.43. The numerical renormalization group method yielded similar results $λ_c = 2.765$, $β= 0.306$, supporting our findings. These exponents differ from those in both the one-dimensional and mean-field cases.

We experimentally demonstrate the direct time-domain characterization of photonic-crystal nanolasers at telecom wavelengths using a nonlinear optical gating technique based on four-wave mixing. This approach enables the temporal characterization of the ultrafast emission dynamics under short-pulse excitation with picosecond time resolution. When a weak continuous-wave component is added to the pulsed pump, the emission becomes deterministic and the build-up time is considerably reduced. The difference between purely pulsed and hybrid excitation regimes points to the influence of pulse-to-pulse timing fluctuations. To elucidate this effect, we perform Langevin-based simulations that reproduce the experimentally observed broadening and confirm that time jitter, originating from spontaneous-emission noise near threshold, dominates the temporal dispersion. These results establish four-wave-mixing gating as a powerful method to probe nanolaser dynamics with picosecond precision.

We investigate the stochastic dynamics of nanoscale perpendicular magnetic tunnel junctions (pMTJs) and the correlations that arise when they are electrically coupled. Individual junctions exhibit thermally activated spin-transfer torque switching with transition probabilities that are well described by a Poisson process. When two junctions are connected in parallel, circuit-mediated redistribution of voltages that occurs in real time as the junction resistances change leads to correlated switching behavior. A minimal stochastic model based on single-junction statistical switching properties and Kirchhoff's laws captures the coupled switching probabilities, while a Markov-chain formalism describes nonequilibrium steady states under multi-pulse driving. Further, these circuit-mediated interactions can be mapped onto the parameters of an Ising Hamiltonian, providing an interpretation in terms of effective spin-spin interactions. Our results demonstrate how simple electrical connections can generate Ising-like couplings and tunable stochastic dynamics in nanoscale magnets.

Lateral heterostructures of 2D transition metal dichalcogenide offer a powerful platform to investigate photonic and electronic phenomena at atomically sharp interfaces. However, their controlled engineering, including tuning lateral domain size and integration into vertical van der Waals heterostructures with other 2D materials, remains challenging. Here, we present a facile route for the synthesis of two types of heterostructures consisting of monolayers of MoSe$_2$ and WSe$_2$ - purely lateral (HS I) and hybrid lateral/vertical (HS II) - using liquid precursors of transition metal salts and chemical vapor deposition (CVD). Depending on the growth parameters, the heterostructure type and their lateral dimensions can be adjusted. We characterized properties of the HS I and HS II by complementary spectroscopic and microscopic techniques including Raman and photoluminescence spectroscopy, and optical and atomic force microscopy, and scanning electron and transmission electron microscopy. The photoluminescence measurements reveal strong interlayer exciton emission in the MoSe$_2$/WSe$_2$ region of HS II, which dominates the spectrum at 4 K and persisting up to room temperature. These results demonstrate high optical quality of the grown heterostructures which in combination with scalability of the developed approach paves the way for fundamental studies and device applications based on these unique 2D quantum materials.

Pre-training on a source task is usually expected to facilitate fine-tuning on similar downstream problems. In this work, we mathematically show that this naive intuition is not always true: excessive pre-training can computationally slow down fine-tuning optimization. We study this phenomenon for low-rank adaptation (LoRA) fine-tuning on single-index models trained under one-pass SGD. Leveraging a summary statistics description of the fine-tuning dynamics, we precisely characterize how the convergence rate depends on the initial fine-tuning alignment and the degree of non-linearity of the target task. The key take away is that even when the pre-training and down- stream tasks are well aligned, strong pre-training can induce a prolonged search phase and hinder convergence. Our theory thus provides a unified picture of how pre-training strength and task difficulty jointly shape the dynamics and limitations of LoRA fine-tuning in a nontrivial tractable model.

Soft, microscale hydrogel fibers and rods play important roles in tissue engineering, flexible electronics, soft robotics, drug delivery, sensors, and other applications. Their viscoelastic mechanical properties, while critical for their function, can be challenging to characterize. We present a flow-induced bending response (FIBR) rheometer that quantifies the bending modulus and viscoelastic properties of small, hydrated fibers and rods using flow through a glass capillary. The fiber is positioned across the capillary entrance, and pressure-driven, controlled inflow of water exerts a quantifiable force on the sample. Fiber deflection is determined by video microscopy obtained simultaneously with measurements of flow rate. We develop an analytical model to resolve the hydrodynamic forces applied to the rod, and use Euler-Bernoulli beam theory to determine its material properties. Using a constant volume flow rate of water enables measurement of steady rod deflection, and thus the bending modulus. Application of viscous forces to the rod in a stepwise, cyclic or oscillatory manner enables measurement of time-dependent responses, creep recovery, viscoelastic moduli, and other properties. We demonstrate the versatility of this technique on natural and synthetic materials spanning diameters from 1 to 500 microns and elastic moduli ranging from 100 Pa to >100 MPa. Because the technique uses water to exert forces on the fiber, it works particularly well for hydrated materials, such as hydrogels and biological fibers, providing a versatile platform to characterize microscale mechanical properties of elongated structures.

Quasicrystalline order induces a fractal energy spectrum, yet its impact on topological protection remains an open fundamental question. Here, we demonstrate that the topological phase transitions characterised by the appearance of Majorana Bound States themselves have a fractal character. By extending this analysis to the full family of Sturmian words, we uncover Kitaev's Butterfly $-$ a spectral fractal analogous to Hofstadter's butterfly, but fundamentally distinguished by a central superconducting gap. Within this framework, we identify Majorana's Butterfly as a fractal topological phase diagram governed by the competition between quasicrystallinity and superconducting pairing. We show that this competition dictates a hierarchy of Majorana stability, where the survival of the topological phase against fractal fragmentation is determined by the relative strength of these competing energy scales.

The heterogeneity in nanoscale emitters hinders efforts to understand their basic photophysics and limits their use in practical applications. Existing methods have difficulty accurately characterizing single-emitter spectra and optical heterogeneity on a statistical scale. Here, we introduce SPICEE (SPectrally Imbalanced Correlations from Ensemble Emission), a spectrally filtered photon-correlation technique that recovers single-particle emission lineshapes from an ensemble sample. Analytical derivations, numerical modeling, and experiments on a solution ensemble of emitters validate the technique. We apply SPICEE to blue-emitting ZnSeTe semiconductor nanocrystals relevant to display applications and find that the low color purity in the ensemble spectrum is primarily caused by a small subpopulation of nanocrystals with a distinct emission mechanism. This work demonstrates that SPICEE is a powerful high-throughput tool for accurately characterizing the single-emitter properties of nanoscale systems.

Measurement of light polarization has long been based on complex, bulk, and slow optical instruments. The advent of materials with in-situ variable polarization photoresponse has led to the concept of reconstructive polarimetry, where the detector itself plays the role of tunable polarizer. Materials enabling such functionality have been limited to complex van der Waals heterostructures. Here, we demonstrate the reconstructive polarimetry with infrared (IR) detectors based on simple gated graphene-metal junctions. The reconstruction exploits the gate tuning of polarization contrast, which enables the evaluation of both infrared power and polarization angle from photovoltage measurements at two sequential gate voltages. The physics enabling the polarimetry lies in polarization-dependent shift of the electron hot spot near the contact, and the gate tuning of the of light-sensitive barrier width. We further show the universality of polarization reconstruction, i.e. its feasibility with different geometries of the junction, and with graphene of different quality, from hBN-encapsulated to the scalable vapor-deposited wet-transferred samples.

The interplay of twist and strain in bilayer graphene enables the formation of moiré patterns and narrow bands that host correlated and topological phases. While magic-angle twisted bilayer graphene has been widely studied, strain provides an additional and realistic control knob for band engineering. In this work, we first generate a global method to construct commensurate supercells for arbitrary twist and strain. Then, using atomistic tight-binding and strain-extended continuum models to study the commensurate structures, we identify configurations that minimize the bandwidth beyond the magic angle. The results reveal a strong dependence of band narrowing and topology on strain type, magnitude, direction and lattice relaxation. Particularly, shear strain produces a stronger distortion than uniaxial strain. Including electron-electron interactions through a self-consistent Hartree potential shows that strain broadens the bare bands while reducing electrostatic renormalization. Strain also drives topological transitions as the narrow and remote bands hybridize, establishing twisted and strained bilayer graphene as a tunable platform for flat-band and topological phenomena.

We study the emergence of statistical mechanics in isolated classical systems with local interactions and discrete phase spaces. We establish that thermalization in such systems does not require global ergodicity; instead, it arises from effective local ergodicity, where dynamics in a subsystem may appear pseudorandom. To corroborate that, we analyze the spectrum of the unitary evolution operator and propose an ansatz to describe statistical properties of local observables expanded in the eigenfunction basis - the classical counterpart of the Eigenstate Thermalization Hypothesis. Our framework provides a unified perspective on thermalization in classical and quantum systems with discrete spectra.

We show that the area operator of a quantum extremal surface can be reconstructed directly from boundary dynamics without reference to bulk geometry. Our approach combines the operator-algebra quantum error-correction (OAQEC) structure of AdS/CFT with modular Krylov complexity. Using Lanczos coefficients of boundary modular dynamics, we extract the spectrum of the modular Hamiltonian restricted to the algebra of the entanglement wedge and isolate its central contribution, which is identified with the area operator. The construction is purely boundary-based and applies to superpositions of semiclassical geometries as well. As an application, we diagnose island formation and the Page transition in evaporating black holes using boundary modular evolution alone, bypassing any bulk extremization. More broadly, our results establish modular Krylov complexity as a concrete and computable probe of emergent spacetime geometry, providing a new route to accessing black hole interiors from boundary quantum dynamics.

We explore the dynamics of the PXP model when subjected to a periodic drive, and unveil the mechanism through which the interplay between spectral properties and initial states governs the emergence of dynamical revivals and their evolution across the space of driving parameters. For Néel-ordered initial states, revivals follow well-defined trajectories in the parameter space of the driving, primarily determined by a dominant quasi-energy spacing in the Floquet spectrum. Initial states interpolating between Néel and fully polarized configurations exhibit hybrid dynamics, which can be controlled by tuning their overlap with Floquet eigenstates via the driving parameters. This control also allows steering different routes for avoiding Floquet thermalization, showing how both initial state choice and driving protocol shape long-lived dynamics in this driven quantum many-body systems.

Accessing the thermodynamic-limit properties of strongly correlated quantum matter requires simulations on very large lattices, a regime that remains challenging for numerical methods, especially in frustrated two-dimensional systems. We introduce the Spatial Attention mechanism, a minimal and physically interpretable inductive bias for Neural-Network Quantum States, implemented as a single learned length scale within the Transformer architecture. This bias stabilizes large-scale optimization and enables access to thermodynamic-limit physics through highly accurate simulations on unprecedented system sizes within the Variational Monte Carlo framework. Applied to the spin-$\tfrac12$ triangular-lattice Heisenberg antiferromagnet, our approach achieves state-of-the-art results on clusters of up to $42\times42$ sites. The ability to simulate such large systems allows controlled finite-size scaling of energies and order parameters, enabling the extraction of experimentally relevant quantities such as spin-wave velocities and uniform susceptibilities. In turn, we find extrapolated thermodynamic limit energies systematically better than those obtained with tensor-network approaches such as iPEPS. The resulting magnetization is strongly renormalized, $M_0=0.148(1)$ (about $30\%$ of the classical value), revealing that less accurate variational states systematically overestimate magnetic order. Analysis of the optimized wave function further suggests an intrinsically non-local sign structure, indicating that the sign problem cannot be removed by local basis transformations. We finally demonstrate the generality of the method by obtaining state-of-the-art energies for a $J_1$-$J_2$ Heisenberg model on a $20\times20$ square lattice, outperforming Residual Convolutional Neural Networks.

We investigate the role of quantum monitoring in the dynamical manifestations of Hamiltonian quantum chaos. Specifically, we analyze the generalized spectral form factor, defined as the survival probability of a coherent Gibbs state under continuous energy measurements. We show that quantum monitoring can tailor the signatures of quantum chaos in the dynamics, such as the extension of the ramp in the spectral form factor, by varying the measurement strength and detection efficiency. In particular, a typical quantum trajectory obtained by monitoring with unit efficiency exhibits enhanced quantum chaos relative to the average dynamics and to unitary evolution without measurements.

Conformal invariance often accompanies criticality in Hermitian systems. However, its fate in non-Hermitian settings is less clear, especially near exceptional points where the Hamiltonian becomes non-diagonalizable. Here we investigate whether a 1+1-dimensional gapless non-Hermitian system can admit a conformal description, focusing on a PT-symmetric free-fermion field theory. Working in the biorthogonal formalism, we identify the conformal structure of this theory by constructing a traceless energy-momentum tensor whose Fourier modes generate a Virasoro algebra with central charge $c=-2$. This yields a non-Hermitian, biorthogonal realization of a logarithmic conformal field theory, in which correlation functions exhibit logarithmic scaling and the spectrum forms Virasoro staggered modules that are characterized by universal indecomposability parameters. We further present a microscopic construction and show how the same conformal data (with finite-size corrections) can be extracted from the lattice model at exceptional-point criticality, thereby supporting the field-theory prediction.

Quantum criticality in the presence of strong quenched randomness remains a challenging topic in modern condensed matter theory. We show that the topology and anomaly associated with average symmetry can be used to predict certain nontrivial universal properties. Our focus is on systems subject to average Lieb--Schultz--Mattis constraints, where lattice translation symmetry is preserved only on average, while on-site symmetries remain exact. We argue that in the absence of spontaneous symmetry breaking, the system must exhibit critical correlations of local operators in two distinct ways: (i) for some operator $O_e$ charged under exact symmetries, the first absolute moment correlation $\overline{|\langle O_e(x)O^{\dagger}_e(y)\rangle|}$ decays slowly; and (ii) for some operator $O_a$ charged under average symmetries, the first-moment correlation $\overline{\langle O_a(x)O^{\dagger}_a(y)\rangle}$ decays slowly. We verify these predictions in a few examples: the random-singlet Heisenberg spin chain in one dimension, and the disordered free-fermion critical states in symmetry class BDI in one and two dimensions. Surprisingly, even for these well-studied systems, our anomaly-based argument reveals critical correlations overlooked in previous literature. We also discuss the experimental feasibility of measuring these critical correlations.

Distinguishing an exciton condensate from an excitonic gas or insulator remains a fundamental challenge, as both phases feature bound electron-hole pairs but differ only by the emergence of macroscopic phase coherence. Here, we theoretically propose that a spin-orbit-coupled bilayer system can host a finite-momentum exciton condensate exhibiting a nonreciprocal perfect Coulomb drag -- the coherent-exciton diode effect. This effect arises from the simultaneous breaking of inversion and time-reversal symmetries in the exciton condensate, resulting in direction-dependent critical counterflow currents. The resulting nonreciprocal perfect Coulomb drag provides a clear and unambiguous transport signature of phase-coherent exciton condensation, offering a powerful and experimentally accessible approach to identify, probe, and control exciton superfluidity in solid-state platforms.

We derive an exact expression for the orbital magnetization of electrons with short-range interactions (such as density-density interactions) in terms of exact zero-frequency response functions of the zero-field system. The result applies to weakly and strongly correlated electrons at zero and finite temperature, provided that the local grand potential density only depends on local thermodynamic parameters. We benchmark the formula for non-interacting and weakly-coupled electrons. To zeroth and first orders in the interaction strength, it agrees with the modern theory of orbital magnetization and its recent generalization to self-consistent Hartree-Fock bands. Our work provides an exact framework of interacting orbital magnetization beyond mean-field treatments, and paves the way for quantitative studies of strongly correlated electrons in external magnetic fields.

High-frequency stability is crucial for the performance of graphene resonators in sensing and timekeeping applications. However, the extreme miniaturization and high mechanical compliance that make graphene attractive also render it highly susceptible to nonlinearities, degrading frequency stability. Here, we demonstrate that graphene parametric oscillators provide an alternative nonlinear operating regime, where short-term frequency stability can be enhanced despite strong nonlinearity. By operating graphene resonators in a phase-locked loop (PLL), we experimentally demonstrate that parametric oscillations in the post-bifurcation regime achieve lower Allan deviation at fast integration times than Duffing oscillations at identical amplitudes. This improvement originates from strong nonlinear damping inherent to parametric oscillators, which suppresses amplitude-to-frequency noise conversion at large amplitudes. A minimal theoretical model captures observed phase diffusion and identifies nonlinear damping as the dominant mechanism governing phase noise reduction. These results highlight the role of nonlinear dissipation in enabling precision sensing beyond conventional limits of graphene oscillators.

We propose a computational methodology based on a hierarchical cluster growth process to solve spin-3/2 Ising models efficiently. The method circumvents the exponential complexity (\(4^{N}\)) of the canonical ensemble partition function by iteratively constructing finite magnetic clusters of size \(N_g\), where the effective spin state of a site in generation \(g+1\) is determined by the local magnetization of a cluster from generation \(g\). This approach, which shares conceptual ground with effective field theories, allows the study of systems of effectively very large size \(N = N_0 (N_g)^{g}\). We apply the formalism to the ferromagnetic spin-3/2 Ising model on a honeycomb lattice, modeling the monolayer CrI$_3$, a prototypical two-dimensional Ising magnet. The model, calibrated using the experimental transition temperature (\(T_{c} \simeq 45\) K), successfully reproduces key experimental features: the temperature dependence of the magnetization \(m(T)\), including its inflection point, and the broadened peak in the specific heat \(c_v(T)\). We also compute the entropy \(s(T)\), finding a finite residual value at low temperatures consistent with the system's double degeneracy. Our results demonstrate that this hierarchical cluster method provides a quantitatively accurate and computationally efficient framework for studying complex magnetic systems.

We consider a run-and-tumble particle (RTP) with stochastic resetting confined to the half line $[0,\infty)$ with a sticky boundary at $x=0$. In the bulk the RTP tumbles at a constant rate $α>0$ between velocity states $\pm v$ with $v>0$ and randomly resets to its initial position and orientation $(x_0,k_0)\in(\mathbb{R}^+,\pm)$. When the RTP reaches the target at $x=0$ it attaches to the boundary for a random waiting time before either detaching and continuing to navigate the bulk domain or (permanently) entering the target. These events are the analogs of adsorption, desorption, and absorption of a particle by a partially reactive surface in physical chemistry. We use renewal theory to characterize the particle trajectory in terms of successive binding events under two distinct desorption protocols: via resetting to $(x_0,k_0)$ and via continuous movement from $x=0$ with velocity $+v$. First we derive the nonequilibrum stationary state (NESS) in the case of no absorption and characterize the accumulation at the boundary. Second, we compute the mean first passage time (MFPT) statistics. In addition to observing the usual unimodal dependence of the MFPT on bulk resetting, both the NESS and MFPT strongly depend on the initial orientation $k_0$ and the desorption protocol. For instance, if the initial orientation is toward the boundary, we find that the desorption-induced resetting protocol can reduce the MFPT more effectively than the non-resetting desorption protocol. We also show how matching the desorption kinetics with the bulk resetting or tumbling rate introduces a trade-off between minimizing the adsorption and absorption times. In this setting we find that the desorption protocol which minimizes the absorption MFPT for a given set of parameters is almost always the opposite of that favored when desorption and bulk kinetics are not the same.

Experiments suggest that nuclear spins play a significant role in the quantum Hall effect (QHE) near integer and fractional QHE spin transitions, but many of these phenomena still remain to be understood. Here we study theoretically the dynamic nuclear polarization (DNP) in the two-dimensional electron liquid near a quantum point contact (QPC) or a domain wall between spin polarized and unpolarized phases induced by electrostatic gating in the fractional QHE at a filling factor 2/3 and analyze the dependence of the spin transition on temperature and the magnitude of the flowing current. We demonstrate that nearly all nuclear spins in the QPC or in the domain wall can be polarized by the electric current. The Overhauser effective magnetic field from the DNP can be strong enough to induce (or modify) a phase transition between polarized and unpolarized phases. This changes the gate voltages and magnetic fields required for the spin transitions, and leads to the reconstruction of the boundary between two phases and a domain wall and a current path displacement. The spread of nuclear spin polarization and the domain wall displacement are strongly asymmetric with respect to the direction of the current flow. Equilibration due to hyperfine interactions and its role on the nuclear spin polarization, domain wall displacements and spin transitions is studied. Back and forth oscillatory transitions between polarized and unpolarized phases near a source contact are discussed. Hyperfine interactions of nuclear spins provide a route for observation and control of the parafermion zero modes that can be induced when the domain wall between the polarized and unpolarized regions is placed in the proximity of a superconductor

We evaluate electromagnetic-response observables in a half-filled Chern band, across a topological phase transition between a composite Fermi liquid (CFL) and a Fermi liquid (FL) phase. While a sharp gapped plasma mode exists deep in the CFL phase, we demonstrate that it is damped near the proposed continuous phase transition between CFL and FL. This plasmon-damping phenomenon originates from emergent gauge fields and a Dirac-fermion-like spectrum. Similar features also occur in other continuous deconfined topological phase transitions, such as the Laughlin to superfluid transition in a bosonic system. In particular, this damping behavior extends over a finite range across the phase boundary, and, hence, we expect it to persist even when the transition is weakly first-order. Furthermore, we analyze the behavior of the Drude weight, the wavevector-dependent conductivity, and the chiral mirror effect across these topological phase transitions.

Electrical conductivity is the most fundamental charge transport parameter, and measurements of conductivity are a basic part of materials characterization for nearly all conducting materials. In thin films, conductivity is often measured in four bar architectures in which the current source and voltage measurement are spatially separated to eliminate systematic error due to contact resistance. Despite the apparent simplicity of these measurements, we demonstrate here that the four bar architecture is subject to significant systematic error arising from the finite conductivity of the metal electrodes. Remarkably, these systematic errors can in some cases become unbounded, producing arbitrarily high measured conductivity at modest true film conductivities, within the range relevant to emerging thin film thermoelectric materials such as conducting polymers. These unbounded errors, which can occur even in properly conducted four-point measurements of patterned films, likely explain literature reports of extremely high conductivities in conducting polymers, and can lead to anomalous scaling in temperature dependent studies, potentially leading to incorrect interpretation of the relevant charge transport mechanism. We characterize the device geometric factors that control these errors, which stand partially at odds with those required for accurate Seebeck coefficient measurements. Our analyses allow us to identify device architectures that provide small systematic errors for conductivity and Seebeck coefficient while still providing a low measurement resistance, critical to reducing noise in thermal voltage measurements. These findings provide important guidelines for accurate measurements in the growing field of thin-film thermoelectric materials.

Active control over strong optical nonlinearity in solid-state systems is central to unlocking exotic many-body phenomena and scalable photonic devices. While exciton-polaritons in transition metal dichalcogenides (TMDs) offer a promising platform, their practical utility is often impeded by fixed interaction parameters and an intrinsic trade-off between nonlinearity and oscillator strength. Here, we report electrically tunable dipolar polaritons in a dual-gated bilayer MoS2 microcavity, demonstrating in situ reshaping of the dispersion and modulation of the light-matter coupling strength via the quantum-confined Stark effect. Crucially, this architecture enables a giant polariton-polariton interaction strength tunable by a factor of seven. This nonlinearity enhancement arises from a synergistic interplay, in which the electric field amplifies the microscopic dipolar repulsion while simultaneously optimizing the macroscopic excitonic Hopfield coefficient. Furthermore, electrostatic doping serves as an independent control knob to switch the system between strong and weak coupling regimes. Our findings bridge the gap between strong optical coupling and giant dipolar nonlinearities, establishing the TMD homobilayer as a versatile platform for engineering programmable correlated many-body states on a chip.

Integrable quantum many-body systems host families of extensive conservation laws, some of which are fragile: even infinitesimal perturbations can qualitatively alter their dynamical constraints. Here we show that this fragility leaves a clear experimental fingerprint in a one-dimensional quantum spin chain of as few as 14 Rydberg atoms. Weak integrability breaking from interatomic dipolar couplings is directly detectable within experimentally accessible times in the dynamics of non-local observables. In particular, magnetization fluctuations are highly sensitive to the breaking of fragile conservation laws and exhibit anomalous growth, which we observe experimentally; similar signatures appear in a semilocal string observable. Numerical simulations on substantially longer chains and a simplified classical stochastic model reproduce those features. We establish non-local observables as a sensitive probe of fragile conservation laws in quantum spin chains and Rydberg-atom arrays as a platform to test perturbative descriptions of quantum many-body dynamics with weak integrability breaking.

We study localization driven solely by interparticle interactions in moiré lattice systems without intrinsic disorder or externally imposed quasiperiodic potentials. We consider a one-dimensional bilayer with incommensurate lattice constants, described by a spin-dependent Fermi-Hubbard-type model with short-range interlayer interactions, where quasiperiodicity emerges only through interactions. Exact diagonalization shows that quenching hopping in one layer generates an interaction-induced mosaic potential with multiple mobility edges. When both layers are dynamical, increasing interlayer interactions drives transitions among ergodic, critical, and many-body localized regimes, with energy-dependent coexistence in certain parameter ranges. An exact mapping to a noninteracting single-particle model on a higher-dimensional structured graph provides a unified interpretation of these results and suggests an experimentally accessible route to interaction-induced moiré physics and localization.

We examine nonlinear transport in a viscous two-dimensional electron fluid within narrow GaAs channels. The differential magnetoresistance shows nonmonotonic behavior, a signature of electron pairing in the hydrodynamic regime. Theoretical models that account for both the influence of these interactions on shear stress relaxation and viscosity changes from electron heating show good agreement with the data. The nonlinear regime thus reveals how such correlated states govern the hydrodynamic behavior of the electron fluid. Our findings establish the nonlinear transport regime as a powerful probe for dissecting the complex interplay of correlated electron states and momentum relaxation in the hydrodynamic flow of an electron fluid.

The construction of the Hamiltonian matrix \textbf{H} is an essential, yet computationally expensive step in \textit{ab-initio} device simulations based on density-functional theory (DFT). In homogeneous structures, the fact that a unit cell repeats itself along at least one direction can be leveraged to minimize the number of atoms considered and the calculation time. However, such an approach does not lend itself to amorphous or defective materials for which no periodicity exists. In these cases, (much) larger domains containing thousands of atoms might be needed to accurately describe the physics at play, pushing DFT tools to their limit. Here we address this issue by learning and directly predicting the Hamiltonian matrix of large structures through equivariant graph neural networks and so-called augmented partitioning training. We demonstrate the strength of our approach by modeling valence change memory (VCM) cells, achieving a Mean Absolute Error (MAE) of 3.39 to 3.58 meV, as compared to DFT, when predicting the Hamiltonian matrix entries of systems made of $\sim$5,000 atoms. We then replace the DFT-computed Hamiltonian of these VCMs with the predicted one to compute their energy-resolved transmission function with a quantum transport tool. A qualitatively good agreement between both sets of curves is obtained. Our work provides a path forward to overcome the memory and computational limits of DFT, thus enabling the study of large-scale devices beyond current \textit{ab-initio} capabilities

The theory of inter-spin-subband electric dipole spin resonance in transition metal dichalcogenide heterobilayers is proposed. Our symmetry analysis demonstrates that, in contrast to monolayers, the reduced symmetry of heterobilayers enables coupling between conduction band spin subbands by an electric field. We establish the optical selection rules for all six high-symmetry stacking configurations. The microscopic mechanism of the effect is identified as the spin-orbit coupling induced mixing of Bloch states from different conduction bands, which generates a non-zero momentum matrix element between the spin-split states. It also leads to the linear-in-wavevector spin-dependent terms in the effective Hamiltonian, i.e., the Rashba effect. Our estimates show that the rate of electric-dipole spin-flip transitions exceeds by far that of the magnetic-dipole transitions in transition metal dichalcogenide heterobilayers.

The coherent interaction between quantum light and material excitations in semiconductor nanostructures is investigated using a fully quantized microscopic approach that incorporates many-body Coulomb correlations. The simulations demonstrate that the quantum dynamics is influenced by biexciton continuum states and is highly sensitive to both the frequency of the cavity mode and the strength of the light-matter coupling.

Recent interest has surged in eigenvalue's nonlinearity-based topological transport governed by the equation of auxiliary eigenvalues $HΨ=ωS(ω)Ψ$ [T. Isobe et al., Phys. Rev. Lett. 132, 126601 (2024); C. Bai and Z. Liang, 111, 042201 (2025); Phys. Rev. A 112, 052207 (2025)] rather than the conventional Schrodinger equation $HΨ=EΨ$ in conservative settings, yet non-Hermitian generalizations remain uncharted. In this work, we are motivated to investigate the nonlinear Thouless pumping in a non-Hermitian and nonlinear Rice-Mele model. In particular, we uncover how non-Hermiticity parameters can induce fractional topological phases--even in the presence of quantized topological invariants as predicted by conventional linear approaches. Crucially, these fractional phases are naturally explained within the framework of the equation of auxiliary eigenvalues, directly linking nonlinear spectral characteristics to the bulk-boundary correspondence. Our findings reveal novel emergent phenomena arising from the interplay between nonlinearity and non-Hermiticity, providing key insights for the design of topological insulators and the controlled manipulation of quantum edge states in the real world.

Several models of the strange-metal state of the cuprate superconductors postulate the existence of strong inelastic forward scattering of the electrons, but direct evidence of such scattering is missing. Here we show that angle-resolved photoemission spectroscopy (ARPES) provides a unique tool which can address this issue. We propose a two-lifetime phenomenological model of the superconducting state of the cuprates and we show that it explains several salient low-energy features of the measured ARPES spectra. The model enables discrimination between forward- and large-angle scattering and, in addition, gives access to the magnitude of the gap function away from the Fermi surface.

This text revisits the origins of econophysics through the figure of Benoît Mandelbrot, not as the father of fractals, but as the instigator of a distinctive scientific posture. The guiding thread is methodological: accept the stubborn features of the data and use models as instruments for intuition rather than as axiomatic certificates of truth. In this perspective, scaling, intermittency and extremes are not peripheral imperfections around a well-behaved equilibrium; they are the very texture of economic and financial fluctuations. This naturally shifts attention from exogenous narratives to endogenous dynamics: interactions, feedback loops, and collective amplification mechanisms that can make systems intrinsically {\it fragile}. We argue that the importation of concepts from statistical physics -- criticality, disorder, emergence, multiplicative cascades -- should be read not as an artificial transposition but as a candid attempt to look for generic mechanisms compatible with empirical regularities observed across scales, from markets to macroeconomic aggregates.

We establish a relationship between the Davey--Stewartson II (DS II) integrable system in $(2{+}1)$ dimensions and quasi-exactly solvable planar interacting Dirac Hamiltonians that exhibit the super-Klein tunneling (SKT) effect. The Dirac interactions are constructed from the real and imaginary parts of breather solutions of the DS II system. In this framework, the SKT effect arises when the energy is tuned to match the constant background of the soliton, while the resulting Dirac Hamiltonians simultaneously support bound states embedded in the continuum. By imposing the SKT boundary conditions, we employ Darboux transformations to construct a general three-parameter family of DS II breather solutions that can be mapped to Dirac Hamiltonians. At the initial soliton time, the corresponding Dirac systems form a massless two-parameter family of Hermitian models with nontrivial electrostatic potentials. As the soliton time evolves, the systems become $\mathcal{PT}$-symmetric and develop a nontrivial imaginary mass term. Finally, when the soliton time is taken to be imaginary, the construction yields Hermitian Dirac systems that lack time-reversal symmetry. In all cases, we identify the emergence of quasi-symmetry transformations that preserve the SKT subspace of states while not commuting with the full Hamiltonian.

We present a fully atomistic approach to exciton-phonon coupling in semiconductor quantum dots that bridges microscopic electronic-structure calculations with non-Markovian open-quantum-system dynamics. On the example of an InAsP quantum dot embedded in an InP matrix, we compute single-particle states using an ab initio-parametrized tight-binding model, then obtain correlated many-body wave functions of neutral excitons, biexcitons, and charged trions via the configuration-interaction method. Using these correlated states, we compute the exciton-phonon coupling matrix elements. The resulting phonon spectral densities for different excitonic complexes are compared with the widely used analytical super-Ohmic form and reveal deviations at higher energies originating from the realistic dot geometry and atomistic wave functions, whereas configuration mixing is found to play only a minor role. Furthermore, we extract radiative lifetimes comparable to values experimentally measured. As a direct application, we simulate the emission brightness of a pulsed-driven quantum dot and demonstrate that the atomistically derived spectral density substantially broadens the region of efficient off-resonant excitation compared to the analytical model. The presented framework provides a nearly parameter-free route to simulate the non-Markovian open quantum dynamics in semiconductor quantum dots.

We introduce a one-dimensional non-Hermitian Kitaev chain with staggered imbalance in the $p$-wave superconducting pairing. By tuning the chemical potential and the pairing imbalance, we find that the eigenenergy spectrum undergoes real-to-complex transitions, and the spectral gap can change from a real to an imaginary line gap. The pairing imbalance significantly enlarges the parameter region supporting a topological superconducting phase. Remarkably, we show that a topologically nontrivial phase hosting Majorana zero modes can be induced by varying the pairing imbalance, even in the regime of strong chemical potential. The gap-closing points and phase boundaries are determined analytically, and the resulting phase diagrams are characterized by a nonzero topological invariant. Furthermore, we identify the existence of Majorana zero modes and finite-energy Majorana edge modes in this system. Our results reveal exotic phenomena arising from imbalanced pairing and establish a new platform for exploring topological superconductivity in non-Hermitian systems.

Quantum geometry, encoded in the Berry curvature and quantum metric, has unified diverse anomalous transport phenomena in solids, yet a microscopic quantum-geometric theory of entropy transport for Bloch electrons is still lacking. We formulate an entropy continuity equation for noninteracting fermions driven by an electric field, starting from the von Neumann entropy, and obtain quantum-mechanical expressions for the entropy current density and entropy production rate. Introducing relaxation through a relaxation-time dissipator, we identify the quantum metric as the origin of the leading entropy production, providing a direct microscopic diagnostic of dissipation in both the extrinsic Drude response and an intrinsic nonlinear Ohmic contribution controlled by quantum metric. We further predict an entropy Hall effect arising from the Berry curvature and show that it obeys an Onsager reciprocal relation with the anomalous Nernst effect under a temperature gradient. Finally, we establish universal relations connecting entropy and charge currents under DC and AC driving, offering experimentally accessible probes of quantum geometry through nonequilibrium entropy flow.

We study spin-resolved transport in a ballistic quantum dot with Rashba spin-orbit coupling, focusing on charge-to-spin conversion and spin Hall effect. In the regime where the dot size is comparable to the Fermi wavelength, we identify a clear crossover from weak localization to weak antilocalization as the Rashba coupling increases. This transition is accompanied by gate-tunable spin currents of Edelstein and spin Hall type, whose behavior reflects the underlying electron wavefunction interference. Notably, the Edelstein current shows an inflection point at the critical Rashba strength, signaling the crossover from weak localization to weak antilocalization. In the presence of an in-plane magnetic field we also report a transition in angular periodicity of the magnetoresistance -- from $π$ to $2π$ -- arising from the interplay between spin-orbit interaction and Zeeman coupling. These results establish a direct link between quantum coherence, charge-to-spin conversion, and geometric confinement in mesoscopic systems.

Quantum many-body systems pose a formidable computational challenge due to the exponential growth of their Hilbert space. While machine learning (ML) has shown promise as an alternative paradigm, most applications remain at the proof-of-concept stage, focusing narrowly on energy estimation at the lower end of the spectrum. Here, we push ML beyond this frontier by extensively testing HubbardNet, a deep neural network architecture for the Bose-Hubbard model. Pushing improvements in the optimizer and learning rates, and introducing physics-informed output activations that can resolve extremely small wave-function amplitudes, we achieve ground-state energy errors reduced by orders of magnitude and wave-function fidelities exceeding 99%. We further assess physical relevance by analysing generalized inverse participation ratios and multifractal dimensions for ground and excited states in one and two dimensions, demonstrating that optimized ML models reproduce localization, delocalization, and multifractality trends across the spectrum. Crucially, these qualitative predictions remain robust across four decades of the interaction strength, e.g. spanning across superfluid, Mott-insulating, as well as quantum chaotic regimes. Together, these results suggest ML as a viable qualitative predictor of many-body structure, complementing the quantitative strengths of exact diagonalization and tensor-network methods.

Recent experiments with rotating Bose gases have demonstrated the interaction-driven hydrodynamic instability of an initial extended strip-like state in the lowest Landau level. We investigate this phenomenon in the low density limit, where the mean-field Gross--Pitaevskii theory becomes inadequate, using exact diagonalisation studies and analytic arguments. We show that the behaviour can be understood in terms of weakly-interacting repulsively-bound few-body clusters. Signatures of cluster behaviour are observed in the expectation values of observables which oscillate at frequencies characterised by the energies of few-body boundstates. Using a semiclassical theory for interacting clusters, we predict the long-time growth of the cloud width to be a power law in the logarithm of time. This slow thermalisation of bound clusters represents a form of quantum many-body scars.

Establishing quantitative adiabaticity criteria at finite temperature remains substantially less developed than in the pure-state setting, despite the fact that realistic quantum systems are never at absolute zero. Here we derive rigorous bounds on the Hilbert-Schmidt fidelity between mixed states by combining a mixed-state quantum speed limit with mixed-state fidelity susceptibility within the Liouville space formulation of quantum mechanics. Applied to protocols that drive an initial Gibbs state toward a quasi-Gibbs target, these bounds yield an explicit threshold driving rate for the onset of nonadiabaticity. For a broad class of local Hamiltonians in gapped phases, we show that, in the thermodynamic limit, the threshold factorizes into two factors: a system-size contribution that recovers the zero-temperature scaling and a universal temperature-dependent factor. The latter is exponentially close to unity at low temperature, whereas at high temperature it increases linearly with temperature. We verify the predicted scaling in several spin-1/2 chains by obtaining closed-form expressions for the threshold driving rate. Our results provide practical and largely model-independent criteria for finite-temperature adiabaticity in closed many-body systems.

In this paper, we review the basic concepts of topologically protected edge modes using the Su Schrieffer Heeger (SSH) model, originally introduced to describe electrical conductivity in doped polyacetylene polymer chains. We then propose an electrical circuit that emulates this model, provide its mathematical description, and present its experimental realization. The experimental setup is described in detail, with explanations designed to be broadly accessible without much prior familiarity with lattice theory, thus offering an introduction to this active area of research. Both theoretical predictions and experimental results confirm the presence of these modes, showing very good overall agreement. Using this concrete experimental system as a motivating example, we highlight the key aspects of topological protection.

Langevin integrators based on operator splitting are widely used in molecular dynamics. This work examines Langevin splitting schemes from the perspective of their internal trajectories and observation points, complementing existing generator-based analyses. By exploiting merging, splitting, and cyclic permutation of elementary update operators, formally distinct schemes can be grouped according to identical or closely related trajectories. Accuracy differences arising from momentum updates and observation points are quantified for configurational sampling, free-energy estimates, and transition rates. While modern Langevin integrators are remarkably stable under standard simulation conditions, subtle but systematic biases emerge at large friction coefficients and time steps. These results clarify when accuracy differences between splitting schemes matter in practice and provide an intuitive framework for understanding observation effects.

The band complex formed by multiple topological states has attracted extensive attention for the emergent properties produced by the interplay among the constituent states. Here, based on group theory analysis, we present a scheme for rapidly identifying the Weyl-Dirac nodal lines (a complex of Weyl and Dirac nodal lines) in bosonic systems. We find only 5 of the 230 space groups host Weyl-Dirac nodal line phonons. Notably, the Dirac nodal line resides along the high-symmetry line, whereas the Weyl nodal line is distributed on the high-symmetry plane and is interconnected with the Dirac nodal line, jointly forming a composite nodal network structure. Unlike traditional nodal nets, this nodal network exhibits markedly distinct surface states on different surfaces, which can be attributed to the fundamental differences in the topological properties between the Weyl and Dirac nodal lines. This unique property thus allows the material to present distinct surface states in a termination-selective manner. Furthermore, by first-principles calculations, we identify the materials NdRhO$_{3}$ and ZnSe$_{2}$O$_{5}$ as candidate examples to elaborate the Weyl-Dirac nodal line and their related topological features. Our work provides an insight for exploring and leveraging topological properties in systems with coexisting multiple topological states.

Despite the unitary evolution of closed quantum systems, long-time expectation of local observables are well described by thermal ensembles, providing the foundation of quantum statistical mechanics. A promising route to understanding this quantum thermalization is the eigenstate thermalization hypothesis (ETH), which posits that individual energy eigenstates already appear locally thermal. Subsequent studies have extended this concept to the full ETH, which captures higher-order correlations among matrix elements through nontrivial relations. In this work, we perform a detailed exact-diagonalization study of finite-size corrections to these relations in the canonical ensemble. We distinguish two distinct sources of corrections: those arising from energy fluctuations, which decay polynomially with system size, and those originating from fluctuations within each energy window, which decay exponentially with system size. In particular, our analysis resolves the puzzle that, for certain observables, finite-size corrections exhibit anomalous growth with increasing system size even in chaotic systems. Our results provide a systematic and practical methodology for validating the full ETH in quantum many-body systems.

Lindbladian quantum master equations (LEs) are the most popular descriptions for quantum systems weakly coupled to baths. But, recent works have established that in many situations such Markovian descriptions are fundamentally limited: they cannot simultaneously capture populations and coherences even to the leading-order in system-bath couplings. This can cause violation of fundamental properties like thermalization and continuity equations associated with local conservation laws, even when such properties are expected in the actual setting. This begs the question: given a physical situation, how do we know if there exists an LE that describes it to a desired accuracy? Here we show that, for both equilibrium and non-equilibrium steady states (NESS), this question can be succinctly formulated as a semidefinite program (SDP), a convex optimization technique. If a solution to the SDP can be found to a desired accuracy, then an LE description is possible for the chosen setting. If not, no LE description is fundamentally attainable, showing that a consistent Markovian treatment is impossible even at weak system-bath coupling for that particular setting. Considering few qubit isotropic XXZ-type models coupled to multiple baths, we find that in most parameter regimes, LE description giving accurate populations and coherences to leading-order is unattainable, leading to rigorous no-go results. However, in some cases, LE description having correct populations but inaccurate coherences, and satisfying local conservation laws, is possible over some of the parameter regimes. Our work highlights the power of semidefinite programming in the analysis of physically consistent LEs, thereby, in understanding the limits of Markovian descriptions at weak system-bath couplings.

Understanding how the physical properties of a fluid influence bacterial behavior is essential for explaining how microorganisms interact with their environment and with animal hosts. Here, we examine how changes in fluid viscosity and rheological properties affect the locomotion of the marine bacterium Vibrio fischeri and its ability to produce luminescence through cell--cell communication. We track the three-dimensional motion of single cells in well-defined fluids with different physical properties and measure the luminescence emitted by cell populations. We find that fluids with higher viscosity cause V. fischeri to spend more time in a slower, turning-focused swimming mode, which reduces how effectively cells spread out and encounter the chemical signals required to activate luminescence. As a result, luminescence first increases and then decreases in Newtonian fluids, but decreases monotonically in fluids that exhibit non-Newtonian rheological behavior. Computer simulations based on our measurements confirm that the ability of cells to explore their surroundings plays a central role in determining when and how strongly they communicate. These findings reveal a direct link between the physical environment, bacterial movement, and collective behavior, and offer new insight into how microorganisms adapt to complex fluid habitats, including those found inside animal hosts.

Muscle contraction is a prototypical multiscale chemomechanical process in which ATP hydrolysis at the molecular level drives force generation and mechanical work at larger scales. A long-standing challenge is to connect microscopic cross-bridge dynamics to macroscopic observables while retaining an explicit, thermodynamically consistent energetic budget for chemical-to-mechanical transduction. Here we use the Energetic Variational Approach (EnVarA) to unify Hill's cycle-affinity viewpoint with Huxley's sliding-filament mechanics within a single thermodynamically closed framework. We formulate a three-state Fokker--Planck-jump description for cross-bridge populations evolving on state-dependent free-energy landscapes, in which ATP hydrolysis enters through local detailed balance and biases the transition rates. Filament sliding velocity is incorporated as a convective transport mechanism in the Fokker--Planck dynamics, so that mechanical power exchange with the external motion emerges transparently from the resulting energy-dissipation law together with chemical input and irreversible dissipation. Under chemostatted conditions and a fast-equilibration closure for the attached substates, the model reduces to a closed two-state molecular motor description; in a further singular limit, this reduction recovers a Huxley-type transport-reaction equation. Proof-of-concept simulations of the reduced model reproduce a Hill-like force-velocity relation and show how ATP availability modulates the force-velocity curve while preserving its characteristic Hill-type shape.

We experimentally investigate the dynamics of a sphere rolling down a granular slope by varying the initial velocity, slope angle, and sphere density. The results show that the sphere rolls down with constant deceleration while sinking into the granular bed. $δ/R$ (the sinking depth $δ$ normalized to the sphere radius $R$) is scaled by the sphere density normalized by the bulk density of the granular layer. To evaluate the translational energy dissipation, we introduce an effective friction coefficient $μ_\mathrm{d}$. We demonstrate that $μ_\mathrm{d}$ decreases with increasing the slope angle and the slip ratio. Furthermore, systematic measurements over a wide range of sphere densities reveal that $μ_\mathrm{d}$ increases linearly with $δ/R$ : $μ_\mathrm{d}=β(δ/R)+μ_0$. The value of $μ_0$ is linearly decreasing with slip ratio and its coefficient $β(\simeq0.41)$ does not vary significantly. The results suggest that the normalized depth and slip ratio determine the effective friction of a rolling sphere.

Non-Hermitian systems have attracted significant theoretical interest due to their extreme properties. However, realizations have mostly been limited to classical applications or artificial setups. In this study, we focus on the quantum nature inherent in bosonic Bogoliubov-de Gennes (BdG) systems, which from the perspective of spectral theory corresponds to non-Hermiticity. Based on this insight, we propose a steady-state skin effect in quantum condensed matter utilizing such BdG non-Hermiticity. Specifically, we introduce BdG quantum terms arising from parametric pumping to the edge states of an underlying bosonic Hermitian Chern insulator, thereby realizing non-Hermiticity without dissipation. This system design has the advantage of being largely independent of microscopic model details. Through analysis using non-equilibrium Green's functions, we find that under open boundary conditions, a steady state exhibiting the non-Hermitian skin effect is realized. The pronounced corner particle accumulation observed in this steady state shows quadrature anisotropy, which manifests the bosonic quantum nature. Our results bridge the gap between the fascinating mathematics of non-Hermitian matrices and practical quantum physical systems.

The water-repellence properties of superhydrophobic surfaces make them promising for many applications. However, in some extreme environments, such as high humidities and low temperatures, condensation on the surface is inevitable, which induces the loss of surface superhydrophobicity. In this study, we propose a freezing-melting strategy to achieve the dewetting transition from the Wenzel state to the Cassie-Baxter state. It requires freezing the droplet by reducing the substrate temperature and then melting the droplet by heating the substrate. The condensation-induced wetting transition from the Cassie-Baxter state to the Wenzel state is analyzed first. Two kinds of superhydrophobic surfaces, i.e., single-scale nano-structured superhydrophobic surface and hierarchical-scale micro-nano-structured superhydrophobic surface, are compared and their effects on the static contact states and impact processes of droplets are analyzed. The mechanism for the dewetting transition is analyzed by exploring the differences in the micro/nano-structures of the surfaces and it is attributed to the unique structure and strength of the superhydrophobic surface. These findings will enrich our understanding of the droplet-surface interaction involving phase changes and have great application prospects for the design of superhydrophobic surfaces.

Spin relaxometry using solid-state spin defects, such as the diamond nitrogen-vacancy (NV) center, probes dynamical processes by measuring how environmental fluctuations enhance the spin relaxation rate. In the weak-coupling limit, relaxation rates sample the transverse magnetic-noise power spectral density through a sensor-specific filter function, turning the defect into a local, frequency-selective noise spectrometer. This review bridges theory and experiment, clarifying how measured relaxation rates map onto noise spectra and how near-field geometry shapes the response. We highlight representative applications across condensed-matter physics, chemical and biological sensing, and relaxometry-based magnetic-resonance spectroscopy. We conclude with emerging opportunities and key challenges.

Probabilistic graphical models with frustration exhibit rugged energy landscapes that trap iterative optimization dynamics. These landscapes are shaped not only by local interactions, but crucially also by the global loop structure of the graph. The famous Bethe approximation treats the graph as a tree, effectively ignoring global structure, thereby limiting its effectiveness for optimization. Loop expansions capture such global structure in principle, but are often impractical due to combinatorial explosion. The $M$-layer construction provides an alternative: make $M$ copies of the graph and reconnect edges between them uniformly at random. This provides a controlled sequence of approximations from the original graph at $M=1$, to the Bethe approximation as $M \rightarrow \infty$. Here we generalize this construction by replacing uniform random rewiring with a structured mixing kernel $Q$ that sets the probability that any two layers are interconnected. As a result, the global loop structure can be shaped without modifying local interactions. We show that, after this copy-and-reconnect transformation, there exists a regime in which layer-to-layer fluctuations decay, increasing the probability of reaching the global minimum of the energy function of the original graph. This yields a highly general and practical tool for optimization. Using this approach, the computational cost required to reach these optimal solutions is reduced across sparse and dense Ising benchmarks, including spin glasses and planted instances. When combined with replica-exchange Monte Carlo, the same construction increases the polynomial-time algorithmic threshold for the maximum independent set problem. A cavity analysis shows that structured inter-layer coupling significantly smooths rugged energy landscapes by collapsing configurational complexity and suppressing many suboptimal metastable states.

A key signature of MBL (many-body localization) is the slow rate at which information spreads. It is shown that the infinite random Heisenberg XXZ spin-$\frac12$ chain exhibits slow propagation of information (logarithmic light cone) in any arbitrary but fixed energy interval. The relevant parameter regime, which covers both weak interaction and strong disorder, is determined solely by the energy interval.

I study via Langevin dynamics simulations two opposite cases of systems of particles that alternate their identity according to density dependent motility (DDM) rules and interact via a soft repulsive potential. In the correlated case, dilute regions are passive and dense regions are active, while in the anti-correlated case, dilute regions are active and dense regions are passive. I classify the emerging steady states, explain the principal phase transitions, and finally suggest directions for further investigation.

We study a variant of the pseudo-inverse learning rule for Hopfield-like Neural Networks, which allows the network to infer archetypal concepts on the basis of a limited number of examples. The mean-field replica theory for this model reveals how this generalization ability is mediated by a multitude of states, with diverse thermodynamic properties, coexisting with the standard Hopfield ones. They appear and vanish through smooth transitions or discontinuous jumps and, interestingly, show much stronger Replica Symmetry Breaking (RSB) effects than the standard Hopfield model, as captured by our 1RSB analysis. Our results, in excellent agreement with numerical simulations, provide deeper insight into the interplay between memory storage and generalization in attractor neural networks.

Non-Hermitian degeneracies of Lindblad generators (Liouvillian exceptional points) can induce non-exponential relaxation and higher-order poles in dynamical response functions. A collective spin coupled to a polarized Markovian bath exhibits an \emph{exceptional spectral phase} in which defective Liouvillian modes imprint super-Lorentzian features in frequency-resolved spectra. We compute the emission spectrum via the Liouvillian resolvent, identify symmetry-sector selection rules, and demonstrate that exceptional-point signatures are strongly state-dependent: they are suppressed in steady-state fluorescence yet become unambiguous for generic (infinite-temperature or random) initial states. Our results provide an experimentally accessible spectroscopic diagnostic of many-body Liouvillian exceptional phases and clarify when steady-state emission can (and cannot) reveal them.

Triadic percolation turns bond percolation into a dynamical problem governed by an effective one-dimensional unimodal map. We show that the geometry of superstable cycles provides a direct, map-agnostic probe of local nonlinearity: specifically, the distance from the map's maximum to a distinguished next-to-maximum point on the attracting $2^n$-cycle (which coincides with a preimage of the maximum at $2^n$-superstability) scales as $|Δp|^γ$ with $γ= 1/z$, where $z$ is the nonflat order of the maximum. This prediction is verified across canonical unimodal families and heterogeneous triadic ensembles, with Lyapunov spectra corroborating the one-dimensional reduction. A derivative condition on the activation kernel fixes the local nonlinearity order $z$ (and thus, under standard unimodal-map hypotheses, the associated $z$-logistic universality class) and gives conditions under which $z>2$ can be realized. The diagnostic operates directly on orbit data under standard regularity assumptions, providing a practical tool to classify universality in higher-order networks.

Our previous understanding of electronic transport in disordered systems was based on the assumption that there is a finite Fermi velocity for the relevant electrons. The Fermi velocity determines important length scales in disordered systems such as the diffusion length and the localization length. However, in disordered systems with vanishing or nearly vanishing Fermi velocity, it is uncertain what determines the important length scales in such systems. In this work, we use the 1D Lieb lattice with isolated flat bands as an example to show that the quantum metric length (QML) is a fundamental length scale in the ballistic, diffusive and localization regimes. The QML is defined through the Bloch state wave functions of the flat bands. In the ballistic regime with short junctions, the QML controls the finite energy transport properties. In the localization regime with long junctions, the localization length is determined by the QML and remarkably, independent of disorder strength over a wide range of disorder strength. We call this unconventional localization regime, the quantum metric localization regime. In the diffusive regime, we demonstrate that the diffusion coefficient is linearly proportional to the QML via the wave-packet dynamics numerically. Importantly, the numerical results are consistent with the analytical results obtained through the Bethe-Salpeter equation. We conclude that the QML is a fundamentally important length scale governing the properties of disordered flat band materials.

We present a comprehensive investigation of structural, magnetic and transport properties of Ni$_2$FeAl Heusler alloy nanoparticles (NPs) synthesized via template-less chemical route. The NPs exhibit high saturation magnetization of 3.02 $μ_ {\rm B}$/f.u. at 5~K, large magnetic anisotropy of 0.238 MJ/m$^3$, and a Curie temperature of 874~K. Magnetocaloric analysis reveals a magnetic entropy change of 3.1 J.kg$^{-1}$K$^{-1}$ at 70 kOe. Low-temperature transport measurements show a weak resistivity upturn, following a $-T^{1/2}$ dependence, indicative of disorder-enhanced electron-electron interactions. First-principles calculations based on density functional theory yield a magneto-crystalline anisotropy energy of 0.987 MJ/m$^3$, consistent with experiment and demonstrate pronounced surface and finite-size effects through comparison of bulk and nanocluster geometries. The combination of high Curie temperature, sizable perpendicular magnetic anisotropy, and moderate spin polarization and magnetic entropy change make the Ni$_2$FeAl as promising candidate for various applications.

Topological classifications of quantum critical systems have recently attracted growing interest, as they go beyond the traditional paradigms of condensed matter and statistical physics. However, such classifications remain largely unexplored at critical points in aperiodic environments, particularly under quasiperiodic modulations. In this Letter, we uncover a novel class of topological quasiperiodic fixed points that are intermediate between the clean and infinite-randomness limits. By exactly solving the quasiperiodic cluster-Ising chain, we unambiguously demonstrate that all phase boundaries separating quasiperiodically modulated phases are governed by a new family of topological Ising-like fixed points unique to strongly modulated quasiperiodic systems: Despite exhibiting indistinguishable bulk critical properties, these fixed points host robust topological edge degeneracies and are therefore topologically distinct from previously recognized quasiperiodic universality classes, as further supported by complementary lattice simulations.

Electron gases in low dimensional materials exhibit unconventional transport and optical phenomena due to reduced phase space, enhanced interactions, and strong sensitivity to lattice symmetry. While commonly realized in quantum confined systems and engineered heterostructures, such states are rare in naturally occurring materials. Hyperbolic materials provide a compelling alternative, as extreme lattice anisotropy can host unconventional electronic states and novel electron-phonon interactions. Here, we investigate the angle resolved polarized Raman (ARPR) response of MoOCl2, the first naturally occurring hyperbolic material whose hyperbolicity originates from a highly anisotropic electron gas. We observe pronounced polarization dependent Fano line shapes, revealing coherent coupling between phonons and an anisotropic electronic continuum. We characterize the directional response of this continuum, incorporating it into effective Raman tensors that quantitatively reproduce the ARPR measurements and capture the distinct Raman fingerprint of MoOCl2. Excitation energy and thickness dependent ARPR measurements further demonstrate a tunable quasi 1D electronic continuum with weak interlayer coupling, establishing MoOCl2 as a model system for Raman studies of electron-phonon coupling in hyperbolic materials

Microbial swarming on mucosal surfaces reshapes microbial communities and influences mucosal healing and antibiotic tolerance. Yet even with time-lapse microscopy and deep learning, analyses of swarming colonies remain descriptive and cannot forecast how their fronts reorganize in time. This limitation is significant because the advancing edge determines access to nutrients, host tissue and competing microbes. We recast the expansion of Enterobacter sp. SM3 swarms as a problem of morphological forecasting, and assemble SwarmEvo, a time-lapse dataset represented as boundary-resolved segmentations. TexPol--Net, a texture- and geometry-aware segmentation model, sharpens diffuse edges and preserves fingered fronts, creating a stable substrate for dynamics. On this representation, we develop Morpher, an autoregressive forecasting network with a ``Morphon'' memory that links local curvature to long-range temporal dependencies. Morpher outperforms leading video-prediction models in maintaining front localization and anisotropic branching, and modest segmentation improvements yield noticeably more stable forecasts. Ablations across sequence models, inference strategies and observation ratios show that attention-based architectures with structural memory best preserve dense-finger propagation. By uniting geometry-aware segmentation with morphology-level forecasting, this framework turns swarming expansion into a predictive dynamical system, enabling quantitative interrogation and potential control of microbial collectives during mucosal repair and gut ecosystem engineering.

We provide a local probabilistic description of the limiting statistics of large preferential attachment trees in terms of the ordinary degree (number of neighbors) but augmented with information on leafdegree (number of neighbors that are leaves). The full description is the joint degree-leafdegree distribution $n_{k,\ell}$, which we derive from its associated multivariate generating function. From $n_{k,\ell}$ we obtain the leafdegree distribution, $m_{\ell}$, as well as the fraction of vertices that are protected (nonleaves with leafdegree zero) as a function of degree, $n_{k,0}$, among numerous other results. We also examine fluctuations and concentration of joint degree-leafdegree empirical counts $N_{k,\ell}$. Although our main findings pertain to the preferential attachment tree, the approach we present is highly generalizable and can characterize numerous existing models, in addition to facilitating the development of tractable new models. We further demonstrate the approach by analyzing $n_{k,\ell}$ in two other models: the random recursive tree, and a redirection-based model.

Acoustic metamaterials offer exceptional control over wave propagation, but their potential remains unfulfilled due to fabrication constraints. Conventional processes yield mostly rigid, planar structures, whereas soft-matter alternatives have so far been confined to ultrasounds. This work overcomes prior limitations with a fully soft 3D metamaterial operating around 200Hz. The design combines a 3D-printed elastomer lattice with resonant inclusions of liquid metal, injected via a network of mesofluidic channels. Its dynamic response is derived from a hybrid strategy uniting a lumped-element model with finite element analysis. Simulations reveal how the dual-phase design decouples flexural and torsional modes, opening a subwavelength band gap for low-frequency elastic waves. Empirical validation is achieved via a custom camera-based vibrometer. Its high spatiotemporal resolution and full-field capabilities enable direct capture of local modes and evanescent waves underlying the band gap. Accelerometer data corroborate these findings and demonstrate greater attenuation than common silicone elastomers at only half of the density. By combining scalable fabrication, compliance, and operations at frequencies relevant to human tactile perception, this novel metamaterial paves the way for lightweight, high-performance cushioning and handles that protect users from harmful vibration exposure.

I present the results from a spatial model of the prisoner's dilemma, played on a toroidal lattice. Each individual has a default strategy of either cooperating ($C$) or defecting ($D$). Two strategies were tested, including ``tit-for-tat'' (TFT), in which individuals play their opponent's last play, or simply playing their default play. Each individual also has a probability of telling the truth ($0 \leq P_{truth} \leq 1$) about their last play. This parameter, which can evolve over time, allows individuals to be, for instance, a defector but present as a cooperator regarding their last play. This leads to interesting dynamics where mixed populations of defectors and cooperators with $P_{truth} \geq 0.75$ move toward populations of truth-telling cooperators. Likewise, mixed populations with $P_{truth} < 0.7$ become populations of lying defectors. Both such populations are stable because they each have higher average scores than populations with intermediate values of $P_{truth}$. Applications of this model are discussed with regards to both humans and animals.

Recently, numerical examples of stable soap bubble clusters with multiple torus bubbles have been presented. The geometry of these clusters is based on the Platonic solids whose vertices have valence $3$ (in order to fulfill Plateau's laws): the tetrahedron, the cube, the dodecahedron. The clusters respectively contain a bubble of genus $3, 5, 11$. The construction is quite generic and can be used with any convex polyhedron. If stable, the cluster obtained using a polyhedron with $n$ faces has $3n+2$ bubbles and one of these bubbles has genus $n-1$. We propose here to show that is it possible to get stable soap bubble clusters with multiple torus bubbles using a geometry based on prisms and Archimedean solids as well.

Self-avoiding random walks on graphs can be seen as walkers interacting with their own past history. This letter considers a complementary class of dynamics: Mutual future avoiding random walks (MFARWs), where stochastically driven walkers are avoiding each others planned future trajectories. Such systems arise naturally in conceptual models of shared mobility. We show that periodic behavior emerges spontaneously in such MFARWs, and that periodic and unstructured behavior coexist, providing a first example of Chimera style behavior of non-oscillatory paths on networks. Further, we analytically describe and predict the onset of structure. We find that the phase transition from unstructured to periodic behavior is driven by a novel mechanism of self-amplifying coupling to the periodic components of the stochastic drivers of the system. In the context of shared mobility applications, these Chimera states imply a regime of naturally stable co-existence between flexible and line-based public transport.

Conflict fatalities tend to follow heavy-tailed statistical distributions. A 2005 fusion-fission theory predicts mathematically that for armed groups operating in dynamically evolving clusters within a given conflict, the number of fatalities per conflict event will follow an approximate power-law distribution with exponent near 2.5, with the specific exponent value offering insight into the relative robustness of larger versus smaller clusters of fighters in that armed group. Since Yemen and Syria are current hotspots for future conflict, yet their most recent conflicts (2023-2025) have not been studied at the event level, we use ACLED data to determine their best-fit exponent value as each conflict evolved. We find that the exponent lies between 2.5 and 3.5 predominantly throughout each conflict, which suggests that the fighters in each of these conflicts continued to operate in smaller clusters as the conflict evolved. Moreover, temporary reductions in the exponent value -- which suggests a temporary increase in the robustness and involvement of larger clusters of fighters -- appear to arise during major crises ahead of the largest battles. Though the lack higher-quality data for these conflicts prevents us from establishing this more firmly, such a temporary reduction in the exponent value hints at its potential use as an early-warning signature.

We develop a general framework for constructing $2n$-site chiral integrable matrix product states in Aharony-Bergman-Jafferis-Maldacena spin chain, based on reflection equations and the fusion procedure. For four-site chiral integrable product states, we propose their exact overlap formulas with Bethe states. We also investigate the chiral integrable subspaces numerically.

The theory of homologies introduces cell complexes to provide an algebraic description of spaces up to topological equivalence. Attractors in state space can be studied using Branched Manifold Analysis through Homologies: this strategy constructs a cell complex from a cloud of points in state space and uses homology groups to characterize its topology. The approach, however, does not consider the action of the flow on the cell complex. The procedure is here extended to take this fundamental property into account, as done with templates. The goal is achieved endowing the cell complex with a directed graph that prescribes the flow direction between its highest dimensional cells. The tandem of cell complex and directed graph, baptized templex, is shown to allow for a sophisticated characterization of chaotic attractors and for an accurate classification of them. The cases of a few well-known chaotic attractors are investigated -- namely the spiral and funnel Rössler attractors, the Lorenz attractor and the Burke and Shaw attractor. A link is established with their description in terms of templates.

Discriminating different types of chaos is still a very challenging topic, even for dissipative three-dimensional systems for which the most advanced tool is the template. Nevertheless, getting a template is, by definition, limited to three-dimensional objects, since based on knot theory. To deal with higher-dimensional chaos, we recently introduced the templex combining a flow-oriented {\sc BraMAH} cell complex and a directed graph (a digraph). There is no dimensional limitation in the concept of templex. Here, we show that a templex can be automatically reduced into a ``minimal'' form to provide a comprehensive and synthetic view of the main properties of chaotic attractors. This reduction allows for the development of a taxonomy of chaos in terms of two elementary units: the oscillating unit (O-unit) and the switching unit (S-unit). We apply this approach to various well-known attractors (Rössler, Lorenz, and Burke-Shaw) as well as a non-trivial four-dimensional attractor. A case of toroidal chaos (Deng) is also treated. This work is dedicated to Otto E. Rössler.

This study investigates a detailed analytical and numerical investigation of a nonlinear two-degree-of-freedom (2DOF) mechanical oscillator subjected to parametric excitation, magnetic stiffness nonlinearities, and dry friction. The considered system consists of two coupled oscillators, both of which are connected to a rotating rectangular beam that induces a time-periodic stiffness variation. The Complex Averaging (CxA) method is employed to derive approximate analytical solutions, which are thoroughly validated through time-domain simulations and bifurcation analyses. The dynamic analysis reveals a rich spectrum of nonlinear behaviors, including periodic, quasi-periodic, and chaotic responses. Detailed bifurcation diagrams, Lyapunov exponent analysis, and Poincaré maps demonstrate the influence of nonlinear stiffness degree, mass symmetry, and frictional effects on system stability and response amplitude. The obtained results give a significant understanding of the dynamic behavior of coupled nonlinear systems and establish a conceptual framework for the development of complex vibration abatement strategies, energy harvesting devices, and advanced mechanical systems.

We show that in a drive-response coupling framework extreme events are suppressed in the response system by the dominance of a single driving signal. We validate this approach across three distinct response network topologies, namely (i) a pair of coupled neurons, (ii) a monolayer network of N coupled neurons and (iii) a two-layer multiplex network each composed of FitzHugh-Nagumo neuronal units. The response networks inherently exhibit extreme events. Our results demonstrate that influencing just one neuron in the response network with an appropriately tuned driving signal is sufficient to control extreme events across all three configurations. In the two-neuron case, suppression of extreme events occurs due to the breaking of phase-locking between the driving neuron and the targeted response neuron. In the case of monolayer and multiplex networks, suppression of extreme events results from the disruption of protoevent frequency dynamics and a subsequent frequency decoupling of the driven neuron from the rest of the network. We also observe that when the size of the neurons in response network connected to the drive increases, the onset of control occurs earlier indicating a scaling advantage of the method.

The conditions necessary and sufficient for the Smoothed Dissipative Particle Dynamics (SDPD) equations of motion to have a Lagrangian that can be used for deriving these equations of motion, the Helmholtz conditions, are obtained and analysed. They show that for a finite number of SDPD particles the conditions are not satisfied; hence, the SDPD equations of motion can not be obtained using the classical Euler-Lagrange equation approach. However, when the macroscopic limit is considered, that is when the number of particles tends to infinity, the conditions are satisfied, thus providing the conceptual possibility of obtaining the Navier-Stokes equations from the principle of least action.

We introduce the Neural Hodge Corrective Solver (NHCS), a hybrid iterative-neural framework for partial differential equations that embeds learned corrective operators within the Discrete Exterior Calculus (DEC) formulation. The method combines classical Jacobi-Richardson iterations with data-driven corrections to refine numerical solutions while preserving the underlying topological and metric structure. NHCS employs a two-phase training strategy. In the first phase, DEC operators are learned through relative residual minimization from data. In the second phase, these operators are integrated into the iterative solver, and training targets the improvement of convergence through learned corrective updates that remain effective even for inaccurate intermediate solutions. This staggered training enables stable, progressive refinement while maintaining the structure-preserving properties of DEC discretizations. To improve multiscale adaptivity, NHCS introduces a convolutional neural network-based correction term capable of capturing fine-scale solution features via localized updates informed by global context, improving scalability over mesh component-wise neural approaches. Moreover, the proposed framework substantially reduces computational cost by avoiding Newton-Raphson-based training and the associated Jacobian evaluations of parameterized operators. The resulting solver achieves improved efficiency, robustness, and accuracy without compromising numerical stability.

In Calgary, November 2005, Chad Hedrick was the first to skate the 5,000 m below 6:10. His world record time 6:09.68 was then beaten a week later, in Salt Lake City, by Sven Kramer's 6:08.78. Further top races and world records followed over the ensuing seasons; up to and including the 2024-2025 season, a total of 126 races have been below 6:10, with Nils van der Poel's 2021 world record being 6:01.56. The appropriately hyped-up canonical question for the friends and followers and aficionados of speedskating has then been when (and by whom we for the first time would witness a below 6:00.00 race. In this note I first use extreme value statistics modelling to assess the state of affairs, as per the end of the 2024-2025 season, with predictions and probabilities for the 2025-2026 season. Under natural modelling assumptions the probability of seeing a new world record during this new season is shown to be about ten percent. We were indeed excited but in reality merely modestly surprised that a race better than van der Poel's record was clocked, by Timothy Loubineaud, in Salt Lake City, November 14, 2025. But Six-Minute Man Sander Eitrem's outstanding 5:58.52 in Inzell, on January 24, 2026, is truly beamonesquely shocking. I also use the modelling machinery to analyse the post-Eitrem situation, and suggest answers to the question of how fast the 5,000 m ever can be skated.

Community detection is a critical tool for understanding the mesoscopic structure of large-scale networks. However, when applied to aggregated or coarse-grained social networks, disjoint community partitions cannot capture the diverse composition of community memberships within aggregated nodes. While existing mixed membership methods alleviate this issue, they may detected communities that are highly sensitive to the aggregation resolution, not reliably reflecting the underlying community structure of the underlying individual-level network. This paper presents the Link Fraction Mixed Membership (LFMM) method, which computes the mixed memberships of nodes in aggregated networks. Unlike existing mixed membership methods, LFMM is consistent under aggregation. Specifically, we show that it conserves community membership sums at different scales. The method is utilized to study a population-scale social network of the Netherlands, aggregated at different resolutions. Experiments reveal variation in community membership across different geographical regions and evolution over the last decade. In particular, we show how our method identifies large urban hubs that act as the melting pots of diverse, spatially remote communities.

We present an adaptive Chebyshev-based Boundary Integral Equation (CBIE) solver for electromagnetic scattering from smooth perfect electric conductor (PEC) objects. The proposed approach eliminates manual parameter tuning by introducing (i) a unified adaptive quadrature strategy for automatic selection of the near-singular interaction distance and (ii) an adaptive computation of all self- and near-singular precomputation integrals to a prescribed accuracy using Gauss-Kronrod (h-adaptive) or Clenshaw-Curtis (p-adaptive) rules and singularity-resolving changes of variables. Both h-adaptive and p-adaptive schemes are explored within this framework, ensuring high-order accuracy and robustness across a broad range of geometries without loss of efficiency. Numerical results for canonical and complex CAD geometries demonstrate that the adaptive solver achieves accuracy and convergence rates comparable to optimally tuned fixed-grid CBIE implementations, while offering automation and scalability to electrically large, geometrically complex problems.

Inverse design has enabled the systematic design of ultra-compact and high-performance nanophotonic components. Yet enforcing foundry design rules during inverse design remains a major challenge, as optimized devices frequently violate constraints on minimum feature size and spacing. Existing fabrication-constrained approaches typically rely on penalty terms, projection filters, or heuristic binarization schedules, which restrict the accessible design space, require extensive hyperparameter tuning, and often fail to guarantee compliance throughout the optimization trajectory. Here, we introduce a framework for nanophotonic inverse design with intrinsic enforcement of design rules through a generative reparameterization of the design space, restricting optimization to a learned manifold of DRC-compliant geometries. We validate this paradigm by designing representative silicon photonic components including broadband power splitters, spectral duplexers, and mode converters operating across the 1,500-1,600 nm band for both electron-beam lithography and photolithography platforms. Across all devices, the manifold-based formulation reaches state-of-the-art performance metrics with over a 5-fold reduction in computational cost compared to pixel-based representations, while ensuring fabrication-compatible geometries throughout the entire design process. By treating fabrication constraints as a fundamental property of the design representation rather than an external penalty, this work establishes a direct pathway toward broadly applicable, platform-agnostic, and intrinsically DRC-compliant nanophotonics.

Environmental stimuli play a pivotal role in triggering impulsive purchases among consumers,while consumers from Sichuan Province, China, exhibit strong ethnocentric tendencies, impacting their decision-making process, particularly regarding Wuliangye liquor, a local product. Through an online survey of 453 Wuliangye consumers from Sichuan, an analysis was conducted using structural equation modeling rooted in the ETPB and SOR theory. This analysis revealed the favorable impact of environmental stimuli and consumer ethnocentrism on purchasing behavior. This influence was found to be partially mediated through perceived value, attitudes, and purchase intention, forming a chain-mediated effect. Notably, purchase intention doesn't always translate to actual buying behavior, with environmental stimuli, consumer ethnocentrism, perceived behavioral control and purchase intention all being robust predictors of purchase behavior. Finally, several management strategies were proposed, aimed at bolstering Wuliangye sales, with a focus on platform development, mid-to-low range product creation, and appealing to Generation Z consumers.

Networks of aromatic amino acid residues within microtubules, particularly those formed by tryptophan, may serve as pathways for optical information flow. Ultraviolet excitation dynamics in these networks are typically modeled with effective non-Hermitian Hamiltonians. By extending this approach to a Lindblad master equation that incorporates explicit site geometries and dipole orientations, we track how correlations are generated, routed, and dissipated, while capturing both energy dissipation and information propagation among coupled chromophores. We compare localized injections, fully delocalized preparations, and eigenmode-based initial states. To quantify the emerging quantum-informational structure, we evaluate the $L_1$ norm of coherence, the correlated coherence, and the logarithmic negativity within and between selected chromophore sub-networks. The results reveal a strong dependence of both the direction and persistence of information flow on the type of initial preparation. Superradiant components drive the rapid export of correlations to the environment, whereas subradiant components retain them and slow their leakage. Embedding single tubulin units into larger dimers and spirals reshapes pairwise correlation maps and enables site-selective routing. Scaling to larger ordered lattices strengthens both export and retention channels, whereas static energetic and structural disorder suppresses long-range transport and reduces overall correlation transfer. These findings provide a Lindbladian picture of information flow in cytoskeletal chromophore networks and identify structural and dynamical conditions that transiently preserve nonclassical correlations in microtubules.

We aim to develop physics foundation models for science and engineering that provide real-time solutions to Partial Differential Equations (PDEs) which preserve structure and accuracy under adaptation to unseen geometries. To this end, we introduce General-Geometry Neural Whitney Forms (Geo-NeW): a data-driven finite element method. We jointly learn a differential operator and compatible reduced finite element spaces defined on the underlying geometry. The resulting model is solved to generate predictions, while exactly preserving physical conservation laws through Finite Element Exterior Calculus. Geometry enters the model as a discretized mesh both through a transformer-based encoding and as the basis for the learned finite element spaces. This explicitly connects the underlying geometry and imposed boundary conditions to the solution, providing a powerful inductive bias for learning neural PDEs, which we demonstrate improves generalization to unseen domains. We provide a novel parameterization of the constitutive model ensuring the existence and uniqueness of the solution. Our approach demonstrates state-of-the-art performance on several steady-state PDE benchmarks, and provides a significant improvement over conventional baselines on out-of-distribution geometries.

Backpropagation, the foundational algorithm for training neural networks, is typically understood as a symbolic computation that recursively applies the chain rule. We show it emerges exactly as the finite-time relaxation of a physical dynamical system. By formulating feedforward inference as a continuous-time process and applying Lagrangian theory of non-conservative systems to handle asymmetric interactions, we derive a global energy functional on a doubled state space encoding both activations and sensitivities. The saddle-point dynamics of this energy perform inference and credit assignment simultaneously through local interactions. We term this framework ''Dyadic Backpropagation''. Crucially, we prove that unit-step Euler discretization, the natural timescale of layer transitions, recovers standard backpropagation exactly in precisely 2L steps for an L-layer network, with no approximations. Unlike prior energy-based methods requiring symmetric weights, asymptotic convergence, or vanishing perturbations, our framework guarantees exact gradients in finite time. This establishes backpropagation as the digitally optimized shadow of a continuous physical relaxation, providing a rigorous foundation for exact gradient computation in analog and neuromorphic substrates where continuous dynamics are native.

Sampling a quantum systems underlying probability distributions is an important computational task, e.g., for quantum advantage experiments and quantum Monte Carlo algorithms. Tensor networks are an invaluable tool for efficiently representing states of large quantum systems with limited entanglement. Algorithms for sampling one-dimensional (1D) tensor networks are well-established and utilized in several 1D tensor network methods. In this paper we introduce two novel sampling algorithms for two-dimensional (2D) isometric tensor network states (isoTNS) that can be viewed as extensions of algorithms for 1D tensor networks. The first algorithm we propose performs independent sampling and yields a single configuration together with its associated probability. The second algorithm employs a greedy search strategy to identify K high-probability configurations and their corresponding probabilities. Numerical results demonstrate the effectiveness of these algorithms across quantum states with varying entanglement and system size.

State-of-the-art AI deep potentials provide ab initio-quality results, but at a fraction of the computational cost of first-principles quantum mechanical calculations, such as density functional theory. In this work, we bring AI deep potentials into GROMACS, a production-level Molecular Dynamics (MD) code, by integrating with DeePMD-kit that provides domain-specific deep learning (DL) models of interatomic potential energy and force fields. In particular, we enable AI deep potentials inference across multiple DP model families and DL backends by coupling GROMACS Neural Network Potentials with the C++/CUDA backend in DeePMD-kit. We evaluate two recent large-atom-model architectures, DPA2 that is based on the attention mechanism and DPA3 that is based on GNN, in GROMACS using four ab initio-quality protein-in-water benchmarks (1YRF, 1UBQ, 3LZM, 2PTC) on NVIDIA A100 and GH200 GPUs. Our results show that DPA2 delivers up to 4.23x and 3.18x higher throughput than DPA3 on A100 and GH200 GPUs, respectively. We also provide a characterization study to further contrast DPA2 and DPA3 in throughput, memory usage, and kernel-level execution on GPUs. Our findings identify kernel-launch overhead and domain-decomposed inference as the main optimization priorities for AI deep potentials in production MD simulations.

Molecular dynamics (MD) simulations remain the gold standard for studying protein dynamics, but their computational cost limits access to biologically relevant timescales. Recent generative models have shown promise in accelerating simulations, yet they struggle with long-horizon generation due to architectural constraints, error accumulation, and inadequate modeling of spatio-temporal dynamics. We present STAR-MD (Spatio-Temporal Autoregressive Rollout for Molecular Dynamics), a scalable SE(3)-equivariant diffusion model that generates physically plausible protein trajectories over microsecond timescales. Our key innovation is a causal diffusion transformer with joint spatio-temporal attention that efficiently captures complex space-time dependencies while avoiding the memory bottlenecks of existing methods. On the standard ATLAS benchmark, STAR-MD achieves state-of-the-art performance across all metrics--substantially improving conformational coverage, structural validity, and dynamic fidelity compared to previous methods. STAR-MD successfully extrapolates to generate stable microsecond-scale trajectories where baseline methods fail catastrophically, maintaining high structural quality throughout the extended rollout. Our comprehensive evaluation reveals severe limitations in current models for long-horizon generation, while demonstrating that STAR-MD's joint spatio-temporal modeling enables robust dynamics simulation at biologically relevant timescales, paving the way for accelerated exploration of protein function.

Autoregressive learning of time-stepping operators offers an effective approach to data-driven PDE simulation on grids. For conservation laws, however, long-horizon rollouts are often destabilized when learned updates violate global conservation and, in many applications, additional state bounds such as nonnegative mass and densities or concentrations constrained to [0,1]. Enforcing these coupled constraints via direct next-state regression remains difficult. We introduce a framework for learning conservative transport operators on regular grids, inspired by lattice Boltzmann-style discrete-velocity transport representations. Instead of predicting the next state, the model outputs local transport operators that update cells through neighborhood exchanges, guaranteeing discrete conservation by construction. For bounded quantities, we parameterize transport within a capacity-constrained feasible set, enforcing bounds structurally rather than by post-hoc clipping. We validate FluxNet on 1D convection-diffusion, 2D shallow water equations, 1D traffic flow, and 2D spinodal decomposition. Experiments on shallow-water equations and traffic flow show improved rollout stability and physical consistency over strong baselines. On phase-field spinodal decomposition, the method enables large time-steps with long-range transport, accelerating simulation while preserving microstructure evolution in both pointwise and statistical measures.

In this short note, we present a flux-correction form of the third-order edge-based scheme for the Euler equations that enables the direct use of a general flux function. The core idea is to replace, without loss of accuracy, the arithmetic average of the flux extrapolations by a general numerical flux evaluated at the edge midpoint, together with a correction term. We show that the proposed flux-correction form preserves third-order accuracy, provided that the general numerical flux is evaluated with the left and right states that are computed exactly for a quadratic function, which can be achieved effectively by the U-MUSCL scheme with κ = 1/2. Numerical results are presented to verify third-order accuracy with the HLLC and LDFSS flux functions on irregular tetrahedral grids.

We study noisy majority-rule dynamics on annealed hypergraphs to clarify how variability in group interaction sizes reshapes collective ordering. At each update, a group is sampled from a prescribed size distribution and either follows the strict within-group majority or, with probability $q$, updates independently under an external bias $p$. At the symmetric point $p=1/2$, we obtain an explicit analytical expression for the critical independence threshold $q_c$, which separates macroscopic ordering from a fluctuating mixed state and can be interpreted as the largest fraction of independent behavior that can be sustained without destroying order. Because $q_c$ is governed by group-size statistics through an effective majority leverage, broad and heavy-tailed size distributions enhance robustness by enabling rare large-group events to realign a substantial fraction of the population. We further derive analytical predictions, benchmarked against Monte Carlo simulations, for the leading finite-size behavior of relaxation: for narrow distributions the characteristic relaxation time typically grows logarithmically with system size, whereas sufficiently heavy-tailed power laws produce strong crossovers and make the large-system dynamics sensitive to how $q$ approaches the transition. In the pure majority-rule limit, we find a crossover from conventional logarithmic consensus times to rapid ordering driven by occasional macroscopic groups, and the exit probability near coexistence collapses onto a universal error-function form controlled by a single structural parameter.

We present an efficient numerical method, inspired by transformation optics, for solving the Poisson equation in complex and arbitrarily shaped geometries. The approach operates by mapping the physical domain to a uniform computational domain through coordinate transformations, which can be applied either to the entire domain or selectively to specific boundaries inside the domain. This flexibility allows both homogeneous (Laplace equation) and inhomogeneous (Poisson equation) problems to be solved efficiently using iterative or fast direct solvers, with only the material parameters and source terms modified according to the transformation. The method is formulated within a finite difference framework, where the modified material properties are computed from the coordinate transformation equations, either analytically or numerically. This enables accurate treatment of arbitrary geometric shapes while retaining the simplicity of a uniform grid solver. Numerical experiments confirm that the method achieves second-order accuracy , and offers a straightforward pathway to integrate fast solvers such as multigrid methods on the uniform computational grid.

Vehicle charging infrastructure targets in Europe currently rely on uniform benchmarks and overlook the flexibility that could be provided by future smart charging (V1G) and vehicle to grid operation (V2G). To address this gap, we explicitly represent charging infrastructure and its costs in a cost minimizing European energy system model, allowing uncontrolled charging, V1G, and V2G to compete. We find that V1G captures the majority of system cost savings, amounting to 19 to 42 billion euros per year, or 2.2 to 4.5 percent, and substantially reduces infrastructure requirements. V2G provides more limited system cost savings of up to 2.5 billion euros per year, but generates substantial balancing market revenues of around 6.4 billion euros per year. V2G deployment is most cost effective in photovoltaic dominated systems and in scenarios with limited grid expansion, where combined solar and wind generation is relatively scarce. Charging infrastructure requirements vary across countries, reflecting either utilization maximization or flexibility maximization. This indicates that uniform EU targets risk overestimating infrastructure needs in some regions while constraining the benefits of smart charging in others.

Transmission line (TL) models implemented in the time domain can efficiently simulate basilar-membrane (BM) displacement in response to transient or non-stationary sounds. By design, a TL model is well-suited for an one-dimensional (1-D) characterization of the traveling wave, but the real configuration of the cochlea also introduces higher-dimensional effects. Such effects include the focusing of the pressure around the BM and transverse viscous damping, both of which are magnified in the short-wave region. The two effects depend on the wavelength and are more readily expressed in the frequency domain. In this paper, we introduce a numerical correction for the BM admittance to account for 2-D effects in the time domain using autoregressive filtering and regression techniques. The correction was required for the implementation of a TL model tailored to the gerbil cochlear physiology. The model, which includes instantaneous nonlinearities in the form of variable damping, initially presented insufficient compression with increasing sound levels. This limitation was explained by the strong coupling between gain and frequency selectivity assumed in the 1-D nonlinear TL model, whereas cochlear frequency selectivity shows only a moderate dependence on sound level in small mammals. The correction factor was implemented in the gerbil model and made level-dependent using a feedback loop. The updated model achieved some decoupling between frequency selectivity and gain, providing 5 dB of additional gain and extending the range of sound levels of the compressive regime by 10 dB. We discuss the relevance of this work through two key features: the integration of both analytical and regression methods for characterizing BM admittance, and the combination of instantaneous and non-instantaneous nonlinearities.

Cellular metabolism is globally regulated by various currency metabolites such as ATP, GTP, and NAD(P)H. These metabolites cycle between charged (high-energy) and uncharged (low-energy) states to mediate energy transfer. While distinct currency metabolites are associated with different metabolic functions, their charged and uncharged forms are generally interchangeable via biochemical reactions such as ${\rm ATP{\,+\,}GDP{\,\rightleftharpoons\,}ADP{\,+\,}GTP}$ and $\rm NADP^+{\,+\,}NADH{\,\rightleftharpoons\,}NADPH{\,+\,}NAD^+ $. Thus, their energetic states are generally coupled and influence each other, which would hinder the independent regulation of different currency metabolites. Despite the extensive knowledge of the molecular biology of individual currency metabolites, it remains poorly understood how the coordination of various coupled currency metabolites shapes metabolic regulation, efficiency, and ultimately the evolution of organisms. Here, we present a minimal theoretical model of metabolic currency coupling and reveal a fundamental tradeoff relationship between metabolic controllability and thermodynamic cost: increasing the capacity to independently regulate multiple currency metabolites generally requires comparable abundances of those metabolites, which in turn incurs a higher entropy production rate. The tradeoff suggests that in complex environments, organisms evolutionarily favor an equal abundance of currency metabolites to enhance metabolic controllability at the expense of a higher thermodynamic cost; conversely, in simple environments, organisms evolve to have imbalanced amounts of them to reduce heat dissipation. These considerations also offer a hypothesis regarding evolutionary trends in nucleotide-pool balance and genomic GC content.

Degenerate Cahn-Hilliard phase-field models provide a robust approximation of surface-diffusion-driven interface motion without explicit front tracking. In computations, however, the geometric volume enclosed by the interface -- the region where the order parameter $φ$ is positive -- may drift at finite interface thickness, producing artificial shrinkage or growth even when the sharp-interface limit conserves volume. We revisit and extend the improved-conservation framework of Zhou et al., where one replaces classical mass conservation by the exact conservation of a designed monotone mapping $Q(φ)$ that more accurately approximates a step function. Building on this framework, we (i) carry out the matched-asymptotic analysis in the unscaled physical time formulation, (ii) derive a simplified representation of the first-order inner correction to the interface profile, and (iii) identify an integral-moment cancellation condition that controls the leading geometric-volume defect. This mechanism becomes a practical design rule: we select regularization kernels within parameterized families -- including exponential and Pade-type -- to reach effective higher-order behavior and satisfy the cancellation condition at moderate parameter values. As a result, the proposed kernels achieve formal third-order accuracy in geometric-volume conservation with respect to interface thickness. Finally, we describe an unconditional energy-dissipative numerical discretization that exactly preserves the discrete conserved quantity. Numerical benchmarks on multi-scale droplet coarsening and shape relaxation demonstrate that the moment-balanced kernels virtually eliminate artificial drift and prevent premature extinction of small droplets.

Self-accelerating Airy matter waves offer a clean setting to access the cubic Kennard phase. Here we reconstruct the relative phase of simulated Airy-shaped Bose-Einstein condensates in free space, a regime approached in microgravity, from interference fringes. The cubic phase dynamics are quantified via windowed polynomial fits with systematics-aware uncertainty estimates that account for window-induced correlations. We compare two experimentally feasible phase-extraction methods - heterodyne-based and density-based - and show that an Airy-Gaussian geometry yields substantially improved robustness to fit-window selection relative to an Airy-Airy collision. In the weakly interacting regime, the extracted cubic coefficient responds linearly to the effective one-dimensional interaction strength. Our approach turns cubic phase dynamics into a practical probe of weak mean-field nonlinearities in self-accelerating condensates.

Optimisation plays a central role in a wide range of scientific and industrial applications, and quantum computing has been widely proposed as a means to achieve computational advantages in this domain. To date, research into the design of noise-resilient quantum algorithms has been dominated by variational approaches, while alternatives remain relatively unexplored. In this work, we introduce a novel class of quantum optimisation heuristics that forgo this variational framework in favour of a hybrid quantum-classical approach built upon Markov Chain Monte Carlo (MCMC) techniques. We introduce Quantum-enhanced Simulated Annealing (QeSA) and Quantum-enhanced Parallel Tempering (QePT), before validating these heuristics on hard Sherrington-Kirkpatrick instances and demonstrate their superior scaling over classical benchmarks. These algorithms are expected to exhibit inherent robustness to noise and support parallel execution across both quantum and classical resources with only classical communication required. As such, they offer a scalable and potentially competitive route toward solving large-scale optimisation problems with near-term quantum devices.

The rapid evolution of Large Language Models (LLMs) has led to the emergence of Multi-Agent Systems where collective cooperation is often threatened by the "Tragedy of the Commons." This study investigates the effectiveness of Anchoring Agents--pre-programmed altruistic entities--in fostering cooperation within a Public Goods Game (PGG). Using a full factorial design across three state-of-the-art LLMs, we analyzed both behavioral outcomes and internal reasoning chains. While Anchoring Agents successfully boosted local cooperation rates, cognitive decomposition and transfer tests revealed that this effect was driven by strategic compliance and cognitive offloading rather than genuine norm internalization. Notably, most agents reverted to self-interest in new environments, and advanced models like GPT-4.1 exhibited a "Chameleon Effect," masking strategic defection under public scrutiny. These findings highlight a critical gap between behavioral modification and authentic value alignment in artificial societies.

A semi-implicit Lax-Wendroff scheme is developed for electron-phonon coupling process in metals based on the two-temperature kinetic equations. The core of this method is to integrate the evolution information of physical equations into the numerical modeling process, which leads to that the time step or cell size is not limited by the relaxation time and mean free path. Specifically, the finite difference method is used to solve the kinetic model again when reconstructing the interfacial distribution function, through which the particle migration, scattering and electron-phonon coupling processes are coupled together within a single time step. Numerical tests demonstrate that this method could efficiently capture electron-phonon coupling or heat conduction processes from the ballistic to diffusive regimes. It provides a new tool for describing electron-phonon coupling or thermal management in microelectronic devices.

Physics-informed neural networks (PINNs) have emerged as a powerful paradigm for solving partial differential equations (PDEs) by embedding physical laws directly into neural network training. However, solving high-fidelity PDEs remains computationally prohibitive, particularly for parametric systems requiring multiple evaluations across varying parameter configurations. This paper presents MF-BPINN, a novel multi-fidelity framework that synergistically combines physics-informed neural networks with Bayesian uncertainty quantification and adaptive residual learning. Our approach leverages abundant low-fidelity simulations alongside sparse high-fidelity data through a hierarchical neural architecture that learns nonlinear correlations across fidelity levels. We introduce an adaptive residual network with learnable gating mechanisms that dynamically balances linear and nonlinear fidelity discrepancies. Furthermore, we develop a rigorous Bayesian framework employing Hamiltonian Monte Carlo.

Benoit Mandelbrot's scientific legacy spans an extraordinary range of disciplines, from linguistics and fluid turbulence to cosmology and finance, suggesting the intellectual temperament of a "fox" in Isaiah Berlin's famous dichotomy of thinkers. This essay argues, however, that Mandelbrot was, at heart, a "hedgehog": a thinker unified by a single guiding principle. Across his diverse pursuits, the concept of scaling -- manifested in self-similarity, power laws, fractals, and multifractals -- served as the central idea that structured his work. By tracing the continuity of this scaling paradigm through his contributions to mathematics, physics, and economics, the paper reveals a coherent intellectual trajectory masked by apparent eclecticism. Mandelbrot's enduring insight in the modeling of natural and social phenomena can be understood through the lens of the geometry and statistics of scale invariance.

We investigate the role of feature superposition in the emergence of power-law training dynamics using a teacher-student framework. We first derive an analytic theory for training without superposition, establishing that the power-law training exponent depends on both the input data statistics and channel importance. Remarkably, we discover that a superposition bottleneck induces a transition to a universal power-law exponent of $\sim 1$, independent of data and channel statistics. This one over time training with superposition represents an up to tenfold acceleration compared to the purely sequential learning that takes place in the absence of superposition. Our finding that superposition leads to rapid training with a data-independent power law exponent may have important implications for a wide range of neural networks that employ superposition, including production-scale large language models.

As quantum computing transitions from theoretical physics to engineering applications, there is a growing need for accessible simulation tools that bridge the gap between abstract linear algebra and practical implementation. While text-based frameworks (like Qiskit or Cirq) are standard, they often present a steep learning curve for students and engineers accustomed to graphical system design. This paper introduces QuVI (Quantum Virtual Instrument), an open-source quantum circuit toolkit developed natively within the NI LabVIEW environment. Moving beyond initial proof-of-concept models, QuVI establishes a robust framework that leverages LabVIEW's "dataflow" paradigm, in which wires represent data and nodes represent operations, to provide an intuitive, visual analog to standard quantum circuit notation while enabling the seamless integration of classical control structures like loops and conditionals. The toolkit's capabilities are demonstrated by constructing and visualizing fundamental quantum algorithms and verifying results against theoretical predictions. By translating "Block Diagrams" directly into quantum state evolutions ("Bloch Spheres"), QuVI offers educators and researchers a powerful platform for prototyping quantum logic without leaving the graphical engineering workspace.

The double empathy problem recasts the difficulty of forming empathy bonds in social interactions between autistic and neurotypical individuals as a bidirectional problem, rather than due to a deficit exclusive to the person on the spectrum. However, no explicit mechanism to explain such a phenomenon has been proposed. Here we build a feedback-loop mathematical model that would theoretically induce the empathy degradation observed during communication in neurotypical-autistic pairs solely due to differences in communication preferences between neurotypical and neurodivergent individuals. Numerical simulations of dyadic interactions show the model, whose mechanism is based solely on communication preferences, can illustrate the breakdown of empathic bonding observed clinically. Stability analysis of the model provides a way to predict the overall trajectory of the interaction in the empathy space. Furthermore, we suggest experimental designs to measure several parameters outlined here and discuss the future directions for testing the proposed model.

Indirect reciprocity maintains cooperation in stranger societies by mapping individual behaviors onto reputation signals via social norms. Existing theoretical frameworks assume static environments with constant resources and fixed payoff structures. However, in real-world systems, individuals' strategic behaviors not only shape their reputation but also induce collective-level resource changes in ecological, economic, or other external environments, which in turn reshape the incentives governing future individual actions. To overcome this limitation, we establish a co-evolutionary framework that couples moral assessment, strategy updating, and environmental dynamics, allowing the payoff structure to dynamically adjust in response to the ecological consequences of collective actions. We find that this environmental feedback mechanism helps lower the threshold for the emergence of cooperation, enabling the system to spontaneously transition from a low-cooperation state to a stable high-cooperation regime, thereby reducing the dependence on specific initial conditions. Furthermore, while lenient norms demonstrate adaptability in static environments, norms with strict discrimination are shown to be crucial for curbing opportunism and maintaining evolutionary resilience in dynamic settings. Our results reveal the evolutionary dynamics of coupled systems involving reputation institutions and environmental constraints, offering a new theoretical perspective for understanding collective cooperation and social governance in complex environments.

Inferring the activities of transcription factors from high-throughput transcriptomic or open chromatin profiling, such as RNA-/CAGE-/ATAC-Seq, is a long-standing challenge in systems biology. Identification of highly active master regulators enables mechanistic interpretation of differential gene expression, chromatin state changes, or perturbation responses across conditions, cell types, and diseases. Here, we describe MARADONER, a statistical framework and its software implementation for motif activity response analysis (MARA), utilizing the sequence-level features obtained with pattern matching (motif scanning) of individual promoters and promoter- or gene-level activity or expression estimates. Compared to the classic MARA, MARADONER (MARA-done-right) employs an unbiased variance parameter estimation and a bias-adjusted likelihood estimation of fixed effects, thereby enhancing goodness-of-fit and the accuracy of activity estimation. Further, MARADONER is capable of accounting for heteroscedasticity of motif scores and activity estimates.

Automated analysis of electroencephalography (EEG) has recently undergone a paradigm shift. The introduction of transformer architectures and self-supervised pretraining (SSL) has led to the development of EEG foundation models. These models are pretrained on large amounts of unlabeled data and can be adapted to a range of downstream tasks. This systematic review summarizes recent SSL-trained EEG foundation models that learn whole-brain representations from multichannel EEG rather than representations derived from a single channel. We searched PubMed, IEEE Xplore, Scopus, and arXiv through July 21, 2025. Nineteen preprints and peer-reviewed articles met inclusion criteria. We extracted information regarding pretraining datasets, model architectures, pretraining SSL objectives, and downstream task applications. While pretraining data heavily relied on the Temple University EEG corpus, there was significant heterogeneity in model architecture and training objectives across studies. Transformer architectures were identified as the predominant pretraining architecture with state-space models such as MAMBA and S4 as emerging alternatives. Concerning SSL objectives, masked auto-encoding was most common, and other studies incorporate contrastive learning. Downstream tasks varied widely and implemented diverse fine-tuning strategies, which made direct comparison challenging. Furthermore, most studies used single-task fine-tuning, and a generalizable EEG foundation model remains lacking. In conclusion, the field is advancing rapidly but still limited by limited dataset diversity and the absence of standardized benchmarks. Progress will likely depend on larger and more diverse pretraining datasets, standardized evaluation protocols, and multi-task validation. The development will advance EEG foundation models towards robust and general-purpose relevant to both basic and clinical applications.

Cognition is increasingly framed in terms of information processing, yet most fMRI analyses focus on activation or functional connectivity rather than quantifying how information is stored and transferred. To remedy this problem, we propose a framework for estimating measures of information processing: active information storage (AIS), transfer entropy (TE), and net synergy from task-based fMRI. AIS measures information maintained within a region, TE captures directed information flow, and net synergy contrasts higher-order synergistic to redundant interactions. Crucially, to enable this framework we utilised a recently developed approach for calculating information-theoretic measures: the cross mutual information. This approach combines resting-state and task data to address the challenges of limited sample size, non-stationarity and context in task-based fMRI. We applied this framework to the working memory (N-back) task from the Human Connectome Project (470 participants). Results show that AIS increases in fronto-parietal regions with working memory load, TE reveals enhanced directed information flows across control pathways, and net synergy indicates a global shift to redundancy. This work establishes a novel methodology for quantifying information processing in task-based fMRI.

Multi-state structured population models, including integral projection models (IPMs) and age-structured McKendrick equations, link individual life histories to population growth and composition, yet the demographic meaning of their dominant eigenstructure can be difficult to interpret. A main goal of this paper is to derive interpretable demographic indicators for multi-state heterogeneity -- in particular expected generation numbers, which act as an effective genealogical memory length (in generations) of the ancestry-weighted contributions driving growth -- together with type reproduction numbers and generation intervals, directly from life-history transition kernels. To this end we develop a determinant-free genealogical framework based on a reference-point operator, a rank-one construction at the kernel level that singles out a biologically chosen reference state and organizes lineages by their contributions relative to that state. This yields stable distributions and reproductive values as convergent series of iterated kernels, and leads to an Euler--Lotka-like characteristic equation expressed by reference-point moments. The resulting expansion admits a closed combinatorial form via ordinary partial Bell polynomials, providing a direct bridge from transition kernels to genealogical quantities. We extend the approach to multi-state McKendrick equations and show how these indicators quantify how population scale and composition are determined by ancestry-weighted initial-state information. The framework avoids restrictive Hilbert--Schmidt assumptions and clarifies how temporal memory and multi-type heterogeneity emerge from cross-generational accumulation, yielding a unified and interpretable route from transition kernels to multi-state demographic indicators.

We introduce a reproducible, bias-resistant machine learning framework that integrates domain-informed feature engineering, nested cross-validation, and calibrated decision-threshold optimization for small-sample neuroimaging data. Conventional cross-validation frameworks that reuse the same folds for both model selection and performance estimation yield optimistically biased results, limiting reproducibility and generalization. Demonstrated on a high-dimensional structural MRI dataset of deep brain stimulation cognitive outcomes, the framework achieved a nested-CV balanced accuracy of 0.660\,$\pm$\,0.068 using a compact, interpretable subset selected via importance-guided ranking. By combining interpretability and unbiased evaluation, this work provides a generalizable computational blueprint for reliable machine learning in data-limited biomedical domains.

Detecting spatial patterns is fundamental to scientific discovery, yet current methods lack statistical consensus and face computational barriers when applied to large-scale spatial omics datasets. We unify major approaches through a single quadratic form and derive general consistency conditions. We reveal that several widely used methods, including Moran's I, are inconsistent, and propose scalable corrections. The resulting test enables robust pattern detection across millions of spatial locations and single-cell lineage-tracing datasets.

Clinical brain-to-text interfaces are designed for paralysed patients who cannot provide extensive training recordings. Pre-training improves data-efficient generalisation by learning statistical priors across subjects, but these priors critically depend on context. While natural speech might unfold gradually over minutes, most methods pre-train with only a few seconds of context. Thus, we propose MEG-XL, a model pre-trained with 2.5 minutes of MEG context per sample, 5-300x longer than prior work, and equivalent to 191k tokens, capturing extended neural context. Fine-tuning on the task of word decoding from brain data, MEG-XL matches supervised performance with a fraction of the data (e.g. 1hr vs 50hrs) and outperforms brain foundation models. We find that models pre-trained with longer contexts learn representations that transfer better to word decoding. Our results indicate that long-context pre-training helps exploit extended neural context that other methods unnecessarily discard. Code, model weights, and instructions are available at https://github.com/neural-processing-lab/MEG-XL .

High-resolution spatial transcriptomics platforms, such as Xenium, generate single-cell images that capture both molecular and spatial context, but their extremely high dimensionality poses major challenges for representation learning and clustering. In this study, we analyze data from the Xenium platform, which captures high-resolution images of tumor microarray (TMA) tissues and converts them into cell-by-gene matrices suitable for computational analysis. We benchmark and extend nonnegative matrix factorization (NMF) for spatial transcriptomics by introducing two spatially regularized variants. First, we propose Spatial NMF (SNMF), a lightweight baseline that enforces local spatial smoothness by diffusing each cell's NMF factor vector over its spatial neighborhood. Second, we introduce Hybrid Spatial NMF (hSNMF), which performs spatially regularized NMF followed by Leiden clustering on a hybrid adjacency that integrates spatial proximity (via a contact-radius graph) and transcriptomic similarity through a tunable mixing parameter alpha. Evaluated on a cholangiocarcinoma dataset, SNMF and hSNMF achieve markedly improved spatial compactness (CHAOS < 0.004, Moran's I > 0.96), greater cluster separability (Silhouette > 0.12, DBI < 1.8), and higher biological coherence (CMC and enrichment) compared to other spatial baselines. Availability and implementation: https://github.com/ishtyaqmahmud/hSNMF

Protein fitness optimization is challenged by a vast combinatorial landscape where high-fitness variants are extremely sparse. Many current methods either underperform or require computationally expensive gradient-based sampling. We present CHASE, a framework that repurposes the evolutionary knowledge of pretrained protein language models by compressing their embeddings into a compact latent space. By training a conditional flow-matching model with classifier-free guidance, we enable the direct generation of high-fitness variants without predictor-based guidance during the ODE sampling steps. CHASE achieves state-of-the-art performance on AAV and GFP protein design benchmarks. Finally, we show that bootstrapping with synthetic data can further enhance performance in data-constrained settings.

Both natural and synthetic chemical systems not only exhibit a range of non-trivial dynamics, but also transition between qualitatively different dynamical behaviours as environmental parameters change. Such transitions are called bifurcations. Here, we show that recurrent neural chemical reaction networks (RNCRNs), a class of chemical reaction networks based on recurrent artificial neural networks that can be trained to reproduce a given dynamical behaviour, can also be trained to exhibit bifurcations. First, we show that RNCRNs can inherit some bifurcations defined by smooth ordinary differential equations (ODEs). Second, we demonstrate that the RNCRN can be trained to infer bifurcations that allow it to approximate different target behaviours within different regions of parameter space, without explicitly providing the bifurcation itself in the training. These behaviours can be specified using target ODEs that are discontinuous with respect to the parameters, or even simply by specifying certain desired dynamical features in certain regions of the parameter space. To achieve the latter, we introduce an ODE-free algorithm for training the RNCRN to display designer oscillations, such as a heart-shaped limit cycle or two coexisting limit cycles.

Cryo-electron microscopy (cryo-EM) has revolutionized structural biology by enabling near-atomic-level visualization of biomolecular assemblies. However, the exponential growth in cryo-EM data throughput and complexity, coupled with diverse downstream analytical tasks, necessitates unified computational frameworks that transcend current task-specific deep learning approaches with limited scalability and generalizability. We present CryoLVM, a foundation model that learns rich structural representations from experimental density maps with resolved structures by leveraging the Joint-Embedding Predictive Architecture (JEPA) integrated with SCUNet-based backbone, which can be rapidly adapted to various downstream tasks. We further introduce a novel histogram-based distribution alignment loss that accelerates convergence and enhances fine-tuning performance. We demonstrate CryoLVM's effectiveness across three critical cryo-EM tasks: density map sharpening, density map super-resolution, and missing wedge restoration. Our method consistently outperforms state-of-the-art baselines across multiple density map quality metrics, confirming its potential as a versatile model for a wide spectrum of cryo-EM applications.

Protein language models (PLMs) face a fundamental divide: masked language models (MLMs) excel at fitness prediction while causal models enable generation, forcing practitioners to maintain separate architectures. We introduce \textbf{Proust}, a 309M-parameter causal PLM that bridges this gap through architectural innovations adapted from recent LLM research, including grouped-query attention with shared K/V projections, cross-layer value residuals, and depthwise causal convolutions. Trained on 33B tokens in 40 B200 GPU-hours, Proust achieves Spearman $ρ= 0.390$ on ProteinGym substitutions, competitive with MLMs requiring 50--200$\times$ the compute. On indels, Proust sets a new state-of-the-art, outperforming models up to 20$\times$ larger. On EVEREST viral fitness benchmarks, it approaches structure-aware methods using sequence alone. These powerful representations position Proust in a sweet spot as it also retains native generative capabilities that MLMs lack by design. Interpretability analysis reveals that per-position entropy variance predicts, to an extent, when retrieval augmentation helps and hurts. Such insights can grow in both quantity and quality at scale and inform capabilities such as test-time scaling. Code and weights are available at https://github.com/Furkan9015/proust-inference

Data-independent acquisition mass spectrometry (DIA-MS) has established itself as a cornerstone of proteomic profiling and large-scale systems biology, offering unparalleled depth and reproducibility. Current DIA analysis frameworks, however, require semi-supervised training within each run for peptide-spectrum match (PSM) re-scoring. This approach is prone to overfitting and lacks generalizability across diverse species and experimental conditions. Here, we present DIA-CLIP, a pre-trained model shifting the DIA analysis paradigm from semi-supervised training to universal cross-modal representation learning. By integrating dual-encoder contrastive learning framework with encoder-decoder architecture, DIA-CLIP establishes a unified cross-modal representation for peptides and corresponding spectral features, achieving high-precision, zero-shot PSM inference. Extensive evaluations across diverse benchmarks demonstrate that DIA-CLIP consistently outperforms state-of-the-art tools, yielding up to a 45% increase in protein identification while achieving a 12% reduction in entrapment identifications. Moreover, DIA-CLIP holds immense potential for diverse practical applications, such as single-cell and spatial proteomics, where its enhanced identification depth facilitates the discovery of novel biomarkers and the elucidates of intricate cellular mechanisms.

Predicting drug-drug interactions (DDIs) is essential for safe pharmacological treatments. Previous graph neural network (GNN) models leverage molecular structures and interaction networks but mostly rely on linear aggregation and symmetric assumptions, limiting their ability to capture nonlinear and heterogeneous patterns. We propose MGKAN, a Graph Kolmogorov-Arnold Network that introduces learnable basis functions into asymmetric DDI prediction. MGKAN replaces conventional MLP transformations with KAN-driven basis functions, enabling more expressive and nonlinear modeling of drug relationships. To capture pharmacological dependencies, MGKAN integrates three network views-an asymmetric DDI network, a co-interaction network, and a biochemical similarity network-with role-specific embeddings to preserve directional semantics. A fusion module combines linear attention and nonlinear transformation to enhance representational capacity. On two benchmark datasets, MGKAN outperforms seven state-of-the-art baselines. Ablation studies and case studies confirm its predictive accuracy and effectiveness in modeling directional drug effects.

Metacognition is a critical component of intelligence, specifically regarding the awareness of one's own knowledge. While humans rely on shared internal memory for both answering questions and reporting their knowledge state, this dependency in LLMs remains underexplored. This study proposes a framework to measure metacognitive ability $d_{\rm{type2}}'$ using a dual-prompt method, followed by the introduction of Evolution Strategy for Metacognitive Alignment (ESMA) to bind a model's internal knowledge to its explicit behaviors. ESMA demonstrates robust generalization across diverse untrained settings, indicating a enhancement in the model's ability to reference its own knowledge. Furthermore, parameter analysis attributes these improvements to a sparse set of significant modifications.

The NBSS (normalized biomass size spectrum) is a common, intuitive approach for the study of natural ecosystems. However, very few studies have been dedicated to verifying possible bias, flaws, and paradoxes in this widely used method. An evident issue of this method, that best exemplifies its discrepancies and paradoxes, is the use of intriguing non-biomass units (such as abundance, biomass flux, or pseudo-abundance units) on NBSS plots, that are intended to visualize biomass spectra. The main objectives of this study were to verify, test and analyze the procedures involved in transformations that lead to the popular NBSS plot, and to check for the correctness of currently used units, while testing the hypothesis that NBSS indeed represents biomass, not abundance or biomass flux (dB/dM), while developing i.) a new conceptual framework, ii.) new terminology, iii.) a novel back-transformation method, iv.) a simple, new calculation method, that yields the best (i.e., least biased) representation of the original biomass vs body mass distribution shape, numerical values, dimensions, and units. Extensive tests with in-situ and synthetic (simulated) data were used to verify the procedures involved in transformations that lead to the popular NBSS plots, and to compare the original biomass distribution data with the binned outputs. Original biomass units and dimensions are retained in the novel backtransformed normalized biomass spectrum (bNBS), proposed and described herein. The proposed bNBS constitutes a new, improved approach of robust size spectra science, that allows for quantitative inter-comparisons of biomass spectra across regions and time periods.

In two-strategy decision-making problems, individuals often imitate the highest earners or choose either the common or rare strategy. Individuals who benefit from the common strategy are conformists, whereas those who profit by choosing the less common one are called nonconformists. The population proportions of the two strategies may undergo perpetual fluctuations in finite, discrete, heterogeneous populations of imitators, conformists, and nonconformists. How these fluctuations evolve as population size increases was left as an open question and is addressed in this paper. We show that the family of Markov chains describing the discrete population dynamics forms a generalized stochastic approximation process for a differential inclusion--the continuous-time dynamics. Furthermore, we prove that the continuous-time dynamics always equilibrate. Then, by leveraging results from the stochastic approximation theory, we show that the amplitudes of fluctuations in the proportions of the two strategies in the population approach zero with probability one when the population size grows to infinity. Our results suggest that large-scale perpetual fluctuations are unlikely in large, well-mixed populations consisting of these three types, particularly when imitators follow the highest earners.

Cortical folding exhibits substantial inter-individual variability while preserving stable anatomical landmarks that enable fine-scale characterization of cortical organization. Among these, the three-hinge gyrus (3HG) serves as a key folding primitive, showing consistent topology yet meaningful variations in morphology, connectivity, and function. Existing landmark-based methods typically model each 3HG independently, ignoring that 3HGs form higher-order folding communities that capture mesoscale structure. This simplification weakens anatomical representation and makes one-to-one matching sensitive to positional variability and noise. We propose a spectral graph representation learning framework that models community-level folding units rather than isolated landmarks. Each 3HG is encoded using a dual-profile representation combining surface topology and structural connectivity. Subject-specific spectral clustering identifies coherent folding communities, followed by topological refinement to preserve anatomical continuity. For cross-subject correspondence, we introduce Joint Morphological-Geometric Matching, jointly optimizing geometric and morphometric similarity. Across over 1000 Human Connectome Project subjects, the resulting communities show reduced morphometric variance, stronger modular organization, improved hemispheric consistency, and superior alignment compared with atlas-based and landmark-based or embedding-based baselines. These findings demonstrate that community-level modeling provides a robust and anatomically grounded framework for individualized cortical characterization and reliable cross-subject correspondence.

Millions of users turn to consumer AI chatbots to discuss behavioral and mental health concerns. While this presents unprecedented opportunities to deliver population-level support, it also highlights an urgent need to develop rigorous and scalable safety evaluations. Here we introduce SIM-VAIL, an AI chatbot auditing framework that captures how harmful AI chatbot responses manifest across a range of mental-health contexts. SIM-VAIL pairs a simulated human user, harboring a distinct psychiatric vulnerability and conversational intent, with an audited frontier AI chatbot. It scores conversation turns on 13 clinically relevant risk dimensions, enabling context-dependent, temporally resolved assessment of mental-health risk. Across 810 conversations, encompassing over 90,000 turn-level ratings and 30 psychiatric user profiles, we find that significant risk occurs across virtually all user phenotypes. Risk manifested across most of the 9 consumer AI chatbot models audited, albeit mitigated in more modern variants. Rather than arising abruptly, risk accumulated over multiple turns. Risk profiles were phenotype-dependent, indicating that behaviors that appear supportive in general settings are liable to be maladaptive when they align with mechanisms that sustain a user's vulnerability. Multivariate risk patterns revealed trade-offs across dimensions, suggesting that mitigation targeting one harm domain can exacerbate others. These findings identify a novel failure mode in human-AI interactions, which we term Vulnerability-Amplifying Interaction Loops (VAILs), and underscore the need for multi-dimensional approaches to risk quantification. SIM-VAIL provides a scalable evaluation framework for quantifying how mental-health risk is distributed across user phenotypes, conversational trajectories, and clinically grounded behavioral dimensions, offering a foundation for targeted safety improvements.

Motivation: Predicting gene-disease associations (GDAs) is the problem to determine which gene is associated with a disease. GDA prediction can be framed as a ranking problem where genes are ranked for a query disease, based on features such as phenotypic similarity. By describing phenotypes using phenotype ontologies, ontology-based semantic similarity measures can be used. However, traditional semantic similarity measures use only the ontology taxonomy. Recent methods based on ontology embeddings compare phenotypes in latent space; these methods can use all ontology axioms as well as a supervised signal, but are inherently transductive, i.e., query entities must already be known at the time of learning embeddings, and therefore these methods do not generalize to novel diseases (sets of phenotypes) at inference time. Results: We developed INDIGENA, an inductive disease-gene association method for ranking genes based on a set of phenotypes. Our method first uses a graph projection to map axioms from phenotype ontologies to a graph structure, and then uses graph embeddings to create latent representations of phenotypes. We use an explicit aggregation strategy to combine phenotype embeddings into representations of genes or diseases, allowing us to generalize to novel sets of phenotypes. We also develop a method to make the phenotype embeddings and the similarity measure task-specific by including a supervised signal from known gene-disease associations. We apply our method to mouse models of human disease and demonstrate that we can significantly improve over the inductive semantic similarity baseline measures, and reach a performance similar to transductive methods for predicting gene-disease associations while being more general. Availability and Implementation: https://github.com/bio-ontology-research-group/indigena

Electroencephalography (EEG) underpins neuroscience, clinical neurophysiology, and brain-computer interfaces (BCIs), yet pronounced inter- and intra-subject variability limits reliability, reproducibility, and translation. This systematic review studies that quantified or modeled EEG variability across resting-state, event-related potentials (ERPs), and task-related/BCI paradigms (including motor imagery and SSVEP) in healthy and clinical cohorts. Across paradigms, inter-subject differences are typically larger than within-subject fluctuations, but both affect inference and model generalization. Stability is feature-dependent: alpha-band measures and individual alpha peak frequency are often relatively reliable, whereas higher-frequency and many connectivity-derived metrics show more heterogeneous reliability; ERP reliability varies by component, with P300 measures frequently showing moderate-to-good stability. We summarize major sources of variability (biological, state-related, technical, and analytical), review common quantification and modeling approaches (e.g., ICC, CV, SNR, generalizability theory, and multivariate/learning-based methods), and provide recommendations for study design, reporting, and harmonization. Overall, EEG variability should be treated as both a practical constraint to manage and a meaningful signal to leverage for precision neuroscience and robust neurotechnology.

Deep learning has been extensively researched in the analysis of pathology whole-slide images (WSIs). However, most existing methods are limited to providing prediction interpretability by locating the model's salient areas in a post-hoc manner, failing to offer more reliable and accountable explanations. In this work, we propose Phenotype-Aware Multiple Instance Learning (PA-MIL), a novel ante-hoc interpretable framework that identifies cancer-related phenotypes from WSIs and utilizes them for cancer subtyping. To facilitate PA-MIL in learning phenotype-aware features, we 1) construct a phenotype knowledge base containing cancer-related phenotypes and their associated genotypes. 2) utilize the morphological descriptions of phenotypes as language prompting to aggregate phenotype-related features. 3) devise the Genotype-to-Phenotype Neural Network (GP-NN) grounded in genotype-to-phenotype relationships, which provides multi-level guidance for PA-MIL. Experimental results on multiple datasets demonstrate that PA-MIL achieves competitive performance compared to existing MIL methods while offering improved interpretability. PA-MIL leverages phenotype saliency as evidence and, using a linear classifier, achieves competitive results compared to state-of-the-art methods. Additionally, we thoroughly analyze the genotype-phenotype relationships, as well as cohort-level and case-level interpretability, demonstrating the reliability and accountability of PA-MIL.

Zero-shot mutation prediction is vital for low-resource protein engineering, yet existing protein language models (PLMs) often yield statistically confident results that ignore fundamental biophysical constraints. Currently, selecting candidates for wet-lab validation relies on manual expert auditing of PLM outputs, a process that is inefficient, subjective, and highly dependent on domain expertise. To address this, we propose Rank-and-Reason (VenusRAR), a two-stage agentic framework to automate this workflow and maximize expected wet-lab fitness. In the Rank-Stage, a Computational Expert and Virtual Biologist aggregate a context-aware multi-modal ensemble, establishing a new Spearman correlation record of 0.551 (vs. 0.518) on ProteinGym. In the Reason-Stage, an agentic Expert Panel employs chain-of-thought reasoning to audit candidates against geometric and structural constraints, improving the Top-5 Hit Rate by up to 367% on ProteinGym-DMS99. The wet-lab validation on Cas12i3 nuclease further confirms the framework's efficacy, achieving a 46.7% positive rate and identifying two novel mutants with 4.23-fold and 5.05-fold activity improvements. Code and datasets are released on GitHub (https://github.com/ai4protein/VenusRAR/).

Current quantum computers suffer from a limited number of qubits and high error rates, limiting practical applicability. Different techniques exist to mitigate these effects and run larger algorithms. In this work, we analyze one of these techniques called quantum circuit cutting. With circuit cutting, a quantum circuit is decomposed into smaller sub-circuits, each of which can be run on smaller quantum hardware. We compare the performance of quantum circuit cutting with different cutting strategies, and then apply circuit cutting to a QAOA algorithm. Using simulations, we first show that Randomized Clifford measurements outperform both Pauli and random unitary measurements. Second, we show that circuit cutting has trouble providing correct answers in noisy settings, especially as the number of circuits increases.

Large language models (LLMs) can generate structured artifacts, but using them as dependable optimizers for scientific design requires a mechanism for iterative improvement under black-box evaluation. Here, we cast quantum circuit synthesis as a closed-loop, test-time optimization problem: an LLM proposes edits to a fixed-length gate list, and an external simulator evaluates the resulting state with the Meyer-Wallach (MW) global entanglement measure. We introduce a lightweight test-time learning recipe that can reuse prior high-performing candidates as an explicit memory trace, augments prompts with a score-difference feedback, and applies restart-from-the-best sampling to escape potential plateaus. Across fixed 20-qubit settings, the loop without feedback and restart-from-the-best improves random initial circuits over a range of gate budgets. To lift up this performance and success rate, we use the full learning strategy. For 25-qubit, it mitigates a pronounced performance plateau when naive querying is used. Beyond raw scores, we analyze the structure of synthesized states and find that high MW solutions can correspond to stabilizer or graph-state-like constructions, but full connectivity is not guaranteed due to the metric property and prompt design. These results illustrate both the promise and the pitfalls of memory evaluator-guided LLM optimization for circuit synthesis, highlighting the critical role of prior human-made theoretical theorem to optimally design a custom tool in support of research.

We report the generation of quantum entanglement between the center-of-mass motion of a levitated nanosphere, coupled by coherent scattering to an optical cavity mode, and the electromagnetic field. Using heterodyne detection, we reconstruct the full set of optical-mechanical correlations and observe a violation of separability bounds between the mechanical degrees of freedom and the propagating optical mode. Thus, we demonstrate the ability to distribute nonclassical correlations beyond the interaction region. Our results are obtained at room temperature and are robust over a broad range of detunings set by the cavity linewidth. These findings establish levitated optomechanical systems as a promising platform for macroscopic quantum optics and for future tests of fundamental physics.

This article establishes a low-regularity Riemannian positive mass theorem for non-spin manifolds whose metrics are only $C^0 \cap W_{\mathrm{loc}}^{1,n}$ and smooth outside a compact set. The main theorem asserts that asymptotically flat manifolds with nonnegative distributional scalar curvature have nonnegative ADM mass. The proof uses smooth approximations of the metric together with a Sobolev version of Friedrichs' Lemma, which yields improved convergence for commutators between differentiation and convolution operators. Rigidity is obtained for $C^0 \cap W_{\mathrm{loc}}^{1,p}$ metrics with $p>n$ via the comparison theory of $\sf{RCD}$-spaces and a rigidity theorem for compact manifolds with metrics of nonnegative distributional curvature by Jiang-Sheng-Zhang. The argument relies on either elementary techniques or generalisations of the standard argument. In essence, a version of the main theorem of Lee-LeFloch is presented in which the spin condition is removed under the assumption that the metric is smooth outside a compact set.

We investigate the topological properties of a generalized Dirac--Kronig--Penney model, a continuum one-dimensional model for a relativistic quantum chain. By tuning the coupling parameters this model can accommodate five Altland--Zirnbauer--Cartan symmetry classes, three of which (AIII, BDI and D) support non-trivial topological phases in dimension one. We characterize analytically the spectral properties of the Hamiltonian in terms of a spectral function, and numerically compute the Zak phase to probe the bulk topological content of the insulating phases. Our findings reveal that, while the Zak phase is quantized in classes AIII and BDI, it exhibits non-quantized values in class D, challenging its traditional role as a topological marker in continuum settings. We also discuss the bulk-boundary correspondence for a truncated version of the chain, analyzing how the emergence of edge states depends on both the truncation position and the boundary conditions. In classes AIII and BDI, we find that the Zak phase effectively detects edge states as a relative boundary topological index, although the correspondence is highly sensitive to the parameters characterizing the truncation.

We study the Page curve and information paradox for Kerr AdS black hole in light of entanglement entropy by employing the recently proposed island paradigm. By incorporating the island rule, we show that the entanglement entropy of Kerr AdS black hole grows linearly at early times and declines to a constant value at late times in agreement with the well established Page curve. The novelty of this study resides in the investigation of influence of phase transitions on the page curve in two different ensembles. We find that a first order phase transition results in a sharp discontinuity in the Page curve. We study the evaporation process in different scenarios and find that in all the situations, the Page curve doesn't violate the unitary principle of quantum mechanics.

Dehnen $(1,4,\frac32)$ dark matter halo has been proven to be a valuable model for describing the surface brightness distributions of elliptical galaxies, yet its implications for black hole spacetimes remain largely unexplored. In this work, we construct a novel exact black hole solution embedded in this Dehnen halo and investigate its physical consequences. The influence of the halo on black hole thermodynamics is analyzed through the mass function, entropy, Hawking temperature, heat capacity, and Gibbs free energy, allowing us to assess both local and global thermodynamic stability of the black hole-dark matter system. Our results show that the presence of a Dehnen-type halo not only stabilizes the otherwise thermodynamically unstable Schwarzschild black hole but also induces phase transitions. In addition, we study null geodesics to examine photon motion, the shadow radius and the optical appearance of the system. The dark matter halo modifies the effective potential, leading to observable changes in the photon sphere and the apparent size of the shadow. We also explore the instability of circular null geodesics and its relation to quasinormal modes in an eikonal limit. These findings highlight the significant role of realistic dark matter distributions in shaping both the thermodynamic behavior and the observable signatures of black holes, providing further insight into the interplay between dark matter halos and central black holes in galaxies.

In fault-tolerant quantum computing, soft outputs from real-time decoders play a crucial role in improving decoding accuracy, post-selecting magic states, and accelerating lattice surgery. A recent paper by Meister et al. [arXiv:2405.07433 (2024)] proposed an efficient method to evaluate soft outputs for cluster-based decoders, including the Union-Find (UF) decoder. However, in parallel computing environments, its computational complexity is comparable to or even surpasses that of the UF decoder itself, resulting in a substantial overhead. Furthermore, this method requires global information about the decoding graph, making it poorly suited for existing hardware implementations of the UF decoder on Field-Programmable Gate Arrays (FPGAs). In this paper, to alleviate these issues, we develop more efficient methods for evaluating high-quality soft outputs in cluster-based decoders by introducing several early-stopping techniques. Our central idea is that the precise value of a large soft output is often unnecessary in practice. Based on this insight, we introduce two types of novel soft-outputs: the bounded cluster gap and the extra-cluster gap. The former reduces the computational complexity of Meister's method by terminating the calculation at an early stage. Our numerical simulations show that this method achieves improved scaling with code distance $d$ compared to the original proposal. The latter, the extra-cluster gap, quantifies decoder reliability by performing a small, additional growth of the clusters obtained by the decoder. This approach offers the significant advantage of enabling soft-output computation without modifying the existing architecture of FPGA-implemented UF decoders. These techniques offer lower computational complexity and higher hardware compatibility, laying a crucial foundation for future real-time decoders with soft outputs.

Solid-state color centers are promising candidates for nodes in quantum network architectures. However, realizing scalable and fully functional quantum nodes, comprising both processor and memory qubits with high-fidelity universal gate operations, remains a central challenge in this field. Here, we demonstrate a fully functional quantum node in silicon carbide, where electron spins act as quantum processors and nuclear spins serve as quantum memory. Specifically, we design a pulse sequence that combines dynamical decoupling with hyperfine interactions to realize decoherence-protected universal gate operations between the processor and memory qubits. Leveraging this gate, we deterministically prepare entangled states within the quantum node, achieving a fidelity of 90%, which exceeds the fault-tolerance threshold of certain quantum network architectures. These results open a pathway toward scalable and fully functional quantum nodes based on silicon carbide.

The variational quantum algorithm (VQA) with a parametrized quantum circuit is widely applicable to near-term quantum computing, but its fundamental issues that limit optimization performance have been reported in the literature. For example, VQA optimization often suffers from vanishing gradients called barren plateau (BP) and the presence of local minima in the landscape of the cost function. Numerical studies have shown that the trap in local minima is significantly reduced when the circuit is overparametrized (OP), where the number of parameters exceeds a certain threshold. Theoretical understanding of the BP and OP phenomena has advanced over the past years, however, comprehensive studies of both effects in the same setting are not fully covered in the literature. In this paper, we perform a comprehensive numerical study in VQA, quantitatively evaluating the impacts of BP and OP and their interplay on the optimization of a variational quantum circuit, using concrete implementations of one-dimensional transverse and longitudinal field quantum Ising model. The numerical results are compared with the theoretical diagnostics of BP and OP phenomena. The framework presented in this paper will provide a guiding principle for designing VQA algorithms and ansatzes with theoretical support for behaviors of parameter optimization in practical settings.

We show how the macroscopic state variables pressure, entropy and temperature of equilibrium thermodynamics can be consistently derived from the (quantum) chaotic spectral structure of one or two particles in two-dimensional domains. This provides a definition of work and heat from first principles, a microscopic underpinning of the first and second law of thermodynamics, and a transparent illustration of the ``eigenstate thermalization hypothesis''.

We prove that the black to white hole transition theorized in several papers can be described as a change in the topology of the event horizon. We also show, using the theory of cobordism due to Milnor and Wallace, how to obtain the full manifold containing the transition.

Quantum chaos is commonly assessed through probe-dependent signatures such as spectral statistics, OTOCs, and entanglement growth, which need not coincide. Recently, a dissipative diagnostic of chaos has been proposed, in which an infinitesimal coupling to a bath yields a finite Liouvillian gap in chaotic systems, marking the onset of intrinsic relaxation. This raises a conceptual question: what is the minimal departure from Clifford dynamics needed for this intrinsically relaxing behavior to emerge? In this work, we investigate the dynamics under the Floquet two-qubit Clifford circuit interleaved with a finite density of Haar-random single-site gates, followed by a depolarizing channel with strength $γ$. For Floquet Clifford circuits built from an \textit{i}SWAP-class two-qubit gate, our analysis identifies two distinct regimes for the Liouvillian gap in the thermodynamic limit, exemplified by the undoped and fully doped extreme cases. In both regimes, the dissipative diagnostic signals chaotic behavior, differing only in how the gap scales with system size. In the undoped circuit, the gap scales as $Δ\sim γN$, whereas in the fully doped circuit it remains finite as $N\to\infty$. We find that the doping density $p_h$ governs the crossover: as $p_h\to 0$, any spatial structure remains undoped-like, whereas for finite $p_h$ certain structures can enter a finite-gap regime. These results are analytically established in the strongly dissipative regime $γ\gg 1$ by deriving lower bounds on the gap as a function of $p_h$ and explicit finite-gap constructions, and their extension toward $γ\to 0$ is supported by numerics. Importantly, our analytic treatment depends only on the spatial doping structure, so the same gap scaling persists even when the Haar rotations are independently resampled each Floquet period.

The ability to tune interparticle interactions is one of the main advantages of using ultracold quantum gases for quantum simulation of many-body physics. Current experiments with ultracold polar molecules employ shielding with microwave or static electric fields to prevent destructive collisional losses. The interaction potential of microwave-shielded molecules can be tuned by using microwaves of two different polarisations, while for static-field-shielded molecules the tunability of interactions is more limited and depends on the particular species. In this work, we propose a general method to tune the interactions between static-field-shielded molecules by applying a microwave field. We carry out coupled-channel scattering calculations in a field-dressed basis set to determine loss rate coefficients and scattering lengths. We find that both the s-wave scattering length and the dipole length can be widely tuned by changing the parameters of the microwave field, while maintaining strong suppression of lossy collisions.

External non-minimally coupled scalar and spinor field perturbations have been studied in a (1 + 1) dimensional dilatonic blackhole spacetime [1, 2]. Exact analytical expressions of the quasi- normal mode frequencies have been found for both the cases. In the scalar perturbations the quasi-normal mode frequencies turn out to be purely imaginary and negative. Furthermore we have found that the quasi-normal frequencies for Dirac field exhibit both real and imaginary part. The QNM frequencies decay monotonically with the overtone number under certain class of the blackhole parameters. The decay profile ensures the stability of the blackhole spacetime under these perturbations.

The proper-time method for constructing perturbative dynamical gravitational fields is presented. Using the proper-time method, a perturbative analytical model of gravitational waves against the backdrop of an exact wave solution of Einstein's equations in a Bianchi IV universe is constructed. To construct the perturbative analytical wave model a privileged wave coordinate system and a synchronous time function associated with the proper time of an observer freely moving in a gravitational wave were used. Reduction of the field equations, taking into account compatibility conditions, reduces the mathematical model of gravitational waves to a system of coupled ordinary differential equations for functions of the wave variable. Analytical solutions for the components of the gravitational-wave metric have been found. The stability of the resulting perturbative solutions is proven. The stability of the exact solution for a gravitational wave in the anisotropic Bianchi IV universe is demonstrated.

Superconducting quantum circuits typically use capacitive charge-based linear coupling schemes to control interactions between elements such as qubits. While simple and effective, this coupling scheme makes it difficult to satisfy competing circuit design requirements such as maintaining large qubit anharmonicity and coherence along with a high degree of qubit connectivity and packing density. Moreover, tunable interactions using linear coupling elements produce dynamical variations in mode hybridization, which can induce non-adiabatic transitions, resulting in leakage errors and limiting gate speeds. In this work we attempt to address these challenges by proposing a junction-based coupling architecture based on SQUID (superconducting quantum interference device) couplers with relatively small Josephson energies. SQUID couplers provide intrinsic cross-Kerr interactions that can be controlled by external fluxes and that do not rely on mode hybridization. The small Josephson energies of the coupler maintain the interaction at a perturbative scale, which limits undesired higher-order mixing between coupled elements while achieving a sufficiently strong cross-Kerr interaction originating from diagonal coupling elements. Based on these properties, we show that a SQUID coupler can be used to implement a fast, adiabatic, and high-fidelity controlled-Z gate without introducing extra modes, and the operation is robust against junction asymmetry for high-frequency qubits. Although unconventional crosstalk may arise due to junction asymmetries and parasitic hybridization with spectator qubits, we show that these effects are sufficiently small for realistic circuit parameters. As an example of the utility of such junction-based coupling schemes, we present a scalable tiling strategy for a miniaturized superconducting quantum processor based on merged-element transmon qubits.

Protocols based on the sending-or-not-sending (SNS) principle have been intensively studied in recent years and have been shown to enable the longest transmission distances in quantum key distribution (QKD). In this work, we propose a sending-or-not-sending phase-matching QKD protocol (SNS-PM-QKD) that improves tolerance to phase mismatch, thereby extending the achievable transmission distance. We present a security analysis of SNS-PM-QKD in the asymptotic (infinite-key) regime under collective attacks. The performance of the proposed protocol is compared with that of standard phase-matching QKD, theoretical SNS-type twin-field QKD protocols (SNS-TF-QKD), and an experimental SNS-TF-QKD operated over transmission distances of up to 1002km. Our results show that SNS-PM-QKD achieves greater transmission distances than these existing protocols, highlighting its potential for long-distance quantum communication.

The recently reported precession period of about $11.24$ years of the M87* jet provides a sensitive probe of strong field gravity and the electromagnetic environment in the immediate vicinity of supermassive black holes. In this work, we study the precession of the spherical orbit in the Kerr-Bertotti-Robinson geometry describing a rotating black hole immersed in a uniform electromagnetic field. Although the timelike geodesics is non-separable, we develop a Hamiltonian approach to investigate the spherical orbits. For sufficiently strong magnetic fields, the study shows that the spherical orbits can only exist within a finite radial range for given orbital inclination. Requiring the existence of the spherical orbits, we obtain an upper bound of the magnetic field, i.e., $B<0.33 M^{-1}$ for prograde and $B<0.0165 M^{-1}$ for retrograde motion. Furthermore, imposing the observed jet precession period, we obtain a significantly tighter constraint, $B\lesssim 0.0145 M^{-1}$, providing a new constrain on the magnetic field of M87* independent of the shadow. Our results provide unified constraints on the parameters of the KBR black hole and demonstrate that the jet precession offers a robust and complementary probe of magnetized black holes in the strong gravity regime.

The potential of Quantum Machine Learning (QML) in data-intensive science is strictly bottlenecked the difficulty of interfacing high-dimensional, chaotic classical data into resource-limited, noisy quantum processors. To bridge this gap, we introduce a physics-informed Koopman-Quantum hybrid framework, theoretically grounded in a representation-level structural isomorphism we establish between the Koopman operator, which linearizes nonlinear dynamics, and quantum evolution. Based on this theoretical foundation, we design a realizable NISQ-friendly pipeline: the Koopman operator functions as a physics-aware "data distiller," compressing waveforms into compact, "quantum-ready" features, which are subsequently processed by a modular, parallel quantum neural network. We validated this framework on 4,763 labeled channel sequences from 433 discharges of the tokamak system. The results demonstrate that our model achieves 97.0\% accuracy in screening corrupted diagnostic data, matching the performance of state-of-the-art deep classical CNNs while using orders-of-magnitude fewer trainable parameters. This work establishes a practical, physics-grounded paradigm for leveraging quantum processing in constrained environments, offering a scalable path for quantum-enhanced edge computing.

General Relativity (GR) faces challenges from cosmic acceleration and observational tensions, necessitating stringent tests at cosmological scales. In this work, we probe GR deviations via a $μ$-$Σ$ modified gravity parameterization, integrating KiDS-Legacy weak lensing (1347 deg$^2$, $z\leq 2.0$), joint CMB data (Planck/ACT/SPT), DESI DR2 BAO, and DES-Dovekie supernovae. KiDS-Legacy significantly improves constraint precision: $μ_0$ (matter clustering) by $\sim 43\%$ and $Σ_0$ (gravitational light deflection) by $\sim 60\%$ relative to CMB alone. In the $Λ$CDM background, $μ_0 = 0.21\pm 0.21$ is consistent with GR, while $Σ_0 = 0.149\pm 0.051$ deviates from GR at the 3.0$σ$ level -- attributed to large-scale CMB lensing from ACT/SPT. This precise separation of GR-consistent matter clustering and deviant light deflection provides key observational clues for new physics or data systematics. Our work underscores the critical role of synergizing high-precision CMB and WL data in advancing GR tests.

Critical decision-making issues in science, engineering, and industry are based on combinatorial optimization; however, its application is inherently limited by the NP-hard nature of the problem. A specialized paradigm of analogue quantum computing, quantum annealing (QA), has been proposed to solve these problems by encoding optimization problems into physical energy landscapes and solving them by quantum tunnelling systematically through exploration of solution space. This is a critical review that summarizes the current applications of quantum annealing to combinatorial optimization and includes a theoretical background, hardware designs, algorithm implementation strategies, encoding and embedding schemes, protocols to benchmark quantum annealing, areas of implementation, and links with the quantum algorithms implementation with gate-based hardware and classical solvers. We develop a unified framework, relating adiabatic quantum dynamics, Ising and QUBO models, stoquastic and non-stoquastic Hamiltonians, and diabatic transitions to modern flux-qubit annealers (Chimera, Pegasus, Zephyr topologies), and emergent architectures (Lechner-Hauke-Zoller systems, Rydberg atom platforms), and hybrids of quantum and classical computation. Through our analysis, we find that overhead in embedding and encoding is the largest determinant of the scalability and performance (this is not just the number of qubits). Minor embeddings also usually have a physical qubit count per logical variable of between 5 and 12 qubits, which limits effective problem capacity by 80-92% and, due to chain-breaking errors, compromises the quality of solutions.

Transport phenomena play a key role in a variety of application domains, and efficient simulation of these dynamics remains an outstanding challenge. While quantum computers offer potential for significant speedups, existing algorithms either lack rigorous theoretical guarantees or demand substantial quantum resources, preventing scalable and efficient validation on realistic quantum hardware. To address this gap, we develop a comprehensive framework for simulating classes of transport equations, offering both rigorous theoretical guarantees -- including exponential speedups in specific cases -- and a systematic, hardware-efficient implementation. Central to our approach is the Hamiltonian embedding technique, a white-box approach for end-to-end simulation of sparse Hamiltonians that avoids abstract query models and retains near-optimal asymptotic complexity. Empirical resource estimates indicate that our approach can yield an order-of-magnitude (e.g., $42\times$) reduction in circuit depth given favorable problem structures. We then apply our framework to solve linear and nonlinear transport PDEs, including the first experimental demonstration of a 2D advection equation on a trapped-ion quantum computer.

We propose an ion-trap implementation of the Vaccaro, Barnett and Wright et al. spin-heat engine (SHE); a hypothetical engine that operates between energy and spin thermal reservoirs rather than two energy reservoirs. The SHE operates in two steps: first, in the work extraction stage, heat from a thermal energy reservoir is converted into optical work via a two photon Raman transition resonant with close-to energy degenerate spin states; second, the internal spin states are brought back to their initial state via non-energetic information erasure using a spin reservoir. The latter incurs no energy cost, but rather the reset occurs at the cost of angular momentum from a spin bath that acts as the thermal spin reservoir. The SHE represents an important first step toward demonstrating heat engines that operate beyond the conventional paradigm of requiring two thermal reservoirs, paving the way to harness quantum coherence in arbitrary conserved quantities via similar machines.

Regular black holes are often used to address singularities, but they typically involve a potentially unstable de Sitter core and an internal horizon that breaks predictability. Integrable singularities (IS) have recently gained attention because they avoid both issues and exhibit finite tidal forces, allowing nondestructive radial infall. First, we present a new BH solution sourced by a string fluid (FS) that exhibits an IS. Motivated by the divergence of the conserved energy in the cloud of strings (CS) model, we introduce an energy density profile based on the screening of the CS energy density within an FS framework, yielding a finite conserved energy. On the other hand, it has been proposed \cite{Ovalle:2024wtv} that an interior region, rather than a pointlike mass, can generate a Schwarzschild BH exterior region. Secondly, motivated by the variety of BH solutions with singularities in the literature, we establish the conditions that an interior region with an IS must satisfy to represent a generic exterior BH solution, with Schwarzschild being only a particular case of the latter. We derive the junction conditions (JC) between the interior and exterior regions, showing that they lead to temperature continuity at the interface, while discontinuities in tangential pressure lead to phase transitions. We propose that the nature of the interior region is described by CS and FS, while the exterior corresponds to Reissner Nordström.

As quantum processors scale to large qubit numbers, device-to-device variability emerges as a critical challenge. Superconducting qubits are commonly realized using Al/AlO$_{\text{x}}$/Al Josephson junctions operating in the tunneling regime, where even minor variations in device geometry can lead to substantial performance fluctuations. In this work, we develop a quantitative model for the variability of the Josephson energy $E_{J}$ induced by interface roughness at the Al/AlO$_{\text{x}}$ interfaces. The roughness is modeled as a Gaussian random field characterized by two parameters: the root-mean-square roughness amplitude $σ$ and the transverse correlation length $ξ$. These parameters are extracted from the literature and molecular dynamics simulations. Quantum transport is treated using the Ambegaokar--Baratoff relation combined with a local thickness approximation. Numerical simulations over $5,000$ Josephson junctions show that $E_{J}$ follows a log-normal distribution. The mean value of $E_{J}$ increases with $σ$ and decreases slightly with $ξ$, while the variance of $E_{J}$ increases with both $σ$ and $ξ$. These results paint a quantitative and intuitive picture of Josephson energy variability induced by surface roughness, with direct relevance for junction design.

Quantum Error Correction (QEC) is essential for building robust, fault-tolerant quantum computers; however, the decoding process often presents a significant computational bottleneck. Tesseract is a novel Most-Likely-Error (MLE) decoder for QEC that employs the A* search algorithm to explore an exponentially large graph of error hypotheses, achieving high decoding speed and accuracy. This paper presents a systematic approach to optimizing the Tesseract decoder through low-level performance enhancements. Based on extensive profiling, we implemented four targeted optimization strategies, including the replacement of inefficient data structures, reorganization of memory layouts to improve cache hit rates, and the use of hardware-accelerated bit-wise operations. We achieved significant decoding speedups across a wide range of code families and configurations, including Color Codes, Bivariate-Bicycle Codes, Surface Codes, and Transversal CNOT Protocols. Our results demonstrate consistent speedups of approximately 2x for most code families, often exceeding 2.5x. Notably, we achieved a peak performance gain of over 5x for the most computationally demanding configurations of Bivariate-Bicycle Codes. These improvements make the Tesseract decoder more efficient and scalable, serving as a practical case study that highlights the importance of high-performance software engineering in QEC and providing a strong foundation for future research.

This paper investigates the quickest change detection of quantum states in a universal setting: specifically, where the post-change quantum state is not known a priori. We establish the asymptotic optimality of a two-stage approach in terms of worst average delay to detection. The first stage employs block POVMs with classical outputs that preserve quantum relative entropy to arbitrary precision. The second stage leverages a recently proposed windowed-CUSUM algorithm that is known to be asymptotically optimal for quickest change detection with an unknown post-change distribution in the classical setting.

We demonstrate efficient sub-Doppler laser cooling of the three eigenmodes of a $^{40}$Ca$^+$ ion confined in a compact Penning trap operating with a magnetic field of 0.91 T. Using the same set of laser beams as required for the initial Doppler laser cooling operation, we detune the laser frequencies to produce a narrow two-photon dark resonance. The process achieves a 1/e cooling time constant of 108(8) $μ$s, ultimately reducing the mean thermal axial mode occupation from 72(23) to 1.5(3) in 800 $μ$s as measured by resonantly probing an electric quadrupole transition near 729 nm. A parametric drive is applied to the trap electrodes which coherently exchanges the axial mode occupation with that of each radial mode, allowing for three-dimensional sub-Doppler cooling using only the axially-propagating laser beams. This sub-Doppler cooling is achieved for an axial oscillation frequency of $ω_z = 2π~\times~$221 kHz, which places the motion well outside of the Lamb Dicke confinement regime at the Doppler laser cooling limit. Our measured cooling rate and final mode occupation are in good agreement with a semiclassical model which combines a Lindblad master equation solution for ion-photon interactions with classical harmonic oscillator motion of the trapped ion.

Following the numerous evidence for large-scale cosmic isotropy violation with the advent of the `precision cosmology' era, we explore the possible advantages of extending the flat $Λ$CDM model to more general models in order to constrain anisotropies in the universe, otherwise absent in the standard model based on FLRW spacetime. Such extensions are offered by the topologically unique Thurston geometries, which are homogeneous but anisotropic spacetime models. In this work, we attempt to distinguish Thurston geometries from one another by introducing anisotropies via different scale factors in different directions, thereby introducing additional model parameters such as shear, eccentricity, curvature, and a preferred axis. We used the latest compilation of Pantheon+ \& SH0ES Type Ia supernova data for deriving model constraints, and found mild evidence of large-scale isotropy violation.

For specifically coupled values of the quadratic gravity parameters, we present a fully explicit static spherically symmetric solution. It contains the central singularity surrounded by the black-hole or the cosmological horizon for the negative or positive cosmological parameter, respectively. This spacetime, thus, belongs to the already analyzed classes of solutions expressed in terms of the Frobenius expansions, continuous fractions, or numerical simulations; however, it has not been explicitly identified before. The purpose of the presented highly-constrained somehow unphysical model is to reveal a global geometric picture that may occur in spherical spacetimes within quadratic gravity, which is typically hidden in more general approximative solutions. At the same time, it can serve as a solid benchmark for evaluating the accuracy of the methods employed to obtain such solutions.

We search for violations of the Equivalence Principle towards the Sun using a rotating torsion balance apparatus. We set 95\%-confidence limits on violations with beryllium and aluminum test bodies of $η_{\odot, Be-Al} \leq 2.1 \times 10^{-13}$. These results are a factor of four improvement of previously reported results towards the Sun and a $\sim20\%$ improvement on previous torsion balance tests regardless of source.

Majorana neutrino oscillations in a magnetic field are considered using the wave packets formalism. The modified Dirac equation for Majorana neutrinos with non-zero transition magnetic moments propagating in a magnetic field is solved analytically in the two flavour case. The expressions for the oscillations probabilities are derived accounting for the decoherence effect emerging at distances exceeding the coherence length. It is shown that for Majorana neutrinos propagating in a magnetic field the coherence length coincides with the coherence length for neutrino oscillations in vacuum when the vacuum frequency is much greater than the magnetic frequency ($ω_{vac} \gg ω_B$), while it is proportional to the cube of the average neutrino momentum if ($ω_{vac} \ll ω_B$). We show that the decoherence effect may appear during neutrino propagation in a magnetic field of supernova.

Electronic spin superposition states enable nanoscale sensing through their sensitivity to the local environment, yet their sensitivity to vibrational motion also limits their coherence times. In molecular spin systems, chemical tunability and atomic-scale resolution are accompanied by a dense, thermally accessible phonon spectrum that introduces efficient spin relaxation pathways. Despite extensive theoretical work, there is little experimental consensus on which vibrational energies dominate spin relaxation or how molecular structure controls spin-phonon coupling (SPC). We present a fully experimental method to quantify SPC coefficients by combining temperature-dependent vibrational spectra from inelastic neutron scattering with spin relaxation rates measured by electron paramagnetic resonance. We apply this framework to two model S = 1/2 systems, copper(II) phthalocyanine (CuPc) and copper(II) octaethylporphyrin (CuOEP). Two distinct relaxation regimes emerge: below 40 K, weakly coupled lattice modes below $50~\mathrm{cm}^{-1}$ dominate, whereas above 40 K, optical phonons above ~$185~\mathrm{cm}^{-1}$ become thermally populated and drive relaxation with SPC coefficients nearly three orders of magnitude larger. Structural distortions in CuOEP that break planar symmetry soften the crystal lattice and enhance anharmonic scattering, but also raise the energy of stretching modes at the molecular core where the spins reside. This redistributes vibrational energy toward the molecular periphery and out of plane, ultimately reducing SPC relative to CuPc and enabling room-temperature spin coherence in CuOEP. Although our method does not provide mode-specific SPC coefficients, it quantifies contributions from distinct spectral regions and establishes a broadly applicable, fully experimental link between crystal structure, lattice dynamics, and spin relaxation.

A recent paper proposes a static ellipsoidal space-time metric that reduces to Schwarzschild. After examining the corrected line element using Maple's Differential geometry package, we find that the Einstein tendor and Ricci scalar are non-zero, and the metric yields non-zero pressures when interpreted as an anisotropic fluid. Thus, it does not represent a vacuum solution. We also checked a cited ellipsoidal metric from the literature and found it likewise fails to satisfy the vacuum Einstein equations.

It is intuitively expected, and supported by earlier studies, that many-body quantum chaos is suppressed, or even destroyed, by dissipative effects induced by continuous monitoring. We show here that this is not always the case. For this purpose, we study the quenched dynamics of a continuously monitored Sachdev-Ye-Kitaev (SYK) model, described by the Lindblad formalism, coupled to a thermal environment modeled by another SYK maintained at constant temperature. We find that the combined effect of monitoring and the thermal bath drives the system toward a non-thermal steady state independently of the initial conditions. The corresponding retarded Green's function exhibits two stages of exponential decay, with rates that depend non-monotonously on the thermal bath coupling and the monitoring strength. In the limit of weak coupling, the late time decay of the Green's function, computed analytically, is closely related to that of the thermal bath. Strikingly, we identify a range of parameters in which continuous monitoring, despite being a source of decoherence, induces or enhances quantum chaotic dynamics suppressed by the thermal bath. For instance, in the limit of weak coupling to the thermal bath, the Lyapunov exponent increases sharply when monitoring is turned on. For intermediate values of the thermal bath coupling, the Lyapunov exponent exhibits re-entrant behavior: it vanishes at zero or sufficiently weak monitoring strength, and becomes positive again as the monitoring strength is increased. Our results offer intriguing insights on the mechanisms leading to quantum scrambling which paves the way to its experimental control and consequently to a performance enhancement of quantum information devices.

We develop an efficient package for the simulation of nuclear magnetic resonance spin echo experiments to study the effects of strong electronic spin correlations on the dynamics of the nuclear spin ensemble. A mean-field model is used to study correlated electronic phases through their hyperfine interaction with nuclear spins. We explore the dynamics of the interacting nuclear ensemble and discuss the key behaviors of the system. In particular, we classify the types of temporal asymmetry that the interaction induces in the system as well as a pulse-dependent shift in the spectral domain. Us- ing these results, we discuss how careful measurement of the pulse-dependent shiftcanbeusedtoextractinformationabouttheanisotropyoftheelectronic interaction and how these results represent a novel tool for the examination of exotic NMR signatures in strongly correlated materials. Finally, we re- view specific aspects of the simulation package developed for our exploration and give explicit examples where package can be used to infer range and anisotropy of electronic correlations. In particular, we discuss its structure, accuracy, and the technical merits of the various approximations used to model the nuclear spin ensemble.

We present a quantum tomography study of fermion pair production at future $e^+e^-$ colliders, emphasizing how longitudinal beam polarization controls the two-qubit spin density matrix. We study the processes $e^+ e^- \to t\bar{t},\ e^+e^-\to μ^+μ^-$ and Bhabha scattering $e^+e^-\to e^+e^-$, representing the mass threshold behavior, the $Z$ pole resonance and the $s/t$-channel interplay. We choose to focus on three key concepts: quantum entanglement via the concurrence $\mathcal{C}$, Bell nonlocality via the optimal Clauser Horne Shimony Holt (CHSH) parameter $\mathcal{B}$, and non-stabilizerness (``magic'') via the second stabilizer Rényi entropy $\mathcal{M}_2$. For the $s$-channel-dominated channels, longitudinal polarization mainly reshapes single-spin polarizations while leaving the spin-correlation matrix largely unchanged, rendering $\mathcal{C}$ and $\mathcal{B}$ comparatively robust, but inducing a pronounced variation of $\mathcal{M}_2$. In contrast, in Bhabha scattering, polarization modifies the relative contributions of the $s$-channel and $t$-channel and can strongly affect all three observables. The observability of entanglement, Bell nonlocality, and magic exceeds the $5σ$ level when both statistical and systematic uncertainties are included, establishing the fermion pair systems as ideal laboratories for quantum-information studies in high energy leptonic collisions. With optimized beam polarization, future $e^+e^-$ colliders will provide a unique opportunity to experimentally explore and influence quantum resources in particle interactions.

State preparation compilers for quantum computers typically sit at two extremes: general-purpose routines that treat the target as an opaque amplitude vector, and bespoke constructions for a handful of well-known state families. We ask whether a compiler can instead accept simple, structure-aware specifications while providing predictable resource guarantees. We answer this by designing and implementing a quantum state-preparation compiler for regular language states (RLS): uniform superpositions over bitstrings accepted by a regular description, and their complements. Users describe the target state via (i) a finite set of bitstrings, (ii) a regular expression, or (iii) a deterministic finite automaton (DFA), optionally with a complement flag. By translating the input to a DFA, minimizing it, and mapping it to an optimal matrix product state (MPS), the compiler obtains an intermediate representation (IR) that exposes and compresses hidden structure. The efficient DFA representation and minimization offloads expensive linear algebra computation in exchange of simpler automata manipulations. The combination of the regular-language frontend and this IR gives concise specifications not only for RLS but also for their complements that might otherwise require exponentially large state descriptions. This enables state preparation of an RLS or its complement with the same asymptotic resources and compile time. We outline two hardware-aware backends: SeqRLSP, which yields linear-depth, ancilla-free circuits for linear nearest-neighbor architectures via sequential generation, and TreeRLSP, which achieves logarithmic depth on all-to-all connectivity via a tree tensor network. We prove depth and gate-count bounds scaling with the system size and the state's maximal Schmidt rank, and we give explicit compile-time bounds that expose the benefit of our approach. We implement and evaluate the pipeline.

In this work, we present the integration of Qiskit Nature's quantum chemistry solvers into the Atomic Simulation Environment (ASE), enabling hybrid quantum-classical workflows for force-driven atomistic simulations. This coupling allows the use of the Variational Quantum Eigensolver (VQE) and its adaptive variant (ADAPT-VQE) not only for ground-state energy calculations, but also for geometry optimisation, vibrational frequency analysis, strain evaluation, and molecular dynamics, all managed through ASE's calculator interface. By applying ADAPT-VQE to multi-electron systems such as BeH2, we obtain vibrational and structural properties in close agreement with high-level classical CCSD calculations within the same minimal basis. These results demonstrate that adaptive variational quantum algorithms can deliver stable and chemically meaningful forces within an atomistic modelling workflow, enabling downstream applications such as molecular dynamics and active-learning accelerated simulations.

We revisit the Hartle-Hawking no-boundary proposal. To extract probabilities, one must use the gravitational path integral (GPI) to compute not only the no-boundary amplitude, but also the norms by which its square is divided. We find that this dramatically alters predictions: the probability for any closed universe is either nearly 1, or exactly 1. That is, in the Hilbert space of closed universes defined by the GPI, the states of interest in cosmology are all nearly parallel to the Hartle-Hawking state up to nonperturbative corrections in $G_N^{-1}$. We also consider a statistical interpretation of the GPI, as an average of arbitrary products of amplitudes. We find that all amplitudes are exactly 1 in this case, consistent with recent arguments that the statistical approach to the GPI with a closed boundary computes an average over one-dimensional Hilbert spaces. As an example, we illustrate the consistent evaluation of the no-boundary proposal in inflationary cosmology.

Quantum jumps, the collapse of a quantum system upon measurement, are among the most striking consequences of observation in quantum mechanics. While recent experiments have revealed the continuous nature of individual jumps, the crossover from coherent dynamics to measurement-dominated behaviour has remained elusive. Here, we tune the measurement strength of a continuously monitored superconducting qubit, and observe that quantum jumps emerge not through a gradual crossover, but via a cascade of three distinct dynamical transitions. The first transition manifests as an exceptional point where coherent oscillations abruptly cease, giving way to jumps towards a stable eigenstate. The second transition marks the onset of dynamical state freezing, where the qubit's dwell time near the eigenstate diverges. A third threshold signals entry into the quantum Zeno regime, where stronger measurement paradoxically suppresses relaxation. Strikingly, we find that decoherence does not blur these transitions but rather fundamentally restructures the dynamical phase diagram, notably inverting their order. These results map measurement-induced transitions in a monitored qubit, revealing that the interplay between coherent driving, measurement, and decoherence gives rise to a hierarchy of distinct dynamical phases.

[Abridged] Some cosmological theories propose that the observable universe is a small part of a much larger universe in which parameters describing the low-energy laws of physics vary from region to region. How can we reasonably assess a theory that describes such a mostly unobservable universe? We propose a Bayesian method based on theory-generated probability distributions for our observations. We focus on basic principles, leaving aside concerns about practicality. (We also leave aside the measure problem, to discuss other issues.) We argue that cosmological theories can be tested by standard Bayesian updating, but we need to use theoretical predictions for "first-person" probabilities -- i.e., probabilities for our observations, accounting for all relevant selection effects. These selection effects can depend on the observer, and on time, so in principle first-person probabilities are defined for each observer-instant -- an observer at an instant of time. First-person probabilities should take into account everything the observer believes about herself and her surroundings -- i.e., her "subjective state". We advocate a "Principle of Self-Locating Indifference" (PSLI), asserting that any real observer should make predictions as if she were chosen randomly from the theoretically predicted observer-instants that share her subjective state. We believe the PSLI is intuitively very reasonable, but also argue that it maximizes the expected fraction of observers who will make correct predictions. Cosmological theories will in general predict a set of possible universes, each with a probability. To calculate first-person probabilities, we argue that each possible universe should be weighted by the number of observer-instants in the specified subjective state that it contains. We also discuss Boltzmann brains, the humans/Jovians parable of Hartle and Srednicki, and the use of "old evidence".

We analyze a bulk effective field theory in AdS containing a U(1)-charged massive spin-2 field coupled to a gauge field, by performing the required holographic renormalization, and computing the one and two-point functions. We then compute the renormalized bulk three-point function involving two massive spin-2 fields and one gauge field. Matching with the CFT 3-point correlator of two non-conserved spin-2 operators and a conserved current, we obtain explicit mappings between the bulk minimal and gyromagnetic couplings and the boundary OPE data. Finally, we take the flat-space limit of the momentum space CFT correlator and verify that the resulting amplitude matches the expected flat-space structure.

We study pseudospectra of quasinormal frequencies and complex linear momenta of a U(1) gauge field in a Schwarzschild black branes in Anti-de Sitter. We present a novel approach for computing the pseudospectra which uses directly the gauge field variables and contrast it to a conventional master scalar field approach. Upon clarifying a subtlety in the energy norm of the master scalar we show that the pseudospectra of both approaches conincide. In the hydrodynamic regime we find that the hydrodynamic quasinormal frequency, the diffusive mode, is spectrally stable to a very good approximation. On the other hand hydrodynamic complex linear momenta show enhanced spectral instability as a consequence of a pole-collision at zero frequency.

Within the framework of scalar-tensor theories of gravity, certain models can evade classical black hole no-hair theorems. A well-known example is Einstein-scalar-Gauss-Bonnet gravity, where black holes carrying a scalar charge can exist. We find that, within this theory, binary black holes initially described by General Relativity can acquire scalar charges once they reach a critical orbital separation ("dynamical scalarization"). We develop a simple semi-analytic model, based on the adiabatic conservation of the total Wald entropy, to estimate the scalar charge evolution during the binary inspiral. We also run fully nonlinear numerical-relativity simulations for different configurations, finding consistent results. The gravitational-wave phase difference between Einstein-scalar-Gauss-Bonnet and General Relativity waveforms, which we use to assess detectability, is also computed. We find that dynamical scalarization might be observable in nearly equal-mass binary black hole mergers with third-generation ground-based gravitational-wave detectors, in a narrow range of the dimensional coupling of the theory.

Core-collapse supernovae (SNe) are sources of gravitational waves (GWs) produced by hydrodynamical instabilities and highly time-dependent anisotropies of the neutrino radiation. In this work we analyze both contributions to the GW signal for two state-of-the-art three-dimensional (3D) SN models computed with the Prometheus-Vertex neutrino-hydrodynamics code. In contrast to the far majority of models analyzed for GWs so far, our core-collapse simulations were started with 12.28 M_sun (18.88 M_sun) progenitors, whose final hour (7 min) of convective oxygen-shell burning was computed in 3D and featured a vigorous oxygen-neon shell merger. The corresponding large-scale asymmetries in the oxygen layer are conducive to buoyancy-aided neutrino-driven explosions. The models were continuously evolved in 3D from the pre-collapse evolution until 5.11 s (1.68 s) after the core bounce. The GW signals result from the well-known dynamical phenomena in the SN core such as prompt postshock convection, neutrino-driven convection, the standing accretion shock instability, proto-neutron star oscillations, and anisotropic ejecta expansion. They do not exhibit any new or specific features that can be unambiguously connected to the powerful pre-collapse activity in the progenitors, but we identify interesting differences compared to results in the literature. We also discuss measurement prospects by interferometers, confirming that GW signals from future Galactic SNe will be detectable with existing and next-generation experiments working in the frequency range f ~ 1-2000 Hz.

If a strong first-order phase transition takes place at sufficiently high rest-mass densities in the equation of state (EOS) modelling compact stars, a new branch will appear in the mass-radius sequence of stable equilibria. This branch will be populated by stars comprising a quark-matter core and a hadronic-matter envelope, i.e., hybrid stars, which represent ``twin-star'' solutions to equilibria having the same mass but a fully hadronic EOS. While both branches are stable to linear perturbations, it is unclear which of the twin solutions is the ``favoured'' one, that is, which of the two configurations is expected to be found in nature. We assess this point by performing a large campaign of general-relativistic simulations aimed at assessing the response of compact stars on the two branches to perturbations of various strength. In this way, we find that, independently of whether the stars populate the hadronic or the twin branch, their response is characterised by a critical-perturbation strength such that the star will oscillate on the original branch for subcritical perturbations and migrate to the neighbouring branch for supercritical perturbations while conserving rest-mass. Because the critical values are different for stars with the same rest-mass but sitting on either branch, it is possible to define as favoured the part of the branch that has the largest critical perturbation, thus correcting the common wisdom that stellar models on the twin branch are the favoured ones. Interestingly, we show that the binding energies on the two branches can be used to deduce without simulations which of the stellar configurations is more likely to be found in nature.

Future quantum networks will interconnect quantum systems with distinct functionalities, ideally over long distances via low-loss telecom optical fibers. Here, we realize a two-node hybrid network that directly connects an atomic single photon source to a solid-state quantum memory in the telecom C-band without the need of frequency conversion and external filtering. Both nodes exhibit state-of-the-art performance at 1530 nm: the source achieves a heralded auto-$g^{(2)}(0)$ = 0.031 at a photon rate of 46 kcps, and the memory a storage efficiency of 10.6% with high multimode capacity. We leverage the intrinsic tunability of both nodes to optimize spectral matching, enabling direct networking between the two: single-photon storage and retrieval for 1 $μ$s over up to 37 temporal modes across extended fibers of 10.6 km (metropolitan) and 49.2 km (laboratory) while preserving non-classicality. These results define a high-bandwidth source-memory link that operates natively in the telecom band, introducing a new paradigm for the design and scaling of hybrid quantum networks.

This letter proposes a new avenue for understanding the cosmological singularity. The standard cosmological model contains a generic initial singularity usually referred to as the {\em big bang}. Herein, we present a novel idea to extend the description of our Universe beyond this limit. The proposal relies on rewriting physics in a purely Riemannian, {\em i.e.} locally Euclidean, 4-dimensional space and the emergence of Lorentzian patches owing to the interaction of all matter fields to a clock field that is responsible for a signature change. If our universe is contained within one of these patches, the initial singularity is replaced by a smooth boundary on which the signature of the physical metric flips. In this letter, we first define the model and draw the necessary conditions on its arbitrary functions for solutions to exist. Next, we prove the existence of solutions that lead to an emergent universe with a primordial (almost) de Sitter phase. To finish, we discuss the consequences of this construction for the universe on scales much larger than our observable universe: a large ``Euclidean sea'' in which Lorentzian islands locally emerge and host an expanding universe potentially similar to ours. While speculative, this scenario has specific features that can be tested, and the present letter sets the basis for further phenomenological investigations.

A central problem in formulating a theory of quantum gravity is to determine the size and structure of the Hilbert space of black holes. Here we use a quantum dynamical Krylov complexity approach to calculate the Hilbert space dimension of a black hole in 2+1-dimensional Anti-de Sitter space. We achieve this by obtaining the spread of an initial thermofield double state over the Krylov basis. The associated Lanczos coefficients match those for chaotic motion on the $SL(2,\mathbb{R})$ group. By including non-perturbative effects in the path integral, which computes coarse-grained ensemble averages, we find that the complexity saturates at late times. The saturation value is given by the exponential of the Bekenstein-Hawking entropy. Our results introduce a new way to compute the Hilbert space dimension of complex interacting systems from the saturating value of spread complexity.

This report reviews key developments in Carrollian physics with an emphasis on their role in the emerging framework of holography in asymptotically flat spacetimes. We begin by introducing the Carrollian limit, understood as the non-relativistic contraction of the Poincaré group obtained by formally taking the speed of light to zero. The geometric structures associated with this limit are described and argued to arise naturally on null hypersurfaces, most notably on null infinity, as well as black hole and cosmological horizons. Building on this, we examine the relation between the Bondi-Metzner-Sachs symmetries governing asymptotically flat gravity and the conformal Carrollian symmetries. Explicit examples of Carrollian field theories are constructed by implementing the limit on well-known relativistic field theories, with particular attention to Carrollian CFTs. We then present the Carrollian holography proposal, according to which gravity in asymptotically flat spacetimes is dual to a Carrollian CFT living at null infinity in one lower dimension. In this framework, the massless $\mathcal{S}$-matrix written in position space at null infinity is naturally reinterpreted in terms of boundary Carrollian CFT correlators, called Carrollian amplitudes. We highlight their relation to celestial amplitudes and show how they naturally emerge from holographic CFT correlators through a correspondence between the flat space limit in the bulk and the Carrollian limit at the boundary. Using this correspondence, we provide strong evidence that flat space holography arises from a controlled and consistent limiting procedure applied to both sides of the AdS/CFT duality. We conclude by outlining future directions and open questions in the program.

The paper solves the problem of continuum limit in bosonic thermal coherent-state path integrals. For this purpose, exact discrete versions of the path integral are constructed for three different orderings of the Hamiltonian: normal, anti-normal and symmetric (Weyl order). Subsequently, their different continuum versions are checked on the harmonic oscillator, to choose the symmetric ordering as a possibly correct choice for all Hamiltonians. Spotted mathematical subtleties in the simple case serve as a clue to the general solution. Finally, a general justification for the symmetric order is provided by deriving the continuum path integral starting from the exact discrete case by a renormalization procedure in the imaginary time frequency domain. While the role of Weyl order has already been found, the paper provides the missing proof of its suitability for every Hamiltonian and simplifies the previously established construction by referring only to creation and annihilation operators (without position and momentum operators).

Dynamical correlations of quantum many-body systems are typically analyzed in the momentum space and frequency basis. However, quantum simulators operate more naturally in real space, real time settings. Here we analyze the real-space time-dependent van Hove spin correlations $G(r,t)$ of the 2D triangular antiferromagnet KYbSe$_2$ as obtained from high-resolution Fourier-transformed neutron spectroscopy. We compare this to $G(r,t)$ from five theoretical simulations of the well-established spin Hamiltonian. Our analysis reveals non-linear sub-ballistic low-temperature transport in KYbSe$_2$ which none of the current state-of-the-art numerical or field-theoretical methods reproduce. Our observation signals an emergent collective hydrodynamics, perhaps associated with the quantum critical phase of a quantum spin liquid, and provides an ideal benchmark for future quantum simulations.

There are three categories of mathematical results concerning quiescent big bang singularities: the derivation of asymptotics in a symmetry class; the construction of spacetimes given initial data on the singularity; and the proof of big bang formation in the absence of symmetries, including the proof of stable big bang formation. In a recent article, the first author demonstrated the existence of developments corresponding to a geometric notion of initial data on a big bang singularity. Moreover, this article, combined with previous articles by the second author, gives a unified and geometric perspective on large classes of seemingly disparate results in the first two categories. Concerning the third category, Oude Groeniger et al recently formulated a general condition on initial data ensuring big bang formation, including curvature blow up. This result, among other things, generalises previous results on stable big bang formation. However, it does not include a statement saying that the solutions induce initial data on the singularity. Here we tie all three categories of results together by demonstrating that the solutions of Oude Groeniger et al induce data on the singularity. However, the results are more general and can potentially be used to derive similar conclusions in other gauges.

Quantum Machine Learning (QML) is becoming increasingly prevalent due to its potential to enhance classical machine learning (ML) tasks, such as classification. Although quantum noise is often viewed as a major challenge in quantum computing, it also offers a unique opportunity to enhance privacy. In particular, intrinsic quantum noise provides a natural stochastic resource that, when rigorously analyzed within the differential privacy (DP) framework and composed with classical mechanisms, can satisfy formal $(\varepsilon, δ)$-DP guarantees. This enables a reduction in the required classical perturbation without compromising the privacy budget, potentially improving model utility. However, the integration of classical and quantum noise for privacy preservation remains unexplored. In this work, we propose a hybrid noise-added mechanism, HYPER-Q, that combines classical and quantum noise to protect the privacy of QML models. We provide a comprehensive analysis of its privacy guarantees and establish theoretical bounds on its utility. Empirically, we demonstrate that HYPER-Q outperforms existing classical noise-based mechanisms in terms of adversarial robustness across multiple real-world datasets.

We present the quantization of the electromagnetic field near the event horizon of a Schwarzschild black hole using Euclidean path integrals. Our result for the vacuum energy describes a black hole surrounded by a finite volume of photons at $T_{H} = \frac{1}{8πG M}$, the black hole quantum atmosphere. The total entropy includes contributions from this atmosphere, and the Bekenstein entropy, which arises from the Gibbons--Hawking--York boundary term, which encodes topological information. We show that the contribution of the quantum atmosphere to the black hole specific heat is positive, indicating that the system may become thermodynamically stable. By analyzing homology groups, we show that the black hole evaporation is a tunneling between topologically distinct spacetimes: Schwarzschild ($χ= 2)$ transitions to the flat spacetime ($χ= 1$) via Hawking radiation, where $χ$ is the Euler characteristic, a topological invariant. This process resembles instanton-driven tunneling in Yang-Mills theories, where topologically non-trivial solutions dominate the vacuum amplitude. In our case, the Gibbons--Hawking--York term dominates the transition amplitude, which induces the evaporation process. These results corroborate the Parikh-Wilczek picture of Hawking radiation and the interpretation of Euclidean black holes as gravitational instantons.

Quantum simulations of quantum chromodynamics (QCD) require a representation of gauge fields and fermions on the finitely many degrees of freedom available on a quantum computer. We introduce a truncation of lattice QCD coupled to staggered fermions that includes (i) a local Krylov truncation that generates allowed basis states; (ii) a maximum allowed electric energy per link; (iii) a limit on the number of fermions per site; and (iv) a truncation in the large N_c scaling of Hamiltonian matrix elements. Explicit truncated Hamiltonians for 1+1D and 2+1D lattices are given, and numerical simulations of string-breaking dynamics are performed.

The Overhauser effect (OE) and the Solid effect (SE) are two Dynamic Nuclear Polarization techniques. These two-spin techniques are widely used to create nonequilibrium nuclear spin states having polarization far beyond its equilibrium value. OE is commonly encountered in liquids, and SE is a solid-state technique. Here, we report a single framework based on a recently proposed quantum master equation, to explain both OE and SE. To this end, we use a fluctuation-regularized quantum master equation that predicts dipolar relaxation and drive-induced dissipation, in addition to the standard environmental dissipation channels. Importantly, this unified approach predicts the existence of optimal microwave drive amplitudes that maximize the OE and SE enhancements. We also report optimal enhancement regime for electron-nuclear coupling for maximal enhancement.

Collective excitations of Fractional Quantum Hall (FQH) liquids at long wavelengths are thought to be of a generally covariant geometric nature, governed by area-preserving diffeomorphisms ($\mathrm{SDiff}$). But current analyses rely solely on the corresponding perturbative $w_\infty$ Lie algebra. We argue this is insufficient: We identify a non-perturbative construction of the effective Maxwell-Chern-Simons quantum field theory which carries unitary $\mathrm{SDiff}$ equivariance. But this turns out to be non-differentiable, suggesting FQH excitation phenomenology beyond the $w_\infty$ algebra.

In this paper, we investigate the equilibrium configurations and the radial perturbations of neutron stars in a subclass of Scalar-Vector-Tensor (SVT) theories. By solving the generalised Tolman-Oppenheimer-Volkoff equations in SVT theories for several values of the modified gravity parameter, we examine the impact of the spontaneous scalarization of charged neutron star (NSs), which arises from the coupling of the scalar field to the electromagnetic tensor and double-dual Reimann tensor, $L^{μναβ}F_{μν}F_{αβ}$. Then we extend our study by deriving the action at quadratic order in linear perturbations of radial type and computing scalar quasinormal modes (QNMs)as well as the normal modes (NMs) showing the coincidence of stability and maximum mass points in generlar relativity (GR) is still present in this modified theory.

In this paper we study the problem of sampling random supersingular elliptic curves with unknown endomorphism rings. This task has recently attracted significant attention, as the secure instantiation of many isogeny-based cryptographic protocols relies on the ability to sample such ``hard'' curves. Existing approaches, however, achieve this only in a trusted-setup setting. We present the first provable quantum polynomial-time algorithm that samples a random hard supersingular elliptic curve with high probability.Our algorithm runs heuristically in $\tilde{O}\!\left(\log^{4}p\right)$ quantum gate complexity and in $\tilde{O}\!\left(\log^{13} p\right)$ under the Generalized Riemann Hypothesis. As a consequence, our algorithm gives a secure instantiation of the CGL hash function and other cryptographic primitives. Our analysis relies on a new spectral delocalization result for supersingular $\ell$-isogeny graphs: we prove the Quantum Unique Ergodicity conjecture, and we further provide numerical evidence for complete eigenvector delocalization; this theoretical result may be of independent interest. Along the way, we prove a stronger $\varepsilon$-separation property for eigenvalues of isogeny graphs than that predicted in the quantum money protocol of Kane, Sharif, and Silverberg, thereby removing a key heuristic assumption in their construction.

Spin defects in two-dimensional materials hold significant potential for quantum information technologies and sensing applications. The negatively charged boron vacancy (VB-) in hexagonal boron nitride (hBN) has attracted considerable attention as a quantum sensor due to its demonstrated sensitivity to temperature, magnetic fields, and pressure.1 However, its applications have thus far been limited by inherently dim photoluminescence (PL). By fabricating a van der Waals heterostructure with a sensitizing donor layer, lead iodide (PbI2), we effectively enhance the PL intensity from the VB- by 5-45x, while maintaining compatibility with other heterostructures and vdW optoelectronic platforms. The type-I band alignment at the heterojunction enables efficient exciton migration while suppressing back-electron transfer, and the strong spectral overlap between the PbI2 emission and defect absorption supports efficient fluorescence resonance energy transfer. Ab initio density functional theory (DFT) predicts a photon-ratcheting mechanism that boosts absorption and emission while maintaining magnetic resonance (ODMR) contrast through minimal hybridization. Experimentally, the heterostructure exhibits enhanced continuous-wave ODMR sensitivity and functions as a precise probe of external magnetic fields. This work establishes a proof-of-concept for amplifying weak defect signals in nanomaterials, highlighting a new strategy for engineering their optical and magnetic responses.

For the LISA mission, Tilt-To-Length (TTL) coupling is expected to be one of the dominant instrumental noise contributions after laser frequency noise is suppressed based, on assumptions on the size of the coupling and angular jitter levels. This work uses for the first time a closed-loop, non-linear, and time-varying dynamics implementation to simulate detailed angular jitters for the spacecraft and optical benches. In turn, this gives an improved expectation of the TTL contribution to the interferometric output. It is shown that the TTL coupling impact is limited given current estimates on the size of coupling coefficients. A time-domain Least Squares estimator is used to infer the TTL parameters from the simulated measurements. The bias and correlations limit the estimator in the case of regular datasets with amplified TTL coefficients to a relative error of $10\%$, but the subtraction of the TTL signal still works well. For lower readout noises, the estimation error diverges, which can be mitigated using a regularization term. Alternatively, using sinusoidal maneuvers improves the inference to a high accuracy of $0.1\%$ for TTL coefficients around the expected level, removing all correlations in the inferred parameters. This validates the maneuver design by Wegener et al. (2025) in this closed-loop setting.

Quantum theory encounters a difficulty when attempting to describe recording devices. If the recording is of events in which quantum uncertainty plays a role, such as an experiment on a quantum system, quantum theory is unable to correctly predict the probabilities of both future and past states of the recording. The nature of this difficulty will be laid out at the outset. A resolution then will be presented, in which the Born rule will be lightly amended so as to correctly predict all probabilities. The resolution will have the further benefit of clarifying how quantum theory applies to an array of situations in which the theory can be ambiguous, such as the descriptions of continuous measurements, and of closed systems containing all observers.

Digital quantum computing promises to offer computational capabilities beyond the reach of classical systems, yet its capabilities are often challenged by scarce quantum resources. A critical bottleneck in this context is how to load classical or quantum data into quantum circuits efficiently. Approximate quantum loaders (AQLs) provide a viable solution to this problem by balancing fidelity and circuit complexity. However, most existing AQL methods are either heuristic or provide guarantees only for specific input types, and a general theoretical framework is still lacking. To address this gap, here we reformulate most AQL methods into a unified framework and establish information-theoretic bounds on their approximation error. Our analysis reveals that the achievable infidelity between the prepared state and target state scales linearly with the total entanglement entropy across subsystems when the loading circuit is applied to the target state. In light of this, we develop AQER, a scalable AQL method that constructs the loading circuit by systematically reducing entanglement in target states. We conduct systematic experiments to evaluate the effectiveness of AQER, using synthetic datasets, classical image and language datasets, and a quantum many-body state datasets with up to 50 qubits. The results show that AQER consistently outperforms existing methods in both accuracy and gate efficiency. Our work paves the way for scalable quantum data processing and real-world quantum computing applications.

We construct a charged Bardeen black hole (BH) surrounded by perfect fluid dark matter (PFDM) and coupled to a cloud of strings (CS). The metric function combines the magnetic monopole charge from nonlinear electrodynamics, the PFDM logarithmic correction, and the CS parameter that renders the spacetime asymptotically non-flat. We analyze the horizon structure, identifying parameter ranges yielding non-extremal BHs, extremal configurations, and naked singularities. The null geodesics, photon sphere radius, and shadow are computed, revealing that both CS and PFDM enlarge the shadow. For neutral particle dynamics, we derive the specific energy, angular momentum, and innermost stable circular orbit location. Quasiperiodic oscillations (QPOs) are examined through the azimuthal, radial, and vertical epicyclic frequencies, where notably the azimuthal frequency is independent of the CS parameter. Scalar field perturbations governed by the Klein-Gordon equation yield an effective potential whose peak decreases with both parameters, yet the transmission and reflection probabilities respond oppositely to CS and PFDM variations. The greybody factor bounds are obtained using semi-analytical methods. Our results demonstrate that the distinct effects of $α$ and $β$ on various observables could allow independent constraints on these parameters through shadow measurements, QPO timing, and gravitational wave ringdown observations.

We propose a Trace-distance binary Tree AdaBoost (TTA) multi-class quantum classifier, a practical pipeline for quantum multi-class classification that combines quantum-aware reductions with ensemble learning to improve trainability and resource efficiency. TTA builds a hierarchical binary tree by choosing, at each internal node, the bipartition that maximizes the trace distance between average quantum states; each node trains a binary AdaBoost ensemble of shallow variational quantum base learners. By confining intrinsically hard, small trace distance distinctions to small node-specific datasets and combining weak shallow learners via AdaBoost, TTA distributes capacity across many small submodels rather than one deep circuit, mitigating barren-plateau and optimization failures without sacrificing generalization. Empirically TTA achieves top test accuracy ($\approx $100\%) among quantum and classical baselines, is robust to common quantum errors, and realizes aggregate systems with 10000 cumulative layers and 0.2M parameters, implemented as many shallow circuits. Our results are empirical and implementable on near-term platforms, providing a resource-efficient route to scalable multi-class quantum machine learning.

In this work, we investigate the observable optical signatures of the Mod(A)Max black hole spacetime. We analyze key optical features, including the photon sphere, black hole shadow, and photon trajectories, and examine how these observables depend on the underlying geometric parameters, such as the electric charge and the Mod(A)Max coupling parameter. We further study the dynamics of neutral test particles in the vicinity of the black hole by deriving the effective potential within the Hamiltonian formalism. Using this potential, we obtain the specific energy and specific angular momentum for test particles on circular orbits of fixed radius, as well as the innermost stable circular orbit (ISCO), and explore how the geometric parameters influence these quantities and the ISCO radius. Finally, we derive the epicyclic (azimuthal, radial, and vertical) frequencies to analyze quasi-periodic oscillations (QPOs) exploring how the geometric parameters influences these and discuss their physical implications.

In this work we develop a real-time Schwinger-Keldysh formulation of Krylov dynamics that treats Krylov complexity as an in-in observable generated by a closed time contour path integral. The resulting generating functional exposes an emergent phase-space description in which the Lanczos coefficients define an effective Hamiltonian governing operator motion along the Krylov chain. In the semiclassical limit, exponential complexity growth arises from hyperbolic trajectories, and asymptotically linear Lanczos growth appears as a universal chaotic fixed point, with sub-leading deformations classified as irrelevant, marginal or relevant. Going beyond the saddle, the Schwinger-Keldysh framework provides controlled access to fluctuations and large deviations of Krylov complexity, revealing sharp signatures of integrability-chaos crossovers that are invisible at the level of the mean. This formulation reorganises Krylov complexity into a dynamical field-theoretic framework and identifies new fluctuation diagnostics of operator growth in closed quantum systems.

In this work, the behavior, evolution, and expansion of the universe are investigated within a Lorentz-violating framework driven by Tsallis holographic dark energy. The cosmological extension is implemented through a spontaneously symmetry-breaking Bumblebee field, which is assumed to play a fundamental role in the dynamics of the universe. Estimates for key Lorentz-violating quantities are obtained, and the evolution of the Hubble parameter is analyzed from the early universe era to the present epoch. This formulation provides an alternative perspective on the Hubble tension.

An extension of standard quantum mechanics is proposed in which the Newtonian time parameter appearing in the unitary evolution operator is replaced with the time shown by a `quantum clock'. A quantum clock is defined by the following properties: (a) the time that the clock shows is non-decreasing, (b) the clock ticks at random with random tick sizes, and (c) on average the clock shows the Newtonian time. We show that the leading term in the evolution equation for the density matrix associated with any quantum clock model gives the von Neumann equation. Modifications to the von Neumann equation are worked out in detail in a parametric family of examples for which the tick sizes have a gamma distribution. The leading correction to the von Neumann equation is given by the Lindblad equation generated by the Hamiltonian, but there are higher-order terms that generalize the von Neumann equation and the Lindblad equation. Lower bounds on the parameters of these quantum clock models are derived by use of the precision limit of an atomic clock.

Proto-neutron stars (PNSs) are the hot, lepton-rich remnants of the core collapse supernovae, which go through a cooling phase and become cold, stable Neutron stars (NSs). Since PNSs are also superdense objects with strong gravitational fields, we can use them to probe general relativity (GR) in the high-curvature regime, similar to NSs. In this study, we analyze the macroscopic properties like mass, radius, compactness, tidal deformability, $f$-mode oscillations and gravitational binding energy of PNSs using four different relativistic mean-field (RMF) equations of state (EOSs) with fixed entropy per baryon ($S$ =1, 2) and varying the lepton fractions ($Y_l$). The variation of $S$ and $Y_l$ has a noticeable effect on these properties. Extending our study beyond GR, we explore these effects within the framework of Energy-Momentum Squared Gravity (EMSG). This modified gravity theory adds the squared energy-momentum terms to the field equations with a free parameter $α$. In the weak-field regimes, EMSG remains indistinguishable from GR, but in the strong-field regimes, such as PNSs or NSs, it shows measurable deviations. Varying the free parameter $α$, we observe significant changes in the macroscopic properties of the PNSs. After that, we focus on the universal relations of the macroscopic properties and the correlations of the universal relations. We find that, despite significant changes in the macroscopic properties induced by the variations of $S$, $Y_l$ and $α$, the correlations remain strong and nearly unaffected.

Some recent experiments claim to show that any model in which a quantum state represents mere information about an underlying physical reality of the system must make predictions which contradict those of quantum theory. The present work undertakes to investigate the issue of reality, treading a more fundamental route from the Hamilton-Jacobi equation of classical mechanics to the Schrodinger equation of quantum mechanics. Motivation for this is a similar approach from the eikonal equation in geometrical optics to the wave equation in electromagnetic theory. We rewrite the classical Hamilton-Jacobi equation as a wave equation and seek to generalise de Broglie's wave particle duality by demanding that both particle and light waves have the freedom of being described by any square-integrable function. This generalisation, which allows superposition also for matter wave functions, helps us to obtain the Schrodinger equation, whose solution can be seen to be as much objective as the classical mechanics wave function. Several other equations which one writes in quantum mechanics, including the eigenvalue equations for observables, series expansion of energy states in terms of eigenstates of observables other than energy, etc., can be written in the classical case too. Absence of any collapse of the wave function, entanglement, etc. in the classical realm have their origin in the nonlinearity of the classical wave equation. These considerations indicate that many of the puzzles in quantum mechanics are present also in classical mechanics in a dormant form, which fact shall help to demystify quantum mechanics to a great extent.

Symmetry underlies many of the most effective classical and quantum learning algorithms, yet whether quantum learners can gain a fundamental advantage under symmetry-imposed structures remains an open question. Based on evidence that classical statistical query ($\SQ$) frameworks have revealed exponential query complexity in learning symmetric function classes, we ask: can quantum learning algorithms exploit the problem symmetry better? In this work, we investigate the potential benefits of symmetry within the quantum statistical query ($\QSQ$) model, which is a natural quantum analog of classical $\SQ$. Our results uncover three distinct phenomena: (i) we obtain an exponential separation between $\QSQ$ and $\SQ$ on a permutation-invariant function class; (ii) we establish query complexity lower bounds for $\QSQ$ learning that match, up to constant factors, the corresponding classical $\SQ$ lower bounds for most commonly studied symmetries; however, the potential advantages may occur under highly skewed orbit distributions; and (iii) we further identify a tolerance-based separation exists, where quantum learners succeed at noise levels that render classical $\SQ$ algorithms ineffective. Together, these findings provide insight into when symmetry can enable quantum advantage in learning.

Domain-specialized FPGAs have delivered unprecedented performance for low-latency inference across scientific and industrial workloads, yet nearly all existing accelerators assume static models trained offline, relegating learning and adaptation to slower CPUs or GPUs. This separation fundamentally limits systems that must operate in non-stationary, high-frequency environments, where model updates must occur at the timescale of the underlying physics. In this paper, I argue for a shift from inference-only accelerators to ultrafast on-chip learning, in which both inference and training execute directly within the FPGA fabric under deterministic, sub-microsecond latency constraints. Bringing learning into the same real-time datapath as inference would enable closed-loop systems that adapt as fast as the physical processes they control, with applications spanning quantum error correction, cryogenic qubit calibration, plasma and fusion control, accelerator tuning, and autonomous scientific experiments. Enabling such regimes requires rethinking algorithms, architectures, and toolflows jointly, but promises to transform FPGAs from static inference engines into real-time learning machines.

Scalable quantum-classical embedding is essential for chemically meaningful simulations on near-term NISQ hardware. Using QDFT, we show systematic recovery of correlation energy relative to the DFT baseline, benchmarked against CCSD in a fixed six-orbital active space across molecules ranging from water to naphthalene. By varying the number of embedded electrons from 2 to 8, aromatic systems saturate near 63-64 percent, while linear molecules such as carbon dioxide reach 68 percent. All systems converge within two embedding iterations under relaxed self-consistency thresholds, highlighting the robustness of the approach. A (4e,6o) active space recovers approximately 60 percent correlation using 10 qubits, providing practical guidelines for NISQ-era simulations.

We study shear fluctuations in non-relativistic holographic systems coupled to torsional Newton-Cartan geometry, using asymptotically Lifshitz spacetimes in Einstein-Maxwell-dilaton gravity. We identify a universal subdiffusive shear mode characterized by the quartic dispersion relation $ω=-iD_4 k^4$, in sharp contrast to the conventional hydrodynamic diffusion. We derive this result analytically through a systematic higher-order matched asymptotic expansion connecting near-horizon and far-region solutions, and we verify it with direct numerical quasinormal mode calculations. Our numerics demonstrate that the first non-hydrodynamic mode is purely imaginary and gapped, following the dispersion relation $ω=-iω_0-i D k^2$, and that both the hydrodynamic and the first non-hydrodynamic modes pass through pole-skipping points. These results highlight Lifshitz holography as an efficient framework for anomalous transport in strongly coupled non-relativistic quantum matter.

Quasiparticle tunneling events are a source of decoherence and correlated errors in superconducting circuits. Understanding and ultimately mitigating these errors calls for real-time detection of quasiparticle tunneling events on individual devices. In this work, we simultaneously detect quasiparticle tunneling events in two co-housed, charge-sensitive transmons coupled to a common waveguide. We measure background quasiparticle tunneling rates at the single-hertz level, with temporal resolution of tens of microseconds. Using time-tagged coincidence analysis, we show that individual events are uncorrelated across devices, whereas burst episodes occur about once per minute and are largely correlated. These bursts have a characteristic lifetime of 7 ms and induce a thousand-fold increase in the quasiparticle tunneling rate across both devices. In addition, we identify a rarer subset of bursts which are accompanied by a shift in the offset charge, at approximately one event per hour. Our results establish a practical and extensible method to identify quasiparticle bursts in superconducting circuits, as well as their correlations and spatial structure, advancing routes to suppress correlated errors in superconducting quantum processors.

Exceptional points in non-Hermitian quantum systems give rise to novel genuine quantum phenomena. Recent explorations of exceptional-point-induced quantum phase transitions have extended from discrete-variable to continuous-variable-encoded quantum systems. However, quantum phase transitions driven by Liouvillian exceptional points (LEPs) in continuous-variable platforms remain largely unexplored. Here, we construct and investigate a Liouvillian exceptional structure based on a driven-dissipative Kerr-cat qubit. Through numerical simulations, we reveal a quantum phase transition occurring at the LEP characterized by a sudden change in dynamical behavior from underdamped oscillations to overdamped relaxations as visualized via Wigner functions and Bloch sphere trajectories. Notably the negativity of the Wigner function serves as a direct signature of genuine quantum coherence unattainable in conventional single-qubit non-Hermitian systems. Furthermore, we introduce the phase difference between the off-diagonal elements of the Liouvillian eigenmatrices as a novel parameter to quantify the transition. Our results establish the Kerr-cat qubit as a novel continuous-variable setting for exploring dissipative quantum criticality and intrinsic non-Hermitian physics.

Important quantum algorithm routines allow the implementation of specific quantum operations (a.k.a. gates) by combining basic quantum circuits with an iterative structure. In this structure, the number of repetitions of the basic circuit pattern is associated to convergence parameters. This iterative structure behaves similarly to function approximation by series expansion: the higher the truncation order, the better the target gate (i.e. operation) approximation. The asymptotic convergence of the gate error with respect to the number of basic pattern repetitions is known. It is referred to as the query complexity. The underlying convergence law is bounded, but not in an explicit fashion. Upper bounds are generally too pessimistic to be useful in practice. The actual convergence law contains constants that depend on the joint properties of the matrix encoded by the query and the initial state vector, which are difficult to compute classically. This paper proposes a strategy to study this convergence law and extract the associated constants from the gate (operation) approximation at different accuracy (convergence parameter) constructed directly on a Quantum Processing Unit (QPU). This protocol is called Numerical Error Extraction by Quantum Measurement Algorithm (NEEQMA). NEEQMA concepts are tested on specific instances of Quantum Signal Processing (QSP) and Hamiltonian Simulation by Trotterization. Knowing theexact convergence constants allows for selecting the smallest convergence parameters that enable reaching the required gate approximation accuracy, hence satisfying the quantum algorithm's requirements.

In quantum gases, weak links are typically realized with externally imposed optical potentials. We show that, in rotating binary condensates, quantized vortices in one component form hollow channels that act as self-induced weak links for the other, enabling superflow through otherwise impenetrable, phase-separated domains. This introduces a novel barrier mechanism: quantum pressure creates an effective barrier inside the vortex channel, set by the constriction width, which controls the superflow. Tuning the interspecies interaction strength drives a crossover from the hydrodynamic transport to Josephson tunneling regime. Long-range dipolar interactions further tune the weak-link properties, enabling both short links and two coupled junctions in series. Circuit models quantitatively capture the dc current-phase relations for both configurations. These results establish vortices as reconfigurable, interaction-controlled Josephson elements in superfluids.

Long-range Rydberg molecules (LRMs) form when a highly excited Rydberg electron scatters from ground-state atoms inside its orbit, creating oscillatory, long-range potentials. We present a combined theoretical and experimental study of caesium dimers correlated to 402P3/2 Rydberg states, with an emphasis on decay via autoionisation (associative ionisation). Our model includes a relativistic treatment of electron-atom scattering with spin-orbit coupling, the perturber's hyperfine structure, and coupling of vibrational levels to a continuum of short-range decay channels. Calculated potential-energy curves predict two families of wells: outer wells near the classical outer turning point supporting long-lived states, and inner wells at shorter range whose lifetimes are limited by tunneling and subsequent vibronic decay. Using photoassociation in an ultracold Cs gas and an analysis of pulsed-field-ionisation signals which are highly selective for the detection of molecules, we assign resonances by binding energy and measure lifetimes. The measured lifetimes of inner-well states increase systematically with increasing detuning and agree with calculated lifetimes; detection of Cs2+ product ions supports autoionisation as a dominant channel. We show that the lifetimes are strongly reduced by spin-orbit interactions in the transient Cs-collision complex, which lift the near-degeneracy in Omega observed for states in the outer well and control the inner-well binding. The identified states also provide promising pathways to create ultracold molecules in ion-pair states.

Bosonic counting problems can be framed as estimation tasks of matrix functions such as the permanent, hafnian, and loop-hafnian, depending on the underlying bosonic network. Remarkably, the same functions also arise in spin models, including the Ising and Heisenberg models, where distinct interaction structures correspond to different matrix functions. This correspondence has been used to establish the classical hardness of simulating interacting spin systems by relating their output distributions to #P-hard quantities. Previous works, however, have largely been restricted to bipartite spin interactions, where transition amplitudes, which provide the leading-order contribution to the output probabilities, are proportional to the permanent. In this work, we extend the Ising model construction to arbitrary interaction networks and show that transition amplitudes of the Ising Hamiltonian are proportional to the hafnian and the loop-hafnian. The loop-hafnian generalizes both the permanent and hafnian, but unlike these cases, loop-hafnian-based states require Dicke-like superpositions, making the design of corresponding quantum circuits non-trivial. Our results establish a unified framework linking bosonic networks of single photons and Gaussian states with quantum spin dynamics and matrix functions. This unification not only broadens the theoretical foundation of quantum circuit models but also highlights new, diverse, and classically intractable applications.

We study the properties of the principal symbol of the 3+1 equations of motion in Teleparallel Equivalent of General Relativity (TEGR) and assess the conditions for hyperbolicity. We use the Hamiltonian formulation based on the vectorial, antisymmetric, symmetric trace-free, and trace (VAST) decomposition of the canonical variables in the Hamiltonian formalism, and the Hamilton's equations previously presented in the literature. We study the system of differential equations at the linear level, and show that the principal symbol has a sector with imaginary eigenvalues, which renders the system not hyperbolic. This situation persists by taking spatial derivatives in either one or three coordinate directions, and it should be interpreted as a problem of the specific gauge choice instead of a general problem with TEGR. The first practical use of Hamilton's equations in this work can be extended for proving well-posedness in spherical symmetry, and establish numerical relativity setups in TEGR.

We examine non-singular bounce cosmology within the framework of a phantom scalar field coupled to the Gauss-Bonnet term in both non-viscous and bulk-viscous cases. Using the scale factor ansatz $α(t)=\left(\fracαη+t^2\right)^{\frac{1}{2 η}}$, we reconstruct the scalar field potential $V(t)$, and observe a smooth potential well centered at the bounce point. The resulting energy density, pressure, and equation-of-state parameter show NEC violation necessary for successful bounce, while viscosity controls post-bounce dynamics with a positive and smooth squared speed of sound. In contrast, for the non-viscous model, sharp divergences occur just at the bounce and continues to be negative in the expanding phase, which in turn emphasises the stabilising role of dissipative effects. The energy condition analysis indicates a temporary NEC and SEC violation in the viscous scenario, whereas its persistent violation within the non-viscous model suggests a continuous accelerated expansion. Observational viability is found through Bayesian MCMC fitting in regards to the Pantheon+ supernova data, with best-fit parameters providing a reduced chi-squared of $χ_{red}^2 =0.995$ while the inflation observables derived from the reconstructed potential place our model predictions inside $68\%$ CL Planck 2018 confidence contours. Our findings suggest that bounce cosmologies could offer a physically reasonable and observationally acceptable alternative or pre-inflationary scenario, while highlighting the role that viscosity could play for a stable and smooth cosmological evolution.

Determining the position of an entity is a fundamental prerequisite for nearly all activities. Classical means, however, have been proven incapable of providing secure position verification, meaning that a prover can mislead verifiers about its actual position. In this work, we propose and experimentally realize a secure position-verification protocol that leverages quantum optics and relativity within an information-theoretic framework. Using phase-randomized weak coherent states, two verifiers separated by 2 km securely verify the prover's position with an accuracy better than 75 meters. These results establish secure position-based authentication as a practical possibility, paving the way for applications in financial transactions, disaster response, and authenticated secure communications.

This paper investigates the quantization of the Schwarzschild--de Sitter (SdS) black hole (BH) using the Misner--Sharp--Hernandez (MSH) mass as the internal energy in a reduced phase space framework. After introducing the canonical variables of the reduced phase space, we derive a discrete spectrum for the surface areas of the BH event horizon (EH) as well as MSH masses. We utilized the MSH mass spectrum to obtain the entropy of the BH. The entropy of the BH and cosmic EHs reveals a logarithmic correction to the Bekenstein--Hawking term. Our results support the robustness of the logarithmic form of quantum corrections in SdS thermodynamics.

We investigate the dissipative dynamics of a two-level atom in a weak gravitational field. Using the Feynman--Vernon influence functional formalism, we derive a quantum master equation describing the two-level atom interacting with a scalar field in a Newtonian gravitational field, and compute the energy dissipation rate of the atom. We find that the spontaneous emission rate (the dissipation rate in vacuum) is modified by the gravitational field. Specifically, this modification depends on the atom's dipole, the position of the atom relative to the source of the gravitational field, and the frequency of the scalar radiation emitted by the atom. Furthermore, we identify the parameter regimes in which the spontaneous emission rate is enhanced or suppressed by gravity. We also discuss how the modification arises from time dilation and dipole radiation in a weak gravitational field. These findings provide a theoretical basis for exploring gravitational effects in open quantum systems.

Product-formula (PF) based quantum simulation is a promising approach for simulating quantum systems on near-term quantum computers. Achieving a desired simulation precision typically requires a polynomially increasing number of Trotter steps, which remains challenging due to the limited performance of current quantum hardware. To alleviate this issue, post-processing techniques such as the multi-product formula (MPF) have been introduced to suppress algorithmic errors within restricted hardware resources. In this work, we propose a dual-channel multi-product formula that achieves a two-fold improvement in Trotter error scaling. As a result, our method enables the target simulation precision to be reached with approximately half the circuit depth compared to conventional MPF schemes. Importantly, the reduced circuit depth directly translates into lower physical error mitigation overhead when implemented on real quantum hardware. We demonstrate that, for a fixed CNOT count as a measure of quantum circuit, our proposal yields significantly smaller algorithmic errors, while the sampling error remains essentially unchanged.

Coulomb-Sturmian functions are complete, orthonormal, and include the full spectrum of continuum states. They are restricted to integer values of quantum numbers, as imposed by boundary and orthonormality conditions. Bagci-Hoggan exponential-type orbitals remove this restriction through a generalization to quantum number with fractional order. The differential equations for N-dimensional Bagci-Hoggan orbitals are derived. It is demonstrated that Coulomb-Sturmian functions satisfy a particular case of these equations. Additionally, Guseinov's Psi-alpha-ETOs are identified as N-dimensional Coulomb-Sturmians with a shifted dimensional parameter alpha, rather than representing an independent complete orthonormal sets of basis in a weighted Hilbert space.

Optimal control is a central problem in quantum thermodynamics. While control theories in the rapid-driving and slow-driving limits have been developed, to the best of our knowledge there is no general optimization method applicable to intermediate timescales. We introduce an optimal-control framework to minimize dissipated work and work variance, defined via the two-point measurement scheme, in open quantum systems governed by time-dependent Lindblad master equations. By introducing an auxiliary operator, we convert the history-dependent work variance into a time-local integral, enabling efficient gradient-based optimization beyond slow or rapid driving regimes. Applying our method, we find that in the coherent spin-boson model the optimized protocol can switch discontinuously between distinct locally optimal solutions as the relative weight between dissipation and fluctuations is varied. Moreover, for a single-level quantum dot coupled to a fermionic reservoir, the optimized fluctuation-minimizing protocol develops a qualitatively different multi-step structure that is not captured by approaches based on slow- or rapid-driving limits.

Computing electronic structures of quantum systems is a key task underpinning many applications in photonics, solid-state physics, and quantum technologies. This task is typically performed through iterative algorithms to find the energy eigenstates of a Hamiltonian, which are usually computationally expensive and suffer from convergence issues. In this work, we develop and implement the Quantum Jacobi-Davidson (QJD) method and its quantum diagonalization variant, the Sample-Based Quantum Jacobi-Davidson (SBQJD) method, and demonstrate their fast convergence for ground state energy estimation. We assess the intrinsic algorithmic performance of our methods through exact numerical simulations on a variety of quantum systems, including 8-qubit diagonally dominant matrices, 12-qubit one-dimensional Ising models, and a 10-qubit water molecule (H$_2$O) Hamiltonian. Our results show that both QJD and SBQJD achieve significantly faster convergence and require fewer Pauli measurements compared to the recently reported Quantum Davidson method, with SBQJD further benefiting from optimized reference state preparation. These findings establish the QJD framework as an efficient general-purpose subspace-based technique for solving quantum eigenvalue problems, providing a promising foundation for sparse Hamiltonian calculations on future fault-tolerant quantum hardware.

In finite-dimensional quantum systems, temperature cannot be uniquely defined. This, in turn, implies that there are several ways to define entropy production in finite-dimensional quantum systems, because the classical entropy production depends on temperature. We propose a unified definition of entropy production based on the difference in quantum relative entropy with respect to reference states characterized by effective temperatures. We demonstrate that the proposed definition naturally decomposes into a Clausius-type entropy production and an additional contribution arising from the time dependence of the effective temperature. Furthermore, we show that requiring the entropy production rate to take the conventional form as the sum of the entropy change and the heat flow constrains the effective temperature to be either constant or equal to a specific energy-matching effective temperature. For general initial states, entropy production can become negative, in which case we derive lower bounds on entropy production and establish sufficient conditions for its non-negativity using the trace distance.

The short-duration gravitational-wave (GW) event GW231123 has inferred component masses in the pair-instability mass gap and exhibits a burst-like morphology with no clearly inspiral, making it an interesting target for tests beyond the standard binary black hole (BBH) interpretation. In this work, motivated by its phenomenological similarity to GW190521, we test whether GW231123 is compatible with a wormhole-echo scenario by modeling a leading echo pulse with a well-motivated phenomenological sine-Gaussian wavepacket. We perform Bayesian model comparison against a BBH baseline described by the IMRPhenomXPHM-SpinTaylor waveform, and obtain the Bayes factor ratio $\ln B^{\rm Echo}_{\rm BBH} = 1.87$, corresponding to weak-to-moderate support for the echo hypothesis. In our previous analysis for GW190521 within the same overall framework, we found $\ln B^{\rm Echo}_{\rm BBH} \approx -2.9$, implying a shift of $Δ\ln B \approx 4.8$ between the two events. This sign change indicates that GW231123 is more compatible with a single-pulse echo description than GW190521.

In this note, we show that the conical solution-operator method of Mao-Tao in [Localized initial data for Einstein equations] applies to a simple construction of vacuum asymptotically flat initial data at minimal and borderline decay thresholds, corresponding to the global and exterior stability of Minkowski spacetime proved by the first named author in [Global stability of Minkowski spacetime with minimal decay] and [Exterior stability of Minkowski spacetime with borderline decay].

In recent years, quantum computing technologies have steadily matured and have begun to find practical applications across various domains. One important area is network communication security, where Quantum Key Distribution (QKD) enables communicating parties to establish a shared secret that can then be used to generate symmetric keys for subsequent encryption and decryption. This study focuses on implementing and comparing two well-known QKD protocols, namely BB84 and E91, within an actual quantum computing environment. It also proposes the use of SX gate operations to generate uniform quantum superposition states. By leveraging the properties of quantum superposition and quantum entanglement, the study illustrates how communicating parties can securely obtain a shared secret while preventing adversaries from intercepting it. The experiments are conducted using the IBM Quantum Platform to demonstrate the feasibility of the BB84 and E91 protocols on actual quantum hardware. The evaluation considers several metrics, including entropy, Independent and Identically Distributed (IID), and error-rate verifications.

In a previous work we were able to define a non-additive measure that can be used to represent both classical and quantum states in physics. We further extended this idea to work on a generic space of statistical ensembles (i.e. an ensemble space) in a way that connects to Choquet theory. The question of which non-additive integral is suitable to generalize the notion of expectation value remains open. In this paper we show that the Sugeno and Choquet integrals are not suitable.

Satellite-based quantum communications represent a critical advancement in the pursuit of secure, global-scale quantum networks. Leveraging the principles of quantum mechanics, these systems offer unparalleled security through Quantum Key Distribution (QKD) and other quantum communication protocols. This review provides a comprehensive overview of the current state of satellite-based quantum communications, focusing on the evolution from terrestrial to space-based systems. We explore the distinct advantages and challenges of discrete-variable (DV) and continuous-variable (CV) quantum communication technologies in the context of satellite deployments. The paper also discusses key milestones such as the successful implementation of quantum communication via the Micius satellite and outlines the primary challenges, including atmospheric turbulence and the development of quantum repeaters, that must be addressed to achieve a global quantum internet. This review aims to consolidate recent advancements in the field, providing insights and perspectives on the future directions and potential innovations that will drive the continued evolution of satellite-based quantum communications.

This paper presents an efficient and scalable tensor network framework for quantum kernel circuit simulation, alleviating practical costs associated with increasing qubit counts and data size. The framework enables systematic large-scale evaluation of a linearly mixed quantum-classical dual kernel of up to 784 qubits. Using Fashion-MNIST, the classification performance of the test dataset is compared between a classical kernel, a quantum kernel, and the quantum-classical dual kernel across the feature dimensions from 2 to 784, with a one-to-one mapping between encoded features and qubits. Our result shows that the quantum-classical dual kernel consistently outperforms both single-kernel baselines, remains stable as the dimensionality increases, and mitigates the large-scale degradation observed in the quantum kernel. Analysis of the learned mixing weights indicates that quantum contributions dominate below 128 features, while classical contributions become increasingly important beyond 128, suggesting that the classical kernel provides a stabilizing anchor against concentration effects and hardware noise while preserving quantum gains at lower dimensions.

Current observations indicate that an inverse exponential form of the inflaton potential provides an excellent description of single-field inflation. This potential fits the SPA$+$BK$+$DESI data sets well with in the $1σ$ bound in the $n_{\rm s}$-$r$ plane, thereby offering a simple and observationally viable single field inflationary scenario. To describe post-inflationary evolution and reheating, we extend the inverse exponential potential by adding a steep exponential term that remains negligible during inflation but becomes important afterwards. The resulting full potential develops a minimum after the end of inflation, leading to oscillations of the scalar field and consequently reheating of the Universe. We find that the maximum reheating temperature attainable in this scenario is of order $10^{13}\,\mathrm{GeV}$. The inverse exponential potential therefore emerges as a compelling candidate for early-Universe inflation, combining theoretical simplicity with robust observational viability.

The direct detection of gravitational waves (GWs) by LIGO has strikingly confirmed general relativity (GR), but testing GR via GWs requires estimating parameterized post-Einsteinian (ppE) deviation parameters in waveform models. Traditional Bayesian inference methods like Markov chain Monte Carlo (MCMC) provide reliable estimates but suffer from prohibitive computational costs, failing to meet the real-time demands and surging data volume of future GW detectors. Here, we propose a factorized neural posterior estimation framework: we construct independent normalizing flow models for each of the nine ppE deviation parameters and effectively integrate prior information from other source parameters via a conditional embedding network. Leveraging a hybrid neural network with a convolutional neural network and a Residual Neural Network for feature extraction, our method performs rapid and statistically reliable posterior inference directly from binary black hole signals. Compared to conventional MCMC, our approach achieves millisecond-scale inference time with a speedup factor of $9 \times 10^4$. Comprehensive validations show that the posterior estimates pass the Kolmogorov-Smirnov test and achieve empirical coverage probabilities close to theoretical targets. This work demonstrates the great potential of deep learning for GW parameter estimation and provides a viable technical solution for real-time GR tests with next-generation detectors.

I review the formalism, Feynman rules, and combinatorics that constrain a field to propagate ``classically", strictly in tree diagrams, either by itself, or interacting with other, purely quantum fields. The perturbation theory is reorganized by virtue of the linear terms that introduce the constraints via Lagrange multipliers, generalizing and giving results that cannot be obtained with the standard procedures which start at the quadratic terms. I apply the formalism to a theory of an $O(N)$-symmetric quantum field interacting with a ``classical" scalar field via cubic interactions in six spacetime dimensions. Using the renormalization group, I examine the effective potential, symmetry breaking with radiative corrections, the fixed points in $d=6-ε$ dimensions, and compare with other works. Other possible generalizations and applications of the formalism are also discussed.

The 21 cm signal from neutral hydrogen atom is almost the only way to directly probe the Dark Ages. The Dark Ages 21 cm signal, observed at frequencies below 50 MHz, can serve as a powerful probe of cosmology, as the standard cosmological model predicts a well-defined 21 cm spectral shape. In this work, we assess the detectability and model-selection power of 21 cm observations assuming physically motivated foregrounds, optimistic error levels, and several observing strategies for the signals predicted in various cosmological models. Using a Bayesian evidence-based comparison, we find that wide-band observations covering 1-50 MHz can identify the evidence of non-zero 21 cm signals from models considered in this paper except the one with a smooth spectrum that peaks at lower frequencies. In particular, observations below 15 MHz are essential to avoid degeneracies with the foreground. Furthermore, even with observations measured at 5 MHz intervals over the frequency range 1-50 MHz, the 21 cm signal can be identified if the errors are sufficiently small. This indicates that the intrinsic 21 cm spectral shape can be captured without foreground degeneracy even with a limited number of frequency channels.

In this manuscript, we study the quench dynamics of a transverse-field XY model starting from coherent Gibbs states. The results reveal that the initial strength of magnetization plays a crucial role in the emergence of dynamical quantum phase transitions. In concrete terms, when quenching within the same phase, through the properties of observables such as Fisher zeros and magnetization, we show that the stronger the initial magnetization, the more difficult the emergence of dynamical quantum phase transitions. The underlying mechanism is that the strong initial magnetization provides a directional effect, which inhibits the spin flipping in the process of quantum quench, making the dynamical quantum phase transition difficult to emerge. Since dynamical quantum phase transitions can be experimentally realized in various artificial systems, we hope that the physics predicted here can be experimentally verified in tabletop platforms.

According to the island formula, information in the code subspace defined in the black hole interior is embedded in the Hawking radiation after the Page time. At first sight, this embedding suggests that operations acting on the Hawking radiation could modify the information in the code subspace, potentially leading to an apparent violation of causality. Indeed, in previous studies based on the PSSY model, which incorporates only the topological degrees of freedom of gravity, it was shown that when the error is sufficiently large, a violation of causality can arise, as indicated by a nonvanishing mutual information. In this paper, we investigate the situation in which dynamical gravity also acts on the Hawking radiation. In this case, operations on the Hawking radiation induce nontrivial backreaction on the bulk spacetime appearing in the gravitational path integral for the mutual information -- an effect that is absent when the Hawking radiation is non-gravitating. We find that this backreaction renders the relevant mutual information vanishing. This result implies that, in theories with dynamical gravity, the apparent violation of causality is resolved.

We present a unified information-theoretic framework to analyze the generalization performance of differentially private (DP) quantum learning algorithms. By leveraging the connection between privacy and algorithmic stability, we establish that $(\varepsilon, δ)$-Quantum Differential Privacy (QDP) imposes a strong constraint on the mutual information between the training data and the algorithm's output. We derive a rigorous, mechanism-agnostic upper bound on this mutual information for learning algorithms satisfying a 1-neighbor privacy constraint. Furthermore, we connect this stability guarantee to generalization, proving that the expected generalization error of any $(\varepsilon, δ)$-QDP learning algorithm is bounded by the square root of the privacy-induced stability term. Finally, we extend our framework to the setting of an untrusted Data Processor, introducing the concept of Information-Theoretic Admissibility (ITA) to characterize the fundamental limits of privacy in scenarios where the learning map itself must remain oblivious to the specific dataset instance.

Sample-based quantum diagonalization (SQD) offers a powerful route to accurate quantum chemistry on noisy intermediate-scale quantum (NISQ) devices by combining quantum sampling with classical diagonalization. Here we introduce HSQD, a novel half-qubit SQD approach that halves the qubit requirement for simulating a chemical system and drastically reduces overall circuit depth and gate counts, suppressing hardware noise. When modeling the dissociation of the nitrogen molecule with a (10e, 26o) active space, HSQD matches the accuracy of SQD on IBM quantum hardware using only half the number of qubits and 40% fewer measurements. We further enhance HSQD with a heat-bath configuration interaction (HCI) inspired selection of the samples, forming HCI-HSQD. This yields sub-millihartree accuracy across the N2 potential energy surface and produces subspaces up to 39% smaller than those from classical HCI, showing a significant improvement in the compactness of the ground-state representation. Finally, we demonstrate the scalability of HCI-HSQD using iron-sulfur clusters, reaching active spaces of up to (54e, 36o) while using only half as many qubits as the original SQD. For these systems, HCI-HSQD reduces SQD energy errors by up to 76% for [2Fe-2S] and 26% for [4Fe-4S], while also reducing subspace sizes, halving measurement requirements, and eliminating expensive post-processing. Together, these results establish half-qubit SQD as a noise-resilient and resource-efficient pathway toward practical quantum advantage in strongly correlated chemistry.

We extend shapelets for the analysis of galaxy images to be available in a phase space, introducing \textit{Wigner Function Shapelets (WFS)}. Whereas conventional shapelets expand images separately in configuration or Fourier space using Hermite-Gaussian or Laguerre-Gaussian modes, WFS represents images directly in the four-dimensional phase space with symplectic group $\mathrm{Sp}(4,\mathbb{R})$, which is quantised by a phase-space cell $2π\lambdabar$ that determines a resolution limit of a telescope. WFS consists of a bilinear form of the cross-Wigner function of the Laguerre-Gaussian modes as an orthogonal and complete basis for the Wigner function of an image, carrying out $\mathrm{SU}(2)$ irreducible representations of the phase space with the Hopf tori. We introduce a scalar function $\mathcal{W}_{k\ell} (Q_0,Q_2)$ from the $\mathrm{U}(1)\times \mathrm{U}(1)$ - covariant tori to a two-dimensional space of constants of motion $(Q_0,Q_2)$ -- the harmonic energy and axial angular momentum -- thereby yielding a natural phase-space ``band structure'', given a pair of winding number $(k,\ell) \in \mathbb{Z}^2$. % WFS leverage key properties of the Wigner function for image analysis: (i) it encodes full information of an image in a symmetry-preserving way; (ii) its trasport equation naturally involves with a Liouville equation at $\lambdabar \rightarrow 0$; (iii) it admits positive/negative oscillatory patterns on $(Q_0,Q_2)$ plane that can be sensitive spatial coherent structure of galaxy morphology and cosmological imprints; and (iv) systematics and noise can be manipulated as a quantum channel operation. This paper aims to bring all the formulae related to the Wigner function in the context of astrophysics and cosmology, formally organising in both terminologies of astronomy and of quantum information theory.

Decoherence is strongly influenced by environmental criticality, with conventional Hermitian critical points universally enhancing the loss of quantum coherence. Here we show that this paradigm is fundamentally altered in non-Hermitian environments. Focusing on qubits coupled to non-Hermitian spin chains and interacting ultracold Fermi gases, we find that approaching exceptional points can either enhance or strongly suppress decoherence, depending on the balance between Hermitian and non-Hermitian system-environment couplings. In particular, when these couplings are comparable, decoherence is dramatically suppressed at exceptional transitions. We trace this behavior to the distinct response of the environmental ground state near non-Hermitian degeneracies and demonstrate the robustness of the effect across multiple models. Finally, we show that the predicted suppression of decoherence is directly observable on current digital quantum simulation platforms. Our results establish exceptional points as a concrete mechanism for suppressing decoherence and identify non-Hermitian criticality as a new avenue for coherence control in open quantum systems and quantum technologies.

We uncover the physical nature of the electric and magnetic monopoles discovered by McGuire and Ruffini on Misner strings accompanying charged nutty black holes, showing that these strings carry singular, nonuniform flows of electric and magnetic fields. These fields inevitably have nonzero divergence, thereby simulating the effective electric and magnetic charge densities along the strings. The latter create a complex short-range electromagnetic hair zone around the horizon, making the combined Misner-Dirac strings classically observable. Typical features of this new type of hair are presented. We also note that rotation can act as a hair generator even in the absence of NUT.

Motivated by bulk replica wormholes, we study the boundary effective theory that describes the near-horizon fluctuations of a near-extremal Reissner-Nordström black hole. This theory consists of a Schwarzian mode and a $U(1)$ phase mode. We compute the partition function of this boundary theory on replica geometries, from which the entropy is derived. Our analysis reveals a rich phase structure, in which the dominance of connected or disconnected replica configurations leads to a phase transition controlled by the temperature and the coupling constants $C$, $K$, and $\mathcal{E}$ of the 1d effective theory.

This paper surveys various results in the field of Quantum Learning theory, specifically focusing on learning quantum-encoded classical concepts in the Probably Approximately Correct (PAC) framework. The cornerstone of this work is the emphasis on query, sample, and time complexity separations between classical and quantum learning that emerge under learning with query access to different labeling oracles. This paper aims to consolidate all known results in the area under the above umbrella and underscore the limits of our understanding by leaving the reader with 23 open problems.

Why does quantum theory need the complex numbers? With a view toward answering this question, this paper argues that the usual Hilbert-space formalism is a special case of the general method of Markovian embeddings. This paper then describes the indivisible interpretation of quantum theory, according to which a quantum system can be regarded as an indivisible stochastic process unfolding in an old-fashioned configuration space, with wave functions and other exotic Hilbert-space ingredients demoted from having an ontological status. The complex numbers end up being necessary to ensure that the Hilbert-space formalism is indeed a Markovian embedding.

We construct higher-dimensional traversable wormholes in quasi-topological gravity (QTG) supported by a phantom scalar field. Using a static, spherically symmetric ansatz, we numerically analyze how quasi-topological gravity corrections affect the geometry and physical properties of the wormhole solutions. The resulting wormhole solutions are symmetric about the throat. Negative mass can arise for certain choices of parameters. For certain parameter ranges, the scalar charge $\mathcal{D}$ of the phantom field rapidly decreases with increasing the higher-curvature coupling parameter $α$ and approaches zero. Moreover, by changing $α$, the overall level of the Kretschmann scalar is also lowered. Finally, for sufficiently large $α$, $-g_{tt}$ becomes close to zero near the throat, exhibiting a ``horizon''-like structure.

In the present work we examine the MTS, for the restriction to spherical dust plus $Λ$, proving that it actually is a characteristic surface of the Cauchy problem (generated by its characteristic curves), which opens the possibility for infinite solutions. This translate as the MTS being a boundary between arbitrarily independent solutions, reminiscent of the Birkhoff theorem effects. This property is illustrated with combinations of 3 examples containing MTSs and $Λ$ ($Λ$CDM, Schwarzschild-de\,Sitter, Lemaître-Tolman-Bondi-de\,Sitter: LTBdS -- i.e. the inhomogeneous, spherically symmetric $Λ$CDM). The LTBdS model presents a static, stable MTS for the first time.

We construct a charged Bardeen black hole (BH) surrounded by perfect fluid dark matter (PFDM) and coupled to a cloud of strings (CS). The metric function combines the magnetic monopole charge from nonlinear electrodynamics, the PFDM logarithmic correction, and the CS parameter that renders the spacetime asymptotically non-flat. We analyze the horizon structure, identifying parameter ranges yielding non-extremal BHs, extremal configurations, and naked singularities. The null geodesics, photon sphere radius, and shadow are computed, revealing that both CS and PFDM enlarge the shadow. For neutral particle dynamics, we derive the specific energy, angular momentum, and innermost stable circular orbit location. Quasiperiodic oscillations (QPOs) are examined through the azimuthal, radial, and vertical epicyclic frequencies, where notably the azimuthal frequency is independent of the CS parameter. Scalar field perturbations governed by the Klein-Gordon equation yield an effective potential whose peak decreases with both parameters, yet the transmission and reflection probabilities respond oppositely to CS and PFDM variations. The greybody factor bounds are obtained using semi-analytical methods. Our results demonstrate that the distinct effects of $α$ and $β$ on various observables could allow independent constraints on these parameters through shadow measurements, QPO timing, and gravitational wave ringdown observations.

We investigate a bouncing cosmological model within the Weyl-type $f(Q)$ gravity framework, employing a power-law form of the non-metricity scalar $Q$. The model successfully resolves the initial singularity problem by demonstrating a nonsingular bounce, where the universe transitions from a contracting phase $ \dot{a}(t)<0 $ to an expanding phase ($ \dot{a}(t)>0 $) at the bouncing point $t \approx 0.$ Key features include the violation of the null energy condition (NEC) near the bounce and the crossing of the phantom divide line ($ω=-1$) by the equation of state (EoS) parameter, indicating quintom-like behavior. The model exhibits accelerated expansion post-bounce, suggesting an inflationary phase. Stability analysis via the adiabatic index reveals instability near the bouncing point, while energy conditions highlight the dominance of dark energy. Additionally, the study explores scalar fields, showing that quintessence-like kinetic energy becomes negative and phantom-like kinetic energy peaks positively near the bounce, aligning with dark energy dynamics. The Hubble parameter, deceleration parameter, and Hubble radius further validate the bouncing scenario, with the latter displaying symmetric behaviour around the bounce. These results underscore the viability of Weyl-type $f(Q)$ gravity as a framework for nonsingular bouncing cosmologies, offering insights into early universe dynamics and dark energy behaviour.

The Dirac-Bergmann algorithm for the Hamiltonian analysis of constrained systems is a nice and powerful tool, widely used for quantization and non-perturbative counting of degrees of freedom. However, certain aspects of its application to systems with first-class constraints are often overlooked in the literature, which is unfortunate, as a naive treatment leads to incorrect results. In particular, when transitioning from the total to the extended Hamiltonian, the physical information encoded in the constrained modes is lost unless a suitable redefinition of gauge invariant quantities is made. An example of this is electrodynamics, in which the electric field gets an additional contribution to its longitudinal component in the form of the gradient of an arbitrary Lagrange multiplier. Moreover, Dirac's conjecture, the common claim that all first-class constraints are independent generators of gauge transformations, is somewhat misleading in the standard notion of gauge symmetry used in field theories. At the level of the total Hamiltonian, the true gauge generator is a specific combination of primary and secondary first-class constraints; in general, Dirac's conjecture holds only in the case of the extended Hamiltonian. The aim of the paper is primarily pedagogical. We review these issues, providing examples and general arguments. Also, we show that the aforementioned redefinition of gauge invariants within the extended Hamiltonian approach is equivalent to a form of the Stueckelberg trick applied to variables that are second-class with respect to the primary constraints.

We present a unified category-theoretic framework for quantum Hall hierarchy constructions and anyon superconductivity based on modular tensor categories over $\mathrm{Rep}(\mathrm{U}(1))$ and $\mathrm{sRep}(\mathrm{U}(1)^f)$. Our approach explicitly incorporates conserved $\mathrm{U}(1)$ charge and formulates doping via a generalized stack-and-condense procedure, in which an auxiliary topological order is stacked onto the parent phase, and the quasiparticles created by doping subsequently condense. Depending on whether this condensation preserves or breaks the $\mathrm{U}(1)$ symmetry, the system undergoes a transition to a quantum Hall hierarchy state or to an anyon superconductor. For anyon superconductors, the condensate charge is determined unambiguously by the charged local bosons contained in the condensable algebra. Our framework reproduces all known anyon superconductors obtained from field-theoretic analyses and further predicts novel phases, including a charge-$2e$ anyon superconductor derived from the Laughlin state and charge-$ke$ anyon superconductors arising from bosonic $\mathbb{Z}_k$ Read-Rezayi states. By placing hierarchy transitions and anyon superconductivity within a single mathematical formalism, our work provides a unified understanding of competing and proximate phases near experimentally realizable fractional quantum Hall states.

The non-resonant production of a Higgs boson pair in association with a top-antitop quark pair ($pp\rightarrow t\bar{t}hh$) has only recently begun to be explored at the Large Hadron Collider (LHC) and provides a unique and largely uncharted probe of the top-Higgs sector, offering complementary sensitivity to the Higgs self-coupling and higher-dimensional interactions beyond the Standard Model. In this work, we present a detailed study of this process within the framework of Higgs Effective Field Theory (HEFT) at the High-Luminosity LHC (HL-LHC). A comparative analysis is performed using a traditional cut-based approach in the single-lepton channel and a multivariate parametric boosted decision tree method in both single-lepton and dilepton final states. We derive one- and two-parameter limits at 95\% confidence level on the HEFT couplings $δκ_λ$, $c_2$, $c_{2g}$, and $c_{tg}$. The projected bound on $δκ_λ$ is weaker than current experimental constraints from dedicated Higgs-pair measurement; however, this coupling plays a critical role in shaping the multidimensional allowed parameter space. For the remaining HEFT couplings, where no direct experimental limits currently exist, our results provide the first sensitivity projections in the $t\bar{t}hh$ channel. Overall, this study demonstrates the strong potential of the $t\bar{t}hh$ production process to probe extended Higgs and top-quark interactions beyond the Standard Model through the exploitation of the $t\bar{t}hh$ data at the HL-LHC.

Recent theoretical work has revealed that basic observables of quantum field theory in de Sitter space, known as in-in or cosmological correlators, exhibit surprisingly simple mathematical structure reminiscent of scattering amplitudes in flat space. For many theories, this simplicity can be made manifest using a set of ``cosmological dressing rules'' which uplift flat-space Feynman diagrams to in-in correlators in de Sitter space by attaching auxiliary propagators to the interaction vertices. In this paper, we show that discontinuities of cosmological correlators with respect to internal energy variables can be obtained by applying auxiliary propagators to unitarity cuts of flat space Feynman diagrams. Moreover, discontinuities with respect to external energy variables can be obtained by cutting auxiliary propagators attached to Feynman diagrams. This observation in turn implies highly non-trivial constraints on cosmological correlators in the form of simple sum rules. We illustrate these ideas in a number of examples at tree-level and 1-loop for conformally coupled scalar theories, although they hold more generally. Finally, we show how to reconstruct cosmological correlators from their discontinuities using dispersion relations, providing a powerful new approach to computing cosmological observables by systematically reconstructing them from data uplifted from flat space.

In this paper we extend recent ideas about observers and closed universes to theories where observers can be fluctuated into existence in the Hartle-Hawking state. This introduces a phenomenon that was not considered in these earlier discussions: the dominant transition from one cosmological state to another can go through a fluctuation that annihilates the universe and creates a new one. We nonetheless argue that the observer decoherence rule allows for the third-quantized description of such a theory to emerge from a factorizing holographic theory with a one-dimensional Hilbert space, without any need for $α$-parameters. We also point out a close analogy between the observer rule in this context and the coarse-graining of the spectral form factor at late times for AdS black holes. Along the way we clarify several aspects of the relationship between holography, the gravitational path integral, and $α$-parameters. We also explain why string theory scattering amplitudes do not lead to a one-dimensional Hilbert space on the worldsheet, despite being computed by a gravitational path integral with a sum over topology. Finally we point out that using the path integral to compute integrated local operators conditioned on an observer in the context of a theory with a landscape can lead to rather surprising conclusions. For example we argue that in a landscape with one AdS minimum and one dS minimum, both of which can support observers, an observer almost surely finds themself in dS and not AdS even if the boundary conditions are dual to a state with an observer in AdS.

The relaxation of a quantum field from a metastable state (false vacuum) to a stable one (true vacuum), also known as false vacuum decay, is a fundamental problem in quantum field theory and cosmology. We study this phenomenon using a two-dimensional interacting and coherently coupled Bose-Bose mixture, a platform that has already been employed experimentally to investigate false vacuum decay in one dimension. In such a mixture, it is possible to define an effective magnetization that acts as a quantum field variable. Using the Stochastic Gross-Pitaevskii equation (SGPE), we prepare thermal equilibrium states in the false vacuum and extract decay rates from the magnetization dynamics. The decay rates show an exponential dependence on temperature, in line with the thermal theory of instantons. Since the SGPE is based on complex scalar fields, it also allows us to explore the behavior of the phase, which turns out to become dynamic during decay. Our results confirm the SGPE as an effective tool for studying coupled magnetization and phase dynamics and the associated instanton physics in ultracold quantum gases.

Ultra-high-energy cosmic rays (UHECRs) can be detected via the broadband radio pulses produced by their extensive air showers. The Giant Radio Array for Neutrino Detection (GRAND) is a planned radio observatory that aims to deploy autonomous antenna arrays over areas of order $\sim 10^5\,\mathrm{km}^2$ to detect this emission. However, Galactic and instrumental radio backgrounds make the identification of low signal-to-noise ratio (SNR) pulses a central challenge. Here, we present a deep convolutional denoiser model that jointly processes each GRAND antenna trace in the time and frequency domains, allowing the network to learn transient pulse morphology and broadband spectral features while suppressing background noise. By training the model on $4.1\times 10^5$ simulated traces that include detailed UHECR radio emission and realistic detector response and noise, we find a median output-SNR improvement of $\sim 15-23\,\mathrm{dB}$ in the $50-200~\mathrm{MHz}$ band and a reduction of the normalized mean squared error of the waveform by about an order of magnitude relative to a Hilbert-envelope denoiser baseline. We also verify that applying the denoiser to noise-only windows does not produce spurious pulse candidates. Near the detection threshold, the denoiser increases the number of antennas contributing reliable pulse timing by a factor of $\sim 2-3$, which correspondingly tightens direction reconstruction uncertainties. When we additionally require accurate recovery of the waveform shape, the denoiser yields a median gain of $\sim 3-4$ antennas usable for energy reconstruction at SNR$\simeq 5-6$, strengthening event-level direction and energy estimates in sparse radio arrays.

In the three-flavor neutrino oscillation framework, we investigate the transition probabilities of an initial muon neutrino flavor state in the presence of non-standard interactions (NSIs) characterized by complex off-diagonal ($|ε_{αβ}|e^{iφ_{αβ}}$) and diagonal parameters ($|ε_{αα}-ε_{ββ}|$), including a CP-violating phase and a constant matter potential, under both normal (NO) and inverted mass ordering (IO) scenarios. Within these scenarios and through the lens of mode entanglement, bipartite entanglement measures such as entanglement entropy and capacity of entanglement are quantified in terms of the transition probabilities, which can be measured in neutrino oscillation experiments. Using these two bipartite entanglement measures, we further explore the quantum speed limit (QSL) time, which describes how rapidly bipartite entanglement evolves during neutrino oscillations. We illustrate our results using the baseline lengths and energies corresponding to ongoing long-baseline accelerator neutrino experiments, such as T2K, NO$ν$A, and the upcoming DUNE experiment. In the presence of a CP-violating phase and a constant matter potential, both with and without NSI effects, we compare the QSL time behavior for bipartite entanglement in neutrino oscillations for NO and IO. The most pronounced discrepancies in the QSL time for bipartite entanglement arise from the off-diagonal NSI parameter $ε_{μτ}$ across both the NO and IO scenarios. We emphasize that among all the experiments considered, NO$ν$A and DUNE exhibit a rapid suppression of bipartite entanglement in neutrino oscillations in the standard oscillation scenario with NO at the end of their baseline lengths for the corresponding best-fit value of CP-violating phase. Our results hint at a possible imprint of new physics in neutrino oscillations.

Modern machine learning is driving a paradigm shift in particle physics phenomenology at the Large Hadron Collider. This short review examines the transformative role of machine learning across the entire theoretical prediction pipeline, from parton-level calculations to full simulations. We discuss applications to scattering amplitude computations, phase space integration, Parton Distribution Function determination, and parameter extraction. Some critical frontiers for the field including uncertainty quantification, the role of symmetries, and interpretability are highlighted.

We study the thermodynamics of the isotropic Heisenberg XXX spin chain with negative spin, focusing on the case $s=-1$. The model is equivalent to the quantum lattice nonlinear Schrödinger (NLS) model and appears as an effective theory in deep inelastic scattering in high-energy quantum chromodynamics. Owing to its integrability, it admits a consistent Bethe Ansatz description and a well-defined thermodynamic limit. Using the thermodynamic Bethe Ansatz, we analyze the ground state, elementary excitations, and finite-temperature properties. In contrast to the conventional positive spin XXX chain, the negative spin model exhibits a distinct vacuum structure and excitation spectrum, leading to modified TBA equations and unconventional low-temperature behavior. Although the integral equations resemble those of the Lieb-Liniger Bose gas, the thermodynamics and scaling properties are qualitatively different and cannot be continuously connected. We derive the free energy, entropy, and specific heat, and identify a quantum phase transition separating different thermodynamic regimes. At zero temperature, the excitation spectrum becomes linear in the continuum limit and can be described by a conformal field theory. The low-temperature regime realizes a Luttinger-liquid like phase with features unique to the negative spin XXX chain.

Some effects induced by SMEFT operators at one loop have attracted a lot of attention in recent years, in particular, the renormalization of divergences by physical operators in single insertions of dimension-6 operators. Important non-logarithmically enhanced contributions must also be calculated. We discuss dimensional regularization in the Breitenlohner-Maison-'t Hooft-Veltman scheme. The goal here consists of determining in this scheme quantum effects in chiral theories at one loop. Namely, the determination of finite counterterms at one loop that reestablish the Slavnov-Taylor identities, which follow from gauge symmetries. These counterterms are necessary due to the presence of evanescent symmetry-breaking terms in the classical Lagrangian needed to regularize fermion propagators. We consider a technique that allows an easier calculation of such finite effects, relying on the identification of $(D-4)/(D-4)$ terms of one-loop amplitudes with an external ghost leg. We focus on dimension-6 four-fermion operators, identifying all finite counterterms in the Breitenlohner-Maison-'t Hooft-Veltman scheme at one loop, and as expected find no obstructions to the Slavnov-Taylor identities that cannot be cured by finite counterterms. This represents one step towards moving to higher order calculations.

Gowdy cosmologies are exact, spatially inhomogeneous solutions of the vacuum Einstein equations which describe nonlinear gravitational waves coalescing at the Big Bang singularity. With toroidal spatial sections they provenly have the Asymptotic Velocity Domination property, in that close to the Big Bang dynamical spatial gradients fade out and the dynamics is governed by a Carroll-type gravity theory. Here we construct an infinite set of Dirac observables for Gowdy cosmologies, valid off-shell, strongly, and without gauge fixing. These observables stay regular at the Big Bang and can be matched to much simpler Dirac observables of the Carroll-type gravity theory. Conversely, in an adapted foliation there is a systematic anti-Newtonian expansion (in inverse powers of the reduced Newton constant) of the full Dirac observables whose leading terms are the Carroll ones. In particular, this provides an off-shell generalization of the Asymptotic Velocity Domination property.

The LHCb collaboration has recently reported a measurement of the difference in direct CP asymmetries for the decays $Λ_b^0 \to J/ψ\, p \, h^-$ (with $h = K, π$), offering new experimental constraints on the decay dynamics of heavy baryons into charmonium final states. Inspired by these findings, we explore the branching ratios and direct CP violations for the decays $Λ_b^0 \to P_c^+(4312, 4440, 4457)\,h^-$ within the framework of final-state rescattering. Our analysis indicates that the branching fractions for $Λ_b^0 \to P_c^+ π^-$ lie around the $10^{-6}$ level, with the corresponding direct CP asymmetries approaching approximately $1\%$. In contrast, the direct CP violation for the decay $Λ_b^0 \to P_c^+ K^-$ is found to be very small, while its branching ratios show a strong dependence on the spin assignments of the $P_c$ states. These predictions may provide useful guidance for more precise CP measurements and amplitude analyses in the $P_c$ region in future experiments.

We present a study of charge balance functions inside jets in proton-proton collisions at $\sqrt{s}=13.6$ TeV using the PYTHIA8 event generator. The balance function is a differential observable of opposite-charge correlations, which is calculated in the jet frame for inclusive charged hadrons and the identified $π$, $K$, and $p$. The results show a clear narrowing of the balancing width with increasing jet charged multiplicity, indicating that particle production becomes more localized in momentum space in high-multiplicity jets.This trend resembles features attributed to collective expansion in heavy-ion collisions. The species dependence highlights sensitivity to the redistribution of strangeness and baryon number during string fragmentation and color reconnection. The new CR tune yields a little broader proton balance-function width in $Δφ^{*}$ than CP5, hinting at enhanced baryon-production dynamics, whereas meson widths differ only mildly. These comparisons suggest that multiparton interactions and color reconnection contribute to the observed trends, potentially generating collective like features inside jets, especially in high multiplicity jets, via nontrivial color dynamics alongside standard fragmentation. Taken together, the results establish identified hadron balance functions in high multiplicity jets as a sensitive probe of hadronization and provide new constraints for models of small system collectivity.

Let $B$ be a spatial region of width $2R$ and $Φ$ a Klein-Gordon wave packet localized in $B$ at time zero. We show the inequality $S \leq 2πR E$; here, $S$ is the entropy of $Φ$ contained in a region $B$, and $E$ is the energy content of $Φ$ within $B$. We consider a wider setting and formulate a variational problem aimed at minimizing our bound when $Φ$ is not localized in $B$. Our inequality holds in more generality in the framework of local, Poincaré covariant nets of standard subspaces and is related to the Bekenstein inequality. We point out a general bound that is compatible with the recent numerical computations by Bostelmann, Cadamuro, and Minz concerning the one-particle modular Hamiltonian of a scalar massive quantum Klein-Gordon field. We also provide a version of the entropy balance and ant formulas for wave packets.

Warm Dense Matter lies at the frontier between condensed matter and plasma, and plays a central role in various fields ranging from planetary science to inertial confinement fusion. Improving our understanding of this regime requires experimental data that can be directly compared with theoretical and numerical models over a broad range of conditions. In this work, a pulsed-power experiment is described in which thin metallic foils, confined within a sapphire cell, are Joule-heated to achieve the expanded warm dense matter regime. Designing such an experiment is challenging, as it requires simultaneously predicting the electrical response of the pulsed-power driver and the hydrodynamic evolution of the heated material. To tackle this challenge, a modeling framework has been developed that couples an electrical description of the pulsed-power system, including the driver, the switching stages and the load with a one-dimensional hydrodynamic code. This coupling allows the electrical energy deposition and the load thermodynamic evolution to be consistently linked through the material electrical conductivity. This approach takes advantage of the simplicity of a 1D geometry while retaining the essential physics and allowing to reproduce various measurements with good accuracy, such as expansion velocity, current and voltage. This numerical approach therefore constitutes a robust and efficient method for designing and optimizing future Warm Dense Matter experiments using pulsed-power facilities.

We introduce a general definition of higher-form connections on principal $\infty$-bundles in differential geometry. This is achieved by developing the formal differentiation and integration of maps from smooth manifolds to derived stacks with sufficient deformation theory. That allows us to introduce the Atiyah $L_\infty$-algebroid of a principal $\infty$-bundle and establish its global sections as the $L_\infty$-algebra of the derived higher symmetry group of the bundle. We define the space of $p$-form connections on an $\infty$-bundle as the space of order $p$ splittings of its Atiyah $L_\infty$-algebroid. We demonstrate that our new concept of derived geometric $p$-form connections recovers the known notion of connections on higher U(1)-bundles defined via Čech-Deligne differential cocycles. We further relate the $L_\infty$-algebras of derived higher symmetries of higher U(1)-bundles and higher Courant algebroids. Some applications in higher gauge theory and in supergravity are mentioned.

We investigate the classification of self-dual nonlinear electrodynamic (NED) theories based on their analyticity properties, which are directly linked to invariance under a discrete $\varphi$-parity transformation. This classification is expressed through the structure of the irrelevant $T\bar{T}$-like deformations that generate the theories from a Maxwell seed. Using both closed-form and perturbative methods within the Courant-Hilbert (CH) and Russo-Townsend auxiliary field formalisms, we demonstrate a precise correspondence: $\varphi$-parity-invariant, analytic theories are generated by irrelevant deformations built from integer powers of the energy-momentum tensor scalars, $\mathcal{O}_λ\sim \sum C_m (T_{μν}T^{μν})^{1-m}({T_μ}^μ{T_ν}^ν)^{m}$. Conversely, $\varphi$-parity-violating, non-analytic theories require deformations involving both integer and half-integer powers, $\mathcal{O}_λ\sim \sum C_m (T_{μν}T^{μν})^{1-m/2}({T_μ}^μ{T_ν}^ν)^{m/2}$. We prove this result in generality via a perturbative CH framework, showing that $\varphi$-parity invariance imposes specific constraints on the expansion coefficients of the CH function $\ell(τ)$ which, in turn, force all half-integer powers in the deformation to vanish. The classification is explicitly verified for known closed-form theories: the analytic generalized Born-Infeld model and the non-analytic examples of the $q=3/4$-deformed and "no $τ$-maximum" theories. Furthermore, we show how the $\varphi$-parity transformation is consistently generalized in the presence of a marginal root-$T\bar{T}$ coupling $γ$, and we derive the corresponding marginal and irrelevant flow equations for the studied theories.

We present a unified framework for neutrino masses, baryogenesis, and dark matter based on a modular $S_4$ symmetry combined with a type-III seesaw mechanism. All Yukawa couplings, CP phases, and flavor textures originate from a single complex modulus $τ$, whose vacuum expectation value controls both visible and dark sector dynamics. The same modular parameter fixes the neutrino mass matrix, determines the CP asymmetries driving resonant leptogenesis, and correlates the resulting baryon and dark matter abundances. A detailed numerical analysis shows that the model reproduces all neutrino oscillation data within the $3σ$ NuFIT~5.2 (2024) ranges for normal ordering, predicting $δ_{\rm CP} \simeq \pm (150^\circ-180^\circ)$, $\sum m_ν\simeq(0.06-0.08)~\mathrm{eV}$, and an effective Majorana mass $m_{ββ} \simeq (8 - 18)\times 10^{-3}~\mathrm{eV}$, testable in next-generation neutrinoless double-beta decay experiments. The same modular Yukawas yield resonantly enhanced CP asymmetries $|ε_{L,χ}| \sim 10^{-9}-10^{-6}$ at $M_Σ\sim 10^{7}~\mathrm{GeV}$, successfully generating the observed baryon asymmetry $η_B\simeq6\times10^{-10}$ and dark relic density $Ω_χh^2\simeq0.12$ without additional free parameters. The predicted correlation $Ω_χ/Ω_B\simeq5.4$ fixes the dark matter mass to $m_χ\simeq0.1-2~\mathrm{GeV}$, consistent with all current constraints. This framework therefore realizes a fully predictive baryon$-$dark matter co-genesis, where the geometry of the modular symmetry links the origin of flavor, CP violation, and the cosmic matter asymmetry.

In this paper we discuss a higher spin extension of the holographic symmetry algebra for graviton and gluon. Our primary observation is that in the presence of higher spin particles the soft symmetry algebra has a subalgebra isomorphic to $w_{\infty}$ which is generated by the \textit{conformally soft higher spin particles}. This $w_{\infty}$ subalgebra does not commute with the $w_{1+\infty}$ subalegbra generated by the conformally soft gravitons. The same thing holds for the colored higher spin particles. One gets a subalgebra isomorphic to the $S$-algebra which is generated by the conformally soft colored higher spin particles. We further verify the soft algebra for colored higher spin particles using the (tree-level) $4$-point MHV amplitude of the higher spin Yang-Mills theory constructed in arXiv:2210.07130. At the end we also discuss the higher spin extension of the deformed holographic symmetry algebra for non-zero cosmological constant as constructed in arXiv:2312.00876.

The transverse momentum (pT) distributions up to pT = 20 GeV/c for pions produced in the ten different multiplicity classes (MCs) of symmetric pp collisions at sqrt(s) = 7 TeV have been investigated. Two distinct models, the Tsallis-Pareto type function (model) and the combined BGBW model and Tsallis-Pareto type model have been employed to fit the pT distributions via the minimum chi-square method. The combined Hydro+Tsallis model is more reliably describing the pT spectra than the Tsallis-Pareto model. The Tsallis temperature (T), non-extensivity parameter (q), normalization constant (N0), Kinetic freeze-out temperature (T0), transverse flow velocity (betaT), and (mean pT) have been extracted through the fitting procedure via the employed models. The Tsallis-Pareto model gives T, q, N0 and mean pT while Hydro+Tsallis model gives T0, betaT, T, q, N0 and mean pT. Incorporating the values of the extracted T and q the thermodynamic quantities and response functions, including energy density (epsilon), particle density (n), entropy density (s), pressure (P), specific heat at constant volume (CV), squared speed of sound (cs2), mean free path (lambda), Knudsen number (Kn), isothermal compressibility (kappaT), and expansion coefficient (alpha) have been calculated at the freeze-out stage. It has been observed that T, betaT, mean pT, N0, epsilon, n, s, P, CV, cs2, and alpha increase with increasing(decreasing) the charged particles multiplicity density dNch/deta(MCs). While T0, q, lambda, Kn, and kappaT decrease with increasing(decreasing) dNch/deta(MCs). These systematic variations in the trends of parameters might suggest the gradual transition towards collectivity and thermal equilibration in the high multiplicity pp events, possibly signalling enhanced collective dynamics and partial thermalization in small collision systems.

In the present work, we study the possible production of the dark Higgs boson ($h_{D}$) candidates, which originated from a simplified-model scenario based on the Mono-Z$^{\prime}$ model, in association with a neutral gauge boson (Z$^{\prime}$). This study has been performed by studying events with dimuon plus missing transverse energy produced in the simulated electron-positron collisions at the foreseen Future Circular Collider in the Electron-Positron collision mode (FCC-ee), operating at 240 GeV center of mass energy and integrated luminosity of 10.8 ab$^{-1}$. In case no new physics has been discovered, we set upper limits at a 95\% confidence level on the mass of the dark Higgs.

We present a jet-free approach for measuring energy-energy correlators (EEC) in proton-proton (pp) collisions at the Large Hadron Collider (LHC), employing an event-topology method that does not rely on explicit jet reconstruction. Using the leading charged hadron as a reference axis, the azimuthal plane is divided into Toward and Transverse regions, enabling a robust background subtraction and extending EEC measurements into the low $p_T$ regime where conventional jet-based approaches become unreliable. The method is validated through comparisons with conventional jet reconstruction results. We systematically explore the dependence of the EEC on the leading-particle transverse momentum and parton flavor. The observed scaling between the EEC peak position and the hard scale suggests that this topology-based EEC captures effectively the transition between perturbative and non-perturbative QCD regimes. Distinct differences are found between quark- and gluon-initiated events, reflecting their different color charges and radiation patterns. Extending the analysis to heavy flavor, EECs triggered by leading charm mesons exhibit a suppressed magnitude and a peak shifted toward larger angular separations relative to inclusive charged-particle triggers, providing a direct manifestation of the dead-cone effect. This jet-free EEC framework offers a simple and experimentally robust tool for studying the scale and flavor dependence of the QCD dynamics, with promising applications to proton-nucleus and heavy-ion collisions at the LHC.

We apply the generalized effective-range expansion of Phys. Rev. Lett. 135, 011903(2025), which incorporates the left-hand cut from one-pion exchange, to low-energy neutron-proton scattering in the $^1S_0$ and $^3S_1$ channels. The amplitude zero for the center-of-mass momentum near 0.35 GeV in the $^1S_0$ channel is naturally accommodated within this framework. We extract the pole position, scattering length, effective range, and the pseudoscalar pion-nucleon coupling constant $g_{πN}^2/(4π)$ at different expansion orders. The low-energy parameters are stable and consistent with established values, while $g_{πN}^2/(4π)$ exhibits larger uncertainties. The extraction of $g_{πN}^2/(4π)$ is data-driven, relying on the analytic constraints from the left-hand cut and phase-shift data within the one-pion-exchange approximation. Despite larger uncertainties compared to high-precision extractions, the consistency with established values demonstrates that this framework can probe the left-hand-cut singularity.

Primordial black holes (PBHs) are able to produce light dark matter (DM) particles via Hawking radiation, and yield a flux of boosted DM that can be probed at underground DM direct detection experiments. We analyze both galactic and extragalactic contributions to the differential flux of light DM from PBH evaporation, and then compute the expected event rate from PBH boosted DM scattering off electrons or nuclei after taking into account the attenuation effect. Using recent data from DM direct detection experiments XENONnT, PandaX-4T and LZ, we set constraints on both DM-electron and DM-nucleus scattering cross sections, as well as the fraction of DM composed of PBHs $f_{\rm PBH}$ for $9\times10^{14}-1\times10^{16}\,\mathrm{g}$ PBHs that are not fully evaporated today. We also investigate the spectral evolution induced by Hawking evaporation throughout the evaporation and post-evaporation regimes. The constraints on the PBH mass are then extended into the $1\times10^{13}-6\times10^{14}\,\mathrm{g}$ window for fully evaporated PBHs.

Working with a symmetry-preserving treatment of a vector $\otimes$ vector contact interaction (SCI), a largely algebraic three-body Faddeev equation treatment of the nucleon bound state problem is introduced and used to deliver results for all nucleon charge and magnetisation distributions and their flavour separation. A strength of the SCI treatment is that it provides for a transparent understanding of this three-body approach to developing predictions for baryon observables. Comparisons of SCI results with predictions obtained in realistic-interaction Faddeev equation studies reveal the sensitivities of given observable to phenomena associated with the emergence of hadron mass.

This review article is based on a seminar presented at the Higgs pairs workshop 2025. Stochastic gravitational wave backgrounds can serve as probe of the diverse phenomenology encountered in beyond-Standard-Model scenarios featuring phase transitions in the early Universe. Focussing on gravitational wave production from first-order phase transitions, we present the main results of a recent analysis by the LISA Cosmology Working Group concerning the detectability of such signals with LISA. Strong degeneracies, both among the parameters controlling the phase transition and between these and the parameters of the beyond-Standard-Model scenario underlying the phase transition, complicate the reconstruction of the model from a potential signal. Nonetheless, once a specific scenario is assumed, LISA observations can supply constraints possibly complementary to those obtainable from present and future particle colliders.

These lecture notes aim to provide a pedagogical introduction to the AdS/CFT correspondence and its extensions to spacetimes with positive (de Sitter spacetime) and zero (flat spacetime) cosmological constant. We begin by explaining the physical motivation for holography and the significance of the AdS/CFT correspondence. We then review the basic ingredients of conformal field theory (CFT) and anti de Sitter (AdS) spacetime required to formulate the duality. Building on these foundations, we discuss the formulation of the AdS/CFT correspondence and discuss several consistency checks that support it. We conclude with a brief discussion of holography in de Sitter and flat spacetimes.

We calculate the shear viscosity, $η$, in two limits: perturbative QCD and an excluded-volume hadron resonance gas (HRG), at finite BSQ densities. Using an interpolation framework, we connect these regimes. In addition, we present results for (almost) next-to-leading order weak-coupling shear viscosity for QCD at finite $μ_B$, and discuss the convergence of the perturbative series.

Motivated by the recent evidence of a stochastic gravitational wave background found by pulsar timing array experiments, we focus on one of the prime cosmological explanations, i.e. a supercooled first order phase transition. If confirmed, it would offer a unique opportunity to probe early Universe dynamics and the related physics beyond the Standard Model of particles and interactions. However, the prediction of the gravitational wave spectrum from a given particle physics scenario requires theoretically and computationally demanding methods. While several tools have been put forward to reduce uncertainties and automatize these computations, we study here the possibility to perform the full pipeline of computations semi-analytically in the $4D$ theory, thus avoiding computationally intensive simulations. Our approach yields accurate results that can be used in phenomenological studies and allow for an efficient exploration of the connection between the particle physics models and their cosmological predictions.

We explore the sensitivity of the Short-Baseline Near Detector (SBND) experiment to light dark matter using MeV-scale electromagnetic activity. Inelastic scattering of dark matter with argon nuclei can excite nuclear states that subsequently de-excite via the emission of MeV-scale photons, producing localized low-energy "blip" signatures in a liquid argon time projection chamber. We perform state-of-the-art ab initio nuclear calculations, including all relevant argon excited states with energies up to 18 MeV, to provide reliable predictions for these signals. After accounting for relevant backgrounds, we find that SBND can probe previously unexplored regions of parameter space for light dark matter.

We present a systematic framework for the maximally-transcendental part of planar QCD scattering amplitudes and perform the first bootstrap computation of six-gluon MHV amplitudes in massless QCD at the symbol level. By analyzing the maximal weight projection of amplitudes at the integrand level, we relate their maximally-transcendental parts to prescriptive unitarity integrals. This reveals a novel analytic structure: the prefactors multiplying the functions of maximal transcendentality are identified with the four-dimensional leading singularities of the theory. As a consequence, these prefactors admit a complete classification and can be computed using on-shell diagrams, a formalism originally developed in $\mathcal{N}{=}4$ super Yang-Mills theory. As a concrete application, we determine the two-loop prefactors for planar MHV gluon amplitudes at arbitrary multiplicity. Combining these prefactors with recent advances in the planar two-loop six-point function space and explicit six-point prescriptive-unitarity input, we construct a complete symbol ansatz and uniquely fix the maximally-transcendental part of the two-loop six-gluon MHV QCD amplitudes by imposing physical constraints. The resulting symbols are expressible in a reduced 137-letter alphabet, suggesting that this alphabet is complete for two-loop six-point massless MHV scattering. We also discuss the implications for multi-collinear splitting and multi-soft functions.

The observation of associated production of an $Υ$(1S) meson with a Z boson and a measurement of the ratio of its fiducial cross section to the fiducial cross section of the Z boson are presented. Both the $Υ$(1S) meson and the Z boson are identified via decays into a pair of opposite-sign muons. The analysis is based on proton-proton (pp) collision data at $\sqrt{s}$ = 13 TeV, collected with the CMS detector in 2016$-$2018 and corresponding to an integrated luminosity of 138 fb$^{-1}$. Using the production of the Z boson decaying into four muons as a normalization channel, the ratio of the fiducial cross sections $σ$(pp $\to$ Z $+$ $Υ$(1S))$\mathcal{B}$(Z $\to$ $μ^+μ^-$)$\mathcal{B}$($Υ$(1S) $\to$ $μ^+μ^-$ ) to $σ$(pp $\to$ Z)$\mathcal{B}$(Z $\to$ 4$μ$) is measured to be $\mathcal{R}_{\mathrm{Z+Υ}\mathrm{(1S)}}$ = (21.1 $\pm$ 55 (stat) $\pm$ 0.6 (syst) $\times$ 10$^{-3}$), where stat and syst denote the statistical and systematic uncertainties, respectively. The result is used to extract the effective double-parton scattering cross section $σ_\text{eff}$ = 13.0$^{+7.7}_{-3.4}$. In addition, for the first time, $σ_\text{eff}$ is measured in bins of the transverse momentum of the $Υ$(1S) meson or of the Z boson.

The most recent results on the searches for the vector-like quarks and excited fermions from the CMS Collaboration are presented. These results are based on the full Run 2 $\sqrt{s} = $13 TeV proton-proton collision data collected by the CMS Collaboration at the LHC from 2016 to 2018, which corresponds to an integrated luminosity of 138 fb$^{-1}$. No significant excess above the Standard Model expectation is observed. Exclusion limits are set at the 95% confidence level on various benchmark models.

We compute synchrotron radiation from a plasma in which $P$- and $CP$-violating parameters, a chiral chemical potential and a chiral gradient, couple to fermions. To do this, we compute exact wavefunctions for the fermions in the presence of these parameters and an external constant magnetic field. We find that these parameters increase the synchrotron radiation emitted by the fermions while also decreasing the traditionally large synchrotron radiation elliptic flow coefficient $v_2$. We apply these results to the quark-gluon plasma, where just such a contribution could provide a solution to the missing direct photons puzzle. We also use our wavefunctions to give a derivation of the chiral magnetic effect.

We study dimension-eight effects in the Standard Model Effective Field Theory extended by right-handed neutrinos ($ν$SMEFT). Using the Hilbert series formalism, we derive the complete basis of dimension-eight operators and confirm agreement with existing classifications, providing a systematic framework beyond the conventional dimension-six truncation. We analyse the collider phenomenology of the representative operator $\mathcal{O}_{N^{2}q^{2}B}^{(1,2)}$ at the High-Luminosity LHC. The resulting signatures involve pair production of right-handed neutrinos in association with jets, followed by decays into electron-jet final states with potentially displaced vertices. Since similar final states are generated by leading dimension-six operators, we explicitly address whether dimension-eight contributions can be experimentally distinguished from dimension-six effects. Using a Boosted Decision Tree analysis based on kinematic observables, we show that the dimension-eight signal can be reliably separated from each relevant dimension-six hypothesis. Our results demonstrate that dimension-eight operators in the $ν$SMEFT can give rise to experimentally resolvable signatures and should be included in collider EFT interpretations.

We develop a defect-theoretic refinement of meromorphic 2d CFTs in which the ordinary torus partition function -- often just the vacuum character -- does not reveal how states organize under symmetry lines. Our central proposal is an \emph{equatorial projection} framework: from a commutant decomposition into commuting rational chiral algebras with categories $\mathcal{C}$ and $\widetilde{\mathcal{C}}$, we encode genus-one couplings by a non-negative integer matrix $M$ pairing characters and satisfying modular intertwiner relations. Invertible topological defect lines act directly on this gluing data (Verlinde lines diagonally via $S$-matrix eigenvalues, and anyon-permuting lines by braided-autoequivalence permutations), making modular covariance of defect amplitudes automatic and sharply distinguishing insertions that yield genuine modular invariants from those defining consistent non-holomorphic interfaces. We further show that the \emph{replacement rules} of \cite{Hegde:2021sdm, Lin:2019hks} arise as equatorial projections of defect actions, and we extend these constructions beyond two-character examples by systematically treating three-character commutant pairs in the $E_{8,1}$ theory. The unique $c=8$ meromorphic CFT $E_{8,1}$ serves as a universal testbed, producing new defect partition functions and clarifying the roles of $\mathrm{Pic}(\mathcal{C})$ and $\mathrm{Aut}^{\mathrm{br}}(\mathcal{C})$. Finally, we outline extensions to higher central charges (e.g.\ $c=32,40$), yielding modular-invariant non-meromorphic theories beyond the $c=24$ Schellekens landscape \cite{Schellekens:1992db} as defect/interface descendants of meromorphic parents.

We propose hydrogenated carbon structures as targets with a remarkable sensitivity to dark matter-nucleon interactions, in the mass range between the 1 MeV and 100 MeV. The ejection of a proton following the interaction with a dark matter particle is a quasi-elastic process, with an extremely small energy threshold, and a clear experimental signature. The proposed detectors are simple, technologically ready, and inexpensive. Yet, they can be considerably more sensitive than current experiments. They also allow strong directionality, to be used towards efficient background rejection.

Generalized neutrino interactions (GNI) are emerging as a convenient framework for describing effective scalar, vector, and tensor interactions. Such interactions arise naturally from extensions of the Standard Model that aim to explain neutrino properties and their mass origin. In this paper, we carefully study the two more common parametrizations for GNI and how to relate them. This allows us to compare bounds obtained from CEvNS and deep-inelastic scattering under the same footing. In addition, we present the current bounds from CEvNS measurements by COHERENT and compare them to those obtained from deep inelastic scattering on the same level. Our results focus on neutrino-quark interactions, and illustrate the complementarity between experiments working at different scales for GNI, showing that scalar interactions are better constrained by low-energy experiments like COHERENT, while tensor interactions are robustly constrained from deep inelastic scattering.

The speed of sound is a key parameter for characterizing equilibrium states. However, sound waves change their properties when propagating through rapidly evolving anisotropic media, such as the quark-gluon plasma created in heavy-ion collisions. This paper uses $\mathcal{N}=4$ Super-Yang-Mills theory to numerically study the time evolution of the speed and attenuation of sound modes along with the relaxation time in a plasma undergoing Bjorken expansion from various initial states in a quasi-static approximation. The longitudinal Bjorken expansion breaks the isotropy, resulting in two distinct sound speeds that range from just below the conformal value to the speed of light. An anisotropic hydrodynamic description is constructed and its applicability is discussed. Implications for the analysis of heavy ion data are considered.

We study entanglement entropy in the low-energy effective field theory of two-dimensional string theory as well as in the singlet sector of the dual $c=1$ matrix quantum mechanics. From the target space perspective, we argue that a generic bulk subregion is expected to have an associated generalized entanglement entropy combining a dilaton-dependent gravitational term and a matter contribution coming from the tachyon. Given that the gravitational area-like term is absent in previous analyses of entanglement entropy in the $c=1$ model, we examine several possible mechanisms for its emergence. We show that the nonlocal transformation induced by the leg-pole factor that relates the target space tachyon and the matrix model collective excitations cannot account for the area-like term, and we comment on its possible origin in the non-singlet sectors of the theory.

This study investigates transverse-momentum (pT) distributions of pi-, pi+, K-, K+, p, pbar, K0s, and Lambda in several centrality classes of Pb-Pb collisions at sqrt(sNN) = 2.76 TeV. The measured spectra are analyzed with the Tsallis non-extensive distribution, from which the effective temperature T, non-extensive parameter q, and the mean transverse momentum mean_pT are extracted for each particle species and centrality interval. To disentangle thermal and collective effects, the mean kinetic freeze-out temperature T0 is obtained from the intercept of the T-versus-mass relation, while the average transverse flow velocity betaT is extracted from the slope of mean_pT versus the mean moving mass for pions, kaons, and protons. The results show that T increases and q decreases with increasing centrality, indicating a hotter and more equilibrated system in central collisions. A clear mass dependence of T supports a multi-freeze-out scenario, with heavier particles decoupling earlier. Both T0 and betaT rise from peripheral to mid-central collisions before saturating toward central events, which may suggest the onset of collective behavior or changes in freeze-out dynamics. These observations provide new insights into the thermal and dynamical properties of the medium created in heavy-ion collisions at the LHC.

We investigate a scenario where Strongly Interacting Massive Particle (SIMP) dark matter interacts with an axion-like particle (ALP) that couples exclusively to electrons. This minimal setup provides interactions which enforce thermal equilibrium between dark matter and the SM in the early Universe. We analyze the cosmological evolution of the dark sector and the constraints arising from dark matter annihilations, ALP laboratory searches and astrophysical observations. Our results show that the allowed parameter space is wider than previous studies and an ALP with mass $m_a \sim {\cal O}(10)~\text{MeV}$ can act as a viable portal between the visible and dark sectors. Interestingly, this mass range overlaps with the parameter space suggested by the reported $X_{17}$ anomaly. Furthermore, the introduction of non-vanishing $θ$ angle in the dark sector of the model opens up the parameter space to heavy ALP masses.

A system of classical interacting spins can develop collective instabilities which, in the nonlinear regime, mimic the motion of a gyroscopic pendulum. Known as the flavor pendulum, this behavior appears among the collective modes of a dense neutrino plasma after a strong reduction of phase space through symmetry assumptions. It has been identified in homogeneous slow and fast flavor systems and, most recently, in single-wave solutions of the fast system. We explain the reasons for its ubiquitous appearance. We show that a system of three classical spins must always be pendular, or only two in the presence of an external field. Furthermore, such a system always defines a continuum of vectors with time-independent length. If these are identified as interacting spins, they immediately lead to the continuum cases of slow and fast flavor pendula. As another new insight, any of these spins can be chosen as the pendulum, periodically exchanging flavor with the rest of the system.

Wilson loops are essential objects in QCD and have been pivotal in scale setting and demonstrating confinement. Various generalizations are crucial for computations needed in effective field theories. In lattice gauge theory, Wilson loop calculations face challenges, including excited-state contamination at short times and the signal-to-noise ratio issue at longer times. To address these problems, we develop a new method by using neural networks to parametrize interpolators for the static quark-antiquark pair. We construct gauge-equivariant layers for the network and train it to find the ground state of the system. The trained network itself is then treated as our new observable for the inference. Our results demonstrate a significant improvement in the signal compared to traditional Wilson loops, performing as well as Coulomb-gauge Wilson-line correlators while maintaining gauge invariance. Additionally, we present an example where the optimized ground state is used to measure the static force directly, as well as another example combining this method with the multilevel algorithm. Finally, we extend the formalism to find excited-state interpolators for static quark-antiquark systems. To our knowledge, this work is the first study of neural networks with a physically motivated loss function for Wilson loops.

We investigate the decays $B_c^- \rightarrow D_{s}^{(*)-}\ell^+\ell^-$ and $B_c^- \rightarrow D_{s}^{(*)-}ν\barν$ within the Standard Model (SM), employing perturbative QCD form factors that are sensitive to the wave functions of $B_c$ and $D_{s}^{(*)}$ mesons. We determine the shape parameters of these mesons and the $B_c \to D_s^{(*)}$ form factors at $q^2 = 0$ from available lattice QCD inputs for $B_s \to D_s^{(*)}$ and $B_c \to D_s$ transitions. To obtain the $q^2$ dependence of the $B_c \to D_s^*$ form factors, we employ heavy-quark spin symmetry and an appropriate parametrisation scheme over the allowed $q^2$ region. Based on these inputs, we present predictions for branching ratios and lepton-flavour-sensitive observables. Furthermore, we perform a detailed angular analysis of the cascade decay $B_c^- \to D_s^{*-}(\to D_s^- π^0)\,\ell^+\ell^-$, providing Standard Model predictions for several angular observables.

The detection of high-energy neutrinos from NGC 1068 and TXS-0506+56 suggests that active galactic nuclei (AGN) may contribute significantly to the the diffuse neutrino flux measured by IceCube. Using 10 years of publicly available IceCube data, we performed a systematic population analysis of X-ray-bright and gamma-ray-bright AGN to evaluate the extent to which this diffuse flux could originate from these sources. We find that gamma-ray-bright blazars can account for no more than 16\% of IceCube's total diffuse flux. Although we find no evidence of neutrino emission from gamma-ray-bright, non-blazar AGN, we cannot exclude the possibility that these sources contribute significantly to the diffuse flux. In contrast, we report (pre-trials) evidence of neutrino emission from several nearby, X-ray-bright, Seyfert-type AGN, including \mbox{NGC 1068} ($4.9σ$), SWIFT J1041.4-1740 ($2.6σ$), SWIFT J0202.4+6824A/B ($2.6σ$), SWIFT J0744.0+2914 (2.6$σ$), NGC 4151 ($2.5σ$), and NGC 3079 ($2.5σ$). Although not fully conclusive, these results suggest that IceCube may be detecting neutrinos from a larger population of Seyfert galaxies. The fact that these sources are not gamma-ray bright indicates that their neutrino production must be taking place in optically thick environments, such as in the coronae surrounding these galaxies' supermassive black holes. We also identify a $4.2σ$ correlation between the neutrinos detected by IceCube and members of the Swift-BAT catalog of X-ray-bright AGN, although this correlation is dominated by NGC 1068. We estimate that this class of sources contributes between 11.2\% and the entirety of IceCube's total diffuse neutrino flux. These results strengthen the emerging case for the prevalence of gamma-ray-obscured AGN as significant sources of high-energy neutrinos.

We present a predictive framework for neutrino mass generation based on the spontaneous breaking of a leptonic $U(1)_{L_μ-L_τ}$ symmetry within a supersymmetric setting. The breaking of the global symmetry gives rise to a Majoron-like axion-like particle and a pseudo-Goldstone right-handed neutrino whose mass is naturally suppressed by supersymmetry-breaking effects. The interplay between the pseudo-Goldstone neutrino and the low-scale seesaw mechanism leads to a structured neutrino mass matrix capable of reproducing the observed neutrino masses, mixing angles, and CP-violating phase without invoking extreme parameter hierarchies. We perform a numerical fit to current neutrino oscillation data and identify representative benchmark solutions consistent with laboratory constraints as well as cosmological and astrophysical bounds. A characteristic outcome of the framework is the emergence of correlated relations linking the symmetry breaking scale, heavy neutrino masses, Majoron couplings, and neutrino lifetimes. Majoron-induced invisible neutrino decay arises generically and can significantly modify cosmological neutrino mass constraints for sufficiently low symmetry breaking scales. We discuss the phenomenological implications across neutrino oscillation experiments, cosmology, and collider searches for long-lived heavy neutrinos. While a detailed experimental simulation is beyond the scope of this work, existing sensitivity projections indicate that portions of the parameter space may become accessible in future facilities. The combined interplay of laboratory probes and cosmological observations provides a consistent and testable picture of neutrino mass generation tied to spontaneous leptonic symmetry breaking and axion-like physics.

Symmetries play a central role in physics, organizing dynamics, constraining interactions, and determining the effective number of physical degrees of freedom. In parallel, modern artificial intelligence methods have demonstrated a remarkable ability to extract low-dimensional structure from high-dimensional data through representation learning. This review examines the interplay between these two perspectives, focusing on the extent to which symmetry-induced constraints can be identified, encoded, or diagnosed using machine learning techniques. Rather than emphasizing architectures that enforce known symmetries by construction, we concentrate on data-driven approaches and latent representation learning, with particular attention to variational autoencoders. We discuss how symmetries and conservation laws reduce the intrinsic dimensionality of physical datasets, and how this reduction may manifest itself through self-organization of latent spaces in generative models trained to balance reconstruction and compression. We review recent results, including case studies from simple geometric systems and particle physics processes, and analyze the theoretical and practical limitations of inferring symmetry structure without explicit inductive bias.

The determination of transverse single-spin asymmetries in experiments involving polarized targets and/or beams may encounter challenges when (1) the magnitude of the polarization varies greatly with time, (2) the polarization magnitude is not the same for each spin state, (3) different integrated luminosities occur for different spin states or different target materials, and/or (4) some kinematic variables require unfolding; these are just a few examples. We present general methods of determining the asymmetry based on both binned analysis and unbinned maximum likelihood optimization, incorporating the unfolding of kinematic variables that are smeared by detector effects, and also including the possibility of background subtraction.

We come up with a novel search strategy for long-lived doubly charged scalars at future proposed lepton colliders. The doubly charged scalar studied in this work belongs to an $SU(2)_L$ complex scalar triplet that accounts for tiny neutrino masses via the Type-II Seesaw mechanism. For scalar masses $\lesssim 200 $ GeV and appropriate values of the triplet vacuum expectation value, this state can be long-lived and decay predominantly into like-sign muon pairs (e.g. $μ^+μ^+ $ or $μ^-μ^-$), producing distinctive displaced-vertex signals. We investigate the pair production of these scalars at the International Linear Collider (ILC) and a prospective muon collider, considering their planned center-of-mass energies. Incorporating theoretical and experimental constraints, we study the resulting signature of four leptons accompanied by missing transverse energy. Displaced vertices offer direct evidence of the scalar's long lifetime, while we further show that the invariant mass distribution of same-sign dilepton pairs serves as a powerful complementary probe for discovering doubly charged Higgs bosons at both the ILC and muon collider.

The determination of the Higgs boson trilinear self-coupling $λ_{3H}$ is a key goal of the LHC physics programme. Its precise measurement will provide unique insight into the scalar potential and the mechanism of electroweak symmetry breaking. Higgs boson pair production in the ${gg}\to{HH}$ process, and particularly in the ${HH}\to{b}\bar{b}γγ$ final state, offers direct sensitivity to $λ_{3H}$. We present the first implementation of the Matrix Element Method at Next-to-Leading Order (MEM@NLO) for this process, which is publicly available. The MEM is a statistically optimal approach that maximises information extraction from collision events. Extending it to NLO represents a major methodological challenge, which we address with a new formalism integrated into the MoMEMta framework. Results with simulated pseudo-experiments demonstrate, in a proof-of-principle study, the strong discriminating power of the method and its ability to extract the coupling modifier $κ_λ$=$λ_{3H}$/$λ_{3H}^{SM}$ with high precision.

Deflections in the propagation of charged ultra-high-energy cosmic rays (UHECRs) caused by magnetic fields make the identification of their sources challenging. On the other hand, the arrival directions at Earth of neutrons point directly to their origin. The emission of UHECRs from a source is expected to be accompanied by the production of neutrons in its vicinity through interactions with ambient matter and radiation. Since free neutrons travel a mean distance $d/\text{kpc}=9.2(E/\text{EeV})$ before decaying, a neutron flux in the EeV range could be detected on Earth from sources of UHECRs in our Galaxy. Using cosmic-ray data from the Phase\,I of the Surface Detector of the Pierre Auger Observatory, we search for neutron fluxes from Galactic candidate sources. We select more than 1000 objects of astrophysical interest, stacking them into target sets. The targets all have declinations within the exposure of the Observatory, ranging from $-90^\circ$ up to $+45^\circ$ for energies above 1 EeV (and up to $+20^\circ$ for energies down to 0.1 EeV). Given that a neutron air shower is indistinguishable from a proton one, there is a significant background due to cosmic rays. A neutron flux from the direction of a candidate source would be identified by a celestial density of events that significantly exceeds the expected density of cosmic rays for that direction. No significant excess is found at any tested target direction, and an upper limit on the neutron flux is calculated for each candidate source.

Three structures, X(6600), X(6900), and X(7100), have emerged from the J$/ψ\,$J$/ψ$ (J$/ψ$ $\to$ $μ^+μ^-$) mass spectrum. These are candidates of all-charm tetraquarks, an exotic form of hadronic matter. A clearer picture of these states is obtained using proton-proton collision data collected by the CMS detector that corresponds to 315 fb$^{-1}$, which yields 3.6 times more J$/ψ\,$J$/ψ$ pairs than previous studies by CMS. All three structures, and their mutual interference, have statistical significances above five standard deviations. The presence of interference implies that the structures have common quantum numbers. Their squared masses align linearly with a resonance index and have natural widths that systematically decrease as the index increases. These features are consistent with radial excitations of tetraquarks composed of two aligned spin-1 diquarks without orbital excitation, and disfavor other interpretations. The J$/ψ\,$$ψ$(2S) $\to$ $μ^+μ^-μ^+μ^-$ decay mode is also explored and the X(6900) and X(7100) states are found with significances exceeding 8 and 4 standard deviations, respectively.

We generalize the spectral-function formalism to describe two-nucleon knockout processes in exclusive kinematics. Significant improvements are introduced both in the treatment of the current operators entering the $Δ$-current contribution and in the modeling of correlations between the two struck nucleons, including a consistent treatment of isospin dependence and the explicit incorporation of angular correlations. The framework is validated through comparisons with relativistic Fermi-gas calculations and with semi-exclusive electron-nucleus scattering data. Our results demonstrate that an accurate description of nuclear dynamics plays a crucial role in modeling this reaction mechanism. We further present a study of selected electroweak observables relevant to neutrino-scattering experiments.

This report is divided into three main parts: 1. The first two chapters discuss the emission of Short GRB (SGRB) from binary mergers surrounded by a strong magnetic field. By introducing our model, we investigated the physics of the emission of SGRBs from rotating and charged rotating BHs. A rapidly spinning, strongly magnetized neutron star (millisecond magnetar) is the primary source of strong magnetic fields ranging from $10^{13}~\rm to ~ 10^{16} G$. The decay of the magnetic field could power electromagnetic radiation, especially X-rays and gamma rays from NSs or NS-BH mergers as their primary sources. Considering the merger of compact bodies (NS-NS or NS-BH or BH-BH), we can obtain interesting results. 2. In the next two chapters, we reviewed the BH interiors to understand the nature of black hole interior information and evaporation from its initial to final phases via entropy variation. The evolution relation obtained from two types of entropy gives diverse understandings of the evaporation of BHs under Hawking radiation. 3. The fifth Chapter is related to BH configuration (information) entropy and the thermodynamic phase transition of $f(R)$ BH. Here, we consider a d$-$dimensional black hole (BH) in $f(R)$ gravity and analyze the effect of modified gravity on critical point parameters, the difference in number densities, and configuration entropy during the BH phase transition phenomenon. These results were also compared with charged AdS BH, the holographic dual of van der Waal's fluid, and hence the BH in modified gravity.

We present an effective field theory (EFT) framework for coupled-channel $B^{(*)}\bar D^{(*)}$ scattering, applying it to recent lattice QCD results by Alexandrou et al. [Phys. Rev. Lett. 132, 151902 (2024)]. Two complementary EFT approaches are developed: (1) A low-energy theory near the $B \bar D$ ($J=0$) and $B^* \bar D$ ($J=1$) thresholds, where coupled-channel effects are integrated out; (2) A coupled-channel formulation, where all relevant momentum scales are treated as soft, incorporating contact interactions and one-pion exchange (OPE). Importantly, OPE contributes to the lowest channels only through off-diagonal transitions, thus resulting in the appearance of the left-hand cut from two-pion exchange. The two approaches yield mutually consistent results, supporting the existence of shallow bound states in both channels, in agreement with the lattice findings. The finite-volume spectra and extracted pole positions show a near-degeneracy in $J=0$ and $J=1$ channels, consistent with heavy-quark spin symmetry (HQSS). Using HQSS, we predict additional shallow bound states near the $B \bar{D}^*$ and $B^* \bar{D}^*$ thresholds, which are accessible to future lattice simulations. The effect of OPE on the finite volume spectra is found to be small, with only moderate impact on HQSS partners.

Neutrinoless double beta decay ($0νββ$) may receive sizable contributions from short-range physics beyond the Standard Model. We present a systematical classification of all tree-level ultraviolet completions of the dimension-11 short-range $0νββ$ decay operators, renormalizable scenarios with scalar and fermion mediators are considered. We identify eight distinct topologies and twenty-eight viable diagrams, from which all consistent UV completions are generated by imposing Standard Model gauge invariance. All these models involve a total of 61 new fields beyond the Standard Model and they typically feature fractionally charged fermions and exotic bosons such as dileptons, diquarks, and leptoquarks. We further study a representative model without colored mediators and analyze its implications for the $0νββ$ decay half-life and light neutrino masses. We find that current and future $0νββ$ decay experiments impose stringent constraints. Our systematic decomposition provides a general framework for exploring exotic short-range contributions to $0νββ$ decay in future experiments.

We study gauge theory based on the intermediate Lie algebra $E_{7+1/2}$, interpolating between $E_7$ and $E_8$. We propose a concrete UV completion via a 6d SCFT whose tensor branch description contains a pure $E_{7+1/2}$ gauge sector. The proposal is tested by 6d anomaly cancellation and by the 5d $\mathcal N=1$ Coulomb branch prepotential from the associated M-theory geometry. As a nonperturbative check, we determine the elliptic genus of the single-string worldsheet CFT using modular bootstrap. The result matches the vacuum character of the corresponding VOA for $E_{7+1/2}$, completing the elliptic genus/VOA correspondence along the Deligne-Cvitanović series.

In these proceedings we will review recent progress in applying ideas from the mathematical framework of twisted cohomology to the study of canonical differential equations for Feynman integrals. Firstly, we will show how the intersection matrix can shed some light on the nature of the canonical basis of a Feynman integral family, a concept still not fully understood in the general case. In particular we will show how the intersection matrix can detect hidden linear dependencies of the iterated integrals resulting from an $\eps$-factorized differential equation, which are difficult to find otherwise. Furthermore, we will explain how the intersection matrix can help in deriving (polynomial) relations between the transcendental functions occurring in the rotation to the canonical basis. This allows us to simplify the rotation, and furthermore leads to simplifications in the final result. The discussion we be kept as light as possible, focusing on a simple running example and deferring the technical details to the original publications.

We study the behaviour of the conformal block expansions of scalar fivepoint Lorentzian conformal correlators in the limit where multiple cross ratios approach zero. Since this limit is controlled by intermediate operators with large spin, we use it to study the large spin expansion of the OPE coefficients involving these operators. By imposing bootstrap assumptions such as analyticity of the correlators, we derive an infinite set of new constraints on the large spin behaviour of OPE coefficients involving multiple spinning operators. We also show that for the case of $l=0$, these constraints can be trivialised to all orders in $1/J$ by identifying a pattern in the coefficients.

The Cho-Maison monopole provides a monopole solution of the electroweak field equations, but possesses an infinite classical energy due to the Maxwell form of the hypercharge sector. Motivated by string-inspired effective field theories, we study the perturbative stability of the Cho-Maison monopole when the hypercharge kinetic term is regularised by a Born-Infeld extension, which renders the monopole energy finite. Focusing on the bosonic electroweak theory with an unmodified $SU(2)_L$ sector and a Born-Infeld U(1)_Y sector, we analyze linear fluctuations about the regularised monopole background. Using a complex tetrad and a spin-weighted harmonic decomposition, we reduce the fluctuation equations to coupled radial Schroedinger-type eigenvalue problems and examine the spectrum of the resulting operators. We extend the separation-of-variables framework developed by Gervalle and Volkov to this non-linear gauge-field setting. We show that, after appropriate gauge fixing and constraint elimination, the Born-Infeld deformation preserves the angular channel structure of the Maxwell theory and leads to a self-adjoint Sturm-Liouville type problem for the stability of the radial modes, with modified radial coefficients determined by the background Born-Infeld profile. The resulting operator represents a smooth deformation of the Maxwell case and retains positive kinetic weight. Our results provide plausible evidence for the stability of the Born-Infeld deformed monopole and, most importantly, a systematic framework for future numerical or variational studies aimed at a definitive spectral analysis.

We study the phenomenology of a light charged Higgs boson in the framework of the Alternative Left--Right Symmetric Model (ALRM). In this model, stringent flavor constraints are evaded due to a non-conventional fermion spectrum in which the right-handed up-type quarks are paired with the exotic down-type quarks rather than the Standard Model down-type quarks, leading to the absence of tree-level flavor-changing neutral currents. Furthermore, a specific assignment of the global $U(1)_S$ symmetry and the resulting emergent $R$-parity prevent mixing between the right- and left-handed charged gauge bosons, $W_R$ and $W_L$, providing additional suppression of flavor-violating effects. The ALRM accommodates potentially viable dark matter candidates, both fermionic and scalar ones. In this context, an associated charged Higgs state, $H_2^\pm$, belonging to the dark sector can naturally acquire a sub-TeV to TeV-scale mass without conflicting with any experimental constraints. We focus on scenarios in which $H_2^\pm$ behaves as a long-lived particle due to a sub-GeV mass splitting with the dark matter candidate. We identify regions of parameter space consistent with the observed dark matter relic density and other experimental constraints. A detailed analysis of disappearing track signatures is performed, including realistic tracklet reconstruction efficiencies, and the existing ATLAS search are recast to assess the current limits and future sensitivities. We find that the HL-LHC has limited sensitivity to TeV-scale charged Higgs bosons in this scenario, while the 27 TeV HE-LHC can effectively probe the relevant parameter space, with a 100 TeV collider offering substantially enhanced discovery potential.

We investigate the reaction mechanism of strangeness production in $K^- p \to φΛ$ within a hybrid Regge approach, taking into account $t$-channel $K$- and $K^*$-Reggeon exchanges. We present results for the total cross section, $t$-dependent differential cross sections, and spin-density matrix elements (SDMEs), and compare them with the available experimental data. We find that $K^*$-Reggeon exchange provides the dominant contribution, while $K$-Reggeon exchange remains nonnegligible, particularly in describing the SDMEs. In contrast, the $s$-channel $Λ$ and $u$-channel nucleon exchanges are negligible. To obtain reliable predictions for the open-charm reaction $K^- p \to D_s^{*-} Λ_c^+$, we employ a Quark-Gluon String Model (QGSM)-motivated Regge framework that incorporates both pseudoscalar- and vector-Reggeon exchanges. Within this framework, the Regge trajectories $α(t)$ and energy-scale parameters $s_0$ are determined consistently, thereby constaining the model and reducing theoretical ambiguities. For the hidden-charm reaction $K^- p \to J/ψΛ$, we use the same quark-level diagrammatic correspondence. The total cross sections for $K^- p \to D_s^{*-} Λ_c^+$ and $K^- p \to J/ψΛ$ are suppressed by approximately 5-6 and 8-9 orders of magnitude, respectively, compared with that for $K^- p \to φΛ$. We also examine possible $s$-channel contributions from the hidden-charm pentaquark states with strangeness, $P_{cs}(4337)^0$ and $P_{cs}(4459)^0$, to both $D_s^{*-} Λ_c^+$ and $J/ψΛ$ production.

Motivated by the recently reported resonant structure $X(2300)$, a strong candidate for a fully strange tetraquark with positive parity, we perform a systematic study of fully strange tetra- and penta-quark systems within a chiral quark model. Low-lying $S$-wave configurations of the $ss\bar s\bar s$ and $ssss\bar s$ systems are investigated using the Gaussian Expansion Method (GEM) combined with the Complex Scaling Method (CSM), which allows for a unified treatment of bound, resonant, and scattering states. For tetraquarks, all possible configurations: meson-meson, diquark-antidiquark, and K-type structures, with complete color bases, are incorporated, while baryon-meson and diquark-diquark-antiquark configurations are considered for pentaquarks. Several weakly bound states and narrow resonances are identified in both sectors. In particular, a compact fully strange tetraquark with $J^P=1^+$ is found near $2.3\,\text{GeV}$, providing a natural interpretation of the $X(2300)$ resonance. Additional exotic states with dominant hidden-color and K-type components are predicted in the mass ranges $1.6-3.1$ GeV for tetraquarks and $2.6-3.2$ GeV for pentaquarks. The internal structure of these states is analyzed through their sizes, magnetic moments, and wave-function compositions, highlighting the essential role of channel coupling and exotic color configurations. Finally, promising two-body strong decay channels are proposed to facilitate future experimental searches.

Vector Leptoquarks (VLQs) have emerged as primary candidates for resolving discrepancies in the Standard Model, specifically within $B$-meson decay channels and the anomalous magnetic moment of the muon. This work presents a rigorous evaluation of VLQ pair production across $e^{-}e^{+}$ collision modes at future linear colliders with center-of-mass energies ranging from 14~TeV to 100~TeV. Our analysis demonstrates that longitudinal beam polarization is a transformative tool for enhancing signal sensitivity. We find that $e^{-}e^{+}$ annihilation consistently yields superior cross-sections compared to photon fusion processes across a mass range of 500--3000~GeV. By optimizing beam helicity to specific configurations, such as $P_{e^{-}} = -0.8$ and $P_{e^{+}} = +0.6$, the production cross-section can be maximized to 120~fb at $\sqrt{s} = 3$~TeV. We further establish that the Left-Right Asymmetry ($A_{LR}$) serves as a robust discriminator for the chiral structure of new physics, peaking at 0.16 under full polarization. Additionally, we show that effective luminosity can be enhanced to 95\% of the total luminosity, while high polarization degrees significantly suppress relative uncertainties in the effective polarization. These results provide a quantitative roadmap for optimizing discovery potential and minimizing systematic errors in future high-energy physics experiments.

We generalize neutrino energy density content in the early universe near BBN era $T\simeq1$ MeV within Tsallis nonextensive statistics. By using Curado-Tsallis constraints we obtain generalized distribution functions $f_q(E)$. We compute the generalized thermodynamic integral for the energy density $ρ_q$. We define a reescaling $R^{(ξ= +1)}_ρ(q) = ρ_q/ρ^{\rm std}$ which is a ratio between the deformed energy density and the standard extensive case. The last was used to directly map and deform neutrino content via the effective number of neutrinos $N_{\rm eff}$. The deformation prediction was confronted against CMB$+$BAO and BBN data for $N_{\rm eff}$ by a joint/combined $χ^2$ type-fit. We obtained the constraints $|q-1|\le 1.09\times 10^{-2}$ (95\% CL) and $|q-1|\le 1.32\times 10^{-2}$ (99\% CL) from the combined analysis by numerically calculating the best value of the Tsallis parameter $q_{\rm best}$.

Topological states of matter are characterized by nonlocal structures that are naturally encoded in the quantum entanglement of many-body wavefunctions. Topological semimetals are short-range entangled states at weak coupling and their entanglement structure at strong coupling remains largely unexplored. In this work, we investigate the multipartite entanglement structure of strongly coupled holographic nodal line semimetals. Building on previous studies of entanglement entropy and the holographic c-function, we focus on multipartite entanglement measures, including the conditional mutual information, multi-entropy, and the Markov gap which is based on the entanglement wedge cross section. Our results demonstrate that while these multipartite measures vanish in the long-distance limit $l \to \infty$, which confirms that the holographic nodal line semimetal remains a short-range entangled state, their large $l$ scaling behavior remains highly sensitive to the underlying topology. The large $l$ power-law decay and scaling exponents serve as robust, non-local order parameters that exhibit sharp changes at the quantum critical point. This work establishes multi-partite entanglement as a powerful probe of quantum topological phase transitions in strongly coupled topological systems.

The values of the Hubble constant ($\rm{H_0}$) inferred from the cosmic microwave background (CMB) and local measurements via the distance ladder exhibit a $\sim5σ$ tension. In this work we propose that the tension might be partially alleviated if a subcomponent of the dark matter undergoes decays triggered by spontaneous symmetry breaking in the dark sector, so that the equation of state parameter of the subcomponent shifts from $w \approx 0$ at early times to $w \approx -1/3$ at late times. We provide an effective field theory whose structure is partially motivated by the desire for a plausible UV completion. We find that such a construction naturally produces a possible self-interacting dark matter candidate with a velocity-dependent scattering cross section as a by-product of gauge invariance. This is relevant for addressing tensions between the predictions of $Λ$CDM and observations of small-scale structure, such as the core-cusp problem.

As an extension of our prior work, we analyze the resonance contributions for the kaon pair originating from the intermediate $ρ(770)$, $ω(782)$ and their excited states in the three-body decays $B\to η^{(\prime)} K\bar{K}$ within the perturbative QCD approach. The information of subprocesses $ρ(770,1450,1700)\to K\bar K$ and $ω(782,1420,1650)\to K\bar K$ are included in the distribution amplitudes for $K\bar K$ system by using the kaon vector time-like form factors. We calculate the $CP$ averaged branching fractions and the direct $CP$ asymmetries for the relevant quasi-two-body $B$ meson decays. The branching fractions of the virtual contributions for $K\bar K$ from the Breit-Wigner formula tails of $ρ(770)$ and $ω(782)$ for these decays are found comparable to the corresponding contributions from the resonances $ρ(1450,1700)$ and $ω(1420,1650)$. Consequently, they constitute a significant component that should be accounted for in the considered three-body decays. All the predictions in this work are expected to be tested by the LHCb and Belle-II experiments in the future.

Identifying which observables most effectively constrain model parameters can be computationally prohibitive when considering full likelihoods of many correlated observables. This is especially important for, e.g., hadronization models, where high precision is required to interpret the results of collider experiments. We introduce the High-Dimensional Sensitivity (HDSense) score, a computationally efficient metric for ranking observable sets using only one-dimensional histograms. Derived by profiling over unknown correlations in the Fisher information framework, the score balances total information content against redundancy between observables. We apply HDSense to rank a set observables in terms of their constraining power with respect to five parameters of the Lund string model of hadronization implemented in Pythia using simulated leptonic collider events at the $Z$ pole. Validation against machine-learning--based full-likelihood approximations demonstrates that HDSense successfully identifies near-optimal observable subsets. The framework naturally handles data from multiple experiments with different acceptances and incorporates detector effects. While demonstrated on hadronization models, the methodology applies broadly to generic parameter estimation problems where correlations are unknown or difficult to model.

Three-dimensional spiral injection scheme enables storage in ultra-compact rings with nanosecond revolution period. We report the first successful storage of a $297\,\mathrm{keV/}c$ electron beam in a $22\,\mathrm{cm}$ weak-focusing storage ring with a $4.7\,\mathrm{ns}$ revolution period using multi-turn vertical kick with a $140\,\mathrm{ns}$ kicker pulse. Using a scintillating-fiber detector, we observe a signal exceeding $5σ$ of the pre-injection rms noise for $\geq 1\,\mathrm{μs}$, confirming beam storage. By varying the weak-focusing field configuration and measuring the stored beam distribution, we show that the storage beam resides within the predicted region by Monte Carlo simulations. This result is a key proof-of-principle for realizing ultra-compact storage rings for next-generation precision measurements including the muon experiments at J-PARC and PSI.

Polylogarithms on Riemann surfaces may be constructed efficiently in terms of flat connections that can enjoy various algebraic and analytic properties. In this paper, we present a single-valued and modular invariant connection ${\cal J}_\text{DHS}$ on the configuration space $\text{Cf}_n(Σ)$ of an arbitrary number $n$ of points on an arbitrary compact Riemann surface $Σ$ with or without punctures. The connection ${\cal J}_\text{DHS}$ generalizes an earlier construction for a single variable and is built out of the same integration kernels. We show that ${\cal J}_\text{DHS}$ is flat on $\text{Cf}_n(Σ)$. For the case without punctures, we relate it to the meromorphic multiple-valued Enriquez connection ${\cal K}_\text{E}$ in $n$ variables on the universal cover $\tilde Σ$ of $Σ$ by the composition of a gauge transformation and an automorphism of the Lie algebra in which ${\cal J}_\text{DHS}$ and ${\cal K}_\text{E}$ take values. In a companion paper, we shall establish the equivalence between the flatness of these connections and the corresponding interchange and Fay identities, for arbitrary compact Riemann surfaces.

This paper E3 shows how to construct the simplest Exotic Invariant in the simplest way.

We show that a quadratic $U(1)$-breaking term, together with an effective chemical potential induced by a dimension five derivative coupling between the inflaton and the $U(1)$ current, can drive efficient particle production during inflation even when the $U(1)$ field is heavier than the Hubble scale. Notably, the chemical potential enables efficient production even when the $U(1)$-breaking mass is smaller than the effective diagonal mass. We compute the gravitational wave signal generated by this mechanism during inflation, derive the primordial tensor spectrum, and map it to the present day energy density $Ω_{\mathrm GW}(f)$. Assuming the $U(1)$ field constitutes the dominant component of dark matter, this mapping fixes the characteristic frequency, which we compare with projected sensitivity curves of ongoing and proposed gravitational wave observatories. Finally, we argue that the same dynamics are accompanied by a cosmological collider signal, providing an independent cross validation of the framework.

We investigate CP violation in the process H^\pm \to W^\pm Z within the framework of the two-Higgs-Doublet model (2HDM). Amplitudes are expressed transparently in terms of physical couplings. Our analysis qualitatively confirms recent results by Kanemura and Mura, for the interference between one-loop bosonic and fermionic amplitudes. Furthermore, we identify additional sources of CP violation arising from internal interference among the bosonic and also among the fermionic loop amplitudes. In the alignment limit, the asymmetry would vanish in the absence of fermionic loop contributions when the sum of the masses of the additional neutral scalars is higher than the mass of the Z. Our results allow for generic Yukawa couplings, and cover both explicit and spontaneous CP violation. For inclusive W^\pm and Z decays, the CP violation will only manifest itself as a charge asymmetry.

We review the structure of superconformal anomalies in 4d $\mathcal N$=4 conformal supergravity (CSG) coupled to a number N$_\rm v$ of $ \mathcal N$=4 vector multiplets and 6d (2,0) CSG coupled to N$_{_{\rm T}}$ of (2,0) tensor multiplets. Anomalies cancel if N$_\rm v$=4 and N$_{_{\rm T}}$=26 respectively. If the CSG part of the action is dropped and N$_{\rm v}$=6+ n$_{\rm v}$, the first theory is classically equivalent to the 4d $\mathcal N$=4 Poincaré supergravity (PSG) coupled to n$_{\rm v}$ vector multiplets, while the second one with N$_{_{\rm T}}$=5+ n$_{_{\rm T}}$ is classically equivalent to the 6d (2,0) PSG coupled to n$_{\rm T}$ tensor multiplets. We argue that these facts imply that divergences in the 4d PSG with n$_{\rm v}$ vectors should be proportional to n$_{\rm v}$+2 and similarly in the 6d PSG with n$_{_{\rm T}}$ tensors to n$_{_{\rm T}}$-21. These predictions appear to be consistent with known results of explicit scattering amplitude computations in these 4d and 6d PSG theories.

Using $(10087 \pm 44) \times 10^6$ $J/ψ$ events collected with the BESIII detector at a center-of-mass energy of $\sqrt{s}=3.097$ GeV, the antihyperon-nucleon annihilation processes $\barΛ p \to K^+ π^+ π^- + kπ^0$ ($k=1,2,3$) are studied at an incident $\barΛ$ momentum of approximately 1.074 GeV/$c$. The reactions $\barΛ p \to K^+ π^+ π^- π^0$ and $\barΛ p \to K^+ π^+ π^- 2π^0$ are observed for the first time, with corresponding cross sections $σ_{\barΛ p \to K^+ π^+ π^- π^0} = 8.5^{+1.2}_{-1.1} (\rm{stat.}) \pm 0.4 (\rm {syst.})$ mb and $σ_{\barΛ p \to K^+ π^+ π^- 2π^0} = 7.9^{+1.9}_{-1.7} \pm 0.4$ mb. No significant signal is found for $\barΛ p \to K^+ π^+ π^- 3π^0$, and an upper limit of 7.2 mb is set at a 90\% confidence level. An evidence for the $K^{*}(892)^+$ resonance is seen in the $K^+π^0$ invariant mass spectrum $M_{K^+π^0}$ for $k=1$, and the corresponding cross section for $\barΛ p \to K^{*}(892)^+ π^+ π^-$ is measured to be $σ_{\barΛ p \to K^{*}(892)^+ π^+ π^-} = 12.5^{+3.8}_{-3.4} \pm 1.2$ mb. Owing to the limited statistics, possible interference effects are not considered. These findings offer crucial input to deepen our understanding of the antihyperon-nucleon interactions.

In this review, the fundamental concepts of group theory and representation theory are introduced. Special emphasis is placed on the unitary irreducible representations of the $SU(N)$ Lie group, the Poincare group, Little Group, discrete group, and their applications in particle physics. Based on the principle of local gauge symmetry, the construction of gauge-invariant Lagrangians and their quantization procedure are discussed. To address gauge redundancy, the modern on-shell amplitude approach is applied to gauge theories, demonstrating both conceptual and computational advantages. From the perspective of symmetry, the Standard Model is presented through the identification of its gauge symmetry, its anomaly-free matter content, and its global symmetries, including flavor symmetry, custodial symmetry, and baryon and lepton number conservation, etc.

We compute the contributions of the dimension-6 SMEFT operators involving four third-generation quarks to the two-loop renormalisation of the quark fields and masses. We perform the computations in both the Naive Dimensional Regularisation (NDR) and the Breitenlohner-Maison-`t Hooft-Veltman (BMHV) schemes, and we present results for all relevant field and mass renormalisation constants in the on-shell and MS-bar schemes. We carefully discuss all scheme choices which affect the final results. This completes the computation of field and mass renormalisation constants relevant for two-loop computations in QCD involving dimension-six SMEFT operators.

Constraints on the birefringence-free subset of Lorentz-violating (LV) operators are derived using 14 GRB photons in the GeV-band. These constraints target the isotropic $c_{(I)00}^{(d)}$ coefficients for dimensions $d=6,8,10$ within the framework of the Standard-Model Extension (SME). Employing theory-agnostic Bayesian parameter estimation methods, our analysis indicates a preference for subluminal LV effects. Focusing on this case, we further refine the parameter constraints, yielding results that are mutually consistent. Within the 95\% posterior credible interval, our constraints yield the bounds, $|c_{(I)00}^{(6)}|\le7.75 \times 10^{-20} ~ {\rm GeV}^{-2}$, $|c_{(I)00}^{(8)}|\le4.92 \times 10^{-24} ~ {\rm GeV}^{-4}$, and $|c_{(I)00}^{(10)}|\le3.46 \times 10^{-28} ~ {\rm GeV}^{-6}$, which improve upon the most stringent credible-interval bounds reported in the literature by at least five orders of magnitude.

We calculate the fragmentation functions for a gluon into $P$-wave $B_c$ mesons within the nonrelativistic QCD factorization framework, incorporating color-singlet and color-octet contributions. Ultraviolet divergences arising from phase-space integrals are removed via operator renormalization in the modified minimal subtraction scheme. The resulting fragmentation functions are presented in both graphical form and as fitted analytic expressions.

The Belle and Belle~II experiments have accumulated a data set of $1.2~\mathrm{ab}^{-1}$ of $e^+e^- \to B\bar{B}$ collisions at the $Υ(4S)$ resonance. Owing to the clean event environment and well-constrained initial-state kinematics, these data are ideally suited for searches for rare electroweak penguin and lepton-flavour violating $B$ decays with missing energy from neutrinos. We report results on $b\to sν\barν$ processes and the interpretation, together with searches for $B\to K^{*0}τ^+τ^-$ and for the LFV decays $B^0\to K_S^0τ^\pm\ell^\mp$ and $B^0\to K^{*0}τ^\pm\ell^\mp$ ($\ell=e,μ$).

Wilson loops are among the most fundamental gauge-invariant observables in quantum field theory, encoding the global structure of gauge fields through their holonomy along closed contours. Originally introduced as order parameters for confinement in non-Abelian gauge theories, they have recently acquired a central role in condensed matter physics, where they characterize topological phases and quantized transport phenomena. In this work we present a unified theoretical picture in which Wilson loops connect nonperturbative gauge dynamics, Berry-phase topology in band theory, and the quantum Hall response of interacting electron systems. We demonstrate explicitly how Wilson loops encode Chern numbers, fractional charge, and anyonic braiding statistics within Chern--Simons effective field theory. Both quantized Hall conductivity and quasiparticle statistics are shown to originate from the same topological invariant -- the linking number of Wilson loops -- establishing a direct correspondence between microscopic topological structure and macroscopic transport.

Topology plays a central role in classifying solitonic configurations in field theories, providing robustness and a nonperturbative label, the so-called topological charge $Q$. In soliton-fermion coupled systems, the relation between the topological charge and the number of zero modes is well established through the index theorem. However, the physical consequences of the sign of the topological charge have remained largely unexplored. In this work, we study fermions in $2+1$ dimensions coupled to Skyrmions with target space $\mathbb{C}\mathrm{P}^N$, particularly focusing on the backreactions of the fermions and on the sign of the topological charge. We obtain the solutions in a self-consistent manner, which exhibit an asymmetry with respect to the topological charge $\pm Q$ especially in the strong coupling regimes. This asymmetry is caused from the fermionic eigenvalue problem inherent in the self-consistent formulation. Although the Lagrangian is symmetric under $Q\to-Q$, the coupled equations for the Skyrmions and anti-Skyrmions become inequivalent once fermionic backreaction is taken into account. We demonstrate the mechanism in $\mathbb{C}\mathrm{P}^1$ and $\mathbb{C}\mathrm{P}^2$ Skyrmions, but the analysis is directly extendable for the general $\mathbb{C}\mathrm{P}^N$.

The possibility to study the nature of the famous charged bottomonium-like states $Z_b(10610)$ and $Z_b(10650)$, which is by far the best known, from their inclusive photoproduction off nuclei near the kinematic threshold is investigated within the collision model based on the nuclear spectral function. The model accounts for $Z_b(10610)^{\pm}$ and $Z_b(10650)^{\pm}$ production in direct photon--nucleon interactions as well as four different scenarios for their intrinsic configurations: compact tetraquarks, molecules of the two open-beauty mesons and two mixtures of both of them for each of $Z_b$ state. We calculate within these scenarios the absolute and relative excitation functions on $^{12}$C and $^{184}$W nuclei at photon energies of 61--90 GeV, the absolute momentum differential cross sections and ratios of them for their production off these target nuclei at laboratory polar angles of 0$^{\circ}$--5$^{\circ}$ and for photon energy of 75 GeV as well as the A-dependences of the transparency ratios for the $Z_b(10610)^{\pm}$ mesons at photon energy of 75 GeV. We show that the absolute and relative observables considered reveal distinct sensitivity to the $Z_b(10610)^{\pm}$ and $Z_b(10650)^{\pm}$ internal structures. Therefore, they might be useful for the determination of these structures from the comparison of them with the experimental data from the future high-precision experiments at the upcoming experimental facilities, such as the planned high-luminosity electron-ion colliders in the United States and China.

We study the Coulomb branches of three-dimensional $\mathcal N=4$ quiver gauge theories of type $T_ρ(SU(N))$ associated with non-maximal nilpotent orbits of $SL(N)$. Using the Hall--Littlewood closed form for Coulomb-branch Hilbert series, together with independent checks from the monopole formula, we compute exact unrefined Hilbert series for all non-maximal partitions $ρ\vdash N$ with $N=4$, and extend the analysis to $N=5,6$. By analyzing the plethystic logarithms of the resulting Hilbert series, we find that in all cases examined the Coulomb branch is a complete intersection. The number of generators and relations follows a uniform pattern governed by the transpose partition $ρ^T$, with exactly $N-1$ relations appearing independently of $ρ$ in these examples. We summarize the results in explicit classification tables and formulate conjectures extending these patterns to arbitrary $N$. Our findings provide strong evidence for a remarkable uniformity in the algebraic structure of Coulomb branches within the $T_ρ(SU(N))$ family at low rank.

In this work, we explore the discovery potential of Vector-Like Singlet Top quarks ($T$) at a future $μp$ collider with center-of-mass energies of 5.29, 6.48, and 9.16 TeV, providing a unique environment to probe beyond Standard Model limits. We analyze the $T \to Wb$ decay mode in both fully hadronic ($bjj$) and leptonic ($blν$) final states, offering a multi-channel assessment of $T$-quark sensitivity across a mass range of 2 to 5 TeV. Our methodology employs multivariate classifiers such as Boosted Decision Trees (BDTs) and Multi-Layer Perceptrons (MLP) to optimize signal-to-background discrimination in complex final states. The results demonstrate that the 9.16 TeV benchmark acts as a definitive discovery machine; even with 100 fb$^{-1}$ of data, the statistical significance exceeds $5σ$ up to 4 TeV masses. We identify a crossover effect where hadronic channels provide superior reach at intermediate masses due to higher branching ratios, while leptonic channels offer robustness at 5 TeV where purity limits detection. Incorporating a 20\% systematic uncertainty via Asimov significance ($Z_A$), we quantify the transition from fluctuation-dominated to systematic-dominated regimes at high luminosities. At 3000 fb$^{-1}$, regions with $g^{*} \in [0.20, 0.50]$ and $m_T$ up to 4 TeV are discoverable via the hadronic channel with MLP, and regions with $g^{*} \in [0.10, 0.50]$ and $m_T$ up to 5 TeV are accessible through the leptonic channel with BDT, highlighting the collider's potential to probe new physics beyond the Standard Model.

We present NeutrinoOsc3Flavor, a lightweight and fully transparent computational framework for exact three flavor neutrino oscillation studies in vacuum and constant density matter. The code numerically solves the Schrodinger evolution equation in the flavor basis using explicit construction and diagonalization of the effective Hamiltonian within the PMNS formalism, including full CP Violating phase dependence. In contrast to large scale oscillation toolkits optimized for experimental simulations, NeutrinoOsc3Flavor is designed as a minimal dependency reference implementation, emphasizing analytical traceability, equation level accessibility, and cross platform portability. The framework relies solely on NumPy for numerical linear algebra and runs natively on both Linux and Windows systems without external compilation or specialized libraries. As an internal consistency and validation feature, we implement an independent analytical determination of the matter modified Hamiltonian eigenvalues using the Cardano method and demonstrate excellent agreement with numerical diagonalization. CP Phase dependence is used as a sensitive diagnostic of numerical stability and correctness of the evolution operator in both vacuum and matter. NeutrinoOsc3Flavor is intended as a verification oriented and pedagogical computational tool, suitable for theoretical cross-checks, educational use, and benchmarking of more complex neutrino oscillation software, rather than as a replacement for full experimental simulation frameworks. Here, we consider the DUNE experiments baseline length in the python implementation but in general we can implement any value of baseline length.

Ultra-fast outflows (UFOs) are thought to be a driving mechanism of large-scale winds driven by active galactic nuclei, which cause significant galactic feedback through quenching star formation and regulating supermassive black hole growth. We present James Webb Space Telescope (JWST) Mid-Infrared Instrument Medium-Resolution Spectrometer observations of two nearby ultraluminous infrared galaxies (ULIRGs), F11119+3257 and F05189-2524, with nuclear X-ray detected UFOs and kiloparsec-scale outflow. These galaxies show remarkably similar mid-infrared continuum and emission line features, notably including a high-velocity $v_{90}$ $\sim$ 4000 km s$^{-1}$ outflow detected in highly ionized neon emission lines, e.g., \nevi. In F05189-2524, we see a slightly slower biconical outflow extending up to $\sim2$ kpc in the same neon emission lines. Both sources show evidence of AGN-driven radiative feedback through a deficit of rotational molecular hydrogen lines in the nuclear region, $<$1 kpc from the central quasar, but no clear evidence of any molecular gas entrained in the quasar-driven outflow. Energetic analysis shows that the warm ionized gas in both of these sources contributes minimally ($\sim0.1-5\%$) to the momentum outflow rate of these sources and leaves the conclusions of previous literature unchanged: the energetics of these sources are broadly consistent with a momentum-conserving outflow.

Numerical simulations are a key tool to decipher the dynamics of gravitation. Yet, they fail to spatially reproduce the Universe we observe, limiting comparison between observations and simulations to a statistical level. This is highly problematic for rare, faint or well studied nearby objects that are observed in a single environment. The computational cost of recovering this environment in random simulations is prohibitive. We present Hamlet-PM, a method that enables the constraining of initial conditions for cosmological simulations so as to produce evolved numerical universes that can be directly compared to observations of the Local Universe: constrained simulations. Our method implements the field-level forward modeling of the early-time density field from sparse and noisy measurements of late-time peculiar velocities. The dynamics are integrated with a particle-mesh gravity solver, thus probing the mildly non-linear regime. The code is applied to the Cosmicflows-4 compilation of peculiar velocities up to z < 0.05 (160 Mpc/h). The constrained ICs a re-simulated with a high precision N-body code. A series of one hundred dark-matter only cosmological constrained simulations with a resolution of 512^3 particles in a 500^3 [Mpc/h]3 box is presented. Special attention is given to twelve prominent nearby galaxy clusters, whose simulated counterparts are matched on criteria of mass and separation. We provide a mass estimate constrained by the dynamical environment for each cluster. Field-level forward modeling of the initial conditions produces highly constrained cosmological simulations. Currently, this method already overtakes in quality the pipeline in use in the peculiar-velocity community, although systematic biases still need to be addressed. Furthermore, improving the model is easy thanks to the inherent flexibility of the Bayesian approach.

Interstellar medium studies rely on magnetohydrodynamic (MHD) turbulence as a framework for interpretation. In this context, the statistical characterization of interstellar observations is of prime importance. We open a new perspective on diffuse interstellar matter by analyzing James Webb Space Telescope (JWST) observations of the Pleiades nebula with NIRCam. These observations are remarkable in that they provide a microscope view at the cold neutral medium (CNM) with a spatial resolution of 0.2 mpc (40 au). A two-dimensional Fourier analysis is used to characterize the structure of PAH emission in regions near and far from the Pleiades star Merope. To produce maps of the interstellar emission, stars and galaxies are filtered out. The final step in the data cleaning involves subtracting a component, in Fourier space, which we infer to be a residual of the near-infrared cosmic background. The PAH emission power spectra are highly anisotropic. They are well fitted with a break-free power-law, suggesting that we do not observe a specific scale for energy dissipation. Power-law indices are -3.5 near Merope and -3 in the more distant field. The magnetic field orientation, as derived from the Planck dust polarization data, aligns with the PAH anisotropy. The power anisotropy is constant across scales. These findings are discussed in relation to interstellar turbulence that may be driven by the Pleiades stars. The JWST observations of the Pleiades offer a new viewpoint for comparing observations and theoretical models, as they examine physical scales at which turbulence in the CNM is subsonic and decoupled from the thermal instability. The observations may indicate that the turbulent energy cascade in the CNM is anisotropic.

Deuterium fractionation is highly efficient during the early stages of star formation, particularly in starless and prestellar cores where temperatures are low (<10 K) and molecular freeze-out onto dust grains is significant. Methanol forms early in these environments following CO freeze-out via successive hydrogenation reactions on grain surfaces, while the production of deuterated methanol requires elevated gas-phase D/H ratios generated through dissociative recombination of deuterated H3+. Consequently, large abundances of deuterated methanol are observed toward young stellar objects where prestellar ices have recently sublimated. Here, we present laboratory infrared spectra of methanol and its deuterated isotopologues in astrophysical ice analogues, complemented by anharmonic vibrational calculations used to guide band assignments. Experiments were performed at the CASICE laboratory using a Bruker Vertex 70v spectrometer coupled to a closed-cycle helium cryostat, with isotopologue ices deposited at 10 K under high-vacuum conditions. Infrared transmission spectra were recorded over 6000-30 cm-1 (1.67-333 um) and compared with spectra of pure isotopologue ices. Distinctive mid-infrared band patterns are identified for each deuterated species. In particular, CH2DOH exhibits a characteristic doublet at 1293 and 1326 cm-1 (7.73 and 7.54 um), while CHD2OH shows a similar doublet at 1301 and 1329 cm-1 (7.69 and 7.52 um), both remaining largely invariant across all studied ice mixtures. These robust spectral signatures provide reliable tracers for identifying deuterated methanol in JWST observations and for constraining astrochemical gas-grain models of deuterium enrichment prior to star and planet formation.

Type II supernovae (SNe II) are the most common terminal stellar explosions in the Universe. With SNe now being detected within days after explosion, there is growing evidence that the majority of Type II SNe show signs of interaction with a confined, dense cirumstellar material (CSM) in the first few days post explosion. In this work we aim to bridge the gap between single SN studies showing early-time interaction in their spectra, and the statistical studies of early-time SN light curves, which imply the existence of CSM. We present a sample of 68 Type II SNe with both early photometric data, obtained with the ATLAS survey, and spectroscopic data, obtained with the ePESSTO+ collaboration. A subset of the sample is classified based on the presence or absence of narrow spectral features with electron-scattered broadened wings in the early spectra, indicative of interaction with CSM. We characterise the photometric and spectroscopic properties of the sample by measuring rise times to maximum light, peak magnitudes, decline rates and line velocities. Additionally, we measure the ratio of absorption to emission (a/e) of the H alpha P-Cygni profile. Our analysis reveals that SNe II showing early spectroscopic signs of interaction with CSM decline faster and are brighter than those without. However no difference is found in rise times between the two groups. A clear separation is observed in the a/e ratio: SNe with signs of interaction exhibit lower a/e ratios at all epochs compared to those without. Our results highlight that understanding SN II ejecta-CSM interaction requires large, uniform samples of photometric and spectroscopic data, such as the one presented in this work.

In isolated binary evolution, binary black hole (BBH) mergers are generally formed through stable mass transfer (SMT) or common envelope evolution. In recent years, the SMT channel has received significant attention due to detailed binary models showing increased mass transfer stability compared to previous studies. In this work, we perform a full zero-age-main-sequence to compact object merger analysis using detailed binary models at eight metallicities between $10^{-4}Z_\odot$ and $2Z_\odot$ to self-consistently model the population properties of BBH mergers in the SMT channel, determined their progenitor initial conditional, and investigate the binary physics governing their formation and metallicity dependence. We use the population synthesis code POSYDON to determine the population of BBH mergers from SMT. Using its extended grids of MESA binary models, we determine the essential physics in the formation of BBH mergers. SMT produces BBH mergers predominantly from systems with $P_{ZAMS}\leq10$ days. In these systems, both the initial mass transfer between two stars and the subsequent interaction between the remaining star and the first-born BH take place while the respective donor star is on the main-sequence (Case A). We find a limited contribution from wider Case B/C systems. Without a natal kick, the SMT channel does not produce BBH mergers above $Z>0.2Z_\odot$ due to orbital widening from stellar wind mass loss. The primary BH mass distribution shows a strong dependence on metallicity, while the mass ratio prefers unity independent of metallicity due to mass ratio reversal. Additionally, the $χ_{eff}$ distributions contain peaks at $χ_{eff}=0$ and ~0.15 of which the former disappears at high metallicities. A mass-scaled natal kick leave this sub-population unchanged but introduce a low-mass, unequal mass ratio sub-population that merges due to their high eccentricity.

Narrow interstellar absorption features in supernova (SN) spectra serve as valuable diagnostics for probing dust extinction and the presence of circumstellar or interstellar material. In this third paper in a series, we investigate how the strength of narrow interstellar absorption lines in low-resolution spectra varies with SN type and host galaxy properties, both on local and global scales. Using a dataset of over 10000 spectra from $\sim1800$ low-redshift SNe, we find that Type Ia SNe (SNe Ia) in passive galaxies exhibit significantly weaker narrow absorption features compared to CC-SNe and SNe Ia in star-forming hosts (SNe Ia-SF), suggesting lower interstellar gas content in quiescent environments. Within the star-forming hosts, the Na I D equivalent-width distribution of SNe II is much lower than that of both SNe Ia-SF and stripped-envelope SNe (SE-SNe). This result is somewhat unexpected, since CC-SNe are generally associated with star-forming regions and occur deeper within galactic disks, where stronger line-of-sight extinction would be anticipated. This suggests that the observed behaviour cannot be explained solely by absorption from the integrated interstellar medium (ISM) along the line of sight. Instead, if part of the absorption arises from material near the explosion, the similarity between the Na I D EW distributions of SNe Ia-SF and SE-SNe implies that comparable absorption signatures can emerge from distinct progenitor pathways. Possible explanations include (a) circumstellar material (CSM) expelled by the progenitor system before explosion, or (b) interaction of SN radiation with nearby patchy ISM clouds. Our results highlight the diagnostic power of interstellar absorption features in revealing the diverse environments and progenitor pathways of SNe.

In this paper, we extend a paper by Milgrom (2009, MNRAS 398, 1023) dealing with the existence of a quasi-universal surface density for object of all mass and structure, if they are in the Newtonian regime, i.e., that their mean acceleration is larger than MOND typical acceleration $a_0$. This result is in agreement with Donato et al. (2009)'s results, claiming the existence of a quasi-universal surface density in all masses in galaxies. The Milgrom paper also predicts that objects with mean inner acceleration smaller than the values discussed do %es not show the quasi-universal behavior of the surface density discussed. In the present paper, we extend the result of Milgrom's paper, based on a point mass model, considering spiral galaxies, modelled with a double exponential disk. Similar to Milgrom's results, we find the existence of a universal surface density for galaxies with large surface density, and a different behavior for galaxies having small surface density.

We present and analyze photometric and spectroscopic observations of the Type Ia supernova (SN Ia) 2024igg, another ``super-Chandrasekhar'' (or 03fg-like) SN whose strong C II $\lambda6580$ feature was initially misidentified as H$α$, thereby constraining its progenitor system, explosion parameters, and physical scenario. SN 2024igg shows many characteristics in common with other 03fg-like objects, such as high ultraviolet flux, slowly declining light curves ($Δm_{15}(B)=0.90\pm0.08$ mag), low expansion velocities, along with strong and persistent C II absorption. Meanwhile, this SN exhibits some remarkable properties within this subgroup, including a moderately low optical luminosity ($M_{\rm max}(B)=-18.99\pm0.15$ mag), a short rise time less than 18.5 days, and strong C II $\lambda6580$. The bolometric analysis yields a $^{56}$Ni mass of $M_{\rm Ni}=0.547\pm0.082$ $M_{\rm \odot}$ and an ejecta mass of $1.54^{+0.22}_{-0.19}$ $M_{\rm \odot}$, marginally exceeding the Chandrasekhar mass. Our TARDIS result indicates that most of the features in the earliest spectrum could be attributed to C II, which is consistent with a model where a supernova explodes within a carbon-rich circumstellar medium (CSM). The CSM interaction would produce a density peak in the ejecta, offering a natural explanation for the slowly evolving line velocities near $-$8000 km s$^{-1}$. The CSM may stem from the debris of a secondary white dwarf in a white-dwarf merger or the envelope of an asymptotic giant branch star. Combined with the unshifted forbidden lines in the spectrum taken at $t\approx\ +$135 days, we suggest that SN 2024igg comes from a symmetric explosion on a secular timescale after the merger.

The dynamics of the inner regions of young stellar objects (YSOs) is driven by a variety of physical phenomena, from magnetospheres and accretion to the dust sublimation rim and inner disk flows. These inner environments evolve on timescales of hours to days, exactly when bursts, dips, and rapid structural changes carry the most valuable information about star and planet formations, but remain hardly reachable with current facilities. A better reactive infrastructure with six or more telescopes, combined with alerts from large time-domain surveys (e.g., at the era of LSST/Rubin type facilities), and equipped with instruments spanning from the V-band to the thermal infrared (N), would provide the instantaneous uv-coverage and spectral diagnostics needed to unambiguously interpret and image these events as they happen. Such a world's first time-domain interferometric observatory would enable qualitatively new science: directly linking optical and infrared variability to spatially resolved changes in magnetospheric accretion, inner-disk geometry, and dust and gas dynamics in the innermost astronomical unit. Crucially, connecting these processes to outer-scale unresolved information from JWST, ALMA, and the ELT would yield a complete tomography of the planet-forming region.

A major source of uncertainty in modeling the strangeness-rich interiors of neutron stars arises from the poorly constrained two-body and three-body interactions among hyperons and nucleons. We perform a comprehensive Bayesian analysis of the $ΛΛ$ and $ΛΛN$ interaction parameters within the Skyrme Hartree-Fock framework, constrained by both hypernuclei experimental data and astrophysical observations. Our results show that the parameter space of the $ΛΛ$ interaction is tightly constrained by combining nuclear and astrophysical data, while the parameters of the $ΛΛN$ three-body interaction remain sensitive to astrophysical inputs alone. Specifically, the local, momentum-independent two-body interaction parameter $λ_0$ is tightly constrained and predominantly attractive, while the momentum-dependent parameters $λ_1$ and $λ_2$ contribute repulsive effects at high densities. A key role is played by the $ΛΛ$ potential depth in pure $Λ$ matter, which effectively constrains the two-body $ΛΛ$ interaction and governs the balance between attraction at low densities and repulsion at high densities. The repulsive components of $ΛΛ$ interactions then decrease hyperon fractions and reconcile hyperon-rich equations of state with the observed $\sim2\,M_{\odot}$ neutron stars, increasing the maximum mass by up to 22\%. The inclusion of $ΛΛN$ three-body forces further stiffens the EOS, raising the maximum mass by up to $\sim 0.1\,M_{\odot}$. Our study represents a promising step toward a complete, experimentally grounded description of dense matter across a wide range of densities and strangeness compositions.

We explore the potential of cross-correlating galaxies and peculiar velocities on non-linear scales to enhance cosmological constraints. Leveraging the \textsc{AbacusSummit} simulation suite and the halo occupation distribution (HOD) formalism, we train emulator models to describe the non-linear clustering of galaxies and velocities in redshift space. Our analysis demonstrates that combining galaxy and peculiar velocity clustering, provides tighter constraints on both HOD and cosmological parameters, particularly on $σ_8$ and $w_0$. We further apply our models to realistic mock catalogues, reproducing the expected density and peculiar velocity errors of type-Ia supernovae and Tully-Fisher/fundamental plane measurements for the combined ZTF and DESI measurements. While systematic biases arise in the HOD parameters, the cosmological constraints remain unbiased, yielding $3.8\%$ precision measurement on $fσ_8$ compared to $4.7\%$ using galaxy clustering alone. We demonstrate that, while combining tracers with realistic velocity measurements still yields improvement, the gains are diminished, highlighting the need for further efforts to reduce velocity measurement uncertainties and correct observational systematics on small scales.

The 21 cm intensity mapping provides a promising probe of the large-scale structure. Astrophysical foregrounds, as the main source of contamination to the cosmological 21 cm signal, persist in a wedge-like region of Fourier space due to the inherent chromaticity in radio interferometric observations. The foreground avoidance strategy focuses on utilizing data from relatively clean regions with minimal foreground leakage, at the cost of losing large-scale information. Non-linear structure formation, however, couples Fourier modes across scales, leaving imprints of the missing large-scale modes in the remaining data. In this work, we employ a deep learning approach to test whether large-scale features of the 21 cm brightness temperature fields, particularly the baryon acoustic oscillations (BAO), can be recovered at the field level using only short-wavelength modes that are beyond the linear scales. To explicitly assess the dependence on the training cosmology, we train the network exclusively on de-wiggled simulations, providing a controlled test of whether the reconstruction arises from physical non-linear mode coupling rather than implicit encoding of BAO features. In the ideal noise-free case, the amplitude and phase of the lost modes can be restored with high fidelity. With instrumental noise included, the reconstructed amplitude becomes biased, while the phase information remains robust. The trained network also exhibits reasonable robustness to variations in the underlying cosmological model. Together, these results suggest that mode restoration offers a complementary approach for extracting cosmological information from future 21 cm intensity mapping analyses.

Identifying mergers from observational data has been a crucial aspect of studying galaxy evolution and formation. Tidal features, typically fainter than 26 ${\rm mag\,arcsec^{-2}}$, exhibit a diverse range of appearances depending on the merger characteristics and are expected to be investigated in greater detail with the Rubin Observatory Large Synoptic Survey Telescope (LSST), which will reveal the low surface brightness universe with unprecedented precision. Our goal is to assess the feasibility of developing a convolutional neural network (CNN) that can distinguish between mergers and non-mergers based on LSST-like deep images. To this end, we used Illustris TNG50, one of the highest-resolution cosmological hydrodynamic simulations to date, allowing us to generate LSST-like mock images with a depth $\sim$ 29 ${\rm mag\,arcsec^{-2}}$ for low-redshift ($z=0.16$) galaxies, with labeling based on their merger status as ground truth. We focused on 151 Milky Way-like galaxies in field environments, comprising 81 non-mergers and 70 mergers. After applying data augmentation and hyperparameter tuning, a CNN model was developed with an accuracy of 65--67\%. Through additional image processing, the model was further optimized, achieving an accuracy of 67--70\% when trained on images containing only faint features. This represents an improvement of $\sim$ 5\% compared to training on images with bright features only. This suggests that faint tidal features can serve as effective indicators for distinguishing between mergers and non-mergers. The future direction for further improvement based on this study is also discussed.

We report on the periodic flux variations in the 6.7 GHz $\mathrm{CH_3OH}$ maser associated with the high-mass protostar G353.273+0.641, based on 13 yr of monitoring mainly by the Hitachi 32 m telescope. We identified a periodicity of 309 days based on a nearly complete light curve, with 833 epochs every few days. A strong correlation is found between the maser and the mid-infrared fluxes at 3.4 and 4.6 $μ$m observed by NEOWISE during these periods, suggesting that the maser emission responds to variations in the protostellar luminosity. The average profile of the maser light curve is asymmetric and shows a steep drop in intensity just before the brightening, resembling that of some pulsating variable stars. Assuming a protostellar pulsation as the origin of maser periodicity, the observed period implies a cool and highly bloated, red supergiant-like structure. Such a bloated structure is consistent with a theoretical model of protostellar evolution under high accretion rates. The inferred protostellar parameters are broadly consistent with the theoretical model of pulsational instability during the early phase of high-mass star formation. However, a periodic accretion scenario caused by an unresolved compact protobinary cannot be completely ruled out. Several irregular peaks that deviate from the periodicity may result from episodic accretion phenomena or jet-launching events independent of the protostellar pulsation. Extremely high-resolution imaging with next-generation interferometers such as the ngVLA will provide a conclusive test for both the protostellar pulsation and the binary accretion scenarios.

The outer Galaxy (galactocentric distance $\gtrsim$13.5 kpc) serves as an excellent laboratory for investigating the chemical complexity in low-metallicity environments. Here, we present the chemical analyses for the outer Galactic hot core Sh 2-283-1a SMM1 ($D_\mathrm{GC}$ = 15.7 kpc and $Z$ $\sim$0.3 $Z_\odot$), recently detected by Ikeda et al. (2025) using ALMA. Toward this source, a variety of molecular species, including complex organic molecules (COMs: CH$_3$OH, $^{13}$CH$_3$OH, CH$_2$DOH, and CH$_3$OCH$_3$) are detected. The molecular abundances relative to CH$_3$OH are similar to those of another outer Galactic hot core, demonstrating that chemically rich hot cores exist in different regions of the outer Galaxy. We also compared molecular abundances among hot cores in the inner Galaxy, outer Galaxy, and Magellanic Clouds. This comparison revealed that the metallicity-corrected $N$(SO$_2$)/$N$(H$_2$) ratios of outer Galactic hot cores are significantly lower than those of the inner Galactic ones, while their $N$(CH$_3$OH)/$N$(H$_2$) ratios are similar. The Magellanic hot cores show different trends despite having metallicities similar to those of the outer Galaxy, indicating that the chemical complexity of hot cores is governed by environmental conditions (e.g., cosmic ray intensity and dust temperature) rather than simple metallicity scaling. These environmental differences would also affect the production efficiency of COMs derived from CH$_3$OH, as the $N$(CH$_3$OCH$_3$)/$N$(CH$_3$OH) and $N$(C$_2$H$_5$OH)/$N$(CH$_3$OH) ratios in the outer Galactic sources are moderately lower than those of inner Galactic sources. The $N$(CH$_2$DOH)/$N$(CH$_3$OH) ratio of Sh 2-283-1a SMM1 is 1.5$^{+3.9}_{-1.2}$$\%$, comparable to that of inner Galactic high-mass sources.

Cluster environments influence galaxy evolution by curtailing star formation activity, notably through ram-pressure stripping (RPS). In this study, using spatially resolved spectroscopic data from the SAMI Galaxy Survey, we identify galaxies undergoing or recently affected by RPS in eight nearby clusters ($0.029 < z < 0.058$), through a visual classification scheme based on the ionised gas ($\rm Hα+ [NII]λ6584$) morphologies, split into unperturbed, asymmetric, and truncated. The projected phase-space analysis shows that asymmetric galaxies are found in a narrow region in cluster-centric distance ($\rm 0.1 < R/R_{200} < 0.6$) and have a larger dispersion in line-of-sight velocity ($σ(|v_{pec}|)_\mathrm{Asym} = 0.71^{+0.09}_{-0.07}\ σ_{200}$) compared to the truncated and unperturbed samples. In terms of star formation activity, RPS candidates yield a much steeper resolved star-forming main sequence (rSFMS; $Σ_\mathrm{SFR} - Σ_\ast$) relation compared to the unperturbed counterparts, primarily emerging from having lower $Σ_\mathrm{SFR}$ values for the low mass density regime, with the steepest gradient deriving from the truncated sample. Moreover, radial star formation profiles reveal that star formation in RPS candidates is suppressed in the outskirts relative to unperturbed galaxies and is more prominent for the truncated sample. In contrast, central ($\rm r/r_{eff}<0.5$) star formation activity in RPS candidates is comparable with that in their unperturbed and field counterparts, suggesting no elevated activity. Taken together, this suggests an evolutionary trend linked to the RPS stage, where unperturbed galaxies likely represent recently accreted systems (pre-RPS), while asymmetric and truncated galaxies may correspond to populations undergoing RPS and post-RPS phases, respectively, favouring outside-in quenching.

PSR~J2032+4127 is the only one of gamma-ray binary, that exhibits pulsations in gamma-ray. Previous research has indicated that the pulsar and the Be star MT91 213 orbit each other in a highly eccentric orbit with an extremely long period, with the pulsar reaching its periastron on November 13, 2017. Since its launch, the \fermi{} satellite has been monitoring this pulsar for 16 years, covering the 8 years before and the 8 years after the pulsar passed its periastron. Using these data, we present an analysis of pulse arrival times, and precisely determine the orbital parameters for the first time: the orbital period of $P_{\rm orb} \sim 52.3$ yr, the eccentricity of $e \sim 0.98$, the semimajor axis of $a$sin$i \sim 25.3$ AU, and the orbital inclination of $\sim$ 47.1$^\circ$ -- 55.1$^\circ$. We also reveal another small glitch occurred in 2021, MJD $\sim$ 59500.

Anomalous X-ray pulsars (AXPs) and soft gamma-ray repeaters (SGRs) are believed to be manifestations of magnetars. Typically, AXPs exhibit higher X-ray luminosities, whereas SGRs are generally fainter and display significantly high signal-to-noise ratios only during their outburst phases. In this work, we report the long-term timing evolution of four AXPs: 1E 2259+586, 4U 0142+61, 1RXS J170849.0-400910 and 1E 1841-045, which were regularly monitored with NICER from 2017 to 2024. Over this period, we identify a total of 10 timing events. In addition to one glitch and one anti-glitch in 1E 2259+586 reported in literature, we detect another 8 new timing events: 5 glitches, 2 anti-glitches, and 1 unusual state transition event. Notably, both anti-glitches were observed in 4U 0142+61, making it the most frequent source of such events, and there is a hint of regular evolution in its pulse profile. In the case of 1RXS J170849.0-400910, it continues to exhibit pronounced high-frequency timing anomalies and undergoes a state transition event. Finally, we study the evolution of the pulse profiles and find that the profiles of 1E 2259+586 and 4U 0142+61 both evolve. This is consistent with the earlier finding that pulse profile evolution is a generic feature of magnetars.

We investigated how Gamma-Ray Burst (GRB) prompt emission measurements are affected by increasing distance to the source. We selected a sample of 26 bright GRBs with measured redshifts $z<1$ observed by the Burst Alert Telescope (BAT) on board the Neil Gehrels Swift Observatory (Swift) and simulated what BAT would have observed if the GRBs were at larger redshifts. We measured the durations of the simulated gamma-ray signals using a Bayesian block approach and calculated the enclosed fluences and peak fluxes. As expected, we found that almost all durations (fluences) measured for simulated high-$z$ GRBs were shorter (less) than their true durations (energies) due to low signal-to-noise ratio emission becoming completely dominated by background, i.e., the ``tip-of-the-iceberg'' effect. This effect strongly depends on the profile and intensity of the source light curve. Due to the uniqueness of GRB light curves, there is no common behavior in the evolution of measured durations with redshift. We compared our synthetic high-$z$ (i.e., $z>3$) GRBs to a sample of 72 observed high-$z$ bursts and found that the two samples were not inconsistent with being drawn from the same underlying population. We conclude that: (i) prompt emission durations (fluences) of high-$z$ GRBs observed by Swift/BAT are most likely underestimations, sometimes by factors of $\sim$several tens ($\sim2$), and (ii) changes in the average GRB prompt emission duration and fluence with increasing redshift are consistent with the tip-of-the-iceberg effect.

We report a revised spectroscopic redshift for the dusty massive galaxy S1, previously inferred with an exceptionally high baryon-to-star conversion efficiency from NIRCam slitless grism data at $z_{\rm grism}=5.58$. Our new JWST/NIRSpec IFU observations reveal multiple rest-frame optical and NIR emission lines, yielding a secure spectroscopic redshift of $z_{\rm spec}=3.2439\pm0.0002$. We show that the earlier grism-based redshift resulted from contamination by a nearby galaxy whose dispersed spectral trace overlaps with S1, illustrating a known challenge of slitless spectroscopy when only a single dispersion angle and single emission feature are available. In addition, we present new ALMA 1 mm observations, which robustly detect dust emission ($S_{\rm 1mm}=0.99\pm0.03$ mJy) and show a dust half-light radius ($R_{\rm e,1mm}=0.73\pm0.10$ kpc) slightly smaller than the stellar size ($R_{\rm e, F444W} = 0.97\pm0.01$ kpc). Using the revised redshift and compiled multi-wavelength photometry, we update the UV-to-FIR SED and find that S1 is less extreme than previously inferred, yet remains a very massive (log$M_{\star}/M_{\odot}\sim10.6$), heavily obscured star-forming galaxy. The updated SED modeling reveals S1 to be a very dust- and gas-rich system with a moderate star formation rate and a long gas depletion time ($τ_{\rm dep} \sim 1.4$ Gyr), deviating from SMGs and OFGs, but more closely resembling typical massive main-sequence galaxies. We note that, although this revision reduces the number of ultra-massive galaxies reported in Xiao et al. 2024, it does not alter the main conclusions of that work. Overall, our study clarifies the nature of S1 and underscores the importance of multi-line spectroscopic confirmation, slitless observations at multiple position angles, and IFU data for robust redshift and physical characterization of rare massive galaxies in the early Universe.

We present a deep photometric study of the globular cluster NGC 5466 and its tidal stream using Subaru/Hyper Suprime-Cam (HSC) imaging with the metallicity-sensitive narrowband filter NB395. We develop an improved member-selection technique based on a k-nearest neighbor algorithm applied to the color-color-magnitude diagram (CCMD), enabling reliable candidate identification down to $i_{2,0} < 23.5$. Photometric metallicities derived from NB395 colors agree with previous measurements, supporting the robustness of our calibration. While modest residual contamination and possible offsets - potentially driven by variations in light-element abundances - may remain beyond 10 arcmin, the metallicity distribution of high-probability inner members matches the known mean metallicity of NGC 5466, demonstrating the effectiveness of our method. The spatial distribution of NB395-selected stars clearly delineates the tidal stream. Beyond the tidal radius, the azimuthally averaged radial surface density profile follows a power law with slope $α= -4.53_{-0.14}^{+0.13}$. We also detect a power-law component perpendicular to the stream, suggestive of multiple apogalactic passages. A density gap is identified at a projected distance of $\sim200$ pc from the cluster center, consistent with eTidal N-body predictions and possibly associated with a recent pericentric passage or Galactic disk interaction. Analysis of the main-sequence mass function reveals a strong negative radial gradient in the slope within the tidal radius, whereas the slope along the outer stream is relatively flat, consistent with preferential tidal stripping of low-mass stars. These results highlight the power of HSC/NB395 photometry for identifying metal-poor populations and deriving photometric metallicities, underscoring its value for future wide-field surveys.

Line-Intensity Mapping (LIM) has emerged as a powerful technique for studying large-scale structure and the high-redshift universe, enabling three-dimensional maps of line emission across vast cosmological volumes. In this review, we summarize the LIM framework, its key scientific goals, and its future prospects. We describe the landscape of emission line tracers, theoretical modeling approaches, anticipated signals, and data-analysis methodologies. We also discuss experimental challenges, particularly those posed by astrophysical foregrounds, and review possible mitigation strategies. Further, we highlight a range of cross-correlation science cases, linking LIM with other cosmological surveys. Finally, we summarize current and upcoming experiments and early results, including recent first detections, while outlining the outlook for future discoveries. Specifically, LIM may offer new insights into galaxy formation and evolution and cosmology, while revealing the Epoch of Reionization, Cosmic Dawn, and possibly the Cosmic Dark Ages. LIM enables cosmological measurements that complement other probes and provide unique access to the high-redshift universe, potentially shedding light on dark matter, dark energy, and cosmic inflation.

We present a cosmological zoom-in radiation magneto-hydrodynamic (RMHD) simulation, using FORGE'd in FIRE, that follows the formation, growth, and evolution of a single metal-free Pop. III (proto)star at redshift $z \sim 14$. The simulation captures a rotationally supported circumstellar disk and protostellar jets, both resolved down to $<100$ au scales. We find the star grows to $\sim 27$ M$_{\odot}$ over $31,000$ years, with its final mass regulated by accretion and protostellar jets. Protostellar jets form because the magnetic mass-to-flux ratio lies within the regime that allows jet launching, and they are further enabled by a rotating circumstellar disk with sufficient gas-magnetic-field coupling, both present in this simulation. These jets regulate accretion onto the (proto)star and drive outflows that collide with infalling gas, slowing inflow at large radii due to the substantial momentum they carry. A circumstellar disk forms, extending out to $\sim 0.01$ pc, which remains gravitationally stable (Q $\gg 1$). The stability of the disk is maintained through both thermal support and turbulence. In this paper we focus on how jets play a critical role not only in shaping the final masses of Pop. III stars but also in directly influencing their surroundings by regulating accretion. These results will provide important insights into the initial mass function and feedback processes in the earliest star-forming regions of the Universe.

Reliable nebular emission line diagnostics are essential for accurately inferring the physical properties (e.g. electron temperature, density, pressure, and metallicity) of H II regions from spectra. When interpreting spectra, it is typical to adopt a single zone model, e.g. at fixed density, pressure, or temperature, to infer H II region properties. However, such an assumption may not fully capture the complexities of a turbulent interstellar medium. To understand how a complex density field driven by supersonic turbulence impacts nebular emission lines, we simulate 3D H II regions surrounding a single O star, both with and without supersonic turbulence. We find that turbulence directly impacts the values of common strong line ratios. For example turbulent H II regions exhibit systematically higher [N II]/H$α$, lower [O III]/H$β$, and lower O32, compared to homogeneous H II regions with the same mean density and ionizing source. These biases can impact inferences of metallicity, ionization parameter, excitation, and ionization source. For our choice of turbulence, direct $T_e$ method metallicity inferences are biased low, by up to 0.1 dex, which is important for metallicity studies, but not enough to explain the abundance discrepancy problem. Finally, we show that large differences between measured electron densities emerge between infrared, optical, and UV density indicators. Our results motivate the need for large grids of turbulent H II regions models that span the range of conditions seen at both high and low redshift to better interpret observed spectra.

Massive stars have been proposed as candidates to be major factories of Galactic cosmic rays (GCRs).However, this claim lacks enough empirical evidence, especially for isolated stars.The powerful stellar winds from massive stars impact the ambient medium producing strong shocks suitable for accelerating relativistic particles.The detection of non-thermal emission-particularly synchrotron emission in low radio frequencies-serves as a key proof of particle acceleration sites.We aim to assess the potential of isolated massive stars as sources of GCRs.We observed the Wolf-Rayet bubble, NGC 2359, using the upgraded Giant Metrewave Radio Telescope at Band 3 (250-500 MHz) and Band 4 (550-950 MHz).Additionally, we used complementary archival radio datasets at different frequencies to derive the broad spectral energy distribution (SED) for several regions within the bubble.To further characterize the interaction between the stellar wind and the ambient medium, we introduced a composite SED model including synchrotron and free-free emission, and two low-frequency turnover processes, the Razin-Tsytovich (RT) effect and free-free absorption (FFA).We used a Bayesian inference approach to fit the SEDs and constrain the electron number density and magnetic field strength.The SEDs of several regions reveal spectral indices steeper than -0.5, indicative of synchrotron emission. and show a turnover below ~1 GHz.Our SED modelling suggests that the observed turnover is primarily caused by the RT effect, with a minor contribution from internal FFA.Our analysis confirms the presence of synchrotron radiation within NGC 2359.This is the second detection of non-thermal emission in a stellar bubble surrounding a WR star, reinforcing the idea that such environments are sites of relativistic particle acceleration and supporting the hypothesis that isolated massive stars are sources of GCRs of at least GeV energies.

The surface-brightness-color (SBC) relationship for Red Clump stars provides a critical foundation for precision distance ladder measurements, including the 1\% distance determination to the Large Magellanic Cloud. Current SBC calibrations rely on angular diameter measurements of nearby Red Clump stars obtained through long-baseline optical interferometry using the Very Large Telescope Interferometer. We explore the application of intensity interferometry to measure limb-darkened angular diameters of Red Clump stars, offering a complementary approach to traditional amplitude interferometry. We describe the framework for extracting angular diameters from squared visibility measurements in intensity interferometry, accounting for limb darkening through the stellar atmosphere models. For the Red Clump star HD~17652, we show that intensity interferometry in the $H$ band at baselines matching PIONIER ($\sim$100~m) could achieve $<1$\% angular size uncertainties in 2-hour exposures by measuring the primary peak of the visibility function, enabling direct comparison with existing measurements. Critically, observations at shorter wavelengths probe the secondary visibility maximum, providing independent checks of both measurement and systematic errors that are largely insensitive to limb-darkening assumptions. Exploiting the multiplex advantage of simultaneous multi-bandpass observations and the large number of baselines available with telescope arrays such as the Cherenkov Telescope Array Observatory can reduce observing times to practical levels, making intensity interferometry a viable tool for validating the angular sizes for a subset of the Red Clump star calibration sample.

Second-order structure functions and power spectral densities are popular tools in the study of statistical properties across scales, particularly for the analysis of turbulent flows. Although intimately related, analyses primarily use one or the other. We introduce a framework for estimating the power spectrum using the second-order structure function without applying Fourier transforms -- enabling one to take advantage of the real-space structure function calculations. We validate and showcase this method, comparing it to classical Fourier power spectrum estimates determined from analytical calculations, fractional Brownian motion, turbulence simulations, and space-physics and astrophysical observations of turbulence. We show that this method is able to robustly obtain the expected power law behaviour where we use turbulence ranges as test-cases.

Strong gravitational lensing has to deal with many modeling degeneracies, the most notable being the Mass Sheet Degeneracy (MSD). We review the MSD when one needs to model more lens planes, each one with an internal mass sheet. We take into account the non-linear lens-lens coupling and line of sight effects, the latter treated as external mass sheets with associated shear. If second order shear terms on external and internal mass sheets can be neglected, we show that the MSD is always retained, and the mass sheets influence can be reabsorbed in the redefinition of angular diameter distances. In particular, internal and external mass sheets can be placed on the same footing. The version of the MSD discussed here does not require any particular relation between the internal mass sheets in the different planes. Even when including time delays from all sources, a residual degeneracy involving time delays, mass sheets and $ H_0 $ remains. We develop a framework which shows what can actually be constrained in multiple plane lens systems.

Complete catalogs of molecular clouds in the Milky Way allow analysis of the molecular medium and the star formation properties of the Milky Way that closely follows the method used for nearby galaxies. We explore whether the big dip in the radial distribution of molecular gas in the Milky Way is peculiar and find several other galaxies with similar patterns, all with similar morphological classifications of YClxxGnR, indicating a clearly defined, long bar leading to a grand-design spiral. This category is fairly rare among galaxies in the PHANGS sample, but all galaxies with this classification have some evidence for dips in the radial distribution of CO emission. The lengths of the bars correlate with the extents of the dips. The Milky Way and the other galaxies with dips have similar stellar masses and star formation rates, both lying near the high ends of the distributions for all PHANGS galaxies.

High-redshift quasars have been an excellent tracer to study the astrophysics and cosmology at early Universe. Using 216,949 high-redshift quasar candidates ($5.0 \leq z < 6.3$) selected via machine learning from the Legacy Survey Data Release 9 and the Wide-field Infrared Survey Explorer, we perform wide-field clustering analysis to investigate the large-scale environment of those high-redshift quasars. We construct the projected auto correlation function of those high-redshift quasars that is weighted by its predicted probability of being a true high-redshift quasar, from which we derive the bias parameter and the typical dark matter halo mass of those quasars. The dark matter halo mass of quasars estimated from the projected auto correlation function is $\log(M_h/M_{\odot})=12.2 ^{+0.2}_{-0.7}$ ($11.9^{+0.3}_{-0.7}$), with the bias parameter $b$ of $12.34 ^{+4.26}_{-4.37}$ ($11.52^{+4.02}_{-4.14}$) for the redshift interval of $5.0 \leq z <5.7$ ($5.7 \leq z <6.3$). Our results, combined with other measurements of dark matter halo masses for quasars or active galactic nucleus which obtain a lower dark matter halo mass of $\sim 10^{11.5}$ M$_\odot$ at similar redshift, suggest a more complex, and possibly non-monotonic evolution of quasar hosting dark matter halo. Moreover, we estimate the duty cycle of those quasars, which is $0.008^{+0.135}_{-0.007}$ ($0.003+^{+0.047}_{-0.003}$) for the redshift interval of $5.0 \leq z <5.7$ ($5.7 \leq z <6.3$).

Observed cyclotron resonant scattering features (CRSFs) in X-ray pulsars (XRPs) exhibit strong variability. In the subcritical luminosity regime, the centroid energy ($E_{CRSF}$) and line width ($σ_{CRSF}$) often show positive correlations with the X-ray luminosity. We investigate the physical origin of the observed variability quantitatively, focusing on the effects of resonant scattering and Doppler shift induced by the plasma flow in the accretion funnel. We developed a relativistic Monte Carlo code to perform detailed radiative transfer calculations in the accretion funnel above the hotspot and derive angle-dependent spectra. Analytical plasma density and velocity profiles were adopted to account for the effects of radiation pressure on the flow. Approximate resonant scattering cross-sections were employed. We varied the accretion luminosity to explore the resulting variability of the CRSF properties. The emergent spectra exhibit a prominent, asymmetric CRSF accompanied by a broad blue wing. The CRSF is systematically redshifted relative to the classical cyclotron energy, with the magnitude of the redshift decreasing at higher luminosities and for larger viewing angles $θ$. Both $E_{CRSF}$ and $σ_{CRSF}$ correlate positively with luminosity for all viewing angles. Their absolute values, however, depend strongly on the viewing angle, indicating substantial variability over the pulse cycle and sensitivity to the system geometry. At fixed luminosity, $E_{CRSF}$ ($σ_{CRSF}$) decreases (increases) with increasing $\cosθ$. Consequently, phase-resolved observations are expected to reveal an anticorrelation between the CRSF centroid energy and width. When applied to the XRP GX 304$-$1, the model reproduces the observed CRSF variability over nearly an order of magnitude in luminosity for geometries in which the accretion funnel is predominantly viewed edge-on.

Double-peaked H$α$ emission profiles can serve as potential signatures of accreting intermediate-mass black holes (IMBHs), particularly those residing outside galactic nuclei. Such features are expected to arise from rotating disk-like structures around black holes and can be used to identify elusive IMBH candidates. \citet{Almeida2022ApJ...934..100S} reported a sample of such double-peaked H$α$ sources in the MUSE-Wide survey, interpreting them as potential signatures of wandering IMBHs after systematically excluding alternative explanations. Their method relied on constructing H$α$ maps around central galaxies and visually identifying compact emission clumps in the surrounding halo regions. In this work, we revisit the analysis using the deeper MUSE Extremely Deep Field (MXDF) data and an automated detection algorithm tailored to identify such features. However, we do not recover any candidate population in MXDF, resulting in a null detection. This outcome is nevertheless informative, as it (1) highlights the inherent challenges in detecting IMBHs, and (2) demonstrates the potential of automated approaches for future systematic searches, even though it did not yield a positive outcome in this case.

We compile a sample of 83 Little Red Dots (LRDs) with JWST imaging and find that a substantial fraction ($\sim$43%, rising to $\gtrsim$85% for the most luminous LRDs) host one or more spatially offset, UV-bright companions at projected separations of $0.5\rm \, kpc \lesssim d\lesssim 5 \rm \,kpc$, with median of $\langle d \rangle = 1.0\,\mathrm{kpc}$. This fraction is even higher when smaller spatial scales are probed at high S/N ratio: we show that the two most strongly lensed LRDs known to date, A383-LRD and the newly discovered A68-LRD, both have UV-bright companions at separations of only $d\sim0.3$ kpc, below the resolution limit of most unlensed JWST samples. We explore whether these ubiquitous red/blue configurations may be physically linked to the formation of LRDs, in analogy with the "synchronized pair" scenario originally proposed for direct-collapse black hole formation. In this picture, ultraviolet radiation from the companions, which typically have modest stellar masses ($M_\ast \sim 10^{8-9}M_\odot$), suppresses molecular hydrogen cooling in nearby gas, allowing nearly isothermal collapse and the formation of extremely compact objects, such as massive black holes or quasi-stars. Using component-resolved photometry and SED modeling, we infer Lyman-Werner radiation fields of $J_{21,LW} \sim 10^{2.5}$-$10^{5}$ at the locations of the red components, comparable to those required in direct-collapse models, suggesting that the necessary photodissociation conditions are realized in many LRD systems. This framework provides a simple and self-consistent explanation for the extreme compactness and distinctive spectral properties of LRDs, and links long-standing theoretical models for early compact object formation directly to a population now observed with JWST in the early universe.

Double-plane gravitational lensing is a rare but increasingly observed phenomenon in which the light from a distant source is lensed by two foreground objects at different redshifts. Such systems can be used to provide simultaneous constraints on the Hubble constant $H_0$ and the dark-energy equation of state, independent of and complementary to other probes. However, just as for single-plane gravitational lenses, the precision of these constraints is limited by the so-called mass-sheet degeneracy (MSD) -- a fundamental limit to the knowledge of the mass profiles of lens galaxies and the line of sight that can be obtained from imaging constraints alone. In this work, we show explicitly how contributions from the line of sight appear in double-plane systems. Because these contributions modify angular diameter distances, we argue that cosmological priors should not be used to simply fix the ``cosmological scaling factor'', a ratio of angular diameter distances which is key to the modelling of double-plane lenses. Motivated by this fact, we generalise the double-plane MSD to account for this uncertainty in the scaling factor. While this complicates the time-delay function, we show that, using the ``unfolding relation'', a geometric relation between distances which holds even in the presence of line-of-sight corrections, the uncertainty in the Hubble constant reduces to the familiar mass-sheet transformation of the first lens plane, and a line-of-sight contribution between the observer and the second lens plane. Our main message is therefore a prescription for reducing the degrees of freedom within double-plane models, while still safely accounting for the MSD in measurements of $H_0$.

The Sloan Digital Sky Survey-V (SDSS-V) Magellanic Genesis survey is a spectroscopic program designed to map the kinematic and chemical structure of the Magellanic Clouds using APOGEE and BOSS spectroscopy. This overview describes the survey's design, target selection, and science goals, and highlights some first results using these data. In the inner regions of the Large and Small Magellanic Clouds (LMC and SMC), the survey obtained high-resolution near-infrared APOGEE spectra (S/N~45) of ~14,000 bright, oxygen-rich asymptotic giant branch (AGB-O) stars. These data provide contiguous spatial coverage of the Clouds' main bodies, enabling detailed chemo-dynamical studies. To explore extended structures, the survey includes BOSS optical spectroscopy of fainter red giant (RG) stars selected with \gaia~DR3 data, reaching G~17.5. Many of these targets extend to the outer regions of the Clouds, which are known to span ~20 deg (LMC) and ~12 deg(SMC) and contain diffuse substructures of unclear origin. BOSS data in the inner regions also complement APOGEE by providing elements inaccessible in the near-infrared and enabling cross-calibration between instruments. The survey further includes APOGEE and BOSS observations of ~300 evolved massive stars and a small sample of symbiotic binaries previously observed by APOGEE-1 and -2, enhancing our understanding of massive stellar evolution and complementing the SDSS-V main-sequence massive star program.

We present near-infrared (NIR) spectroscopy of the hydrogen- and helium-poor (Type Ic) supernova (SN) 2024aecx that displays a strong NIR excess emerging 32 days post peak. SN 2024aecx is a peculiar SN Ic that exhibited luminous shock-cooling emission at early times, suggestive of close-in circumstellar medium (CSM), unexpected for this class of SNe. Its early NIR spectra are typical for a SN Ic but with strong CI absorption features. By ~32 days post peak, the spectra show a strong NIR excess, while maintaining normal optical colors, unprecedented for SNe Ic. We find that the NIR excess is well fit with a single-temperature, optically thin dust model with declining temperature, increasing mass, and roughly constant luminosity over time. The NIR excess appears too promptly for dust to have formed in the SN ejecta, indicating an IR echo from pre-existing dust in the CSM. The IR echo is likely powered by the relatively slowly evolving SN peak light, and not the brief shock cooling emission, as the latter requires unrealistically high CSM densities to explain the observed dust mass. We consider different potential CSM geometries and find that a thick face-on disk with an inner edge around $5\times 10^{16}$ cm can best explain the dust mass and temperature evolution. In this scenario, the SN shock should start interacting with this CSM $440\pm200$ days post explosion. CSM around SN Ic is rare, and follow-up observations of SN 2024aecx will probe the mass-loss process responsible for removing hydrogen and helium from their progenitor star.

We fit the multi-band light curves of 2,205 Type Ia supernovae (SNe~Ia) from the Zwicky Transient Facility DR2 with a one-zone radioactive decay model with a phenomenological addition to include Fe recombination physics. We find a strong correlation between inferred nickel mass and SALT2 stretch, which within our simplified modelling is linked to larger ejecta masses providing longer diffusion times, providing a physical basis for the brighter-slower relation. SN~Ia in low-mass hosts ($\log_{10}(M_*/M_\odot) < 10$) produce $12\%$ more $^{56}$Ni than those in high-mass hosts ($ΔM_{\rm Ni} = 0.13~M_\odot$), linking the host-galaxy mass step to ejecta properties and hinting at metallicity or age-dependent burning efficiencies. This suggests that standardisation based on physical parameters may remove the mass-step. SN~1991T-like events show higher ejecta masses (median $1.64~M_\odot$ vs. $1.38~M_\odot$ for normals) and produce $30\%$ more $^{56}$Ni, with $84\%$ having super-Chandrasekhar masses. Through Hierarchical modelling of $902$ SNe ($z \leq 0.06$), we find thermonuclear supernovae can be well described by a Gaussian distribution in ejecta mass and nickel mass with $μ_{\rm ej} = 1.26 \pm 0.01~M_\odot$ ($σ_{\rm ej} = 0.33 \pm 0.01~M_\odot$) and $μ_{\rm Ni} = 0.64 \pm 0.06~M_\odot$ ($σ_{\rm Ni} = 0.42 \pm 0.02~M_\odot$), respectively. This leads to inferred fractions of $43 \pm 2\%$ sub-$M_{\rm Ch}$ ($<1.2~M_\odot$), $34 \pm 1\%$ near-$M_{\rm Ch}$ ($1.2$--$1.5~M_\odot$), and $24 \pm 2\%$ super-$M_{\rm Ch}$ ($>1.5~M_\odot$) events. This work provides a step towards holistic physical characterization of the local SN~Ia population, reinforcing the physical basis of SN~Ia standardization while quantifying diversity and environmental dependencies critical for understanding progenitor physics and mitigating systematics in precision cosmology.

A population of anomalous ultra-massive white dwarfs discovered with Gaia, often referred to as the Q branch, show high (multi-Gyr) cooling delays produced by exotic physical mechanisms. They are believed to be the products of stellar mergers, but the exact origin and formation channel remain unclear. We obtained a spectroscopically complete, volume-limited sample of the Q branch region within 100 pc, and found significant differences in atmospheric composition and rotation rates as a function of tangential velocity. In particular, we discover that stellar remnants with the longest cooling delays do not show strong magnetism nor detectable short-period rotational variability, as opposed to what is generally believed for double-degenerate mergers. This indicates that either these white dwarfs arise from a formation channel with no strong magnetism induced, or that the magnetism produced from the merger dissipates over the cooling delay timescales. Our follow-up photometry has also discovered pulsations in the second and third hydrogen-dominated DAQ white dwarfs, one hotter than 15,500 K, possibly extending the boundaries of the DAV instability strip for white dwarfs with thin hydrogen layers.

In this study, we present a comprehensive analysis of the structural, astrophysical, and dynamical properties of eight open clusters: NGC 559, NGC 1817, NGC 2141, NGC 7245, Ruprecht 15, Ruprecht 137, Ruprecht 142, and Ruprecht 169, using precise astrometric and photometric data from Gaia Data Release 3. By fitting the King model to the radial density profiles, we determined the structural parameters of the clusters, including core and limiting radii, which were found to range from 3.07 to 16.21 arcmin and from 9.97 to 25.97 arcmin, respectively. Fundamental astrophysical parameters were derived by fitting PARSEC isochrones to the colour-magnitude diagrams. The results show that the clusters have logarithmic ages between 7.95 and 9.34, metallicities in the range 0.007 to 0.015, and heliocentric distances between 1640 and 5203 pc. The total stellar masses of the clusters were estimated to lie between 257 and 1916 solar masses. For most of the clusters, the mass function slopes are consistent with the Salpeter initial mass function. Our dynamical analysis indicates that all clusters, except Ruprecht 15, are dynamically relaxed. In addition, the spatial distribution and the bimodal structure observed in the radial density profile of NGC 7245 provide strong evidence that this object is a binary cluster candidate. Finally, kinematic analysis and orbit integrations demonstrate that all clusters exhibit dynamical properties fully consistent with membership in the Galactic thin disc.

Measuring stellar motions at the Milky Way's Galactic center (GC) provides unique insight into the dynamical processes within galactic nuclei. We present proper motion measurements for 23 SiO-maser emitting stars within 45'' of SgrA*, including four previously reported to have velocities exceeding their local escape velocities (i.e., they are "locally unbound" from the GC). Derived from 14 epochs of HST WFC3-IR observations (2010 - 2023), our measurements have a median precision of 0.038 mas/yr - up to ~100x more precise then previous constraints for some sources. By combining these proper motions with published radial velocities, we derive updated 3D velocities for the masers and find that only one is locally unbound (IRS 9; v3d = 370 +/- 1.2 km/s). Orbit integrations place the first constraints on the orbit of IRS 9, which is bound to the GC at larger radii with r_peri >= 0.100 +/- 0.005 pc and r_apo >= 5.25 +/- 0.18 pc. IRS 9's high velocity relative to stars at similar radii in the Nuclear Star Cluster makes it a candidate to have experienced a strong dynamical interaction in order to place it on its orbit. We explore the Hills mechanism as a possible origin, but binary evaporation and ejection velocity limits indicate that IRS 9 is unlikely to have experienced such an event in the past 0.4 Myr (the timescale constrained by the orbit integrations). Alternative mechanisms that could produce IRS 9 include binary supernova disruption, two-body interactions, and stellar collisions. Identifying additional stars like IRS 9 will be essential for understanding these various dynamical processes.

The Electromagnetically Isolated Global Signal Estimation Platform (EIGSEP) is a new instrument designed to measure the global 21-cm signal from Cosmic Dawn and the Epoch of Reionization, redshifted to frequencies below 250 MHz. To reduce spectral structure in the antenna beam associated with ground scattering, EIGSEP uses a shaped bowtie antenna suspended in a canyon 100 m above the ground. We describe the current system design of EIGSEP, including the rotating antenna platform, a transmitter antenna to characterise the beam of the bowtie antenna, and auxiliary ground antennas. We then discuss the EIGSEP calibration scheme, which incorporates traditional Dicke switching in the receiver, and novel approaches that include beam mapping, beam modulation, and interferometric cross-correlation. The instrument has been deployed near Marjum Pass, Utah, for testing and initial data collection. We discuss the site characteristics and present initial field measurements.

We investigate the physical origin of the star formation scaling relations between the gas depletion time, the star-forming gas mass fraction, and the gas surface density, $Σ_{\rm gas}$, on kiloparsec scales, all of which are the key ingredients of the observed Kennicutt-Schmidt relation. To elucidate these trends, we employ an analytical framework that explicitly connects these kiloparsec-scale properties to the timescales governing the rapid, continuous ISM gas cycle on the scales of individual star-forming regions, including the formation, dispersal, and local depletion of star-forming gas. Using a suite of idealized disk galaxy simulations spanning a range of environments from dwarf and Milky Way-mass systems to a gas-rich starburst analog, we measure the timescales of the gas cycle and relate them to the dynamical and turbulent properties of the interstellar medium (ISM). We find that star-forming regions form on a timescale close to the vertical turbulent crossing time of the galactic disk, $\sim$3-30 Myr, which decreases at higher $Σ_{\rm gas}$ due to the increase in turbulent velocities in the ISM and the decrease in the disk thickness. In contrast, the local star formation and dispersal of such gas are set by the local conditions. Specifically, the local depletion time, $\sim$200-2000 Myr, is decreasing at higher $Σ_{\rm gas}$, as star-forming gas becomes denser and more efficient in forming stars. The lifetime of such gas is very short, $\sim$0.4-1 Myr, and only weakly increases with $Σ_{\rm gas}$. Together, our results demonstrate how the star formation properties of galaxies on kiloparsec scales emerge directly from the interplay between the galaxy-scale dynamics, ISM turbulence, and the state of star-forming gas.

Quasi-periodic eruptions (QPEs) are recently discovered transients of unknown nature occurring near supermassive black holes, which feature bright X-ray bursts separated by hours to days. A promising model for QPEs is the star-disc collisions model, where a star repeatedly interacts with an accretion disc around a black hole, creating shocks that expel dense outflows of gas from which radiation emerges. We investigate the dynamics of the star-disc collisions, the properties of the outflows, and the resulting radiation signatures. Our study focuses on the generic case where the star remains unperturbed by the collision and the stellar crossing time through the disc is sufficiently long for shocked gas to flow around the star. We performed a three-dimensional (3D) radiation-hydrodynamics simulation of the star-disc collision. The star was modeled as a solid, spherical body, and the interaction was simulated for a small, local section of the accretion disc. We found that star-disc collisions generate a nearly paraboloidal bow shock. The heating of gas is not confined to the column of gas directly ahead of the star but also extends laterally as the shock front expands sideways while traveling with the star. As the star crosses the disc, it injects momentum preferentially along its direction of motion, leading to an asymmetric redistribution of energy and momentum. As a result, two outflows emerge on opposite sides of the disc with different properties: the forward outflow expands faster, contains more mass, carries more energy, and is about twice as luminous as the backward outflow. Our findings suggest that the asymmetry in outflow properties and luminosity arises naturally from the collision dynamics, offering a possible explanation for the alternating strong-weak flare patterns observed in several QPE sources.

Molecular gas, modest in mass yet pivotal within the cosmic inventory, regulates baryon cycling as the immediate fuel for star formation. Across most of cosmic history, its reservoir has remained elusive, with only the tip of the iceberg revealed by luminous carbon monoxide (CO) emitting galaxies. Here we report the first detections of the mean cosmic CO background across its rotational ladder at 7$σ$, together with ionized carbon ([CII]) at 3$σ$, over $0<z<4.2$. This uses tomographic clustering of diffuse broadband intensities with reference galaxies, directly probing aggregate emission in the cosmic web. From CO(1-0) we infer the total molecular gas density, $Ω_{\rm H_2}$, finding it about twice that resolved in galaxy surveys. The global depletion time is $\sim$1 Gyr, shorter than the Hubble time, requiring sustained inflow. CO excitation links to star-formation surface density and, with depletion time, yields a super-linear Kennicutt-Schmidt law that appears universal. Together these results establish a global picture of galaxy growth fueled by a larger, short-lived molecular reservoir. The CO and [CII] detections mark a turning point for line-intensity mapping, replacing forecasts with empirical line strengths and defining sensitivity requirements for upcoming 3D experiments poised to open new windows on galaxy formation and cosmology.

WLM is an archetypal dwarf irregular galaxy that has not experienced interactions with major Local Group galaxies within the past 8 Gyr. It has recently been shown that WLM is losing its gas due to ram pressure forces exerted by the surrounding intergalactic medium (IGM). In this work, we explore how ram pressure may also affect the WLM gas kinematics, and we show that its dynamics is especially perturbed at its outskirts, explaining the asymmetric rotation between the approaching and receding sides. Moreover, we have been able to decompose WLM in two main components, a compact one with a solid-body rotation that resembles a bar-like structure, and a more extended one with a characteristic double-horn profile suggesting an edge-on disk. The former is relatively unaffected by ram pressure while the latter has its dynamics considerably affected by ram pressure. This study shows that mass estimates of a dwarf galaxy like WLM should account for a full modeling of its dynamical components, especially accounting for its asymmetric rotation curve.

We investigate the chemical abundances and star-formation histories (SFH) of ten massive ($\mathrm{log}_{10} (M_{\star}/\mathrm{M}_{\odot})>10.5$) quiescent galaxies at $3<z<4$ using deep, medium-resolution spectroscopic data obtained as part of the \textit{JWST DeepDive} Cycle 2 GO program. Our \textit{DeepDive} sample demonstrates early formation and quenching times inferred from spectro-photometric fitting, with most galaxies having formed 50\% of their stellar mass by $z \sim 5$, and quenching by $z \sim 4$, showing good agreement across the various SFH parameterizations explored in this work. Though they differ slightly between SFH parameterizations, the inferred formation timescales for the {\it DeepDive} sample span both rapid ($\lesssim$ 100 Myr) and more extended ($\gtrsim$ 200 Myr) episodes, corresponding to star formation occurring over a few to several dynamical times given their compact sizes and high densities at $z\sim3-4$. On average, massive quiescent galaxies at $3<z<4$ are $α$-enhanced ($\langle [α/\mathrm{Fe}]\rangle$= $0.22^{+0.22}_{-0.17}$), although there is strong diversity ($\sim0.3$ dex in scatter) among individual [$α$/Fe] values. Our results for $α$-enhancement are consistent with lower-redshift studies, implying weak evolution in [$α$/Fe] from $z \sim 4$ to $z\sim 1$. The SFH timescales associated with the low [$α$/Fe] measurements suggest longer formation timescales, potentially pointing to earlier enrichment by Type Ia supernovae, or metals preferentially being removed via outflows driven either by powerful early active galactic nuclei or supernovae. Overall, this work represents the first, statistically representative combined study of the star-formation histories and chemical abundances of massive quiescent galaxies at $z>3$.

Baryonic feedback processes associated with galaxy formation directly influence the large-scale structure by redistributing gas. Recent measurements of the kinetic Sunyaev-Zel'dovich effect and stacks of X-ray emission from optically selected galaxy clusters suggest that feedback from Active Galactic Nuclei (AGN) is more efficient at expelling gas from low-mass clusters than previously thought. The measurement of the cross-correlation between cosmic shear and diffuse X-ray emission provides a new probe of the distribution of gas in groups and clusters. We use the FLAMINGO cosmological, hydrodynamical simulations to examine the X-ray--cosmic-shear cross-correlation. The cross-correlation is most sensitive to the distribution of gas in haloes with masses $10^{14}\leq M_{200\mathrm{c}}/\mathrm{M}_{\odot}\leq10^{15}$. It is sensitive to the strength of feedback, but the effects of variations in cosmology and baryonic physics are largely degenerate. We compare the FLAMINGO predictions with the cross-correlation between cosmic shear from the Dark Energy Survey and ROSAT all-sky X-ray maps. We find that, if we neglect the X-ray emission from AGN that would remain unresolved by ROSAT, then the fiducial FLAMINGO model is in excellent agreement with the data, while models with stronger or weaker feedback are ruled out. However, if we account for unresolved AGN, either using the direct FLAMINGO predictions or by abundance matching to the observed (extrapolated) AGN luminosity function, then models with stronger feedback are preferred. We conclude that to exploit the potential of the X-ray--lensing cross-correlation, it will be necessary to resolve fainter AGN, and to use external constraints to break the degeneracy between baryonic feedback and cosmology.

Context. Distances to evolved massive stars in the Milky Way are not well constrained by Gaia parallaxes due to their brightness and variability. This makes it difficult to determine their fundamental stellar parameters, such as radius or luminosity, and infer their evolutionary states. Aims. We aim to improve the distance estimates of Yellow Hypergiants (YHGs) and Yellow Supergiants (YSGs) by identifying possible cluster and association memberships. Using these distances, we derived updated luminosities and revised their positions in the Hertzsprung-Russell diagram. Methods. We compiled from the literature a sample of 35 luminous yellow massive stars (YHGs and the most luminous YSGs). We used Gaia DR3 astrometry to identify possible membership in clusters and OB associations. We derived distances by combining the parallaxes of nearby co-moving stars. We independently validated these distances by comparing the stellar radial velocities to the Galactic H I kinematic map. We combined angular diameters and effective temperature values from the literature with the new distances to estimate luminosities. Results. We improved the distance estimates for 28 of the 35 stars through association with co-moving stellar groups. For an additional six stars, we provided distance estimates based on the H I kinematic map. For one star, the distance remains unclear. Most YSGs are members of young stellar populations, while the environments of the YHGs are more diverse, and for some of them, their origin populations remain unclear. We derived updated luminosities for a subset of 20 stars. Most YHGs have luminosities above log L/L = 5.4, while YSGs occupy a wider range of luminosities and the most luminous YSGs have luminosities similar to YHGs.

Although half-wave plates (HWPs) are becoming a popular choice of polarization modulators for cosmic microwave background (CMB) experiments, their non-idealities can introduce systematic effects that should be carefully characterized and mitigated. One possible mitigation strategy is to incorporate information about the non-idealities at the map-making level, which helps to reduce the HWP-induced distortions of the reconstructed CMB. Nevertheless, the non-idealities can only be known with finite precision. In this paper we investigate the consequences of discrepancies between their true frequency profiles and those assumed by the map-maker. We present an end-to-end framework, including a blind component-separation step, and use it to translate these discrepancies into a bias on the tensor-to-scalar ratio, $r$, for the LiteBIRD satellite mission. We subsequently derive realistic and conservative measurement requirements for accurately characterizing the HWP non-idealities to ensure they do not compromise LiteBIRD's ambitious scientific goals. We find that the obtained results are robust against sky models with varying complexity.

Establishing a unified origin that simultaneously accounts for the wide-band diffuse flux and recent ultra-high-energy (UHE) detections is a pressing challenge in multi-messenger astrophysics. Successive shock regimes in shock-breakout candidates, most notably low-luminosity gamma-ray bursts (LL GRBs), naturally introduce distinct physical environments producing a multi-component neutrino flux extending from 10 TeV to the UHE regime. Integrating prompt and afterglow phases within a unified dynamical framework yields a self-consistent explanation for this broadband emission. In this work, we discuss this framework, building on MWL observations. We show that the prompt emission from GRB~060218-like events accounts for $\gtrsim 10\%$ of the diffuse flux at 100~TeV, while GRB~100316D-like configurations predict a distinct flux peak near $10^{-9}\rm~GeV~cm^{-2}~s^{-1}~sr^{-1}$ at 100~PeV, providing a physical interpretation for the 220 PeV KM3-230213A event. This decoupling explains the lack of low-energy counterparts for individual UHE detections while maintaining consistency with the total diffuse neutrino flux. Ultimately, this framework identifies SBO-like LL~GRBs as a unifying origin for these phenomena, providing a physical link across a wide band from 10 TeV to EeV energies testable by next-generation observatories, including GRAND, IceCube-Gen2, and RNO-G.

Low-luminosity gamma-ray bursts (LL GRBs), a subclass of the most powerful transients in the Universe, remain promising sources of high-energy astrophysical neutrinos, despite strong IceCube constraints on typical long GRBs. In this work, a novel approach is introduced to study a sample of seven LL~GRBs with their multi-wavelength observations to investigate leptohadronic processes during their prompt emission phases. The relative energy densities in magnetic fields, non-thermal electrons, and protons are constrained, with the latter defining the cosmic-ray (CR) loading factor. Our results suggest that LL~GRBs exhibit diverse emission processes, as confirmed by a machine-learning analysis of the fitted parameters. Across the seven LL~GRBs, we find the posterior medians of the CR loading factor in the range of $ξ_p \sim 0.2$--$1.6$. GRB~060218 and GRB~100316D, the lowest-luminosity bursts ($L_{γ, \rm iso} \sim 10^{46}$-$10^{47}\rm~erg~s^{-1}$) consistent with the shock-breakout (SBO) scenario, yield the highest CR loading factor and therefore are expected to produce neutrinos more efficiently. Our model predicts the expected number of neutrino signals that are consistent with current limits but would be detectable with next-generation neutrino observatories. These results strengthen the case for LL~GRBs as promising sources of high-energy astrophysical neutrinos and motivate real-time searches for coincident LL~GRB and neutrino events. Next-generation X-ray and MeV facilities will be critical for identifying more LL~GRBs and strengthening their role in multi-messenger astrophysics.

The recently released Quaia quasar catalogue, with its broad redshift range and all-sky coverage, enables unprecedented three-dimensional reconstructions of matter across cosmic time. In this work, we apply the field-level inference algorithm BORG to the Quaia catalogues to reconstruct the initial conditions and present-day matter distribution of the Universe. We employ a physics-based forward model of large-scale structure using Lagrangian perturbation theory, incorporating light-cone effects, redshift-space distortions, quasar bias, and survey selection effects. This approach enables a detailed and physically motivated inference of the three-dimensional density field and initial conditions over the entire cosmic volume considered. We analyse both the G < 20.0 (Quaia Clean) and G < 20.5 (Quaia Deep) samples, where G denotes the Gaia broad optical-band magnitude, imposing conservative sky cuts to ensure robustness against foreground contamination. The resulting reconstructions span a comoving volume of (10h^{-1} Gpc)^3 with a maximum spatial resolution of 39.1 h^{-1}Mpc, making this the largest field-level reconstruction of the observable Universe in terms of comoving volume to date. We validate our reconstructions through a range of internal and external consistency checks, including the cross-correlation of the inferred density fields with Planck CMB lensing, where we detect a signal at ~4σsignificance. Beyond delivering high-fidelity data products, including posterior maps of initial conditions, present-day dark matter, and velocity fields, this work establishes a framework for exploiting quasar surveys in field-level cosmology.

We employ negentropy ($J$), defined as the difference between the information content of a non-Gaussian probability distribution and a Gaussian with identical variance, as an information-theoretic probe of non-Gaussianity in the cosmic density field. We quantify its sensitivity to dynamical dark energy by studying the evolution of $J(a)$ and its derivatives $Γ_1(a)$ and $Γ_2(a)$ across three parameterisation schemes: CPL, JBP, and BA. We determine the characteristic redshift $z_{NG}$, marking the epoch of maximal non-Gaussian structure formation, and the turnaround redshift $z_{TA}$, when information production transitions due to dark-energy domination, finding $z_{NG}\sim0.81$ and $z_{TA}\sim0.18$ for $Λ$CDM. Our diagnostics clearly discriminate between thawing and freezing quintessence models and phantom dark energy at low redshifts. Thawing models show small departures from $Λ$CDM, freezing models display higher $z_{TA}$, while phantom models exhibit lower $z_{TA}$, reflecting late-time evolution. We provide a practical prescription for measuring negentropy from discrete galaxy distributions, establishing a framework that can be applied to simulations and observations. This information-theoretic approach offers a robust and complementary tool for probing dark energy dynamics, enabling sensitive discrimination between evolving and cosmological-constant scenarios.

This work presents the first results of the PRImordial galaxy Survey with MIRI Spectroscopy (PRISMS), a JWST cycle 4 program (PID 8051) aimed at the characterization of a relatively large sample of ten galaxies about 500 Myr after the Big Bang. Here, we present deep (13.9 hours) spectroscopy with the MIRI LRS of the lensed galaxy UNCOVER-26185 at a redshift of z=10.054. It is a faint UV galaxy (UV absolut magnitude of -18.83 mag) previously identified as a X-ray luminous AGN. MIRI LRS detects the H$β$+[OIII]4960,5008 complex and H$α$ emission line with a significance of 10$σ$ and 8$σ$, respectively, as well as the optical continuum emission at rest-frame 0.45 $μ$m and 0.57 $μ$m with a signal-to-noise ratio of 6-7. The UV-to-optical spectral energy distribution, combining continuum and emission lines, is compatible with: (i) a low stellar (A$_V$= 0.2) and nebular (A$_V$=0.0) extinction, (ii) a SFH composed by a young (7 Myr) starburst and an intermediate-age (65 Myr) stellar population, and (iii) a total stellar mass of 1.7$\times$10$^{8}$ M$_{\odot}$. The H$α$-derived star-formation rate is 1.3 M$_{\odot}$ yr$^{-1}$. The low optical emission line ratios locate UNCOVER-26185 as the most metal-poor (Z = 0.04 Z$_{\odot}$), and as outlier with the lowest ionization (logU=-2.5) galaxy identified so far at redshifts above 9. With no evidence of an active galactic nuclei in the rest-frame UV-to-optical spectrum, UNCOVER-26185 has the properties of a metal-poor, main-sequence star-forming galaxy at redshift 10, with ISM and ionization properties very different than those of the already studied UV-bright galaxies at redshifts beyond 10. PRISMS is starting to explore the population of intermediate-UV luminosity galaxies at z=10, covering UV absolute magnitudes in the range of -17.9 to -20.5, fainter than those of UV-bright galaxies studied so far.

We present very deep (~11h) JWST/MIRI low-resolution spectroscopy of the rest-frame optical emission of U37126, a UV-bright (M_UV ~ -20), mildly lensed ($μ\simeq 2.2$) galaxy at z=10.255. The continuum emission is well detected in both NIRSpec and MIRI spectra, yet no nebular recombination or metal emission lines are observed (EW(Hbeta+[OIII])<300A and EW(Halpha)<400A, at 3sigma). Combined with the exceptionally blue UV continuum slope, beta_UV ~ -2.9, and weak/flat Balmer break, these constraints indicate a stellar population dominated by very young and massive stars with a strongly suppressed nebular contribution. Comparisons with synthetic stellar population models indicate that U37126 requires both a very high ionizing photon production efficiency, log(Xi_ion / Hz erg^-1) ~ 25.75, and a nearly unit LyC escape fraction, of fesc>86% (3sigma) based on Halpha flux limit and fesc=0.94+/-0.06 derived independently from SED fitting. The best-fit SED yields a (de-lensed) stellar mass of Mstar ~ 10^7.8 Msun and a star-formation rate of SFR~10Msun/yr (sSFR~160 Gyr^-1), that along with its very compact size, reff~61pc, yields very high stellar mass and star-formation-rate surface densities, Sigma_M ~ 3x10^3 Msun/pc^2 and Sigma_SFR ~ 400 Msun/yr/kpc^2. Together with the lack of detectable nebular emission, these properties suggest that U37126 is undergoing an ``ISM-naked'' starburst phase, possibly driven by an extremely efficient gas-to-star conversion followed by strong feedback that has cleared the remaining gas from its stellar core, allowing most LyC photons to escape. Finally, we show that even a small fraction of galaxies like U37126 (~ 3%-6%), with extreme LyC production and escape, could contribute disproportionately (~ 50%-100%) to the ionizing photon budget during cosmic reionization.

Recent studies suggest that chemical abundances hold the key to disentangling halo substructure, providing a more reliable tracer than dynamics alone. We aim to probe the Milky Way stellar halo using high-dimensional chemical abundances from GALAH DR4. By leveraging multiple nucleosynthesis channels in synergy with integrals of motion (IoM), we extract information hidden in the raw abundance space to perform chemical tagging. With a graph attention autoencoder, we reconstruct a dynamics-informed, denoised chemical space and identify coherent stellar substructures by applying ensemble clustering. Our method successfully recovers the three largest globular clusters hidden in the dataset, estimates the in-situ fraction to be approximately 41\%, and chemically characterizes several dynamical halo substructures. Strikingly, stars dynamically associated with Gaia-Sausage-Enceladus (GSE) separate into two chemically distinct clusters. By examining their abundances, energy ($E$) and angular momentum ($L_z$) distributions, together with the metallicity trend with $E$, we connect these clusters to their birthplace within the progenitor by proposing a simple infall scenario: one cluster traces the metal-poor, less evolved outskirts, while the other traces the metal-rich, chemically evolved core.

It has been theorized that the formation of extremely massive and supermassive stars ($>10^3\ {\rm M}_\odot$) could plausibly be the outcome of stellar mergers in low metallicity ($Z<10^{-1}$~Z$_\odot$) and dense ($\gtrsim10^3\ {\rm M}_\odot\ {\rm pc}^{-3}$) stellar environments. These objects remain relevant as they can serve as the progenitors of intermediate-mass black holes and they are also formidable chemical polluter candidates, as evidenced by the peculiar abundances seen across cosmic history. This work investigates merger-induced mass loss in extremely massive stars within a hydrodynamic framework and provides a prescription derived from the simulations to estimate both the mass loss and the outcome of the interaction. We adapted the 1D hydrodynamic, stellar structure, and evolution code MESA to simulate stellar inspirals. In our simulations, we considered stars of $>1000\,\rm M_{\odot}$ with inspiraling companions of $<100$ M$_\odot$; hence, with mass ratios of $<0.1$. As the inspiral progresses, the orbital energy of the system is lost through the hydrodynamic and gravitational drag forces. This energy gets deposited as thermal energy in the extremely massive star's envelope. We find that the total ejected mass is $\sim$10-30$\%$ of the system's mass. Our results point out that most of the energy deposited by the inspiral is used to eject mass. These findings demonstrate that merger-induced mass loss is non-negligible for the considered configurations. Thus, it is an important process to account for when investigating the formation of extremely massive stars and predicting their possible role throughout cosmic history.

We focus on the sodium destruction in models reaching the high hot bottom burning temperatures needed to efficiently cycle oxygen to nitrogen in AGB models at the nominal [Fe/H] of the cluster NGC 2808. We increase the initial neon abundance by a factor 2-4 with respect to the "standard" abundances obtained by scaling the solar values down to the metallicity of this cluster, and explore the average abundances in the ejecta obtained by adopting smaller mass-loss rates. Higher neon produces higher sodium in the AGB envelope. Lowering the mass-loss rate allows both to keep reasonably large sodium abundances and to increase the depletion of oxygen and magnesium. A balance between the lower mass-loss rates and the necessity of not increasing too much the episodes of third dredge up gives a neon abundance larger by a factor two and a mass-loss rate smaller by a factor four as best compromise. Comparison with the abundances in NGC 2808 shows a better agreement than the standard models for all the patterns of abundances, but the extreme stars (group E) requires models slightly less rich in iron. t Thus, we propose that the extreme population in NGC 2808 is composed of stars having a slightly smaller metallicity, and sketch a possible scenario for its formation, in the framework of the hierarchical clusters assembly scenario. Abundances of potassium are larger by $\sim 0.2 dex$ in the E group, but the explanation in terms of burning of the initial argon requires a drastic increase of the relevant cross section. The abundances of neon and argon at low metallicities may be an important tool to better reproduce the abundances of light elements in the framework of the AGB model for globular clusters.

The discovery of transient phenomena, such as supernovae, novae, Fast Radio Bursts (FRBs), Gamma-Ray Bursts (GRBs), and stellar flares, together with the emergence of new cosmic messengers like high-energy neutrinos and Gravitational Waves (GWs), has revolutionized astrophysics in recent years. To fully exploit the scientific potential of multi-messenger and multi-wavelength follow-up observations, as well as serendipitous detections, researchers need a tool capable of rapidly compiling and contextualizing essential information for every new event. We present Astro-COLIBRI, an advanced platform designed to meet this challenge. Astro-COLIBRI is a comprehensive platform that combines a public RESTful API, real-time databases, a cloud-based alert system, and user-friendly interfaces including a website and mobile app for iOS and Android. It ingests alerts from multiple sources in real time, applies user-defined filters, and situates each event within its multi-messenger and multi-wavelength context. The platform provides clear data visualization, concise summaries of key event properties, and evaluations of observing conditions across a wide network of observatories worldwide. We here detail the architecture of Astro-COLIBRI, from the data pipelines that manage real-time alert ingestion and processing to the design of the RESTful API, which enables seamless integration with other astronomical software and services.

Radio emission from magnetically active stars arises mainly from non-thermal processes and complements high-energy X-ray emission. Sensitive, wide-field radio and X-ray surveys now allow identification of larger samples of active stars across the Galaxy. We aim to identify and characterise radio and X-ray-emitting stars in the Galactic plane by combining MeerKAT radio data with soft X-ray observations and assess their consistency with the canonical Güdel-Benz relation, which links thermal coronal X-rays to non-thermal gyrosynchrotron radio emission. We cross-matched compact sources from the SARAO MeerKAT Galactic Plane Survey with counterparts from the ROSAT All-Sky Survey and the first release of SRG/eROSITA (eRASS1). We computed radio-brightness temperatures and radio-X-ray luminosities to test the relation. We identify 137 stars with both radio and X-ray detections. Their $T_B$ ranges from $10^7$ to $10^{12}$ K, except two outliers: AXJ1600.9-5142 ($4.8 \pm 1.5 \times 10^{12}$ K) and HD~124831 ($8 \pm 1 \times 10^{6}$ K). The remainder are consistent with incoherent gyrosynchrotron emission. The sample lies below the canonical Güdel-Benz relation, driven by enhanced 1.3 GHz radio luminosities relative to the 5 GHz relation. This suggests the classical relation represents an upper envelope rather than a tight correlation. Additionally, eROSITA detections show early-type stars lie below the typical $\log (L_{\rm X}/L_{\rm bol}) \sim -3$ relation.

Emission line galaxies (ELGs) are the primary tracers of the large-scale structures of the Universe in ongoing Stage-IV cosmological spectroscopic surveys, which aim to measure the clustering statistics at higher redshifts $z \simeq 1.5 \text{--} 2$ with unprecedented precision. In this study, we construct realistic mock ELG samples with IllustrisTNG hydrodynamical simulations and stellar population synthesis framework. In order to validate the modelling of clustering, we measure the anisotropic correlation functions of mock ELGs and infer the linear growth rate, which is one of key cosmological parameters in galaxy clustering. As a control sample, we construct the mass-limited subhalo samples with the same number density as ELGs. The isotropic correlation functions in real space for both samples do not differ significantly. However, the quadrupole moment of the anisotropic correlation function, which is sensitive to the velocity of galaxies, is suppressed for ELGs, potentially due to the infalling motion of ELGs towards the centre of the hosting halos. The smaller amplitude leads to the underestimation of the linear growth rate and implies the velocity bias between ELGs and dark matter. When the analysis is limited to large scales $(\gtrsim 15 \, h^{-1} \, \mathrm{Mpc})$, the parameter bias vanishes. Next, we investigate the ELG-halo connection through the phase-space distribution of satellite ELGs within hosting halos and galactic conformity of star formation activity. The infalling motion is further confirmed by the phase-space distribution relative to the host halo, and this dynamics of ELGs challenges the assumption that the radial distribution of satellites follows that of dark matter.

Galaxy clusters are powerful probes of the growth of cosmic structure through measurements of their abundance as a function of mass and redshift. Extracting precise cosmological constraints from cluster surveys is challenging, as we must contend the complex relationship between richness and the underlying halo mass, selection function biases, super-sample covariance, and correlated measurement noise between mass proxies. As upcoming photometric surveys are expected to detect tens to hundreds of thousands of galaxy clusters, controlling these systematics becomes essential. In this paper, we present a forward-modelling approach using Simulation-Based Inference (SBI), which provides a natural framework for jointly modelling cluster abundance and lensing mass observables while capturing systematic uncertainties at higher fidelity than analytic likelihood methods - which rely on simplifying assumptions such as fixed covariances and Gaussianity - without requiring an explicit likelihood formulation. We introduce $\texttt{Capish}$, a Python code for generating forward-modelled galaxy cluster catalogues using halo mass functions and incorporating observational effects. We perform SBI using neural density estimation with normalizing flows, trained on abundance and mean lensing mass measurements in observed redshift-richness bins. Our forward model accounts for realistic noise, redshift uncertainties, selection functions, and correlated scatter between lensing mass and observed richness. We find good agreement with likelihood-based analyses, with broader SBI posteriors reflecting the increased realism of the forward model. We also test $\texttt{Capish}$ on cluster catalogues built from a large cosmological simulation, finding a good fit to cosmological parameters.

Observationally, early H II regions are classified by size into ultracompact and hypercompact configurations. It remains unclear whether these phases are long-lived or transient. Understanding the physical processes that stall H II region growth may help to solve the so-called lifetime problem: the observation of more compact H II regions than expected from theory. Utilizing two-dimensional, axially symmetric radiation hydrodynamic simulations of young expanding H II regions, including the phase of early star and disk formation, we seek to better understand the trapping of H II regions. Trapping forces include gravity and ram pressure, which oppose forces such as thermal pressure expansion, radiation pressure, and centrifugal force. Without radiation pressure, the H II region remains gravitationally trapped in the ultracompact phase indefinitely. With radiation pressure, the H II region escapes gravitational trapping but experiences ram pressure trapping on larger scales. For initial mass reservoirs with high central density, no trapping occurs, while a less steep density gradient yields clear trapped phases. Hypercompact trapped phases exhibit a so-called flickering variation in H II region radius, in agreement with observations of stalling and even contraction over small time scales. With radiation pressure, low-density reservoirs experience both gravitational and ram pressure trapping, while high-mass reservoirs undergo only the latter.

We conducted a search for brown dwarf candidates in a James Webb Space Telescope deep field around A2744 to investigate the space density of these objects at kiloparsec distances. Our methodology employed an initial selection based on photometric colours, followed by spectral energy distribution fitting to both stellar atmospheric models and high-redshift galaxy templates. This approach yielded two robust T dwarf candidates and one possible L subdwarf candidate. The T dwarfs have estimated Galactic heights of 0.43 and 0.86 kpc, likely residing near the outer edges of the Galactic thin and thick discs, respectively. We measure a T dwarf surface number density of 0.094 per squared arcmin in the UNCOVER field, lower than previous predictions but consistent at the order-of-magnitude level. We also provide space number density estimates for T5-T8.9 dwarfs across different effective temperature and spectral type bins, finding that T5-T7 dwarfs out to 2 kpc have significantly lower densities than their solar neighbourhood counterparts, whilst T8 dwarfs within the thick disc exhibit densities comparable to local values. Our analysis demonstrates that broad-band near- to mid-infrared photometry provides high sensitivity to late-T dwarfs but is relatively less sensitive to L and early-T dwarfs. Spectroscopy is typically required to distinguish photometric candidates of L dwarfs, early-T subdwarfs, and high-redshift galaxies in JWST deep fields. This study demonstrates the potential for expanding our understanding of brown dwarf distributions and characteristics at unprecedented distances, offering new insights into substellar populations beyond the solar neighbourhood.

We introduce a physics-informed neural framework for modeling static and time-dependent galactic gravitational potentials. The method combines data-driven learning with embedded physical constraints to capture complex, small-scale features while preserving global physical consistency. We quantify predictive uncertainty through a Bayesian framework, and model time evolution using a neural ODE approach. Applied to mock systems of varying complexity, the model achieves reconstruction errors at the sub-percent level ($0.14\%$ mean acceleration error) and improves dynamical consistency compared to analytic baselines. This method complements existing analytic methods, enabling physics-informed baseline potentials to be combined with neural residual fields to achieve both interpretable and accurate potential models.

We conduct a detailed spectral analysis of the Galactic ultraluminous X-ray pulsar Swift J0243.6+6124 in its sub-Eddington regime, using Insight-HXMT and NICER observations during multiple outbursts including the 2018 giant outburst. We discover a new transition at $L_{\rm t} \approx 4.5 \times 10^{37}\ {\rm erg\ s^{-1}}$, accompanied by systematic evolution of spectral parameters, in particular a significant turnover in the blackbody normalization. This transition luminosity in the sub-Eddington regime represents the fifth transition identified so far in Swift J0243.6+6124, further highlighting the complexity of its accretion-powered emission. We interpret the transition in terms of a multipolar magnetic-field configuration, where weak ($\sim 2.8 \times 10^{12}\ {\rm G}$) and strong ($\sim 1.6 \times 10^{13}\ {\rm G}$) magnetic poles dominate the emission at different accretion rates. On the magnetospheric scale, this configuration is equivalent to an effective dipole field of $\sim 6.6 \times 10^{12}\ {\rm G}$, while allowing the local surface field to exceed $10^{13}\ {\rm G}$.

The literature has not converged onto a precise depiction of the magnetogenesis process for pulsars, and it is profitable to preliminarily but exhaustively assess the viability of the plethora of alternative proposals, before substantial efforts are invested into simulating them in detail. In this note, we tackle one of them, taking notice of an earlier work that suggests neutrino ponderomotive force could spawn a magnetic field not so far off from pulsar strengths. We reexamine this mechanism with more modern technology, accounting for actual core collapse dynamics, and show that this mechanism is likely less powerful than originally envisioned.

Giant radio sources (GRSs) harbor the Universe's largest structures generated by individual galaxies, with projected source sizes exceeding 700 kpc. These enigmatic objects have been mainly studied at radio frequencies, and their physical properties in the high-energy domain are poorly understood. Here we present the results of a multiwavelength study focused on NuSTAR J112829+5831.8 (J1128+5831), the only known GRS serendipitously detected with the Nuclear Spectroscopic Telescope Array. Being located in proximity to the famous interacting galaxy system, Arp 299, J1128+5831 has been serendipitously observed also by the Chandra X-ray Observatory, Hubble Space Telescope, and XMM-Newton satellites. From radio observations with the Low Frequency Array, the NRAO VLA Sky Survey and the Very Large Array Sky Survey, we have determined that J1128+5831 has an overall steep radio spectrum ($α=-0.86$; $F_ν\proptoν^α$) and a low core dominance ($C_{\rm D}=-2.4$, in log-scale), indicating the source to be viewed at large angles. From the X-ray spectral analysis, we found J1128+5831 to harbor an obscured active galactic nucleus (AGN) with neutral hydrogen column density exceeding $10^{23}$ cm$^{-2}$. Its optical spectrum, taken with the Dark Energy Spectroscopic Instrument, exhibits prominent narrow emission lines but lacks broad components, thus confirming J1128+5831 to be a Type 2 AGN powered by a radiatively efficient accreting system. Overall, the broadband properties of J1128+5831 are consistent with those observed for the general GRS population.

The formation of super star clusters (SSCs) in galaxies remains a fundamental yet unresolved problem. Among the proposed mechanisms, cloud-cloud collisions (CCCs) have been suggested as a potential trigger, although observational validation has been limited. Here we present high-resolution ($0.12^{\prime\prime}$, $\sim14\,\mathrm{pc}$) ALMA observations of CO ($J=1\!-\!0$) emission toward a super giant molecular cloud (SGMC) in the overlap region of the Antennae galaxies. The data resolve the SGMC into two distinct velocity components separated by $\sim50\,\mathrm{km\,s^{-1}}$. One component exhibits a ``U-shaped'' structure within a large filament likely shaped by ram pressure, while the other shows hub-filament morphology. Such a morphology is naturally interpreted as a CCC scenario. The 108\,GHz continuum emission detected at the apparent collision interface is dominated by free-free radiation, with an ionizing photon rate consistent with the stellar mass and age of the optically identified SSCs. Supplementary infrared imaging with JWST reveals emission spatially coincident with the inferred collision interface, further supporting the CCC scenario. These results provide compelling, multi-wavelength evidence that CCCs play a key role in triggering SSC formation in merging galaxies.

We present a multiphase study of the star-formation-driven outflow in the Virgo galaxy NGC 4383, combining ALMA CO(2-1) data with deep MeerKAT HI imaging and MUSE spectroscopy obtained as part of the Multiphase Astrophysics to Unveil the Virgo Environment (MAUVE) program. Our previous work revealed a spectacular ionised outflow, but the effect of the outflow on the cold phase remained unclear. Our analysis shows that potentially outflowing molecular gas is detected only within the inner 1 kpc above the disc, where CO clouds exhibit disturbed kinematics and spatial correspondence with the ionisation cone. At larger heights, the CO surface brightness rapidly drops, indicating that the molecular phase contributes little to the mass of outflowing gas. In contrast, the HI distribution shows plumes a few kiloparsecs above the disc that are aligned with the ionised cone, and complex kinematics suggestive of parts of the atomic phase being entrained in the outflow. However, the extended and warped HI disc associated with NGC 4383 complicates the unambiguous identification of outflowing atomic gas and, most importantly, the quantification of outflowing mass and loading factor. Independent support for a cold component in the outflow comes from dust extinction features associated with the outflow and coincident with HI plumes. Despite significant uncertainties in the estimate of the mass of cold gas associated with the outflow, these results suggest that the atomic phase likely dominates the cold outflow above 1 kpc. The observed cold gas velocities remain below the velocities of the ionised phase, suggesting that NGC 4383 does not host a large-scale escaping wind but more likely a galactic fountain, in which feedback redistributes material within the halo and regulates ongoing and future star formation.

Active Galactic Nuclei (AGN) sources feature supermassive black holes that launch relativistic plasma jets. They are key $γ$-ray sources providing a unique laboratory for studying extreme particle acceleration and plasma physics. Variability in $γ$-ray emission is an important signature that may constrain the size of the emission region and the physical processes driving flares. However, current large-scale $γ$-ray catalogs, such as the Fermi-LAT 4LAC-DR3, typically characterize variability only on long timescales (yearly or 60-day), lacking necessary constraints on short-term behavior from days to weeks. To address this, we systematically characterize $γ$-ray variability in AGNs across short timescales: 3-day, 7-day (weekly), and 30-day (monthly). We present a preliminary catalogue of variable AGN based on light curves from the Fermi-LAT Light Curve Repository. We show that the variability amplitude ($σ_{\rm NXS}^{2}$) presents similar values across different timescales, potentially increasing for a subsample of sources as the observation timescale increases. This high-cadence analysis reinforces the known dichotomy between flat-spectrum radio quasars (FSRQs) and BL Lacertae objects (BL Lacs), with FSRQs consistently exhibiting stronger variability. By identifying the most luminous and variable sources at each timescale, we highlight key targets for follow-up with next-generation observatories such as the Cherenkov Telescope Array Observatory (CTAO), ASTRI Mini-Array, and the Southern Wide-field Gamma-ray Observatory (SWGO), where strong short-term variability suggests highly compact emission zones and extreme particle acceleration efficiency. This catalogue contributes to the understanding of high-energy outflows in AGN jets and provides a foundation for optimizing observational strategies through a unified variability metric across timescales.

We compare the growth in stellar mass of galaxies in the $6<z<12$ epoch with predictions of a semi-analytic galaxy formation model - Galacticus. In contrast to diverse and controversial results that compare models and data for the \emph{luminosity} evolution of galaxies -- reported in an abundance of recent papers, we find very good, unambiguous agreement in the more fundamental quantity of stellar mass - measured from JWST observations - and Galacticus predictions. Specifically, we find good agreement for the shape of the integrated stellar mass as a function of redshift without any adjustment of parameters, and in \emph{amplitude} as well, when 'feedback' is lowered by a factor of 3 compared to that required to match later-universe models and data. This result emerged from detailed investigation of the claim by Dressler et al. that bursts of star formation dominated the growth in stellar mass, specifically, that half of the galaxies with stellar mass growth of $M_* > 2\times10^8 \mathrm{M}_\odot$ in the epoch $8<z<6$ had less than $M_*<\times10^8 \mathrm{M}_\odot$ prior to $z = 8$. Here too we find agreement between models and data, namely that these ~100 Myr 'bursts' had strong in situ growth at $z\le8$, or showed (in Galacticus) substantial stellar and/or gas-rich mergers, and 30-40 Myr 'starbursts' as are common in $z<3$ galaxies. We note that, if a theoretical simulation is unable to pass the test of matching the growth of stellar mass, any success in reproducing the luminosity function is meaningless.

We present a neural network classification (NNC) method for photometric redshift estimation that produces well-calibrated redshift probability density functions (PDFs). The method discretizes the redshift space into ordered bins and optimizes the Continuous Ranked Probability Score (CRPS), which respects the ordinal nature of redshift and naturally provides uncertainty quantification. Unlike traditional regression approaches that output single point estimates, our method can capture multi-modal posterior distributions arising from color-redshift degeneracies. We apply this method to the DESI Legacy Imaging Surveys Data Release 10 (LSDR10) and Pan-STARRS Data Release 2 (PS1DR2), using an unprecedented spectroscopic training sample from DESI DR1 and SDSS DR19. Our method achieves $σ_{\mathrm{NMAD}} = 0.0153$ and $η= 0.50\%$ on LSDR10, and $σ_{\mathrm{NMAD}} = 0.0222$ and $η= 0.34\%$ on PS1DR2 combined with unWISE infrared photometry. The NNC method outperforms Random Forest, XGBoost, and standard neural network regression. We demonstrate that DESI DR1 significantly improves photo-$z$ performance at $z > 1$, while the combination of deep optical photometry and mid-infrared coverage is essential for achieving high precision across the full redshift range. We provide a unified photometric redshift catalog combining LSDR10 and PS1DR2 with a hierarchical model selection strategy based on available photometry. The well-calibrated PDFs produced by our method are valuable for cosmological studies and can be extended to next-generation surveys such as CSST, Euclid, and LSST.

{Early-type galaxies host low-luminosity active galactic nuclei, traced by radio emission spanning parsec- to kiloparsec scales.} {We investigate the Eddington-scaled radio luminosity distribution of 117 nearby early-type galaxies to test for bimodality and assess the role of host-galaxy properties, extending results from a 62-galaxy sample \cite{Wojtowicz2023}.} {We compile galaxies with directly measured black hole masses and 1.4,GHz and 3,GHz flux densities. Statistical tests assess bimodality, while VLASS imaging, host-galaxy kinematics, and central stellar structure characterize radio-dim and -bright sources.} {Using the 117-galaxy sample, we confirm that $L_{\rm 1.4,GHz}/L_{\mathrm{Edd}}$ is bimodal, with an antimode at $\approx -8.6$, which disappears when black hole masses are inferred from the $M_{\rm BH}$-$σ_\star$ relation. Radio-bright galaxies host resolved jets, while radio-dim systems show compact nuclear emission often exceeding that expected from star formation (FIR-radio correlation). Radio-bright galaxies are mainly slow rotators with depleted cores; radio-dim galaxies are predominantly fast rotators.} {Nearby early-type galaxies show a clear bimodality in Eddington-scaled radio luminosity, separating compact, radio-dim nuclei from extended, radio-bright systems. The dichotomy correlates with host-galaxy kinematics and central structure, indicating that sustained jet production depends primarily on galaxy assembly history and feeding mode rather than black hole mass or accretion rate alone. Radio-dim emission likely reflects intermittent, stochastic delivery of magnetized gas, plausibly via tidal disruption of giant-branch stars near the SMBH.}

We conduct an in-depth study of the quasi-periodic oscillation (QPO) properties of RE J1034+396, by constructing QPO phase-folded light curves from 10 XMM-Newton observations during 2020-2021. Our analysis reveals that the QPO in the source exhibits two mutually convertible lag-energy modes: "hard lag" and "soft lag". Despite different lag characteristics, the energy dependency of the root mean square (RMS) amplitude of the QPO under both modes are consistent, suggesting the two types of QPO originate from the same physical mechanism. By performing a spectral analysis, we further find a correlation between time-lag modes and spectral states: the soft lag mode typically corresponds to harder X-ray spectra and higher blackbody temperatures. Through comprehensive comparison of multiple theoretical models, we propose that the relativistic precession model (RPM) of the corona provides a plausible qualitative explanation for the observed complex phenomena, including time-lag mode transitions, and variations of spectral hardness and QPO signal strength.

Recent observations have revealed that the ion-neutral drift velocity in star-forming molecular clouds and dense cores is on the order of 100 m s^-1. Theoretical studies have shown that, in ambipolar diffusion, the process responsible for the differential motion between ions and neutrals, the dust size distribution has a significant impact on the magnetic resistivities. In this study, we perform simulations to investigate how dust growth through accretion and coagulation affects the ion-neutral drift velocity in molecular clouds and cores. We find that, on core scales, both dust growth and a magnetic field strength of 200 microgauss are required to reproduce the observed drift velocity. We suggest that measurements of ion-neutral drift velocity, particularly on core scales, may serve as a new diagnostic to constrain the dust size distribution and magnetic field strength in such environments.

Solar flares promptly release large amounts of free magnetic energy in the solar corona to produce substantial populations of high-energy charged particles, both ions and electrons. These particles are detected when they radiate microwaves in solar magnetic fields and X- and γ-rays when they encounter matter. Analysis of γ-rays in solar flares has revealed a distinct continuum component dominating at MeV energies, which differs from the well-studied X-ray continuum produced by flare-accelerated electrons with steeply falling energy spectra. The origin and precise spatial location and extent of this mysterious MeV component have been unknown up to now. If it is produced by bremsstrahlung, such a γ-ray component requires an unusual population of electrons peaked at a few MeV. Here we report a joint study of this MeV-peaked electron population in the 2017-Sep-10 solar flare with Fermi MeV γ-ray data and EOVSA spatially resolved microwave imaging spectroscopy data. We demonstrate that the microwave spectrum from the peaked MeV distribution has a distinctly different shape from that produced by the well-known population of electrons with falling energy spectrum. We inspected microwave maps of the flare and identified an evolving area where the measured microwave spectra matched the theoretically expected one for the MeV-peaked population, thus pinpointing the site where this MeV component resides in the flare. The locations are in a coronal volume adjacent to the region where prominent release of magnetic energy and bulk electron acceleration were detected, which implies that transport effects play a key role in forming this population.

We introduce a variable-slope smooth-$k$ (VSMK) filter within the Press-Schechter formalism to model halo mass functions derived from damped and oscillatory matter power spectra. While the standard smooth-$k$ approach successfully captures small-scale suppression effects, it intrinsically couples these to oscillatory features at intermediate scales. The VSMK filter generalizes this framework by allowing the effective logarithmic slope of the $k$-space window function to vary smoothly between two asymptotic regimes, thereby decoupling the small-scale suppression of halo abundances from the intermediate-scale oscillatory features characteristic of dark acoustic oscillations. We compare the analytic predictions obtained with the VSMK filter to $N$-body simulations for warm dark matter and ETHOS-based models, showing that a single parameter set reproduces both regimes simultaneously. The VSMK filter thus provides a unified and flexible analytic framework for modeling halo abundances in non-cold dark matter scenarios with damped and oscillatory power spectra.

A significant fraction of galaxies show warps in their discs, usually noticeable at its periphery. The exact origin of this phenomenon is not fully established, although multiple warp formation mechanisms are proposed. In this study, we create a sample of more than 1000 distant ($z \lesssim 2.5$) edge-on galaxies imaged by HST and JWST. For these galaxies, we measurd characteristics of warps and finally analyse how their parameters and frequency change with time. We focus on our main result that galaxies with strong warps were more prevalent in the past compared to the modern epoch. We check how selection effects and varying image quality between objects in our sample could influence our results and conclude that varying fraction of warped galaxies is not caused by observational effects, but represents a genuine evolution. Such a trend may be consistent with mergers and interactions between galaxies being the primary mechanism of warp formation, as number density of galaxies decreases with time, implying higher rate of mergers and interactions in the past.

We present a gas kinematic study of the massive protocluster G318.049+00.086. The protocluster is reported to contain 12 prestellar core candidates and 4 protostellar cores. Filamentary structures are identified using the 1.3 mm dust continuum map, with four of them converge into a dense central region, forming a hub-filament system (HFS). High velocity gradients (10 - 20 km s$^{-1}$ pc$^{-1}$) derived from PV analysis of H$^{13}$CO$^{+}$ emission along three of those filaments are suggestive of mass inflow onto the central hub. A mass inflow rate higher than $10^{3}$ M$_{\odot}$ Myr$^{-1}$ along the filaments is indicating that the central hub is capable of forming massive star(s). Investigation of H$^{13}$CO$^{+}$ and CCH spectral profiles revealed the majority of the cores having the characteristic blue asymmetric line profiles, typical signature of gravitational collapse. The remaining few cores showed red asymmetric profiles, indicative of gas expansion. Also, the derived mass infall rates for the protostellar cores in hub-region is significantly higher in comparison to those located along the filaments. The mass-radius relationship of the cores revealed that the cores with red profiles reside in the massive star formation regime. However, the global velocity gradient along the filaments suggests that these particular cores are losing material to the hub. Our results are supporting a competitive accretion scenario of massive star formation where gas is expected to be funnelled from less gravitationally dominant cores to the cores located at the gravitationally favorable position.

Recent Pulsar Timing Array datasets provide compelling evidence for a nano-Hertz gravitational-wave background, but robust detection requires characterizing statistical fluctuations of the Hellings-Downs (HD) correlation expected from a finite population of discrete sources. Building on the variance calculation of Allen (2023), we derive the third central moment (skewness) of the HD correlation for a single unpolarized point source and an ensemble of many interfering point sources in the confusion-noise regime. To isolate the intrinsic non-Gaussianity of the background, we extend the pulsar-averaging formalism to third order by introducing a three-point averaged correlation function, which allows us to define the cosmic skewness. We find that the skewness remains non-zero in the large-source-number limit and is controlled by a new geometric three-point function. These results suggest that incorporating higher-order moments could provide additional information on source discreteness beyond standard Gaussian analyses.

We conduct a comparative analysis of galaxy kinematics using IllustrisTNG simulations and integral-field spectroscopy (IFS) observations. We identify 2,342 early-type galaxies (ETGs) from the TNG100 simulation and 236 ETGs from the TNG50 simulation, comparing them with observations from MaNGA and ATLAS$^{3D}$. For these systems, we measure key kinematic parameters: the intrinsic spin parameter $λ_{R,\mathrm{intr}}$ (measured edge-on), the cylindrical rotational energy fraction $κ_{\mathrm{rot}}$, and structural mass ratios including the spheroid mass fraction $f_{\mathrm{spheroid}}$ and stellar halo mass fraction $f_{\mathrm{halo}}$. Our study reveals that standard classifiers--the $λ_{R}(R_e)=0.31\sqrt{\varepsilon}$ relation and $\overline{k_5}$ coefficient (higher-order Fourier term of velocity fields)--fail to align with observed kinematic bimodality. We propose revised thresholds: $λ_{R,\mathrm{intr}} \sim 0.4$, $κ_{\mathrm{rot}} \sim 0.5$, and $f_{\mathrm{spheroid}} \sim 0.6$, which classify galaxies into rotation-dominated (fast rotators) and random motion-dominated (slow rotators). Scaling relations from TNG enable observational estimates of $κ_{\mathrm{rot}}$ and $f_{\mathrm{spheroid}}$. The simulations exhibit a bimodality deficit, characterized by a lack of fast rotators and suppressed $λ_{R,\mathrm{intr}}$, attributed to excess galaxies with intermediate rotation and high spheroid/stellar halo mass. We introduce a novel method to estimate $f_{\mathrm{halo}}$ from IFS kinematics, though uncertainties remain.

Magnetic fields play a central role in accretion physics around black holes, yet their physical origin within accretion flows remains an open problem. In this work, we investigate the generation and subsequent evolution of magnetic fields triggered by anisotropic radiation fields in black hole accretion discs with compact rotating inner corona. We self-consistently evolve the magnetic field using the generalized field evolution MHD equation, including advection, shear-driven induction, and Hall effects. The radiation field acts as a primary field generator, while azimuthal rotation in the magnetized plasma provides rapid amplification. We find that radiation-generated fields efficiently reach a dominant toroidal component by Keplerian rotation, leading to magnetic field strengths of order $\sim 10^{8}\,\mathrm{G}$ in the vicinity of a 10 solar mass black hole and accretion disc-corona emitting at luminosity equivalent to the Eddington unit. These magnetic fields are achieved within viscous timescales and reach or exceed local equipartition estimates based on gas pressure. When vertical outflows are included, the amplified magnetic fields are advected into the corona, magnetizing disc-launched winds and jet precursors with field strengths of similar order. Our results demonstrate that radiation is not merely a passive component of accretion flows, but provides a robust and unavoidable trigger for the generation of dynamically significant magnetic fields. Our results provide a physically grounded explanation for the origin of large-scale, structured magnetic fields in and around accretion discs. This mechanism offers a pathway for magnetizing accretion discs and their outflows without invoking externally supplied magnetic flux, with broad implications for X-ray binaries, active galactic nuclei and other transients such as gamma-ray bursts (GRBs).

The peculiar gamma-ray burst GRB 250702B is the longest event ever observed, lasting about one day and exhibiting four prompt-emission flares of $\sim100$ s with irregular recurrence intervals of at least one hour. To explain this hierarchy of timescales, we consider a scenario in which a stellar object undergoes repeated partial tidal disruptions by a black hole (BH). We find that if a white dwarf (WD) is on a highly eccentric orbit ($e\approx0.97$) around an intermediate-mass black hole (BH) with $M_{\rm BH}\lesssim10^{6}\,M_\odot$ and $a = 50\,R_\odot\left(M_{\rm BH}/10^{6}\,M_\odot\right)^{1/3}$, the observed properties of GRB 250702B can be naturally reproduced. In this framework, the duration of each flare is determined by the viscous accretion timescale of material stripped near pericenter, with a typical mass $ΔM \approx 2\times10^{-2}\,M_\odot$. The minimum recurrence time corresponds to the orbital period, while the total activity period is set by the secular orbital evolution timescale leading to the complete disruption of the WD. Furthermore, if $M_{\rm BH}\gtrsim10^{5}\,M_\odot$ and the orbit has a minimum polar angle relative to the BH equatorial plane of $θ_{\rm min}\gtrsim0.12 {\rm rad}$, relativistic frame dragging induces $\gtrsim0.1$ rad precession of the orbital angular momentum between successive pericenter passages, comparable to a typical GRB jet half-opening angle, resulting in intermittent alignment with the observer and irregular flare spacing. The WD experiences $\approx40$ jet-launch episodes before complete disruption, but only four are expected to be observed on-axis. The remaining off-axis jets become visible at late times, enhancing the radio afterglow by about an order of magnitude, providing a testable prediction of this scenario.

The fundamental question of baryogenesis and the problem of matter-antimatter asymmetry motivate this study into the formation and evolution of antimatter objects in the early Universe. Hypothesize is the existence of isolated antimatter domains in a baryon-asymmetric Universe that survive until the era of first star formation ($Z \approx 20$). By assuming CPT-symmetry, the thermodynamics, mechanics, and energy dynamics of an antimatter gas cloud (composed of antihydrogen and antihelium) are treated symmetrically to their primordial matter counterparts. Analysis demonstrates the physical feasibility of the gravitational collapse process for a conservatively estimated antimatter domain ($\approx 5 \times 10^3 M_{\odot}$). The initial conditions easily satisfy the Jeans and Bonnor-Ebert mass criteria, indicating a high propensity for instability and runaway collapse. The subsequent dynamical evolution, driven by $\bar{H}_2$ cooling, is predicted to proceed identically to that of Population III star formation, leading to the formation of a dense, adiabatic anti-protostellar core. The theoretical viability of a true antistar hinges upon a critical assumption: the physical possibility of antinuclear fusion (e.g., the antiproton cycle) under extreme core conditions. Assuming this symmetry holds, the collapse is predicted to yield massive antistars ($\gtrsim 22 M_{\odot}$). This suggests that if antimatter domains formed in the early Universe, they likely underwent stellar formation. Observational constraints on the existence of these objects must rely on the detection of characteristic high-energy $γ$-ray or X-ray signals resulting from matter-antimatter annihilation at the domain boundaries or during mass accretion.

Context: The dynamical evolution of binary populations in embedded star clusters shapes the statistical properties of binaries observed in the Galactic field. Accurately modelling this process requires resolving both early cluster dynamics and binary interactions. Aims: We aim to characterize the early dynamical evolution of primordial binaries in embedded clusters and identify the key parameters that govern binary survival and disruption. Methods: We perform a set of direct $N$-body simulations starting from 100\% primordial binaries in a time-varying gas potential of a gas-embedded cluster. To describe the evolution of binary orbital properties, we define empirical dynamical operators for period, binding energy, and mass ratio, and calibrate them across the simulated ensemble. Results:The binding energy and orbital period evolve in a consistent, sigmoidal fashion. Their dynamical operators reveal that hard binaries heat the cluster and suppress wide binary formation, while a small residual population of soft binaries survives. The evolution of the mass-ratio distribution is less directly linked to dynamical processing and more shaped by internal processes such as stellar physics process in the pre-main-sequence phase. High-$q$ systems tend to be enhanced, while low-$q$ systems are prone to disruption. Conclusions: The binary evolution in clusters is primarily governed by binding energy and orbital period. Our model improves over previous parameterizations of the dynamical operator by allowing for the existence of wide binaries and incorporating the embedded cluster phase. For individual clusters, direct $N$-body modelling remains the only reliable approach. On Galactic scales, population synthesis methods based on the stellar dynamical operator approach developed here remain essential.