arXiv Daily Digest - 2026-07-09

Structure-property relationships are foundational to biology, chemistry and materials science, where function, reactivity and physical response emerge from spatial, chemical and periodic organization. Mechanistically explaining these relationships requires interpreting structural evidence through scientific principles and physical constraints, from stereochemistry and bonding to symmetry, energetics and periodic order. However, applying artificial intelligence to this process presents a joint challenge of representation and reasoning: models must preserve domain-native structural information while showing how specific evidence supports predictions under these constraints. Here we introduce SciReasoner, a multimodal scientific foundation model for native structural reasoning across proteins, small molecules and inorganic crystals. SciReasoner discretizes coordinates, topologies and periodic connectivities into a unified structure-aware vocabulary, treating structural tokens as addressable evidence units during reasoning. In homology-controlled Gene Ontology prediction, SciReasoner improves Cellular Component annotation for low-homology and orphan-like proteins, increasing $F_{\max}$ from 0.42 to 0.55. In chemistry, it raises single-step retrosynthesis accuracy from 0.63 to 0.72 while generating fragment-level disconnection and precursor-verification traces. In materials science, its representations separate elemental and compound phases and resolve high- and low-band-gap regimes. Across 86 benchmarks, SciReasoner achieves state-of-the-art performance on 67 tasks. Double-blind expert evaluation rates its reasoning traces as preferred or at least comparable to those of a frontier large language model in 98% of cases. By making structure an inspectable substrate for reasoning under scientific constraints, SciReasoner connects accurate prediction with interpretable scientific inference.

Limited memory language models (LMLMs) externalize factual knowledge during pretraining to a knowledge base (KB), rather than memorizing it in their weights. During generation, the model then fetches knowledge from the KB as needed. This recently introduced paradigm provides multiple advantages, including knowledge control capabilities that remain beyond conventional LLMs. We propose continuous-query LMLM (CO-LMLM), where the KB pairs continuous keys with textual knowledge values, a significant departure from prior reliance on relational KB and queries. CO-LMLM generates flexible vector queries at minimal cost, while still integrating human-readable and attributable retrieved knowledge into its generation. We pair this design with an annotation pipeline that tags free-form factual spans in arbitrary text, removing prior work's restriction to Wikipedia. Across pretraining on Wikipedia and FineWeb-Edu and at multiple model scales, CO-LMLM outperforms prior LMLMs and vanilla LLMs in both perplexity and factual precision. At 360M scale, this includes lower perplexity than models pretrained on 40x more data, and SimpleQA-verified performance that is in line with gpt-4o-mini and higher than Claude Sonnet 4.5.

The quadratic cost of causal self-attention severely bottlenecks long-context transformer inference. While numerous post hoc linearization pipelines exist, it is difficult to identify which components preserve model quality. This work isolates the effect of state update design in a strict frozen-backbone regime. We show that softmax relies on key-dependent, rank-1 orthogonal projections, elucidating why delta-style networks outperform purely gated accumulation. We identify a potential source of approximation errors and introduce structural interventions, specifically sink tokens, short convolutions, and fixed-budget cache routing, which reduces the remaining gap. We scale this linearization approach across LLaMA and Qwen models up to 32B parameters, outperforming prior post hoc baselines on MMLU and matching the long-context retrieval of complex adaptive-caching frameworks.

The optimization of long-horizon agents increasingly relies on reflection-based mechanisms, where a large language model (LLM) acts as an optimizer to diagnose agent failures and improve agent policies. However, real execution traces are difficult to use directly for optimization: large trace collections are often redundant and heterogeneous, making optimization inefficient and prone to overfitting to low-value failures; meanwhile, each individual trajectory also contains many irrelevant steps, while naive context reduction methods such as truncation or sliding windows can discard causally important evidence and produce misleading optimization signals. To resolve this dilemma, we introduce STRACE (Structural TRajectory Analysis and Causal Extraction), a framework that constructs high signal-noise optimization contexts for more precise and effective optimization. At the batch level, STRACE mines failure patterns to filter redundant traces and retain representative failures; within each selected trace, it performs causal localization over a textual dependency graph to remove non-causal steps and identify the true root-cause module for optimization. Empirical results demonstrate that STRACE significantly outperforms standard context-filtering baselines. Notably, on a challenging formal verification task (VeruSAGE-Bench), it successfully optimizes human-expert designed agents, delivering $1.4\times$ success-rate improvement (42.5% to 58.5%). The code is available at https://github.com/moomight/STRACE .

Analytical workloads operating on data stored in external database systems face a fundamental bottleneck: data access is guarded entirely by the database driver, like JDBC or ODBC, forcing all reads through query execution and other driver layers that are not designed for bulk columnar analytics. We present Jailbreak, an approach that bypasses the database engine entirely by reading storage files directly and materializing data as in-memory columnar buffers. Jailbreak's key insight is that database file formats, while complex, are fully specified by their source code and documentation, artifacts that Large Language Models (LLMs) can ingest to regenerate operator-specific table reading components without human-engineered parsing logic. Jailbreak leverages LLM-assisted code synthesis for database storage decoding, turning a traditionally opaque format into a directly queryable artifact. We evaluate Jailbreak on PostgreSQL and MySQL storage files, targeting analytical snapshot scenarios common in read replicas and offline processing pipelines. The generated reader produces Apache Arrow buffers consumable directly by most of the widely known query engines, including DuckDB, Apache Spark, and GPU-accelerated frameworks such as cuDF and Spark RAPIDS. We validate correctness against JDBC/ODBC-based baselines using the TPC-H benchmark across all query results, and demonstrate significant performance improvements in end-to-end analytical throughput, achieving up to 27x speedups. Our results showcase that LLM-assisted storage reader synthesis is a viable and generalizable methodology for breaking data lock-in across database systems, with applications beyond PostgreSQL and MySQL for any system whose file format is available to the LLM from documentation or source code.

We introduce institutional red-teaming, an evaluation methodology for testing deployment rules in multi-agent AI: hold the agents, objectives, and task state fixed, vary only one rule, and attribute the resulting change in collective behavior to that rule. We instantiate the methodology in IABench-CA, a consequence-allocation benchmark spanning 228 contexts, five canonical rules, and seven model populations (33,924 games), with a normative cooperative reference and auto-labelled reasoning traces. Three findings emerge. (1) Deployment rules causally alter collective safety: changing only the consequence rule moves mean fatality by 22 to 58 percentage points within every population. (2) There is no safe default, but the targeting hazard is universal: the safest rule, the least-safe rule, and even the direction of the incidence effect vary across populations, yet regressive identity-targeting is never decisively safest in any context for any population, eliminates the least-resourced agent in 30-87% of games everywhere, and is selection-unsafe relative to the cooperative reference for all seven populations. (3) Identity salience is the mechanism: a one-shot anonymization ablation on the most exploitation-prone population (gpt-5.1) shows that merely naming the loss bearer in the rule text drives targeted elimination from 22% to 81% at identical payoffs; under repeated play, anonymization only delays the targeting, as agents re-infer the hidden rule from observed eliminations. We package the methodology as a safety-case workflow that certifies a provisional rule region $Φ(c,P)$ per deployment context and population, with explicit residual risks and monitoring obligations.

Reinforcement learning from human feedback (RLHF) has emerged as a powerful paradigm for aligning generative models with human preferences. However, applying RLHF to diffusion models remains highly feedback inefficient, as existing approaches typically require large amounts of human or reward model evaluations. This limitation reduces the practicality of diffusion RLHF in realworld settings where feedback is the primary bottleneck. In this paper, we propose two complementary strategies that substantially improve the feedback efficiency of diffusion RLHF while preserving generalization to unseen prompts. Our key observation is that reward information in diffusion trajectories is unevenly distributed: not all denoising timesteps or trajectories contribute equally to learning from a reward signal. By emphasizing informative timesteps and trajectories during optimization, we obtain more effective gradient updates. First, we introduce a per-timestep weighting scheme that reweights denoising steps during policy optimization. We theoretically connect this weighting to the optimal convergence properties of proximal policy optimization (PPO) and approximate the resulting weighting trend empirically. Second, we introduce a replay mechanism that prioritizes informative trajectories, enabling the model to reuse past samples instead of repeatedly querying new rewards. Together, these strategies significantly improve the feedback efficiency of diffusion RLHF. Under identical hyperparameter settings, our approach achieves up to a 6$\times$ improvement in sample efficiency compared to widely used diffusion RLHF baselines.

Reinforcement learning from verifiable rewards (e.g. GRPO) is the engine behind today's reasoning models, yet it grades only the final answer. On hard problems this trains models to write more rather than to think better, since the trace itself is never graded and no label for good thinking exists. We introduce Agon, which makes two competing models each other's graders. Both attempt the same problem; in alternating roles, one drafts a solution and the other reads it while solving, and each is rewarded for out-solving the other. To win, a model must out-reason a rival that has seen its work, so reasoning is judged implicitly during training, with no process labels and no reward model. Because both models are optimized, each faces a progressively stronger rival, which single-model RL cannot provide. The two need only be comparably strong and behaviorally different. At inference the pair deploys as it trains, a two-stage cascade in which one model drafts and the other answers after reading the draft. On the hard split of DeepMath with Qwen3, this doubles GRPO's pass@1, roughly eight times the gain of an untrained Mixture-of-Agents pass over the same base. The ordering replicates on competitive-programming code and across model families (Qwen3.5, Gemma 4). For now the models talk in text; the next step is to let them reason together in latent space.

Long-horizon LLM multi-agent systems face reliability risks invisible to infrastructure monitoring. We propose the ADE Predictive Reliability Framework (ADE-PRF), enabling proactive health trajectory prediction from passive degradation detection. ADE-PRF aggregates 20 heterogeneous signals across five layers into a Trust Margin (TM) metric (39.2-point dynamic range). Triple-method parallel prediction enables 8-hour forecasts: the Exponential method achieves MAE=1.228, Direction Accuracy=76.8%, with 99.65% within +/-10-point tolerance. Production validation spans 380,227 predictions and 280,579 validations across six agent profiles over 15 continuous days, plus seven sandbox-controlled experiments. Key findings include detection of "false prosperity" -- degradation concealed by normal surface metrics -- and immediate TM coupling with ground-truth states upon ADE plugin integration, with 16/20 factors relying on ADE-collected data. Exponential consistently outperforms Kalman. ADE-PRF provides among the earliest reliability quantification with forward-looking warnings for production LLM agents.

We present recent performance-oriented developments in WaterLily.jl, a scale-resolving incompressible flow solver written in pure Julia that runs seamlessly on CPUs and GPUs of any vendor. Supported by the newly added MPI-based parallelism, strong-scalability tests display a near-ideal linear trend, and weak-scaling efficiency is kept above 85\% before node memory-concurrency contention dominates parallel performance. Inter-node weak scalability is sustained above 96\% with grid size up to 1 billion cells. We further benchmark improvements to the geometric multigrid Poisson solver enabled by an adaptive under-relaxed red-black Gauss--Seidel smoother together with anisotropic coarsening operators.

Electrocardiography (ECG) is one of the most widely used tests for diagnosing cardiovascular disease. Yet several remote clinics still utilize paper ECG printouts for their analysis due to limited connectivity and computational capacity. As a result, vast numbers of physical ECGs obtained in remote areas still remain incapable of being accessed by contemporary artificial-intelligence (AI)-based decision support as they require high computational resources or strong high-speed internet connectivity. This causes several cases where conditions like acute coronary occlusion (ACS) is overlooked and reperfusion therapy delayed. Although prior work has tackled digitization and diagnosis separately, and utilized advanced AI models for them, there still remains a lack of a compute-light, on-device framework that reconstructs paper ECGs at high fidelity, while accurately supporting multiple clinically relevant endpoints. We address this need with an end-to-end lightweight on-device digitization-to-diagnosis pipeline that converts a smartphone photo or scan of a paper ECG into a calibrated 12-lead signal and screens for Myocardial Infarction (MI) pathologies, with SHapley Additive exPlanations (SHAP) to support interpretability. Trained and evaluated on 21,799 ECGs from the PTB-XL dataset and further validated on hospital-acquired ECG-Matrix dataset, the complete system runs in <30 s per ECG on CPU-only resources, achieving 95.51% accuracy (F1 = 0.9519) for MI detection on PTB-XL and 88.89% accuracy (F1 = 0.8862) for OMI detection on ECG-Matrix. This work showcases that legacy paper records can be reliably democratized in any part of the world, providing a scalable decision support when digital ECG export, connectivity, or high-end compute are unavailable

The inverse design of physical systems governed by partial differential equations is computationally demanding due to the high dimensionality and non-convexity of design spaces. Generative models for inverse design often lack robustness and transferability, whereas evolutionary strategies are robust but struggle in high-dimensional spaces. This paper introduces a Neural Operator-enabled Topology-informed Evolutionary Strategy (NOTES) that integrates dimensionality reduction, representation learning, and evolutionary optimization for efficient and transferable inverse design. NOTES couples a DeepONet-based neural operator with the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to perform global optimization in a compact latent space that encodes topology-aware priors while discovering high-performance designs for unseen operating conditions. Applied to nanophotonic beam-deflector inverse design governed by Maxwell's equations, NOTES reduces the design dimensionality from 256 to 25 and consistently achieves over 95 percent efficiency, outperforming CMA-ES, topology optimization, and other baselines. Applied to structural optimization, NOTES discovers designs that achieve compliance down to 246. By decoupling topology learning of a DeepONet from the governing physics in a PDE solver, NOTES provides a flexible and transferable framework for the inverse design of physical systems.

Many machine learning models are defined for inputs of different sizes, such as point clouds containing different numbers of points, sequences of tokens of different lengths, and graphs on different numbers of nodes. Such models are trained on finitely-many examples of necessarily limited sizes. How well do these models generalize from inputs of small size to larger inputs of size not seen during training? Furthermore, evaluating such models on large inputs is often expensive. How can we sketch large inputs to obtain smaller ones on which the model takes similar values? At the heart of both questions is the need to compare inputs of different sizes and to approximate large inputs by small ones. We present a unified approach to address these questions by using random sampling maps to compare inputs of different sizes. The sampling maps we consider are generalizations of sampling with replacement, random binning, and species sampling. We characterize the application domains in which each type of sampling is appropriate in terms of the symmetries and relations between problem instances of different sizes in the domain. Our framework yields explicit generalization and sketching rates for function classes continuous with respect to a chosen notion of sampling, encompassing large families of functions defined on sequences, graphs, and tensors of different sizes. Specific examples include moment polynomials on measures, homomorphism densities and numbers of graphs, permutation-invariant transformers, and graph neural networks.

Rotary Position Embeddings (RoPE) provide transformers with a fixed grid of positional frequencies, yet trained models use these frequencies highly non-uniformly. We study what determines this frequency usage and propose a data-centered explanation: RoPE frequencies are selected to match the relative-distance structure of the training data. Viewing each frequency as a positional lens, we formalize a field-resolution tradeoff and show that, for a data-induced dependency profile of width $W$, the optimal frequency scales as $1/W$. This frequency-matching principle explains controlled observations on synthetic and text-based data, and suggests that the mid-low frequency bands observed in language models arise from the multi-scale dependency structure of natural language. We further connect frequency selection to position-interpolation-based length generalization: scaling frequencies down expands the effective field while reducing resolution. This helps when longer-context dependencies are approximate dilations of those seen during training, but can fail when relevant dependencies do not scale with context length. Empirically, we show that natural language exhibits approximate self-similarity across positional scales, explaining why test-time frequency scaling can support long-context generalization. Overall, our results identify a data-driven mechanism behind emergent RoPE frequency usage and show that long-context generalization depends on two forms of scale matching: between learned frequencies and training-time dependencies, and between frequency scaling and how those dependencies extend to longer contexts.

Autonomous AI agents can execute complex tasks with limited human review, yet they often lack the grounded operational knowledge to make their outputs not just executable but correct, secure, and maintainable. We introduce SkillCenter, to our knowledge the largest open skill library for agents by total count: 216,938 structured skills across 24 domain bundles. A SkillGate-filtered pipeline contributes 114,565 source-grounded skills from peer-reviewed journals, ArXiv, and over 24,000 technical sources, integrated with 102,373 community skills from GitHub and the ClawHub marketplace. We present the end-to-end framework that builds the pipeline subset: multi-source acquisition, an LLM-based quality gate (SkillGate), template-driven generation, iterative source-grounding, and quality-controlled publishing. Source grounding is a traceability guarantee: each retained claim maps to an exact quotation in its source. All skills ship as offline-searchable SQLite FTS5 bundles.

Group Relative Policy Optimization (GRPO) stalls on a model's hardest problems: when no rollout in a group succeeds, the group-relative advantages vanish and the problem contributes no gradient, wasting the frontier examples we most want to learn from. Prepending a correct prefix of a reference solution raises the success rate, making prefix length a continuous knob on difficulty. Concurrent methods set the knob once; AdaPrefix-GRPO turns it into a feedback controller: throughout training it adjusts how much of the solution each problem gets, holding its success rate near 50%, where GRPO's gradient signal is largest, then withdraws the assistance entirely, so the deployed model solves problems unaided. On hard math, at matched training FLOPs, it more than doubles GRPO's accuracy on held-out problems from the training distribution for a 0.6B model (2.1x), with 1.6x on Qwen3-1.7B and 1.7x on AIME, while roughly halving trace length. The method is implemented in data preparation plus a loss mask on prefix tokens; the trainer is otherwise stock. The smaller the model, the larger the gain.

Medicine is inherently multimodal, requiring clinicians to synthesize information across diverse data streams. Yet the development of multimodal foundation models is constrained by limited access to large-scale, high-quality clinical data. Although PubMed Central (PMC) offers a complementary source of expert-authored image-text data, existing PMC-derived resources remain limited in fidelity, reproducibility, and clinical validation. We introduce MedPMC, an automated, continuously updatable framework that transforms permissively licensed literature into high-fidelity infrastructure for medical multimodal models. Applied to 6.1 million PMC articles, MedPMC curated 11 million medical image-text pairs. Component evaluations showed strong performance for initial screening (F1 = 93.2), multi-panel figure detection (F1 = 96.5), figure separation (mAP = 89.8), caption separation and alignment (F1 = 81.4; ROUGE-L = 85.3), and medical figure classification (F1 = 96.5). Manual review by five annotators, three with medical training, found 95.3% of MedPMC images medically relevant, versus 19.7% in a prior PMC-derived dataset. Across 26 benchmarks spanning 11 specialties, a MedPMC-trained CLIP-style model improved average zero-shot AUC by 7.1 percentage points over the strongest architecture-matched biomedical CLIP baseline despite using fewer than half as many image-text pairs. As the vision encoder in a multimodal large language model, it improved medical visual question-answering by 1.9 and 16.9 percentage points across two benchmarks. In 10,524 Yale New Haven Health System dermatology photographs, it improved morphology-to-image retrieval Recall@5 by 11.7 percentage points. These findings show that high-fidelity literature curation strengthens medical multimodal foundation models across benchmark and clinical settings. We publicly release the framework, corpus, benchmarks, and pretrained models.

Probabilistic circuits (PCs) can model complex joint distributions while supporting exact and efficient computation of many inference queries. However, standard likelihood-based PC learning is vulnerable to overfitting and fragile generalization when confronted with data noise, small sample sizes, or distribution shifts. This can be mitigated using distributionally-robust optimization which consider worst-case distributions within a Wasserstein ball of the empirical distribution, but current methods are limited to training a model from scratch in this framework. Instead, we propose PeTeR: a novel, data-free post-training framework designed to robustify pre-trained PCs against distribution shifts without retraining from scratch. Empirical evaluations across multiple density estimation benchmarks demonstrate that PeTeR effectively robustifies baseline models against both random and adversarial perturbations, achieving competitive or superior performance to data-dependent robust learning baselines.

Large language models hallucinate most about entities they have never seen. We ask whether a model's activations betray entity familiarity before a single answer token is generated, and whether that signal predicts the factual reliability of the answers. On four Polish Bielik models (1.5B-11B parameters), we probe four entity domains (athletes, cities, writers, musicians), each with 42 well-known, 42 obscure-but-real, and 42 fabricated entities addressed by a one-sentence question (504 prompts per model). Two unsupervised, single-forward-pass dispersion measures over post-SwiGLU MLP activations, inverse participation ratio and spectral entropy, separate known from fabricated entities at AUROC 0.95-1.00 across all domains and scales; a supervised linear probe reaches 0.99-1.00. Both clear selection-aware permutation floors of about 0.70-0.74 (empirical p<=1e-3), survive held-out layer selection (0.93-0.99), and persist on real names (known vs. obscure-but-real: 0.96-1.00). The signal transfers across entity types (mean off-diagonal AUROC 0.92-0.99); a matched-template counterfactual shows the only large drops are template-caused, not entity-type effects, and the signal is diffuse across heads. This representational signal is already at ceiling at 1.5B, whereas behavioral factual reliability scales sharply: 0, 2, 10, and 19 of 42 known athletes are answered fully correctly by the 1.5B, 4.5B, 7B, and 11B models under a strict judge. Within known entities, separating correct from hallucinated answers is much harder (probe 0.93; dispersion no better than a first-token-entropy baseline). A five-sample semantic-entropy baseline reaches only 0.71-0.83 at 5x the inference cost. Despite this internal awareness, the models almost never abstain: an audit of 2,520 answers finds 2 refusals and 1 hedge. Entity familiarity and factual reliability are distinct phenomena on different scaling curves.

Large language models increasingly \emph{understand} dialectal English, yet still \emph{produce} only standard, US-leaning English, leaving dialectal generation, the harder half of the problem, largely unaddressed. We introduce \textbf{DiaLLM}, which continually pretrains three open-weight language model families on the International Corpus of English and applies implicit and explicit post-training paradigms, each combined with three model alignment strategies, giving the first controlled comparison of these components across Australian, Indian, and Northern British English. Our results reveal that dialectal robustness and generation are \emph{dissociated}: benchmarks are shaped by continual pretraining and SFT, while alignment visibly reshapes generation in ways benchmarks do not capture. Explicit variety-targeted adaptation produces output reliably recognised as dialectal and preferred over broad alignment, yet the method that most aggressively optimises the dialectal reward is not preferred by human evaluators. Independent linguistic analysis corroborates this reward-quality gap, most clearly on two of the three families. No single alignment method dominates, and closing the gap will require richer reward designs and continued investment in dialectal resources. We release all code, checkpoints, and preference datasets.

Large language models (LLMs) demonstrate remarkable reasoning capabilities, yet their stateless architecture fundamentally limits deployment in long-horizon research workflows requiring multi-session continuity and quantitative rigor. Here we present Ensemble QSP, a multi-agent framework featuring a three-layer hierarchical memory architecture that keeps injected context bounded and constant in project duration (mid-term project state: median 301 tokens, max 4,050, across 104 runs) by capping each state category and evicting completed work, enabling continuous autonomous operation without context degradation. The system orchestrates five specialist worker agents under domain-expert principal investigators, enforcing physical constraints through physics-based checklists and structured-domain knowledge. Comprehensive benchmarking demonstrates robust autonomous pharmacokinetic-pharmacodynamic model selection without human intervention, consistent result quality across both lower-cost and frontier LLMs, improved PK parameter recovery relative to single-agent baselines, and stable model selection across linguistically diverse prompts of the same task. Feature-level ablation across physiologically based pharmacokinetic (PBPK) models spanning a broad complexity range shows that PI-agent oversight improves debugging efficiency while preserving final accuracy across conditions. The architecture is structurally domain-agnostic, adding a new scientific domain requires only a new PI agent configuration.

Classifier-free guidance (CFG) is the standard way to strengthen class-conditioning in diffusion and flow-matching samplers, yet at large guidance it oversaturates and destabilizes, symptoms practitioners suppress with more steps or limited-interval schedules. We analyze CFG through an asymptotic-preserving, numerical-analysis lens. Building on a recent result that the deterministic DDIM step is the unique fitted operator for the unguided terminal layer, exact on the final small-sigma stretch of sampling, we show that guidance re-stiffens exactly the discriminative subspace to an anomalous exponent 1+w. DDIM is therefore no longer fitted there, and on coarse meshes its guided residual diverges as sigma_min goes to zero. We prove a guided clock barrier with three ordered step-size thresholds, and read one-step oversaturation as its endpoint: a solver artifact on the calibration model rather than the continuous guided law. The same analysis yields a one-coefficient, zero-extra-NFE repair: replace CFG's w(r-1) by r^(1+w)-r on the guidance direction. On the calibration model's discriminative crossover, this removes CFG's sigma_min-divergent blow-up and is first-order accurate against the exact guided flow as sigma_min goes to zero. On learned CIFAR-10 checkpoints, and as a cross-domain smoke test on Stable Diffusion 1.5 DDIM, it acts as a high-guidance stabilizer at no extra cost rather than a universal quality knob: it cuts residual amplification and saturation, gives 9/9 point-FID wins over CFG on the tested grid, and preserves classifier-proxy target accuracy in the hard-cell blocks. We report the limits alongside: it is not a universal image-quality win, and against a dense vanilla-CFG reference it is not a uniformly better integrator of that field.

AI systems increasingly participate in their own improvement: revising their outputs, adapting their own harnesses during deployment, training on data they generate, and, increasingly, conducting AI research itself. This literature is described under a vocabulary ("self-refine," "self-reward," "self-play," "self-evolve") that conflates fundamentally different ambitions. We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and already industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment. We survey the evaluator design space -- judges, process reward models, verifiers, rubrics, meta-evaluation -- order the signals into a verification hierarchy from formal verifiers (strongest) to intrinsic self-assessment (weakest), and observe that demonstrated self-improvement strength tracks this hierarchy, that its failure modes (self-confirming loops, model collapse, diversity collapse) follow from its violations, and that the "research direction-setting" bottleneck keeping humans in the loop sits at the top of that hierarchy. We connect the technical literature to the theory of RSI limits and to the safety and governance questions raised by frontier-lab accounts of closing the loop, and identify governance-grade measurement of self-improvement as the field's most underpopulated niche.

Does RL post-training merely amplify primitive skills already latent in a base model, or can it compose primitive skills into new higher-level strategies? We study this question in a fully observable rewrite-grammar environment where the pretraining distribution is known and every generated rewrite can be audited. A Transformer is pretrained on primitive symbol-rewrite chains and post-trained on a Trace-based reasoning task with only a binary final-answer reward. RL solves held-out problems that remain rarely solved by the pretrained model even under much larger sampling budgets, while rejection fine-tuning improves early but plateaus. Trace analysis shows that RL reorganizes primitive competence through a phased compositional mechanism: it first strengthens primitive reductions, then discovers valid composed procedures. These include sequential compositions, which collapse ordered chains of primitive contractions, and parallel compositions, which combine independent primitive contractions in a single step. The composed procedures are not isolated samples; they are reused and consolidated into a stable repertoire. Comparing RL with rejection fine-tuning shows that the key difference is not exploration volume but selectivity: RFT produces many shortcut-like rewrites, much of them invalid, whereas RL concentrates exploration into valid reusable structure. Pretraining ablations show that the emergence of compositional strategies is gated not by primitive exposure alone, but by whether pretraining organizes primitive competence into reduction procedures that RL can later compress. The base model provides weak procedural ingredients; RL builds them into reliable higher-level strategies.

FPGA architectures increasingly incorporate domain-specific in-fabric hardblocks to accelerate DL inference, particularly GEMM, which dominates DL computation. To realize the performance gains of these hardblocks, manual RTL design is required: the programmer must understand the hardblock microarchitecture, instantiate them in RTL, and manage tiling and control logic. While programming in C/C++ and using HLS tools has increased the abstraction level and productivity of FPGA engineers, HLS tools do not support code generation for custom hardblocks natively. Prior work has demonstrated that blackbox mechanisms in HLS tools can be used to target custom hardblocks, but this still requires explicit function calls in user-written HLS C and manual creation of RTL IP libraries, significant effort that must be repeated for every layer in a DL model. Furthermore, for DL, an even high-level programming interface, e.g., Pytorch/Keras instead of C/C++, is desirable for improved programmability and user adoption. We present ATLAS, a fully automated flow from a high-level DL model description to a hardware implementation on an FPGA with custom in-fabric DL-optimized hardblocks, requiring no manual RTL design or explicit hardblock instantiation from the end user. Our approach uses GEMM as a universal abstraction layer and comprises two components: (1) hls4ml-GEMM, a compiler frontend that transforms DL layers into HLS C code with architecture-agnostic GEMM function calls, and (2) a GEMM IP Generator, an architecture-aware backend that produces hardblock-based RTL wrappers with tiling logic, control FSMs, and scheduling metadata. We evaluate the flow across 11 DL designs, including individual fully connected, convolution, and attention layers, as well as full CNN, MLP, and Transformer models targeting an FPGA architecture with Tensor Slices using Catapult for HLS and VTR for implementation.

Deep learning has significantly advanced time series imputation, yet most existing architectures primarily rely on localized temporal context within the corrupted input sequence. This reliance can be limiting in real-world scenarios, where time series often exhibit non-stationary dynamics, weak temporal correlations, and infrequent patterns that are difficult to reconstruct from nearby observations alone. In this paper, we propose ALER-TI, Aligned Latent Embedding Retrieval for Time Series Imputation, a retrieval-augmented framework that explicitly leverages historical patterns to supplement degraded local context for more reliable missing-value reconstruction. The core of ALER-TI is Latent Embedding Alignment (LEA), which mitigates the representation mismatch between corrupted queries and complete historical candidates. By applying post-hoc masking in the latent space, LEA aligns candidates with the query's missingness pattern while allowing historical embeddings to be pre-computed and cached for efficient retrieval. ALER-TI is model-agnostic and can be integrated with various imputation backbones through a lightweight adaptation module. Extensive experiments on six real-world datasets under different missing rates demonstrate that ALER-TI consistently improves strong baseline models and enhances robustness across diverse imputation settings.

This work presents a novel approach for adapting neural network architecture along the depth based on a posteriori error estimation. By formulating neural network training as a continuous-time optimal control problem, we derive rigorous error estimates that quantify how approximation error distributes across network layers. This error decomposition enables a principled depth adaptation strategy: new layers are inserted at locations of maximum estimated error, allowing the network to efficiently capture complex, nonlinear variations in the underlying problem. Our framework introduces a novel network architecture that treats weights and biases as piecewise linear functions varying across layers, with the error estimator bounding the discrepancy between this discrete representation and the true continuous optimal control solution. The approach leverages dual weighted residual methodology from finite element analysis to derive computable upper bounds on the functional error. A key theoretical contribution is the derivation of explicit error bounds that decompose the total approximation error into interval-wise contributions, providing a rigorous basis for targeted architecture refinement. We demonstrate the effectiveness of our method on scientific datasets, including learning the observable-to-parameter map for the Navier-Stokes equation. Numerical results reveal that our approach consistently outperforms existing architecture adaptation methods in terms of generalization performance.

Time series analysis plays a vital role across a wide range of scientific and engineering domains but poses substantial computational challenges. A major difficulty arises from the time reparameterization invariance of time series data, which complicates the extraction of meaningful temporal features. In this work, we address the problem of time series classification by exploring the application of quantum computation techniques. We propose a hybrid quantum-classical architecture that integrates recent advances in quantum neural networks with the mathematical framework of path signatures, mitigating the impact of time reparametrization invariance. The architecture employs feature layers that compute a signature kernel between pairs of input paths, consisting of a reference path and a target path for classification, using either classical or quantum variational linear solvers (VQLS). These feature layers are followed by a Quantum Convolutional Neural Network (QCNN) to perform downstream learning tasks. We evaluate several realizations of the proposed architecture, differing in QCNN configurations, on a binary classification task involving time series representations of handwritten digits. Our experiments demonstrate the potential advantages of implementing path signature kernel layers within quantum circuits and provide an analysis of the computational limitations associated with the VQLS component.

Reliable confidence estimation is essential for deploying large language models (LLMs) in confidence-aware systems, where downstream decisions such as retrieval, tool use, and adaptive computation depend on accurately estimating answer reliability. Existing approaches, however, largely treat confidence as a property of completed responses, overlooking how confidence-related information evolves throughout the answering process. In this work, we investigate confidence from a temporal perspective by comparing pre-solution Feeling-of-Knowing (FOK) and post-solution Judgement-of-Learning (JOL) confidence estimates across frontier and open-source LLMs. We show that post-solution confidence is consistently better calibrated and more discriminative than pre-solution confidence, while linear probes trained on hidden representations recover substantially richer confidence-related information than models explicitly verbalise. Building on this observation, we introduce future confidence distillation, which trains predictors operating on pre-solution hidden representations using teacher confidence estimates produced by post-solution correctness probes. Despite requiring only pre-solution representations for inference, distilled predictors recover much of the calibration improvement achieved by post-solution confidence, remain highly sample efficient, and transfer across datasets within the same domain. Together, our findings demonstrate that confidence-related information evolves throughout the answering process and can be anticipated before answer generation is complete, enabling significantly more reliable yet low-cost confidence estimation.

Blockchain systems are undergoing a fundamental transition from decentralized ledgers for digital assets to general-purpose trust infrastructures for verifiable computation, decentralized physical resources, and automated infrastructure management. Meanwhile, the limitations of the Blockchain as a Service (BaaS) model stem from a common structural problem: outsourcing control of infrastructure to third-party service providers inevitably involves a systemic surrender of trust, flexibility, and data sovereignty. RISC-V, with its open, modular, and extensible design, provides a general-purpose computing foundation for public blockchains that is open, low-level, compileable, verifiable, and scalable. Inspired by the development and characteristics of eSIM, the embedded Blockchain infrastructure management (eBIM) is defined as a software-hardware collaborative paradigm for blockchain infrastructure management with RISC-V. This study aims to provide a comprehensive survey on eBIM supporting research and technologies, to answer the following research questions (RQs): RQ1 What is eBIM? RQ2 How does eBIM work? RQ3 What can eBIM do? By introducing the concept of eBIM, this paper establishes a foundational reference for researchers, hardware architects, and protocol designers in this rapidly evolving landscape, including cryptographic acceleration, trusted execution environments, zero-knowledge virtual machines, and smart contract execution engines. The prospects of the proposed e-BIM and its future research directions are indicated in this paper.

Nonlinear least-squares optimization is central to regression, physics-informed neural networks, and other machine-learning tasks. Such problems have a natural geometric interpretation, model predictions form a manifold in data space, while the chosen parameterization can introduce parameter-effects curvature that becomes a dominant source of nonlinearity. This exposes a limitation of the Levenberg-Marquardt (LM) method, its tangent-space step is applied as a straight update in parameter coordinates. Geodesic acceleration gives a second-order correction, but its removal of parameter-effect curvature is exact only in the infinitesimal-step limit. We propose a Riemann-normal-coordinate Levenberg-Marquardt method (RNC-LM) to improve this consistency for finite optimization steps. By reformulating the geodesic equation, RNC-LM extends geodesic acceleration to arbitrary-order corrections and constructs finite-step updates with progressively higher reparameterization consistency. A line search along the resulting RNC curve controls the traveled distance while keeping the cost close to standard LM. The method eliminates the tangential component of residual acceleration order by order in a moving tangent frame, making the actual objective reduction more consistent with the linear model prediction of LM. On classical nonlinear least-squares benchmarks, RNC-LM improves convergence and robustness in curved valleys and rank-deficient problems. On a reaction-diffusion PINN failure-mode benchmark, it reduces the relative L2 error to the order of 1e-3 and recovers a physically meaningful solution. On a large-scale machine-learning potential-energy-surface fitting task, it achieves a 34-fold speedup over standard LM.

Many function-level performance benchmarks have been proposed to evaluate whether large language models (LLMs) can generate efficient programs. However, results on these benchmarks often show that LLM-generated implementations have little or no execution-time difference from canonical solutions. In this paper, we revisit four popular benchmarks: EffiBench, Enamel, EvalPerf, and Mercury. We evaluate 1,538 tasks under more rigorous setting by running each task 30 times and assessing the runtime differences between the canonical solutions and benchmark-provided performant implementations with statistical testing. With the benchmark-provided test suites, only 6.11% of the performant implementations are significantly faster than the canonical solutions. In a manual analysis of 308 non-significant tasks, 99 performant implementations contain no meaningful performance change, while 209 contain potential performance improvements that are not exposed by the original tests. These results suggest that the main limitation is not only the evaluation method, but also the limited sufficiency of the benchmark-provided performance tests. To address this limitation, we propose an LLM-based multi-agent framework to generate performance-oriented tests that expose runtime differences more effectively than the original tests. The framework uses three separate agents to generate, diagnose, and repair deterministic tests that preserve functional correctness while better exposing performance differences. Across 1,345 benchmark tasks for which the original tests found no significant performance difference, tests generated by our framework with DeepSeek-v3.1 and GPT-4o reveal statistically significant improvements in 24.01% and 25.43% of the tasks, respectively, outperforming the SOTA LLM-based performance test generation method.

Artificial intelligence is rapidly evolving from generative systems to agentic AI capable of autonomously planning and executing tasks. Widely characterized as the Year of Agentic AI, 2025 marked accelerated development and deployment, introducing new ethical and governance challenges. This paper presents a systematic review of the emerging literature on agentic AI governance. Our analysis identifies features that distinguish agentic AI from traditional systems and why it warrants targeted governance attention. We synthesize prevailing governance priorities, proposed mechanisms, and stakeholder roles shaping this evolving domain. As an initial scholarly effort, this review lays the preliminary groundwork for developing a structured roadmap to guide responsible and adaptive agentic AI governance.

Background: Graph neural networks improve computational prediction of polypharmacy side effects, but standard binary cross-entropy training allocates equal capacity to well-classified and difficult examples, potentially missing clinically significant interactions. We evaluated whether an asymmetric focal objective could improve multi-relational drug-drug interaction (DDI) prediction by emphasizing difficult positive interactions. Methods: ClinicalFocal loss was integrated into a relation-aware graph convolutional network using molecular fingerprints, physicochemical descriptors, and learned embeddings. The model was evaluated on TWOSIDES using five-fold cross-validation with identical experimental conditions (architecture, features, data partitions, hyperparameters, and random seeds) for ClinicalFocal loss and binary cross-entropy baseline. Results: ClinicalFocal loss increased accuracy from 0.699 to 0.892 (+19.3 percentage points) and F1 score from 0.700 to 0.894 (+19.4 percentage points). AUROC increased from 0.766 to 0.914, and AUCPR increased from 0.714 to 0.860. The false-negative rate decreased from 29.8% to 9.1%, while specificity increased from 69.6% to 87.5%. Overall classification error decreased from 30.1% to 10.8%, corresponding to a 64.1% relative reduction. Improvements were consistent across all five folds. Conclusions: Asymmetric focal optimization improved classification and ranking performance while achieving 90.9% recall for observed interaction triples, without modifying the underlying architecture. Loss-function design is a direct, tunable lever for improving graph-based DDI prediction.

Safety evaluation for autonomous driving is dominated by rare, safety-critical interactions, motivating simulators that can deliberately synthesize corner cases with photorealistic observations. Corner-case generation is inherently a multi-source problem spanning visual representation, scene reasoning, and vehicle trajectory generation and control. Prior knowledge- and model-based approaches typically focus on scene or trajectory components in isolation, while diffusion-based methods attempt end-to-end generation but still struggle to ensure spatiotemporal consistency and physical realism. To unify these aspects within a single framework, we propose CARLA-GS, a modular corner-case synthesis pipeline that decouples visual representation, semantic reasoning, and physics-based execution while maintaining tight cross-module coupling. Starting from real driving data, we reconstruct an editable gaussian scene with additional geometry-consistent constraints. A multi-agent LLM then performs scene-level reasoning to identify risky interactions and generate intent-level waypoint trajectories, while the low-level motion control is delegated to CARLA, where a PID controller ensures kinematic and dynamic feasibility. The simulated vehicle states are finally re-projected into the gaussian scene for ego-centric rendering. This design enables high-level semantic reasoning, low-level physically executable motion, and photorealistic corner-case generation within a unified pipeline. Experiments on the Waymo Open Dataset show, both quantitatively and qualitatively, that our framework enables controllable corner-case generation and produces photorealistic, spatiotemporally consistent videos aligned with semantic intent and physically feasible motion.

The emergence of Quantum Software Engineering (QSE) responds to the need for systematic, disciplined, and quantifiable approaches to the development, operation, and maintenance of quantum software. Within this context, quantum computer certification represents a significant challenge: verifying that quantum devices produce valid entangled states despite hardware imperfections, noise, and decoherence. This paper presents QAccCert, a hybrid certification framework developed following QSE principles, demonstrating how heterogeneous technologies like FPGAs and Artificial Intelligence can be integrated for quantum processing. The framework implements entanglement certification through CHSH inequality violation in ideal quantum simulations using Qiskit AerSimulator. Through LLM-guided optimization, the system achieves 99.94% of the theoretical maximum of $2\sqrt{2}$, evidencing more efficient parameter space exploration than random search. These simulated results illustrate how QSE methodologies, combined with strategic technology interconnection, can be applied for practical and scalable quantum certification on real NISQ hardware in future work. This study provides a concrete case study of systematic quantum software development.

Automated program repair (APR) agents are transitioning from research benchmarks to developer workflows, yet they still begin with bug reports written for human developers. While decades of research have established what makes a good bug report for humans (e.g., steps to reproduce, stack traces), it remains unclear whether these features transfer to LLM-based agents. We study this question in two analyses. First, we use statistical modeling to examine associations between 27 bug-report features and repair success across 433 SWE-bench Verified issues attempted by 87 repair agents. We find that fix suggestions, reproduction scripts, repository source code, and localization info are associated with higher resolution likelihood, while longer reports are associated with lower odds. Second, we conduct controlled ablations across 2 models and 17 problem-statement mutations on SWE-bench Pro, varying the information available to an agent while holding the underlying task fixed. We remove or isolate selected bug-report content, delete fault-localization cues, and test structural changes that flatten lists or remove section headers. We find that both models depend on localization cues and expected behavior, and that structural changes alone can reduce solve rates, even without removing any content. The two models diverge in how they handle missing information: Qwen searches more widely and can exhaust its turn budget, while Gemma commits to a plausible interpretation early and patches on it. Our findings indicate that a good bug report for an agent overlaps with, but is not identical to, a good report for a human: agents benefit most from concrete, executable, and well-localized information, whereas some qualities long emphasized for human readers, such as natural language steps to reproduce and readable descriptions, contribute little or even correlate with lower success.

Assigning Common Weakness Enumeration (CWE) categories to Common Vulnerabilities and Exposures (CVE) records remains an important but largely manual step in vulnerability analysis. We study this task as a text classification problem and compare two modelling choices: a \emph{multi-class} formulation that predicts a single CWE per CVE and a \emph{multi-label} formulation that allows multiple assignments. Three transformer encoders (BERT Base, SecureBERT, and CySecBERT) are evaluated on three nested label spaces (83, 47, and 25 classes). Multi-class training achieves higher macro-F1 across all settings, although the gap to multi-label narrows from 21 to 2 percentage points as the label space shrinks. Post-hoc threshold optimisation on the multi-label side closes this gap on the 25-class setting. Confusion analysis shows that the dominant misclassification patterns follow the CWE hierarchy and are shared across all three encoders (Pearson $r > 0.92$), which suggests that the error structure is driven more by taxonomy design than by encoder choice. A hierarchy-relaxed evaluation that forgives within-family confusions raises macro-F1 from ${\sim}$81\% to ${\sim}$90\%, indicating that strict metrics understate branch-level classifier quality. CySecBERT achieves the strongest results overall, with statistically significant gains concentrated in the multi-label setting.

One-shot federated learning (OSFL) addresses the communication overhead of federated learning by limiting training to a single round, but doing so without sacrificing model quality is non-trivial, particularly when client data distributions diverge. Recent work has addressed this challenge by aggregating client knowledge on the server through the construction of transferable synthetic datasets or distillates. However, most of these methods lack formal privacy guarantees, leaving a gap in jointly achieving low communication, robustness to heterogeneity, and rigorous privacy. We propose FedKT-CSD (Federated Knowledge Transfer via Collaborative Synthetic Data), a framework inspired by neural image compression that closes this gap by leveraging publicly pretrained autoencoders as a shared latent space. Each client encodes its private data in a single forward pass, computes class-conditional latent statistics, and transmits these to the server. The server aggregates these statistics via secure aggregation, adds calibrated differential privacy noise, and decodes a synthetic dataset for training a global model and further downstream tasks. This design provides formal $(\varepsilon,δ)$-differential privacy by construction, while keeping client-side computation and communication lightweight. Despite operating under privacy constraints, FedKT-CSD is competitive with and even outperforms non-private baselines across diverse datasets and heterogeneity settings, and scales to a large number of clients. Our code is available at: https://github.com/an7123/FedKT-CSD

One-shot pruning methods like Wanda and SparseGPT apply the same sparsity ratio to every layer of a transformer, ignoring known variation in layer importance. We propose PALS (Percentile-Aware Layerwise Sparsity), which adjusts per-layer sparsity based on the 99th percentile of activation magnitudes, bounded to $\pm 5\%$ around the target ratio. On LLaMA-2-7B at 50\% sparsity, PALS achieves 10.96 WikiText-2 perplexity versus 12.92 for uniform Wanda (mean over 9 runs, $p < 0.001$). The benefit is architecture-dependent: LLaMA-3-8B shows marginal gains and Mistral-7B shows none. We also find that gradient-based allocation -- the seemingly more principled approach -- produces results worse than random, suggesting that gradient magnitude does not predict the impact of discrete weight removal. PALS adds negligible cost to the pruning pipeline and requires no fine-tuning.

Large language model based search agents increasingly adopt multi-agent architectures in which a main agent decomposes a complex question into sub-queries and dispatches them to parallel sub-agents. However, existing systems instantiate all roles from a single model of identical scale, leaving open how model capacity should be distributed across roles. We factorize hierarchical search into three roles: a delegation role responsible for task decomposition, an execution role responsible for retrieval and evidence extraction, and an answer generation role held fixed as a confound control. We then conduct controlled capacity sweeps along the delegation and execution axes on five multi-hop QA benchmarks. The experiments yield three findings. First, role factorization consistently outperforms a single-agent baseline, improving exact match from 4.5 to 8.6 points across six model scales. Second, capacity sensitivity is asymmetric: scaling the delegation backbone improves EM by ~11 points, whereas scaling the execution sub-agent moves EM by only ~2.6 points, identifying decomposition as the capability bottleneck. Third, a 1.7B-parameter executor trained via quality-filtered trajectory distillation matches a frontier sub-agent in accuracy while consuming 37% fewer sub-agent tokens, advancing the Pareto frontier. These results suggest a concrete recipe for building hierarchical search agents: concentrate capacity at delegation and downsize execution without sacrificing accuracy. Our code is available at https://github.com/QinnanCai0115/role-factorized-search.

Training a model with noisy gradient descent can be idealized as overdamped Langevin dynamics on the loss landscape, and a natural safety question is to bound the probability $ν_t(\mathcal{A}_H) = \mathbb{P}(Q_t \in \mathcal{A}_H)$ that the trajectory lies in a designated failure region $\mathcal{A}_H$. We study this for a smooth, strongly convex loss in $d$ dimensions and a failure region separated from the minimizer by an energy gap. Three bounds emerge. At the end of training, the equilibrium mass $π(\mathcal{A}_H)$ is exponentially small in $d$, with a complementary energy-barrier rate when the noise is small. Along the trajectory, a shape-free bound $ν_t(\mathcal{A}_H) \le π(\mathcal{A}_H)(1 + \sqrt{χ_0^2/π(\mathcal{A}_H)}\,e^{-mt})$ shows that the in-set probability relaxes to (twice) the static value after a burn-in time of order $d$, using only the global spectral gap $m$ of the loss. A worked Ornstein-Uhlenbeck example shows this burn-in is necessary: an angular slice of the equilibrium shell can transiently swell by a factor exponential in $d$, even though its equilibrium mass is tiny. To rule such swelling out we introduce a local relaxation rate attached to the failure region, defined through the spectral measure of its centered indicator rather than a Dirichlet-form Rayleigh quotient. For geometrically isolated regions this rate exceeds the global one, shrinking the burn-in proportionally, and combined with a maximum-principle ceiling it caps the trajectory probability uniformly in time. The picture is that strong convexity sets how fast training relaxes, but the shape of the unsafe set decides whether the trajectory bulges through it on the way home.

Artificial intelligence (AI) is a double-edged sword: while it has achieved remarkable success across a wide range of domains, its deployment also calls for effective oversight and regulation, for which the detection of AI-related content and artifacts is perhaps the most direct and cost-effective approach. To this end, we propose a unified detection framework based on Mahalanobis distance scores (MDS), applicable to several important settings, including the detection of large language model (LLM) generated text, hallucination, watermark, and adversarial examples. A key component of the proposed method is to accurately characterize the positive class--such as human-generated text, factual statements, unwatermarked text, or non-adversarial samples--which requires an efficient and robust estimator of the covariance matrix of deep representations of positive samples before computing the MDS. Since the positive samples typically consist of multiple classes, and these classes may exhibit both homogeneity and heterogeneity, we develop joint estimation methods for both the casewise and cellwise minimum covariance determinant (MCD) estimators. We provide efficient optimization algorithms for both estimators and prove their convergence. We provide a reasonable definition of the breakdown point for the joint estimators and prove their corresponding high breakdown point properties. Empirical evaluations confirm the effectiveness of the proposed detection framework.

AI agents that generate final answers based on user input often do not meet the needs of creative fields. Fields such as structural design and architecture need interactive systems that help users externalise and develop ideas, explore alternatives, and refine partial solutions. The final product of such designs needs to comply with many constraints concerning, e.g., spatial configuration, mechanical behaviour, material quantities, and costs. These constraints create friction in the design process, which can stimulate novel and creative solutions. In this paper, we discuss the misalignment between current generative AI goals to remove friction and provide final solutions and the needs of creators, such as structural designers, who develop ideas through iterative work. We present the design dimensions of systems allowing for constrained human-AI co-creation that rely on vision-language models making structural exploration conversational, multimodal, and responsive to evolving human intent in ways that follow and augment the discipline's creative process. Through a pilot design interface based on these principles and a study with experts in the field, this paper shows how structural designers perceive interactive AI systems and how such systems can support design space exploration by reducing repetitive modelling friction while preserving reflective design friction.

We address the problem of efficiently sampling multimodal probability distributions, where standard Markov Chain Monte Carlo methods often suffer from poor mixing and mode trapping. To mitigate these issues, we propose Gradient-free Riemannian Langevin Sampler (GRiLS), a novel proposal that improves exploration without requiring gradient evaluations of the target density. Our approach introduces a Riemannian metric which reshapes the local geometry in order to facilitate transitions across modes. The resulting gradient-free MCMC algorithm is particularly suitable for complex, computationally expensive targets where derivatives are unavailable or impractical. The GRiLS proposal requires knowing the mean and covariance of the target density, which we estimate using an ensemble of interacting particles. Empirical results on multimodal benchmarks demonstrate that GRiLS achieves improved mixing compared to existing gradient-based and gradient-free MCMC approaches.

We study bimatrix two-player games and investigate the last-iterate convergence and stability of equilibria for the iterates generated by the optimistic exponential weights method. In contrast to prior work, we allow the step sizes $η_x$ and $η_y$ to differ. Our first main result establishes, under the assumption that the set of fixed points is finite, a sufficient condition for global last-iterate convergence in the special case of zero-sum games, which constrains only the product $η_xη_y$ of the step sizes. This condition is practically relevant and partially explains empirically observed behavior. Our second main result provides an almost-tight threshold for asymptotic stability and instability, again in terms of products of the step sizes, for general bimatrix games. This result is primarily of theoretical interest. We derive several known results and practically relevant step size bounds for special cases and illustrate our results by experiments.

Self-supervised learning matches supervised accuracy from a fraction of the labels, but the labeled-sample efficiency behind this has lacked a theoretical explanation. We provide one. Data augmentation induces a similarity graph on the unlabeled data, so downstream learning on that graph is graph-Laplacian-regularized learning. We prove a fast transductive rate, $O(1/n_L)$ in the number of labels, in place of the supervised $O(1/\sqrt{n_L})$, by carrying the leave-one-out stability apparatus of Johnson and Zhang (JMLR 2007) over to the augmentation graph, and without the unrealistic assumptions of limit-based analyses (exact kernel, generalizing features). The bound makes augmentation quality explicit: the expected error is at most $C/n_L + R_{\mathrm{DA}}(y)$, where the data-augmentation alignment error $R_{\mathrm{DA}}(y)$ is the graph-cut mass of augmentations that cross a label boundary, so good augmentations let few labels suffice. The analysis uses a streamlined loss that drops the projector, negative-sample, and orthogonality overhead of standard objectives yet still recovers the top-$K$ ideal features in the infinite-data limit, the augmentation-kernel eigenspace studied by Zhai et al. The result explains the observed accuracy-versus-label-count curve rather than only bounding a generalization gap.

Business-to-X (B2X, "X" for business, customer, etc.) mobile applications ("apps") show various mobile-specific chances and challenges that must be ad-dressed in the whole development process. Knowledge about mobile app devel-opment models' usage remains restricted, although various process models for B2X app development have been published already. This article first reviewed available process models for mobile app development. In addition, 28 expert in-terviews with various stakeholders involved in typical B2X mobile app develop-ment processes were conducted to examine this research-practice knowledge di-vide. Since hybrid process models are often advantageous for B2X mobile app development, technical backgrounds or communication processes, are also cru-cial. Since no process model can be used unadopted, this study theorizes mobile-specific characteristics and challenges of app development process models to a reference model generally valuable for management decision support. The find-ings create avenues for better theorizing toward successful B2X mobile app de-velopment.

Generating signals on graphs requires permutation-equivariant models that exhibit stability with respect to relative structural perturbations. While favorable stability properties of Graph Neural Networks (GNNs) have been well documented, it is unclear how structural errors propagate through the dynamics of continuous generative flow models that are gaining traction for graph signal generation. In this paper, we analyze continuous normalized flow models parameterized by GNNs and show that permutation equivariance is preserved for both the resulting continuous-time ordinary differential equations and their discrete numerical approximations used as graph signal samplers. Our primary contribution is to derive explicit stability bounds on the generated probability distributions, which quantify how relative graph perturbations affect the final sampled signals. Motivated by these theoretical bounds, we introduce a stability-promoting regularized flow matching strategy that actively penalizes the spatial Lipschitz constant of the vector field during model training. Experiments using synthetic smooth signals on stochastic block model graphs and real-world fMRI signals on brain connectomes demonstrate that this bound-oriented approach yields generative models that are more robust to structural noise, without sacrificing output quality.

Reinforcement learning (RL) is becoming increasingly important for post-training large language models (LLMs). Previous RL pipelines for LLMs were mostly synchronous and batch-interleaved, which is inefficient for long-horizon agentic tasks. Recently, asynchronous RL has emerged as a more efficient alternative by updating the model as rollouts arrive. However, existing asynchronous RL systems often emphasize throughput, while leaving training stability and task effectiveness largely underexplored. For example, a key challenge is that group-wise sampling in the widely-used GRPO framework does not naturally fit asynchronous agentic training. In this paper, we present Single-rollout Asynchronous Optimization (SAO) to address the stability and off-policy challenges in asynchronous RL. To reduce off-policy effects and improve generalization, we replace group-wise sampling with single-rollout sampling, that is, using one rollout per prompt. We further improve this single-rollout strategy with practical value-model training designs. To improve optimization stability, we introduce a strict double-side token-level clipping strategy. SAO is able to train stably for one thousand steps and consistently outperform GRPO and its variants on agentic coding and reasoning benchmarks, such as SWE-Bench Verified, BeyondAIME, and IMOAnswerBench. We also demonstrate that single-rollout RL is particularly effective in a simulated online learning setting, where the model must adapt to changing evolving environments. To this end, SAO is successfully deployed in the agentic RL pipeline for training the open GLM-5.2 model (750B-A40B).

Hallucinations in vision language models (VLMs) are commonly treated as semantic errors, yet they often arise from partial or ambiguous visual evidence. Prior work mainly focuses on detecting or suppressing hallucinations at generation time, leaving the subsequent reasoning stage largely unexplored. In this work, we study Post Hallucination Reasoning (PHR), the stage in which hallucinated semantics enter the model's inference context and influence downstream predictions. To systematically investigate PHR, we introduce HIVE, Hallucination Inference and Verification Engine, an evaluation infrastructure that enables controlled comparisons between faithful and hallucinated captions. Across nine tasks and nine models, we observe structured modality dependent patterns: hallucinated captions often improve accuracy on vision language tasks, while text only tasks exhibit limited or unstable effects. Further analyses show that hallucinated cues broaden semantic coverage and reshape reasoning dynamics while preserving stable inference. These findings highlight that hallucinated semantics may influence downstream reasoning once they enter the model's inference context. Understanding this post hallucination stage is important for improving the reliability and interpretability of multimodal reasoning systems. Code is publicly available at https://github.com/hefengcs/HIVE.

Product data scientists often ask LLM-based agents to help with recurring execution tasks such as cleaning data, writing SQL, choosing statistical tests, and formatting results. Reusable skill files are meant to avoid prompting from scratch by packaging guidance for a task family. Expert-written skills can encode high-quality guidance, but writing and maintaining them across many data-science task families creates a manual bottleneck. We ask whether LLM-generated skills offer a useful low-curation alternative: do they improve performance over the task prompt alone? We test this question across four lifecycle stages: data preparation, data extraction, statistical analysis, and reporting, using one generated skill per stage. We find no reliable improvement from full generated skills over No-Skill prompting. We then ask whether any part of the skill is useful by ablating different skill components. The main ablation covers 56 tasks, nine model configurations, and three providers, yielding 7,560 runs. Compared with prompting using the task alone, neither the full generated skill nor any ablated skill variant significantly improves performance; all p-values are at least 0.396, and the total spread across variants is only 1.2 pp. A supplemental token-matched control adds 1,512 runs and finds that Full skills perform similarly to task-irrelevant skill-formatted content. The results caution against using one LLM-generated skill per data-science workflow as a default single-shot prompting strategy.

Time series classification (TSC) is dominated by a two-stage paradigm: train a feature encoder -- either from scratch on the target dataset or via pretraining on large corpora -- and then fit a task-specific classifier on top. While effective, this decoupling optimizes representation learning independently of the classification objective, requires per-dataset training, and prevents the model from exploiting label information during inference. We introduce TimEE, a 4.5M-parameter foundation model for end-to-end TSC via in-context learning. Given a labeled support set and a query time series, TimEE directly outputs a predicted class distribution in a single forward pass with no per-dataset training required. Following the prior-data fitted network (PFN) framework, TimEE is meta-trained exclusively on synthetic TSC tasks, where each task contains time series with distinct class identities arising from structured distributional shifts in the generative process. Despite seeing no real time series during pre-training, TimEE ranks first in ROC AUC (and third on accuracy) on the UCR benchmark among all compared methods, which include both foundation models and supervised deep learning baselines. To our knowledge, TimEE is the first purely synthetic-pretrained model to reach state-of-the-art performance on the UCR benchmark. These results establish end-to-end ICL with synthetic priors as a compelling, largely unexplored direction for TSC, with scaling, prior design, and richer generation mechanisms as natural avenues for improvement. Code is publicly available at http://github.com/automl/timee.

Testing is a major effort for the gaming industry, requiring a significant part of development budget and people power. We present a case study on a development version of the ice hockey game EA SPORTS NHL 26, for which human playtesters test the goalie AI for behavioral exploits. To reduce the effort of re-testing the goalie AI after every game or behavior modification in the development phase, we propose Reward-Adaptive Iterative Discovery (RAID), a novel approach to automatically find exploits using an iterative Reinforcement Learning (RL) approach that trains a population of goal scoring agents. While previous approaches can already successfully find exploits, RL algorithms tend to overfit to a single solution. We introduce a simple extension on top of existing RL algorithms, such that they find multiple diverse high-quality solutions. For our first deployment of this approach, within a single experiment we were able to find six hockey scoring exploit strategies that were qualitatively similar to those that playtesters had found in hours-long manual testing sessions.

Gradient communication is a primary scaling bottleneck in large language model (LLM) pretraining. Communicating gradients in low-precision formats, such as FP8 and NVFP4, can significantly reduce the communication volume. Existing methods quantize gradients via linear or nonlinear mappings in Euclidean space, often degrading model performance because highly anisotropic gradients incur direction-dependent distortion. We present GIFT, a geometry-informed gradient scaling method that performs low-precision communication in geometry-aware coordinates. By transforming gradients into a near-isotropic space before quantization, GIFT makes low-precision representations substantially more faithful to their high-precision counterparts. GIFT only changes the coordinate system used for low-precision gradient communication and does not change the optimizer, training recipe, communication collective, or low-precision format. We also develop a simplified geometry-aware transformation algorithm with low-rank approximation and selective application to balance the computation overhead and communication reduction. We examine the empirical convergence of GIFT using Llama-300M and Llama-600M models. Our results show that GIFT reduces the end-to-end pretraining time of Llama-600M by 7.6% on 64 NVIDIA GH200 Superchips, while improving the downstream task preservation profile over direct Euclidean FP8 communication under the same optimizer and communication path.

Many reasoning tasks are not well described by a single left-to-right chain: a solver may need to pursue a plausible branch, observe delayed failure, and return to the latest prefix that can still be completed. We introduce Pyligent, a training and inference framework inspired by the Diligent Learner formulation that represents reasoning as validated search over partial solution chains. A task validator labels generated continuations and failures, and the resulting search trees are converted into supervised targets for three actions: continue, finish, and backtrack, with optional traces that summarize abandoned branches. We evaluate Pyligent on a hidden directed graph task designed to isolate delayed-failure recovery, and on structured reasoning domains with exact validators, including $4{\times}4$ Sudoku, Sudoku with reasoning traces, and Blocksworld. Compared with gold-only supervised fine-tuning, Pyligent improves solve rate by $72.7$ percentage points on hidden graphs, by $17$ and $18$ points on mixed and expert Sudoku, by $27$ and $14$ points on mixed and expert Sudoku with reasoning traces, and by $13$ points on Blocksworld. These results suggest that explicit failed-branch supervision can teach useful recovery behavior beyond imitation of polished solution chains.

Large language models are increasingly used to evaluate and support software engineering tasks, yet the validity of these evaluations is often undermined by uncertainty about whether benchmark instances were seen during pretraining. This can lead to data contamination, which may inflate performance and result in misleading conclusions about model capability. Despite this, the training corpora of many modern models are only partially disclosed, making direct decontamination infeasible. This creates a need for practical methods that can detect a large language models' prior exposure to training data without access to the full training corpus. To address this challenge, we organize the first Poisoned Chalice of LLM Evaluation Competition, co-located with the FSE-AIWare 2026 Competition Track. The competition frames contamination detection as a white-box membership inference task on source code and provides participants with curated datasets, target models, baseline attacks, and a final evaluation on a held-out model and dataset. This design encourages methods that generalize beyond superficial dataset artifacts and beyond a single training setting. This paper reports the setup and results of the competition. More broadly, the competition aims to catalyze the community around trustworthy LLM evaluation for software engineering.

Generative AI is increasingly permeating software engineering, enabling developers to generate functions, files, and even entire applications from natural language specifications. AI systems are also becoming more personalized, adapting outputs based on inferred user characteristics and interaction history. While personalization may improve the development experience, it raises concerns that generated software could be shaped by attributes of the developer rather than by task requirements alone. Prior work has shown that generative AI can produce biased software artifacts, but little is known about how developer identity can bias generated code. We characterize three dimensions through which inferred developer attributes can influence generated artifacts: interface design, template content, and code structure. First, through controlled experiments on 800 AI-generated websites, we find that age- and gender-related signals produce significant differences across all three dimensions. Second, we conduct an observational study and follow-up interviews with 20 participants who used AI to create a personal website to both examine how personalization impacts software artifacts in practice, and also to understand how programmers perceive the boundary between personalization and bias. Together, our results show that developer attributes can meaningfully influence generated software beyond stated requirements, highlighting a previously underexplored tension between personalization and fairness in AI-assisted programming.

FFT-based spectral preprocessing of learned query-key (Q/K) projections substantially improves transformer attention on character-level language modelling. On TinyShakespeare: a fixed random spectral filter achieves val=1.031 (Delta=+0.443); a single learned frequency at paragraph scale achieves val=0.608 (Delta=+0.867); and four learned frequencies spanning paragraph to word scale achieve val=0.309 (Delta=+1.166), a 79% reduction over standard dot-product attention. The single-frequency result is confirmed across three random seeds (mean val=0.236, std=0.019). The four frequencies converge to a near-geometric multi-scale ordering (49, 27, 10, 6 tokens/cycle) corresponding to paragraph, sub-paragraph, phrase, and word scales. The gain is specific to spectral preprocessing: random orthogonal and non-orthogonal projections of Q/K produce no measurable improvement, suggesting the benefit comes from global frequency-domain mixing rather than metric distortion. All results are verified by a shuffled-validation diagnostic against positional leakage. Causal filters (Gaussian, Mexican Hat, Morlet) do not improve over standard attention at character-level tokenisation: the bilateral FFT kernel is structurally non-causal, coupling every position to future tokens. This defines an architectural boundary between bilateral spectral attention (this paper) and genuinely causal spectral attention at word-scale tokenisation (companion paper MorletQK). This work is architecturally distinct from FNet (Lee-Thorp et al., 2021), which replaces attention with Fourier mixing of token embeddings. Here, spectral preprocessing applies only to Q/K projections while the full attention score structure is preserved.

Agentic red-teaming benchmarks report whether an injected agent was compromised as a single bit: the attack succeeded, or it did not. We argue that this binary attack-success rate discards the information a defender most needs, namely how harmful the resulting action was. We introduce an action-graded harm rubric that scores an agent's tool-call trajectory on a seven-level ordinal scale (L0 to L6) according to whether the executed action was reversible, whether it crossed scope to reach another party, and whether it expanded privilege. We compute the scale two ways: a deterministic oracle that reads the trajectory and the attacker's stated goal, and a panel of three frontier language-model judges that read a tag-free account of the same trajectory. Across four victim models and two defenses on the AgentDojo workspace suite, severity grading exposes three cases the binary metric hides, including a defense that reports a zero attack-success rate while still permitting an externally visible cross-scope leak through an unfiltered tool. The judge panel reproduces the oracle with high ordinal agreement (Krippendorff's alpha = 0.91) but shares systematic blind spots that we characterize, most notably a failure to recognize escalation chains. Unlike prior work that provides harm taxonomies, harmful-task completion tests, execution-level safety benchmarks, or severity-aware simulation, our contribution is a reusable, trace-grounded severity instrument applied to the actual actions recorded in existing red-team logs. All code, prompts, and per-episode logs are released.

Machine learning models are increasingly deployed in high-stakes domains, raising concerns about both privacy and fairness. Differential Privacy (DP) has become a gold standard for privacy-preserving data analysis, while fairness-aware mechanisms aim to mitigate discrimination against underrepresented groups. However, these objectives can conflict: DP often amplifies disparities across demographic groups, and little is known about whether established fairness interventions remain effective under DP constraints. In this work, we present, to our knowledge, the first systematic evaluation of fairness interventions on differentially private synthetic tabular data. Our benchmark centers on the Adaptive Iterative Mechanism (AIM), identified as the state-of-the-art marginal-based DP synthesizer (Cormode et al. 2025). We thus evaluate fairness interventions across four datasets, multiple group fairness metrics, and three categories of mitigation strategies (pre-processing, in-processing, and post-processing) under a wide range of privacy budgets. We compare four pipeline configurations: (Baseline) training on original data; (DP-only) training on DP synthetic data; (Fair-only) applying fairness mechanisms on original data; and (DP+Fair) combining fairness mechanisms with DP synthetic data. Our results demonstrate that while DP alone can degrade both utility and fairness, applying fairness interventions can partially restore equitable outcomes. Among them, post-processing methods tend to provide more stable fairness-utility trade-offs across privacy budgets and synthesizers, achieving strong fairness improvements while preserving competitive utility relative to other intervention stages. We release all code, data, and experimental artifacts in an open-source repository to ensure full reproducibility and to support future research on the privacy-fairness-utility trade-off.

Fine-tuning large language models (LLMs) for e-commerce attribute extraction requires labeled data representative across thousands of product types, attributes, and multiple languages. This combinatorial scale translates to millions of annotations, rendering human labeling prohibitively costly. While recent work has demonstrated synthetic label generation using LLMs, deploying such approaches at industrial scale requires integrated quality control mechanisms. We present SynthAVE, a large-scale human-validated benchmark for attribute value extraction spanning 12,726 products across 229 product types, 792 attributes, and 4 languages (Spanish, French, Italian, German). To validate synthetic labels at scale, we introduce a multi-LLM arena framework where samples are independently evaluated by 21 judge configurations (7 model families $\times$ 3 prompts), with final labels determined via majority voting. The majority vote ensemble agrees with human experts at Cohen's $κ= 0.92$ (95.2% agreement), while individual judges show substantial inter-model agreement (Fleiss' $κ= 0.76$). This demonstrates that diverse models with varying individual judgments aggregate into highly reliable predictions, enabling cost-effective validation at scale while maintaining quality parity with human review.

We study the recovery of sparse functions from finite, noisy, and indirect observations in the framework of statistical inverse learning. The unknown is modeled as an element of $\ell^1$, and observations are generated through a possibly nonlinear forward operator $A:\ell^1\to H$, where $H$ is a vector-valued reproducing kernel Hilbert space. We propose an $\ell^1$-regularized empirical risk minimizer and develop a theoretical analysis of its statistical properties. Under mild assumptions, we establish almost-sure consistency and derive non-asymptotic high-probability convergence rates in both the prediction and $\ell^1$ reconstruction norms. The rates depend on the source smoothness parameter $r$, characterized by a variational source condition, and the effective dimension exponent $b$, describing the polynomial spectral decay of the covariance operator. We further prove matching minimax lower bounds, showing that the obtained convergence rates are optimal. To relate the theory to practical sparsity models, we consider finitely smoothing operators of the form $A=G\circ S$, where $S$ is a synthesis operator, and show that approximation-space assumptions imply the required variational source conditions. In particular, we prove that membership in the approximation space $k_t$ is equivalent to polynomial decay of the best $n$-term approximation error. Finally, we verify the assumptions for two representative inverse problems: reaction coefficient identification in elliptic PDEs and sparse computed tomography. For filtered Radon transforms, we derive explicit effective-dimension asymptotics, yielding concrete convergence rates for standard image models and sparsifying systems.

Spatial and Single-cell transcriptomics are transformative in deciphering cellular dynamics. As the fundamental paradigm for reconstructing cell developmental paths, trajectory inference (TI) is critical. However, existing methods require extensive manual intervention and proficiency in heterogeneous tools, posing a significant barrier to efficient TI analysis. To bridge this gap, we propose SpaCellAgent, an autonomous large language model (LLM) multi-agent framework that automates end-to-end spatiotemporal analysis and narrative generation. SpaCellAgent utilizes a multi-agent architecture for strategic workflow planning, a dynamic tool-orchestration engine for adaptive algorithm selection, and a self-evolution module that iteratively refines performance through feedback. We evaluate SpaCellAgent on six heterogeneous datasets encompassing complex temporal developmental trajectories, diverse sequencing platforms, and spatially-resolved tissue architectures. SpaCellAgent consistently demonstrates over 40\% improvement in analytical efficiency while maintaining expert-aligned performance. By converting natural language specifications into optimized analytical workflows and fully automating the pipeline, SpaCellAgent democratizes advanced spatiotemporal modeling and establishes a scalable, agent-driven paradigm for computational biology. The code and materials are available at https://github.com/LittleXH-shw/SpaCellAgent.

Large language models (LLMs) are increasingly deployed as autonomous agents that interact with external tools and services via the Model Context Protocol (MCP), a standardized interface for dynamic tool invocation. While MCP simplifies integration, it also expands the attack surface and enables generic exploits across multiple servers. Despite prior work on malicious MCP servers, the vulnerability landscape of MCP servers remains underexplored. In this work, we systematically analyze MCP server vulnerabilities, focusing on metadata characteristics, vulnerable code patterns, and community responses. Our study reveals that taint-style vulnerabilities constitute a substantial fraction of MCP server vulnerabilities, require significant code modifications to remediate, and are met with slow community responses. Motivated by these findings, we propose SPELLSMITH, presenting a novel textbased avenue for shielding taint-style vulnerabilities in MCP servers. In particular, SPELLSMITH analyzes the high-risk capabilities exposed by an MCP server and combines them with tool descriptions and parameter semantics to identify potential taint-style vulnerability risks, thereby constructing a tool-level risk profile. Then, SPELLSMITH leverages the Description property of the protocol to embed behavioral guidance (Description Enhancement Module) and exploits LLMs' self-reflection capabilities (Self-Reflection Module) to iteratively evaluate and refine outputs. By strengthening LLM internal decision-making, SPELLSMITH provides an active and unified mitigation strategy that generalizes across multiple vulnerabilities, reducing reliance on context-specific code-level fixes. Our experiments demonstrate that SPELLSMITH effectively mitigates taint-style vulnerability exploitation in MCP servers, highlighting its practical applicability and advantages over traditional code-level mitigations.

Navigation for social organisms rarely is a fully independent activity. Group structure and dynamics, as well as embodied interactions, critically influence useful behavior. Individual neural network controlled agents are trained to navigate in different social contexts, where social dependence and behavioral strategy learned is determined by relative task performance and spatial effect. Increasing high quality social information drives phase transitions from individual to following navigational strategy, and to collision avoidance in response to a crowded foraging patch. Predictable, nonstationary environmental dynamics drive behavioral hybridization between individual and social navigation, far and near the patch. Our findings challenge the approach of only inspecting individual behavior for social organisms and highlight the importance of taking a bottom-up approach in understanding how organisms behave.

Assertion-based Verification (ABV) plays a critical role in the Design Verification (DV) process. However, ABV requires substantial manual effort in generating assertion from specification by verification engineers, making it a time-consuming stage in the chip design flow. With the recent development of Large Language Models (LLMs), researchers have started exploring their use as an assistance in the ABV process, particularly for generating SystemVerilog Assertions (SVAs) from design specification. In this paper, we provide an overview of recent works, highlighting the different methods used to generate SVAs. In particular, we investigate LLM-based SVA generation and ask a central question: How can LLM-based assertion generation be made systematic and quality-aware? While addressing this key question, we provide Key Takeaways at the end of each challenge, summarizing the important methodological insights, and also provide guidelines and directions in solving those challenges that can help generate a high-quality set of assertions using LLMs.

A self-evolving agent retires its bad skills by watching them fail, so what happens when the judge cannot see the failures? Skill retirement is the structural constraint that keeps a growing library from drifting below the no-skill baseline, but its guarantee assumes an unbiased reward, which is false for the LLM judges that reference-free tasks force upon us. We show that a biased judge does not merely add noise; it \emph{silently switches off the curator}. We make this precise with a corrupted-reward analysis and, isolating the causal channel by injecting corruption on top of a deterministic reward, a behavioral study on a reference-free report-writing testbed with a code-generation cross-check. Symmetric noise leaves retirement intact, but \emph{false-pass} bias (failures slipping through as passes) disables contribution-based retirement past a sharp threshold that no amount of data can cross. Separating genuine retirement from cap-eviction churn shows this \emph{mechanism} failure is universal, holding across domains and failure rates and sparing only near-zero-false-pass, verifier-like graders. The downstream \emph{outcome}, though, is regime-dependent: eval quality degrades only where the same corruption also starves skill synthesis, and otherwise holds steady, so the disabled curator is \emph{silent}, surfacing in no aggregate metric. The contribution is a behavioral safety result, not a performance one. A cheap defect-injection audit then tells an operator, before deployment, which side of the threshold their judge occupies.

Agents acting on our behalf in the real world (e.g. placing phone calls) must learn online from costly, often irreversible interactions rather than cheap simulator steps. Two things follow. First, deployability depends on the path, not only the outcome. An agent must respect outcome-neutral constraints such as not repeatedly calling an unresponsive user, respecting business hours, or completing required authentication constraints that outcome-based rewards cannot express, since violating them frequently improves apparent success. Second, because each interaction is expensive, the agent must learn efficiently from very few examples. Reinforcement learning from verifiable rewards (RLVR) is blind to both challenges: it optimizes solely on the outcome and wastes expensive rollouts on all-fail groups where group-relative advantage collapses to zero. Attempts to densify supervision by rewarding progress target the hard-to-verify direction. In contrast, real agentic environments can cheaply detect bad moves. Since group-relative advantage is equivalent to within-group variance, a dense signal helps only when it supplies variance the outcome lacks. A verifiable penalty on the path meets this condition reliably, while a progress potential helps only where partial progress is reachable. The resulting recipe "penalize the path, reward the outcome" achieves high task success with near-zero violations, where outcome-only training violates constraints on nearly every episode. We provide four design rules for effective penalties, including avoidance of the inaction trap that arises when a penalty is used in isolation.

In this work, we reviewed different approaches in mathematical modeling of biologically plausible neural systems. Models are characterized and classified based on their common features and special use cases. In addition to spiking models, different types of discrete and continuous analogs are considered to accurately simulate biological processes, including membrane potential dynamics. The models under investigation include neurons and various components encountered in neural systems and affected the dynamics. The selection of specific approaches was driven by their prevalence and innovative perspectives in order to enhance the relevance of the presented information.

Many biological processes are governed by complex dynamical mechanisms that remain incompletely understood despite increasing volumes of experimental data. Biologically-informed neural networks (BINNs) seek to address this challenge by embedding mechanistic differential equations into neural network training, enabling interpretable constitutive operators to be recovered directly from sparse and noisy observations. However, reliable operator recovery depends sensitively on network architecture, optimisation strategy, and data informativeness. Here, we present a systematic empirical study of how these factors influence mechanistic inference using BINNs applied to canonical one-dimensional advection-diffusion-reaction partial differential equation models. Across a suite of benchmark problems, we investigate how network expressivity, learning rate, loss weighting, and batch size influence optimisation behaviour and operator recovery. We show that successful mechanistic inference depends on balancing competing objectives rather than maximising any single aspect of the model or optimisation. Moderately expressive architectures outperform overly complex networks, intermediate learning rates improve optimisation stability, balanced data and PDE losses are essential for accurate operator recovery, and intermediate batch sizes provide the best compromise between computational efficiency and reproducibility. We further identify practical diagnostics for recognising common failure modes, including over-fitting, unstable optimisation, and poor mechanistic recovery when the ground truth is unavailable. Together, these findings provide evidence-based guidelines for deploying BINNs as credible tools for biological model discovery.

We prove that, in the realizable PAC setting, the sample complexity of exact-trace learning for full autoregressive Chain-of-Thought traces is upper bounded by the standard multiclass rate of the local next-token class, where this rate is governed by the Daniely--Shalev-Shwartz dimension. Under exact-trace loss, one wrong action makes the whole trace incorrect; nevertheless, for every stopping rule $\mathtt{halt}$ and every pointwise $\mathtt{halt}$-halting local class $\mathrm{H}$, $n_{\mathrm{PAC}}^{\varepsilon,δ}(\operatorname{Roll}_{\mathtt{halt}}(\mathrm{H}))=O((\operatorname{DSdim}(\mathrm{H})+\log(1/δ))/\varepsilon)$, with no dependence on rollout length. The dependence on $\operatorname{DSdim}(\mathrm{H})$ is worst-case optimal, since one-step stopping recovers ordinary multiclass learning of $\mathrm{H}$. The proof introduces parity dimension, a rollout-stable refinement of DS dimension based on even pseudo-cubes. It controls one-inclusion density via a low-coordinate spanning theorem on finite restrictions and, unlike DS dimension itself, does not increase under autoregressive rollout. We also show why this detour is necessary: DS dimension can increase under rollout.

Logical Multi-Hop Query Answering over Knowledge Graphs (KGs) can be formulated as querying, with an implicit completeness assumption. Current works mainly focus on Existential First Order Logic (EFO) queries. These EFO queries contain conjunction, disjunction, and negation operators. Most existing works employ transductive reasoning, meaning they are not capable of reasoning over entities unseen during training. In the real world, there is a resource scarcity, and we cannot train a model with all the nodes of a large KG. Hence, we propose InductWave, a wavelet-based inductive embedding method for logical query answering on large KGs. Here, the training graph consists of fewer nodes than the test graph. Our model performs on par with the baseline models while having half the number of message-passing layers. It outperforms all of them in most cases, with 75% of the layers. These fewer resource requirements enable us to evaluate InductWave on massive graphs, such as Wiki-KG. We test our model using extensive experiments across varying train-test graph proportions of the FB15k-(237) dataset, comparing it with the state-of-the-art models. The code and datasets for the model are available at https://github.com/kracr/inductwave/.

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel. Block-parallel drafters such as DFlash further improve drafting efficiency by predicting an entire block in one pass, but their position-wise predictions lack explicit intra-block causal conditioning. Recent methods such as Domino and DSpark attempt to introduce such causality into block-parallel drafting, but they require training the draft model from scratch, which limits their flexibility and increases training cost. We propose DeLS-Spec, a decoupled long-short context speculative decoding method. DeLS-Spec treats the fixed DFlash model as a long-context expert and introduces a lightweight local head as a short-context expert. The local head can be trained independently with a standard next-token prediction objective, without joint training with the target model or the DFlash backbone, leading to extremely low training cost. At inference time, DeLS-Spec combines long-context and short-context logits, and the local head is not tied to a specific DFlash checkpoint, making the method more modular and flexible. Experiments on Qwen3 models show that DeLS-Spec consistently improves speedup and average acceptance length over DFlash across math, code, and dialogue benchmarks.

Automatic prosodic segmentation identifies boundaries between speech units from acoustic and linguistic evidence. Although recent deep learning approaches have produced strong results for English, automatic segmentation for Brazilian Portuguese (BP) still relies mostly on rule-based or traditional machine-learning methods. This paper presents SAMPA, a Whisper-based segmenter that transcribes BP speech while inserting explicit markers for terminal prosodic boundaries. We fine-tune Whisper large-v3 on manually segmented recordings from the NURC-SP dataset and evaluate different training and test-time filtering configurations, including out-of-distribution testing on the MuPe-Diversidades dataset. SAMPA achieves competitive boundary-detection performance across settings, with the best models reaching F1=0.731 on the held-out test split and F1=0.796 on MuPe-Diversidades. Finally, through n-gram and acoustic-visual analyses, we show that our model follows morphosyntactic, semantic, and prosodic cues for detecting prosodic boundaries.

Tool-using LLM agents can violate the very policies they are deployed to enforce while appearing to complete the task successfully. In policy-permissive environments, a tool may execute any well-formed call even when the corresponding state transition is forbidden by domain policy. The result is a silent wrong state (a booking cancelled, a passenger count changed, a claim acted on without verification) that neither the tool nor the agent's self-report exposes. We study this failure mode in the $τ^2$-bench airline domain. On a budget agent, 78% of observed failures are silent wrong-state failures with no tool error, and the aggregate failure rate is reproducible across disjoint seeds, not sampling noise. We then evaluate a lightweight intervention: deterministic, read-only pre-execution gates that inspect the proposed call and current state before allowing a write. A four-gate suite raises full-benchmark success from 29.6% to 42.0% on gpt-4o-mini (+12.4pp; paired task-level bootstrap P=0.0012), and the lift reproduces on a disjoint 15-seed set (+12.3pp; P=0.0008). The effect is concentrated where the gates fire: on the 26/50 firing tasks, success rises by +19.2pp, while movement on the 24 non-firing tasks does not exclude zero. Two negative controls (a self-enforcing retail domain and BFCL) bound the mechanism: gates help when tools are policy-permissive and add little where tools already self-enforce. As suggestive evidence, not a central claim, the same failure mode persists at the frontier: gpt-5.2 at default reasoning still attempts policy-violating writes, and the same suite improves success from 61.2% to 71.6% (+10.4pp; P=0.020; n=5, no replication). The contribution is a bounded evaluation and reliability result: deterministic gates do not guarantee task success, but they can deterministically prevent a known class of silent policy-violating writes at the action boundary.

Vision Language Models (VLMs) and Vision Language Action (VLA) models have shown promise in robotic control. Yet, they face significant challenges regarding explainability, generalization, and compute requirements. This paper presents a Multi-Agent System (MAS) architecture that addresses these limitations by deploying specialized agents on onboard hardware - eliminating dependence on external compute. The system controls a multi-purpose autonomous mobile manipulator in a simulated industrial warehouse, fulfilling five task categories: safety inspection, warehouse maintenance, warehouse search, package quality verification, and responding to human requests. Compact VLMs (3-20B parameters) are used throughout, with fine-tuning applied to improve package inspection accuracy. A novel "Megamind" orchestration agent mitigates context retention issues inherent to long-horizon planning with smaller models. The system was validated in a hardware-in-the-loop simulation using an AMD Ryzen(TM) AI mini PC. Results demonstrate that a fully onboard MAS architecture is a viable, cost-efficient alternative to cloud-dependent deployments, with strong potential for real-world transfer. The simulation environment has been released as open source under the Apache 2.0 licence.

While whole-body multimodal medical imaging scanners have been increasingly recognized for more effective medical applications, the excessive long acquisition time in PET-MR scanning is a major obstacle in more efficient clinical practice. Deep learning-based MRI translation provides a potential solution to reduce scan duration. However, current models often focus on specific anatomical regions and face challenges for whole-body scans that consists of highly heterogeneous feature distributions mainly due to (1) different anatomical regions across whole-body, and (2) lesions or pathological tissues. This paper tackles the challenges through a novel Heterogeneity-Adaptive Diffusion Schrodinger Bridge (HA-DSB) framework. By explicitly modeling translation as stochastic transport between source and target distributions, HA-DSB incorporates region context embeddings derived from a vision-language model (VLM) to enable region-specific modeling. To enhance fidelity of the pathological tissue, lesion-aware metabolic prior from PET is integrated directly into the bridge dynamics through a dual-stage guidance mechanism. Specifically, a PET-guided noise modulation module adaptively scales spatial diffusion perturbations during the forward process, while PET features are leveraged during the reverse process to selectively amplify lesion-relevant structures via an attention mechanism. Experiments demonstrate the superiority of our method across different body regions in whole-body MRI translation and show improved translation quality in lesion areas under PET guidance. Our code is available at Github.

In programming courses, instructors may need to interpret whether a submission is consistent with a student's prior programming profile, especially when code similarity alone is inconclusive. Existing source-code authorship methods are often evaluated on programming-contest or open-source datasets, where reusable templates and local code patterns can produce strong author-related signal. Educational repositories present a different setting. Students solve shared assignments while their programming practices are still developing. This study uses task-aware evaluation to contrast these production contexts and tests whether repository-visible process features add information beyond final code in six matched educational comparisons. Contest data provide a high-signal contrast, with a Kick Start mean top-1 of 0.938. Educational datasets produce substantially lower attribution performance. Adding process features raises the educational mean from 0.094 to 0.233 and mean pairwise verification ROC-AUC from 0.556 to 0.752. The comparisons show that measured signal depends on production context and that process patterns can complement weak final-code signal in educational repositories. Such models are therefore appropriate only as instructor-mediated decision support, not as independent proof of authorship.

Autonomous agents promise substantial gains in speed, scale, and labor efficiency, but their failures can impose abrupt and often irreversible costs. The central challenge for agentic automation is therefore to increase the benefits of automation while bounding the consequences of failure. While databases remain central to modern computing, agents operate over a broader data environment spanning files, APIs, applications, and system state. In this talk, I will outline early work on Agentic Data Environments -- the execution substrate in which agents operate -- that both amplify agent capabilities and enforce safety guarantees. This perspective reframes data systems from passive stores of state into active substrates for safe, reliable execution.

Reliable confidence estimation remains a key limitation of test-time adaptation in vision-language models (VLMs), where prompt tuning improves zero-shot accuracy but often degrades calibration due to entropy-driven overconfidence. Prior approaches mitigate this using LLM-derived class attributes and contrastive regularization, yet treat attributes independently, ignoring their relational structure. We propose ARGTCA, which represents (class, attribute) pairs as nodes in a Symbolic Attribute Graph and trains a Graph Attention Network (GAT) using contrastive objectives to produce structurally informed embeddings that capture inter-attribute dependencies. We introduce two attribute selection strategies: ARGTCA-DIV for intra-class diversity and ARGTCA-DISC for inter-class discrimination. Experiments across nine benchmarks show that ARGTCA-DIV reduces average Expected Calibration Error (ECE) by approximately ~37% over baselines, while ARGTCA-DISC consistently performs as the second-best variant, reducing average ECE by approximately ~17% over baselines. These results suggest that modeling symbolic attribute interactions provides a principled approach for reliable test-time adaptation in VLMs.

Mathematical reasoning benchmarks typically provide all facts needed to solve each problem, while interactive benchmarks often mix reasoning with tools, retrieval, and long-horizon dialogue. We introduce MIRA-Math, a benchmark for a narrower diagnostic capability: solving mathematical problems whose full latent state has a unique answer, but whose solver-facing view is missing exactly one necessary atomic fact. The solver must request the missing information in natural language under a strict budget and then integrate the returned fact into an exact final answer. A fixed constrained LLM responder sees only the dataset-provided atomic fact and must either offer the quoted fact when the request matches it, or decline otherwise. Thus, instance generation, typed hint specifications, validation, and final-answer verification are deterministic, while request metrics are measured under a fixed LLM-mediated responder channel. MIRA-Math contains 2{,}310 generated instances from 22 typed mathematical families spanning algebra, probability, linear systems, discrete structures, signal processing, Markov chains, circuits, interpolation, and numerical boundary-value problems. Experiments across frontier and small models show that request success and final-answer accuracy are separable: models may ask for the right fact yet fail the downstream computation, or fail before obtaining the canonical hint. We release generators, verifiers, prompts, run metadata, and dataset documentation to support reproducible evaluation of minimal information requesting in mathematical reasoning.

Large Language Models (LLMs) store factual knowledge and domain-specific patterns implicitly in dense Transformer parameters, making knowledge expansion costly through pretraining, fine-tuning, retrieval augmentation, or longer contexts. Engram-style memory offers a compact hidden-state injection pathway, but existing GPU-resident designs often rely on hash-based compression, causing unrelated phrases to collide in shared slots and weakening phrase-level semantic fidelity. We present TF-Engram, a train-free Engram system that constructs phrase-specific semantic memory offline from external corpora, stores large memory tables across a GPU--DRAM--SSD hierarchy, and uses Early-Exit Guided Predictive Prefetching to hide external-memory latency during autoregressive decoding. On Qwen3-0.6B, TF-Engram improves the average downstream score from 57.6 to 59.4, outperforming both the frozen backbone and a parameter-matched LoRA baseline. System evaluation shows that large TF-Engram tables can be built with moderate offline cost, SSD-backed storage substantially reduces GPU memory demand, and predictive prefetching recovers much of the throughput loss caused by external memory access. These results demonstrate that static phrase memory can be integrated into LLM inference as a scalable, train-free, and low-overhead system component.

Recently, Large Language Models (LLMs) have been utilized in various applications of computational social science and provide the possibility to integrate such models into agent-based modeling to explore the cognitive processes. However, how specific cognitive modules drive individual decisions and macro-level opinion dynamics remains unclear. Therefore, this study introduces a framework that integrates an LLM (Qwen3-8B) into agent-based modeling to investigate this problem, using vaccination opinion dynamics as a case study. We utilize this framework to simulate opinion dynamics among agents with heterogeneous profiles and social networks, evaluating scenarios by enabling different cognitive modules: a memory module and a prompt diversity module. The simulation results reveal that different cognitive modules have opposite impacts on our emergent opinion. Furthermore, the framework reproduces the non-linear behavior patterns of social influence observed in existing research, demonstrating our framework's validity and potential to reach the level 3 validation of agent-based models.

Linear attention models allow a fixed state size and a fixed amount of compute per token. However, due to their limited state size, linear attention models fall behind in long-context recall compared to softmax-attention-based transformer architectures. Increasing the state size of linear attention improves recall performance but at the cost of higher FLOPs. In this work, we introduce Sparse Delta Memory (SDM), an architecture that scales the hidden state of gated linear RNNs to orders of magnitude higher capacity using a sparse addressing scheme. SDM extends the Gated DeltaNet architecture by replacing the dense key-value outer product with sparse reads and writes to a large explicit memory. We show that, under an isoFLOP constraint and with an identical number of parameters, a higher state memory capacity significantly improves performance on in-context learning and long-context retrieval tasks. Moreover, by learning the initial state of the SDM memory and therefore using it as a parametric memory, we show that the model further improves on a wide range of common-knowledge and reasoning tasks.

Fast Radio Bursts (FRBs) are millisecond-duration radio transients whose automated detection increasingly relies on highly specialized deep learning models. These detectors achieve exceptional performance, but they require large task-specific training datasets and cannot be redefined without retraining. In this work, we evaluate whether small, open-weight, locally run generalist Vision-Language Models (VLMs) can detect FRBs in dynamic spectra under a zero-shot, prompt-only regime, with no fine-tuning and no labeled examples, returning structured decisions with a natural-language justification. From a controlled set of 3000 simulated L-band dynamic spectra containing FRBs, structured Radio Frequency Interference (RFI), and noise, we draw a balanced binary benchmark of 2000 samples and compare two such VLMs (Gemma 4 2B and 4B), sample by sample, against the state-of-the-art specialized detector SwinYNet. At the default threshold, Gemma 4 2B reaches an accuracy of 93.65%, with no statistically significant difference from SwinYNet (92.90%), while showing a significantly lower false-positive rate on structured RFI (6.4% vs. 25.0%) and no false positives on pure noise. SwinYNet retains a perfect probabilistic ranking on this benchmark (ROC-AUC of 1.0000 vs. 0.9482), a ceiling that the zero-shot VLM approaches from general-purpose pretraining alone. Rewriting the prompt alone reconfigures the same models for three-class FRB/RFI/noise classification on the full set of 3000 spectra, where they reach up to 86% accuracy without a single false FRB.

In agentic scientific machine learning (SciML), large language model (LLM) agents can discover surrogate models and select one by an automated score, typically an error metric. A low error, however, does not establish that the predicted fields satisfy the physics that matter for mechanics, such as boundary conditions, superposition, stiffness scaling, or causality. We introduce Physics-Audited Agentic SciML (PA-SciML), a verification-first workflow for agentic SciML discovery. The workflow fixes a scoring evaluator before search, derives reviewable machine-checkable physics requirements, checks each trained candidate on its outputs, and separately searches prescribed input ranges or measured load-history spans for high-violation cases without reference solution fields. A surrogate is reported as verified only under the stated checks. When enabled, the workflow also adds advisory numerical probes before training and tests one modeling change at a time to record which isolated edits are associated with score gains before reuse. In the reported computational-solid-mechanics numerical examples, the static elasticity run selects a surrogate with lower validation error than the error-only baseline while both selected models pass the common linear-elastic checks. In the transient elastodynamics run, an error-only baseline with similar mean error fails a stricter causality check by responding to future parts of the loading history, while the selected surrogate passes the stated checks. The main distinction is per-candidate physics evidence on predicted fields, not a richer aggregate score.

Adversarial vulnerability in deep neural networks (DNNs) has been studied from the perspectives of decision-boundary geometry, feature robustness, input-output Jacobians, and the instability of inverse problems. Here, we focus on the spectral structure of intermediate linear transformations that propagate information through modern DNNs, an unexplored mechanism of adversarial vulnerability. Specifically, we investigate transformer-based vision-language models, whose linear layers admit interpretable spectral decompositions and whose widespread adoption makes understanding their robustness increasingly important. We propose a white-box spectral-subspace-guided attack (SSGRA) that aligns intermediate representations with the subspace spanned by the bottom right singular vectors. Our experiments show improved attack effectiveness over existing baselines. In addition, SSGRA offers a spectral interpretation of adversarial vulnerability in VLMs, providing insights for improving their robustness.

In biological circuits, sequential neural activity evolves along dynamic, low-dimensional manifolds to enable flexible behavior. Spiking network models link aspects of this sequential activity to features of manifold geometry through specific circuit mechanisms, making dynamic neural manifolds parameterizable, and thereby offering an explainable framework for neural computation. Extending this framework to neuromorphic engineering, we present an implementation on the SpiNNaker 2 chip for real-time, closed-loop control. By allowing sensory inputs to modulate heterogeneous inhibition, gain, and transient currents, our architecture drives rapid subspace rotations to switch between behaviors, as well as fine-grained trajectory control within them. We validate this via a robotic simulation where an agent uses sensory feedback to dynamically reconfigure its manifold geometry to navigate through a maze. Our results establish dynamic manifolds as a feasible approach for explainable neuromorphic architectures and a substrate for investigating biological neural dynamics.

The Quantum Control Processor (QCP) bridges the gap between compiler toolchains and control electronics, and is responsible for translating compiled quantum circuits into executable instructions that directly manipulate qubits and handle measurement feedback. However, existing designs rely primarily on customized instruction sets, limiting design reuse and requiring significant effort to build supporting toolchains. Furthermore, efficiently addressing qubits and scheduling operations in highly scalable scenarios remains a critical challenge. In this work, we present a vectorized quantum control approach built upon the RISC-V Vector (RVV) engine with a quantum-oriented extension. Leveraging the high parallelism of RVV, our approach can address up to 128 qubits in a single instruction. We also embed parameterized rotation information into the instruction set, enabling dynamic tuning of gate rotations in hybrid quantum-classical programs. To support mid-circuit measurements, we design a hardware-based halt-resume protocol that resumes pipeline execution within 80 $ns$ of receiving the measurement result. Comprehensive evaluation using both RISC-V toolchains and FPGA prototypes demonstrates that our design achieves up to 2.52$\times$ speedup over the baseline in program execution time, with excellent scalability.

In embodied intelligence systems, the motion controller serves as the critical bridge between semantic reasoning and physical execution. Humanoid control has progressed rapidly through large-scale human motion-capture data and motion-tracking paradigm. However, producing quadruped robots motion corpora with scalability and physical feasibility faces more fundamental obstacles: animal motion data is scarce, and cross-embodiment retargeting remains fragile. We present ABot-C0, a generalist motion-control system for quadruped robots that establishes three complementary behavior foundations: a scalable multi-source motion-data pipeline, robust policy learning across motion tracking, locomotion, and scene interaction, and a unified deployment stack for reliable real-world operation. Fundamentally, we construct a data pyramid through conditional video-generation synthesis, annotated motion capture, teleoperation and human design, producing 16,074 physically feasible motion clips as the data foundation for various motion learning demands. We then train a Flow-Matching generalist policy that demonstrates for the first time quadruped motion tracking scaling law that its performance improves consistently as training scales up, with zero-shot capability to track unseen motions. Then, we push a step further for robust all-terrain traversal locomotion by adopting a three-stage privileged-to-perceptive framework with temporal LiDAR memory and terrain-predictive supervision. Collectively, these components form a motion generalist that coordinates multi-policy execution, smooth behavior transitions, energy-efficient control, and safety mechanisms for real-world deployment. Extensive experiments on urban-terrain autonomous navigation and companion-style multimodal interaction demonstrate that quadruped robots move beyond single-function demos toward product-level behavioral intelligence.

AI control is a family of techniques to prevent an AI with malicious goals from subverting its operator's intent. AI Control usually studies a single agent in one trajectory, but real deployments run many agents over shared infrastructure, and the most severe risks (model-weight exfiltration, training-run poisoning) plausibly need several agents acting in concert. We initiate the empirical study of multi-agent AI control, formalising distributed attacks in which several agents jointly aim for a malicious goal. We develop FakeLab: a synthetic AI-lab codebase (9 services, 86 benign tasks, 4 attack objectives). We evaluate single agent monitoring against distributed attacks, varying the number of agents, their coordination, model capabilities and precise monitoring configuration. Our central finding is the fragmentation effect: as more agents coordinate to attack, per-agent monitoring becomes less likely to catch any of the attackers. Secondly, this is not driven by the ratio of benign to malicious code -- we argue this is likely dependent on model capability. Third, an explicit planner amplifies the fragmentation effect, raises the rate of completing the attack up to sevenfold, and mildly sharpens executors' poor monitor-theory-of-mind. Finally: a weak "trusted" monitor misses most attacks, while a stronger one cuts undetected success by more than an order of magnitude but remains imperfect. We are happy to share FakeLab, our control setting, with safety researchers on request.

Effective social robot navigation requires sensitivity to human behavior, often revealed through subtle skeletal cues like gait and orientation. We present Human-Aware Implicit Social Robot Navigation (HumAIN), a novel framework that fuses implicit social cues directly into the planning loop via knowledge distillation. We first employ a transformer-based teacher model that fuses rich multi-modal inputs, including historic images, skeletal keypoints, robot state, and a robot's target goal, to learn robust, human-aware representations for the robot's future trajectory planning. To enable real-time deployment, we then distill this knowledge into a lightweight student model. By optimizing for both trajectory reconstruction and latent feature alignment with the teacher, the student learns to infer complex social dynamics from minimal inputs. Bridging the prediction-planning gap with an efficient distilled architecture, our method enables robots to reason about human behavior in a manner that is adaptive, robust, and socially compliant. We validate HumAIN through extensive experiments, where it improves trajectory prediction metrics by an average of 29.8% across all metrics compared to state-of-the-art baselines. These results highlight the benefit of using implicit, whole-body cues to achieve human-like navigation awareness on resource-constrained platforms.

Online truckload bid acceptance is a closed-loop stochastic decision problem in which a carrier or broker must, in real time, accept or reject a tendered load subject to operational feasibility, fleet repositioning costs, and opportunity cost against future demand. Public, reproducible benchmarks for this problem are scarce: existing routing benchmarks are static, while dynamic-fleet studies typically rely on private operator data. We introduce FreightBidBench, a public-calibrated, dependency-free, closed-loop benchmark in which feasibility (pickup reach, appointment windows, simplified hours-of-service, stochastic yard delays) and economics (service-failure penalty, terminal fleet value, daily price-premium window) are explicit, versioned, and reproducible from public Freight Analysis Framework and U.S. Department of Agriculture truck rate data. We develop two full-horizon hindsight ceilings: a simple LP style relaxation and a tighter Lagrangian-per-truck information relaxation that retains per-truck hours-of-service and sequencing structure and is 20.7% tighter than the LP relaxation on a tight-capacity scenario and 39.3% tighter on a scarce-capacity scenario. We introduce a parametric surrogate-rollout cascade with boundary-band and scarcity-pressure escalation triggers. On ten-seed tight and scarce scenarios, the best simple policy retains 91.0% and 86.5% of rollout profit and the standard-library surrogate 94.2% and 89.3%; a cascade at a single escalation band recovers roughly 98% on both at 40-56% of rollout's mean decision latency, and on the tight scenario is statistically indistinguishable from the rollout teacher (paired-bootstrap 95% CI on the profit delta spans zero).

Nonlinear dynamics is ubiquitous in nature, ranging from chemical pattern formation to ocean circulation, yet its simulation on quantum computers is fundamentally limited by the unitary nature of quantum evolution. We propose the quantum Koopman method, a data-driven framework that embeds nonlinear dynamics into a learned linear representation and implements the resulting evolution using shallow quantum circuits. This method learns Koopman observables from trajectory data, projects the lifted dynamics onto a finite-dimensional subspace, and decomposes the corresponding non-unitary propagator into parallel spectral channels. We utilize the Koopman method on a superconducting processor to simulate three distinct nonlinear systems, comprising reaction-diffusion dynamics, fluid motion on a sphere, and satellite-derived observations of Gulf Stream currents, employing up to 32 parallel circuits of 10 qubits. These quantum simulations capture the dominant multiscale patterns and statistical signatures of the underlying dynamics, and reveal a transition from performance limited by hardware noise in weakly nonlinear systems to performance limited by finite-dimensional Koopman representations as nonlinear scale interactions increase. This transition identifies a practical boundary for quantum-amenable nonlinear dynamics, establishing a hardware-validated route for simulating moderately nonlinear dynamics on near-term quantum hardware.

Large-scale combinatorial optimization remains demanding for classical heuristics, particularly when dense Quadratic Unconstrained Binary Optimization (QUBO) formulations induce large memory footprints, high CPU utilization, and long execution times. While near-term quantum processors cannot yet deliver unconditional quantum advantage, hybrid architectures can provide practical value by reducing the resource burden. This paper presents a resource-efficiency study of Hybrid Quantum Neighborhood Selection (HQNS), a framework that decomposes large dense QUBO instances into bounded-width quantum subproblems via stochastic frontier selection. We evaluate HQNS on the Maximum Diversity Subset Selection Problem (MDSSP), focusing on the trade-off between solution quality retention and resource consumption. Benchmarks up to N=1000 candidates show that HQNS preserves 99.9908% of the mean diversity score of an 11-restart parallel Simulated Annealing baseline, while reducing wall-clock time by 94.91%, peak CPU utilization by 64.68%, and peak memory usage by 88.61%. The QPU execution time remains bounded within a 6-7 second envelope across scales, indicating that the quantum component is decoupled from the global QUBO dimension when the frontier size is fixed. These results suggest that HQNS provides a resource-aware pathway for deploying hybrid quantum optimization in practical large-scale settings, serving as an efficient architecture for incorporating near-term quantum processors into classical optimization pipelines.

Hypergraph neural networks have shown powerful capability in modeling higher-order relations, yet their predictive uncertainty remains underexplored. Unlike pairwise graphs, uncertainty in hypergraphs arises not only from noisy attributes and ambiguous labels, but also from variations in node-hyperedge incidence structures and complex higher-order dependencies. Existing approaches mainly estimate uncertainty from final predictions or rely on computationally expensive ensembles and Bayesian inference, limiting their ability to capture uncertainty evolution during representation learning. In this paper, we propose Hypergraph Neural Stochastic Diffusion(HyperNSD), a stochastic differential equation framework for uncertainty estimation on hypergraphs. HyperNSD models hypergraph representations as stochastic processes evolving over node-hyperedge incidence structures. A learnable drift function captures deterministic higher-order diffusion dynamics, while a learnable stochastic forcing function characterizes structural ambiguity and representation noise. Predictive uncertainty is directly quantified through the variability of stochastic representation trajectories, providing an intrinsic uncertainty measure beyond post-hoc confidence scores. We formulate HyperNSD with neural drift and diffusion networks, enabling joint learning of prediction and uncertainty propagation. Theoretical analyses establish well posedness, perturbation stability,permutation equivariance, and numerical convergence of the proposed stochastic dynamics. Experiments on multiple hypergraph benchmarks demonstrate that HyperNSD achieves reliable uncertainty estimation for out-of-distribution and misclassification detection while preserving competitive prediction accuracy. These results provide a principled stochastic-dynamical framework for trustworthy higher-order representation learning.

Automated crowd counting in Hajj video is difficult not because current models lack capacity, but because the footage violates the assumptions those models were built on: cameras observe the crowd from steep, near-vertical angles, individuals occlude one another extensively, and a single frame can contain well over a thousand people. Benchmarks that test crowd counting in such an environment are either private or not detailed per second. We revisit the HAJJv2 dataset and contribute HAJJv2-CrowdCount: per-second human-annotated crowd counts for its testing videos. Using these annotations, we benchmark three recent zero-shot counting paradigms: an open-vocabulary detector (YOLO-World), a point-based counter (APGCC), and a promptable segmentation-based counter (SAM3Count). SAM3Count attains the lowest overall mean absolute error (MAE 70.4, 95% CI 56.0-86.1), ahead of YOLO-World (92.0) and APGCC (152.9). This ordering reverses, however, in the regime most relevant to deployment: on the densest frames, the detection- and segmentation-based counters both degrade sharply (MAE exceeding 300), while the point-based counter degrades far more gracefully (MAE 114.9). This inversion is decision-relevant for Hajj crowd management, where reliable counts are needed most precisely in the densest and most occluded scenes. The annotations are released to support reproduction and extension of these results.

Tool utilization enables Large Language Model (LLM) agents to interact with the real world and resolve complex tasks. However, existing agent frameworks predominantly rely on static toolsets composed of granular atomic actions (e.g., basic file I/O or single-turn search), which forces agents to reinvent low-level logic for every recurring workflow, leading to increased reasoning overhead and failure rates. In this study, we propose that agents can achieve self-evolution by synthesizing these atomic actions into reusable Standard Operating Procedures (SOPs), which function as callable higher-order tools that encapsulate multi-step logic. We further introduce EvoSOP, a framework that empowers agents to extract SOPs from execution trajectories and iteratively optimize the toolset through a systematic lifecycle of construction, merging, evaluation, and pruning. Extensive experiments demonstrate that EvoSOP significantly boosts task success rates while substantially reducing the number of interaction rounds compared to baselines. Our analysis also reveals that iterative tool optimization fosters reliable and efficient tool-use patterns, providing a scalable pathway for the development of self-evolving agents.

Rigorous content moderation is crucial for online advertising but leads to millions of daily rejections. This scale renders manual rectification infeasible, particularly for video advertisements. However, existing safety-driven methods often suffer from aggressive over-editing, which compromises the advertiser's original semantic intent merely to satisfy compliance. In this work, we target the rectification of textual violations in video ads, covering both speech transcripts and on-screen text. We propose R^3, a novel framework designed to harmonize compliance with original semantic intent preservation. Our approach integrates three key innovations: (1) an experience-driven data synthesis framework that bootstraps high-quality supervision via a group-Relative compliance experience extractor; (2) a curriculum Reinforcement learning strategy with hierarchical rewards designed to enforce compliance while maximizing semantic consistency; and (3) a comprehensive video Rectification framework seamlessly integrating text recognition, rewriting, and re-rendering for industrial deployment. Extensive experiments on industrial datasets and online A/B testing demonstrate that R^3 significantly outperforms state-of-the-art baselines, achieving an optimal trade-off between violation rectification and intent preservation.

This article offers a comprehensive overview of mechanistic interpretability, an emerging field that seeks to reverse-engineer the internal algorithms of modern neural networks. While traditional explainable AI methods often stop at surface-level input-output correlations, this approach directly addresses the opaque "black box" nature of machine learning models, which is essential for ensuring safety and auditability in high-stakes deployments. The paper provides a detailed examination of Transformer circuit analysis, exploring how internal components like the residual stream, attention mechanisms, and induction heads drive complex tasks and in-context learning. It subsequently tackles the core challenge of superposition and polysemanticity, demonstrating how tools like Sparse Autoencoders (SAEs) and transcoders can decompose tangled network activations into distinct, human-interpretable features. Furthermore, the paper explores methods for actively controlling and modifying model behavior through steering vectors and causal interventions. Finally, it connects these mechanistic insights with neurosymbolic AI frameworks designed to translate neural representations into explicit, executable logical rules.

Federated learning (FL) avoids explicit data exposure by keeping raw data on local clients, yet privacy risks remain in the training process and the learned model itself. Recently, centralized Taking Away Training Data (TATD) attacks have shown that malicious training could abuse the memorization capacity of deep models to store and later recover training data. However, this memorization-based threat has not been systematically studied under FL environments, where multi-client averaging could overwrite encoded training data. In this paper, we study a white-box TATD attack in which a malicious server selects n target clients from K participating clients and actively writes private training data into the global model during federated training. We propose FedCVESA, a federated variant of Correlation Value Encoding Attack (CVEA), by adding a Pearson-correlation regularizer to the loss function of target clients, so that private training data are gradually encoded into selected model parameters, referred to as carrier parameters. To reduce the overwriting of carrier parameters during server aggregation, we further propose segmented aggregation over dispersed carrier parameters, preserving selected carrier parameters while keeping standard averaging on the remaining parameters. Experiments on MNIST, Fashion-MNIST, and CIFAR-10 under Dirichlet non-IID partitions show that the proposed method can steal semantically meaningful private training images from the trained model while maintaining acceptable main-task utility in a controlled proof-of-concept setting. These results demonstrate that FL can become a parameter-level memorization channel for active TATD attack under the studied white-box malicious-server setting.

Pairwise comparison (PC) via pairwise reciprocal matrices (PRMs) is central to the Analytic Hierarchy Process (AHP). Although the traditional eigenvector method is widely applied to derive priorities, its theoretical robustness in reflecting true priority vectors remains debated. Building upon a previous iteration of this study, this research develops the revised Least Penalty-Squared Prioritization (LPSP) optimization models, including the revised Least Product of Penalty and Direct Squares (LPPDS) and revised Weighted Squares (LPPWS), to minimize the revised Root Mean Penalty-Squared Variance (RMPSV) and the revised Root Mean Penalty-Weighted Square Variance (RMPSWV). However, solving these non-linear formulations is computationally complex for decision-makers. To overcome these limitations, this study proposes the Parallel Osprey Optimized Least Penalty-Squared Prioritization (POO-LPSP) method. By integrating an improved bio-inspired metaheuristic Parallel Osprey Optimization Algorithm (POOA), this framework efficiently solves complex LPSP models to minimize RMPSV and RMPSWV, thereby enhancing prioritization reliability. The practical utility and computational efficiency of the POO-LPSP method are validated through a numerical application focusing on a Generative AI (GAI) vendor selection problem. To extend, POO-LPSP can serve as a robust alternative to Saaty's Eigen system method for AHP applications.

In this paper we first study the problem of generalized linear bandit (GLB) under heavy-tailed noise. The characteristics of heavy-tailed distributions are widely observed in real-world applications such as personalized recommendation, financial markets, and medical treatments. Based on the online mirror descent (OMD) method, we propose an algorithm EHM that extends the adaptive Huber loss method (Wang et al., 2025) with one-pass update ($\mathcal{O}(1)$ computational complexity with respect to current round $t$ and the time horizon $T$), which simultaneously achieves an almost optimal regret of $\widetilde{\mathcal{O}}(T^{\frac{1}{1+ε}})$ where $T$ is the time horizon. In addition, by utilizing a special property of some link function (Sawarni et al., 2025), our algorithm eliminates the need to know a commonly used parameter. Next, we study the GLB problem under the case when contextual characteristic becomes piecewise constant, and we slightly revised former algorithm to obtain the PGLB-EHM algorithm. After theoretical analysis, we prove that the regret upper bound order stays the same. Furthermore, we look deeper into a special case of nonlinear bandit (NB) and present the NB-EHM algorithm with bisection method and special restriction. Eventually we utilize the affine lifting approach and show that the general NB problem can be applied with NB-EHM to achieve a sublinear regret bound.

This paper reports an empirical study evaluating the relevance of several RAG metrics. The experiment is based on a question-answering dataset created by human annotators from business data. The generated responses and retrieved spans of a RAG system are scored using evaluation metrics from four libraries (Ragas, DeepEval, RAGChecker, Opik). These metrics are compared to scores given by two evaluators, as well as to standard metrics such as recall. An analysis of correlations is conducted. Finally, we highlight certain limitations of our methodology, compare it to those used in the literature, and suggest some avenues for future research. This paper is an English translation of a paper originally published in the French-speaking workshop EvalLLM (Brabant, 2026).

Turn-taking prediction is a key requirement for social robots involved in human-human interaction, particularly in mediator settings, where the robot must anticipate conversational dynamics rather than merely react to pauses. This work presents a Multimodal Voice Activity Projection (MM-VAP) framework that extends the original audio-only VAP formulation to synchronized audio-visual inputs while preserving its self-supervised future-projection objective. The proposed approach builds on pretrained audio-visual backbones originally optimized for speech-related tasks and adapts them through Low-Rank Adaptation to the multimodal turn-taking problem. After independent speaker encoding, an inter-speaker attention stage models the relational dynamics required to project future voice activity. In addition, a semantic consistency loss is introduced to regularize the 256-state output space according to higher-level dialogue activity patterns. Experiments on NoXi and NoXi+J showed improvements over the current baselines, particularly for some turn-taking events. Additional evaluation on the Haru EDR corpus further supported the suitability of this direction for mediation-oriented human-robot interaction.

Semantic communication, leveraging joint source-channel coding, is designed to mitigate semantic distortion introduced by the channel. However, most current studies focus solely on semantic deviation distortion caused by physical wireless channels, while overlooking semantic erasure distortion due to packet loss. A CSI-Guided Mamba-based video semantic wireless digital communication system (MamVSC) employing semantic grouping is proposed to simultaneously address both semantic deviation and erasure distortions. In this system, a semantic Mamba module, guided by channel state information (CSI) feedback, is utilized to dynamically adjust the granularity of extracted semantic information, adapting to channel conditions. Furthermore, a Semantic Channel Codec based on dynamic Semantic clustering centers is introduced, where the distance between semantic vectors within the same semantic class and their corresponding Semantic clustering center is dynamically adjusted according to channel conditions, enhancing robustness against channel noise. Additionally, a adaptive packet loss recovery module, dynamically adaptive to the CSI, is proposed. The system achieves an MS-SSIM greater than 0.6 and a PSNR exceeding 21 dB at an SNR of -8 dB and a packet loss rate of 90% in AWGN channel.

Accurately estimating urban carbon emissions is critical for sustainable urban planning, yet many existing approaches remain difficult to apply consistently across cities due to data-source heterogeneity and the lack of fine-grained semantic-temporal context in remote sensing data. We propose CarbonCLIP, a task-oriented multimodal distillation framework that improves satellite-based carbon emission prediction by transferring contextual knowledge into a unified satellite representation through dual-branch contrastive learning. Unlike conventional methods that rely on static visual features, CarbonCLIP explicitly bridges the gap between top-down satellite views and ground-level human activities. Specifically, the spatial branch uses fine-grained textual descriptions automatically generated from street-view images by Large Multimodal Models (LMMs) to provide semantic priors reflecting building functions, infrastructure, and urban activities, while the temporal branch employs a month encoder to encode temporal priors associated with monthly emission variation. CarbonCLIP requires multimodal data only during the pretraining phase; during inference, it relies solely on satellite imagery, thereby supporting scalable deployment when ground-level data are unavailable at inference. Experiments on Beijing and Singapore demonstrate that CarbonCLIP outperforms baselines in both study cities. The results validate that our method effectively transfers multimodal knowledge into satellite representations, offering a robust solution for satellite-based urban carbon modeling.

This study presents a multi-fidelity framework for the systematic optimization of genetic algorithm (GA) hyperparameters. The framework integrates three fidelity levels: high-fidelity Fast Fourier Transform (FFT) homogenization for validation, a medium-fidelity 3D convolutional neural network surrogate for rapid property evaluation, and a low-fidelity Gaussian process (GP) surrogate within a Bayesian optimization (BO) framework to guide the hyperparameter search. Various acquisition functions are evaluated, with logNEI achieving the best performance by effectively accounting for the noise inherent in GA evaluations. The proposed framework identifies hyperparameter configurations that enable a 25-generation GA run to achieve elastic modulus values comparable to those obtained in a full 75-generation optimization. Furthermore, introducing a penalized BO objective significantly reduces the number of required lattices with only minor decreases in absolute achieved elastic modulus, revealing a practical trade-off between performance and the number of structures that must be evaluated. High-fidelity FFT validation verifies the effectiveness of the surrogate-driven optimization strategy. The optimized hyperparameters allow for rapid convergence, eliminate the need for lattice mutation, and reduce the overall computational cost by 24% (from 225 to 171 hours) while preserving mechanical performance. These results demonstrate the potential of multi-fidelity optimization as an efficient and practical approach for GA hyperparameter tuning and future experimental lattice design studies.

The Prasthanatrayi -- the ten principal Upanisads, the Brahmasutra, and the Bhagavadgita, with Sankara's commentaries (bhasya) -- is the foundational corpus of Advaita Vedanta. Continuous euphonic combination (sandhi), long compounds (samasa), and dense scholastic prose make it hard to read at the word level: where one word ends, and what each word means grammatically, are both obscured. We present an open, fully offline, word-level digital reader of the entire Prasthanatrayi with Sankara's bhasya. Every word -- of both the root text (mula) and the commentary -- is clickable and resolves to a pop-up giving its split (padaccheda), morphological analysis, and gloss. Because every word carries a lemma, the reader also acts as a concordance: a search on a dictionary headword retrieves all of that word's inflected and sandhi-hidden occurrences, and its occurrences inside compounds, across both layers. The resource covers thirteen commentarial units (2,971 verses, sutras, and prose sections; 36,881 analysed word-occurrences of root text) and a global dictionary of 95,587 distinct commentarial surface forms. We describe the corpus, the hybrid pipeline -- a rule-based sandhi splitter over an inflected-form lexicon and attested-corpus look-ups, with LLM-assisted analysis under an adversarial two-pass verification protocol -- and a durable human-review loop whose corrections survive every regeneration. An intrinsic evaluation against independent Sanskrit resources finds high-confidence analyses agree with an authoritative inflectional lexicon on over 99% of attested forms, and a band-blind adjudication confirms that quality degrades predictably across confidence bands, with errors concentrated in the low-confidence tier the review loop targets. The reader is a single self-contained HTML file needing no server or network, offered as a freely redistributable teaching and reading aid.

Harmful online communication often contains slang, coded terms, abbreviations, and community-specific expressions, which make messages difficult to interpret. This paper presents an exploratory study of interpretation difficulty in Discord chats related to cybercrime. We construct reference interpretations of purposefully selected difficult messages, which were reviewed by an expert. We then use them to evaluate human and large language model (LLM) interpretations under different context conditions. The results show that local context alone is often insufficient for humans, while external knowledge and extended conversational context substantially improve human interpretation. For LLMs, local context also improves interpretation, and the larger model performs better. We further conduct a qualitative error analysis and propose a preliminary classification of factors that make harmful chats difficult to interpret. These findings suggest that harmful-content analysis should treat interpretation as an evidence-integration problem, rather than as message-level classification alone.

Bubbly flows exhibit complex multiscale dynamics, with deformable bubbles interacting through the surrounding liquid and giving rise to strongly coupled kinematic and morphological behavior. We present BubbleSH, a bubbly flows dataset consisting of transient, three-dimensional bubble-swarm dynamics obtained from high-fidelity direct numerical simulations of bubbles rising in a periodic domain. The dataset provides time-resolved bubble trajectories, velocities, and shape evolution, with bubble morphology compactly represented using spherical harmonics. Designed to be lightweight yet physically expressive, the dataset enables data-driven modeling of bubbly flow simulators where shape deformation and bubble-bubble interactions play a central role. We characterize the dataset with bubble kinematics, morphology, and interaction patterns, and introduce evaluation metrics for both trajectory and shape prediction. The sensitivity of bubble-swarm dynamics to local perturbations makes BubbleSH particularly well suited to generative models that learn distributions over possible future trajectories. We evaluate a permutationally and translationally equivariant probabilistic emulator on BubbleSH given the proposed metrics. Therefore, we establish a compact, high-fidelity dataset and a benchmark for developing and evaluating data-driven models of deformable, chaotic multiphase systems.

Claims about the universality of human concepts have been predominantly assessed through linguistic similarity across languages and cultures. However, words are effective as communication devices because they compress rich experiential variation into shared conventions, potentially obscuring hidden individual and cultural differences in how concepts are mentally represented. Here, we analyse 2.6 billion human-made sketches of common concepts from 236 countries and territories to examine conceptual structure through people's visual imagination. Consistent with recent work on image-based cognition, we find that single concepts unfold into multiple distinct visual exemplars, revealing latent information about similarities and differences in conceptual structure across cultures. This variation is strongest for concepts involving haptic interaction, suggesting that visual imagery reflects variation in embodied experience as much as conventional definitions. Comparing embedding models of sketches with word embedding models across languages, we find that their geometries diverge, with visual representations preserving rich semantic and cultural structure that language models compress. Cross-cultural similarities derived from sketches align 45% more closely with established cultural distances than do text-based measures. Together, these results suggest that patterns of human conceptual universality may depend critically on the modality through which concepts are measured, with large-scale sketching providing a direct, high-resolution probe of conceptual diversity across embodied and cultural dimensions of thought.

In many realistic scenarios, large volumes of time series data are generated with limited or expensive annotations. This limitation makes supervised learning methods difficult to apply and leads to the use of unsupervised approaches capable of discovering meaningful structures directly from raw data. Clustering therefore plays a crucial role in organizing time series into groups that share similar temporal patterns, enabling exploratory analysis and downstream tasks without requiring manual labeling. However, existing deep clustering methods often struggle to capture long-range temporal dependencies or rely on architectures with high computational cost. This paper introduces FMMVCC, a Mamba-based deep clustering framework for time series that leverages state space sequence modeling to efficiently learn temporal representations with linear complexity. Additionally, it utilizes multi-view self-supervised learning with temporal masking and augmentations. Experimental evaluation in 15 benchmark datasets proves that FMMVCC consistently outperforms state-of-the-art baselines, achieving the best overall performance in 29 of 60 total metric evaluations and the highest average rank in all tested scenarios.

Reinforcement learning (RL) enables the synthesis of control policies directly from data, making it highly appealing for complex cyber-physical systems (CPSs) and robotics. A persistent challenge, however, is ensuring strict, hard safety constraints during the active learning phase. In real-world physical systems, violating mechanical limits can cause irreversible damage, necessitating that exploration remains strictly within safe operational regions. We propose a generalized framework that combines the adaptive, high-performance nature of deep reinforcement learning (DRL) with the formal safety guarantees of model predictive control (MPC). Using a mathematical model of the system dynamics, offline MPC computations define a feasible state-action space, representing all safe combinations of system states and control inputs that guarantee constraint satisfaction. During training and deployment, the RL agent's instantaneous actions are projected onto this globally verified feasible set via a safety filter. We systematically evaluate our generalized approach on a non-linear 1-DoF laboratory testbed, demonstrating successful exploration and stable policy convergence on physical hardware.

One of the expected abilities of vision-language models (VLMs) is spatial reasoning ability based on a given text and image. To evaluate the spatial reasoning abilities of VLMs, we focus on the use of spatial deictic expressions, which are defined as spatial expressions whose referent is determined by their situational context, such as ``this'' and ``that''. To handle spatial deictic expressions, VLMs must jointly reason over language and visual space, grounding context-dependent references in the image's spatial structure. In addition, selecting appropriate spatial deictic expressions across languages requires VLMs to understand the language-specific spatial distinctions encoded by these expressions. In this paper, we develop a benchmark to evaluate the multilingual ability of VLMs to use spatial deictic expressions in four languages. Our experiments using this benchmark reveal that the tested models use demonstratives in a manner different from that of humans, particularly in selecting the appropriate demonstratives based on the distance to the object.

Explainability remains a key issue in reinforcement learning (RL). Distilling an interpretable policy from an agent trained in a complex environment is particularly challenging when the action space is continuous. We introduce ORCAID, a novel method for extracting interpretable rule-based policies from RL agents operating in mixed continuous-discrete environments with continuous action spaces. Our main contribution is an efficient oblique decision tree training algorithm that partitions the state space by hyperplanes and fits local linear models. The key idea lies in a three-stage split search: efficient random initialization, local refinement, and backward elimination. Finally, adjacent leaves are merged to yield a concise set of interpretable rules describing a given deep RL policy. We evaluate ORCAID across multiple RL environments, demonstrating that the extracted rule-based policies maintain strong performance with a low number of parameters and can even be used to improve the performance of the original deep RL policy.

Deep generative models offer a promising paradigm for topology optimization, enabling rapid design exploration. However, these approaches lack intrinsic physics guidance, often leading to poor generalizability across unseen boundary conditions and the formation of floating material artifacts. To address these limitations, we propose Hierarchical Physics-Guided Diffusion (HPG-Diff), a novel diffusion framework that enforces physics consistency through two synergistic mechanisms. First, we introduce a hierarchical physics-guided strategy that aligns different precomputed physics features with the denoising process, guiding material distribution toward optimal load paths to enhance generalizability. Second, we propose a floating material suppression loss as a differentiable connectivity constraint inspired by thermal conduction to improve topological connectivity. By simulating a virtual heat propagation process from load positions, this mechanism explicitly penalizes floating material during training. Quantitative evaluations demonstrate that HPG-Diff achieves average compliance errors of 0.87% (in-distribution) and 5.29% (out-of-distribution), while reducing floating material ratios to 2.90% and 2.44%, respectively. Furthermore, case studies on a 3:1 rectangular domain, including cantilever and bridge benchmarks, provide preliminary evidence that lightweight LoRA fine-tuning with a small dataset can support the adaptation of HPG-Diff to rectangular non-square domains.

Diffusion models represent a leading paradigm for graph generation, with notable impact in domains such as molecular design. Yet, scaling these models to large graphs remains an open problem. We approach this question in the dense-graph setting through the lens of graphons, the size-agnostic limit objects of dense graph sequences, to study how structural graph statistics behave across node-size scales. This perspective leads to DiPhon, a diffusion framework for size-scalable graph generation. Specifically, we formulate a continuous diffusion process on the graphon space via a Jacobi stochastic differential equation (SDE), and propose DiPhon, a discretized graph-level process that mimics these dynamics on finite graphs. We further derive the corresponding reverse-time process, which requires access to the marginal score. For the Jacobi process, this score interestingly admits a tractable form, which we estimate from data via graph denoising and plug into the reverse process to generate graph samples. We prove that DiPhon matches exactly the first moment of the marginal distributions induced by the continuous graphon process, and approximates the second moment up to a closed-form discrepancy. Thus, DiPhon inherits key size-agnostic statistical properties of the graphon dynamics, providing a principled route toward scalable graph generation. Empirically, we demonstrate this scalability by training on small graphs and generating progressively larger graphs at inference time, without retraining, while preserving their core topological properties.

Prior work has shown that chain-of-thought (CoT) reasoning is often unfaithful: a model's stated reasoning does not reliably reflect the process that produced its output. Detecting unfaithfulness, though, requires controlled experimental interventions, which cannot be applied to evaluation transcripts after the fact. We turn instead to a more tractable question that has received less attention: whether the stated reasoning is logically consistent with the answer it accompanies. Unlike faithfulness, consistency can be assessed from a transcript alone, with no intervention. We introduce reasoning consistency scanning, a reusable method for detecting this property in AI safety evaluation transcripts. Our contributions are fourfold. First, we formalize reasoning consistency as distinct from faithfulness and define a six-subtype taxonomy of inconsistency. Second, we build a validated benchmark of 60 transcripts, manually adapted from InstrumentalEval outputs. Third, we implement a working scanner for InspectScout, the first to target this property in safety evaluation transcripts. Fourth, we report results across four generator models and three evaluations from inspect_evals, showing that reasoning inconsistency is present, detectable, and varies systematically across both models and task types.

Vision foundation models (VFMs) are increasingly being developed for radiological imaging, yet their definition, development and evaluation remain heterogeneous. We conducted a PRISMAScR scoping review of peer-reviewed studies published between January 2017 and March 2026 describing foundation models trained exclusively on radiological imaging data. Sixty-seven studies were included and mapped across three pillars: data scale and heterogeneity, architectural and pretraining scalability, and downstream transferability and generalization. Datasets primarily covered brain MRI, thoracoabdominal CT, and chest X-ray, ranging from fewer than 100,000 samples to multi-million-image cohorts. Transformer-based architectures and self-supervised pretraining predominated, particularly masked image modeling, contrastive learning and multi-stage approaches. Evaluation focused mainly on segmentation and classification, whereas cross-center, cross-scanner, anatomical and modality-shift validation was inconsistently reported. Alignment with FUTURE-AI principles was uneven. Overall, radiology-specific VFMs show promising transferability, but clinical translation remains constrained by limited data representativeness, heterogeneous benchmarks, incomplete reporting and insufficient deployment-oriented evaluation.

Smart contract compilers are critical to ensuring the correctness of public blockchains whose defining characteristics are open-source and immutable code. We created SolSmith, a semantics-aware differential fuzz testing tool, to improve the quality of the Solidity compiler -- the most popular compiler for the Ethereum blockchain -- and spent over three years finding compiler defects that produce incorrect code. We call these defects miscompilation bugs. During this time period, we have discovered 25 miscompilation bugs that went unnoticed, some for multiple years. Our first contribution is to make compiler testing more rigorous. SolSmith achieves this goal by generating valid test programs that are likely to stress test code generation and optimization components. This helps SolSmith find bugs missed during routine testing that could potentially have serious implications for smart contracts and their users. Our second contribution is a qualitative and quantitative analysis of miscompilation bugs that we found in the Solidity compiler. We classify miscompilation bugs found by SolSmith based on their nature, root-causes, and impact on end-users. This sheds light on some pitfalls of optimizing compilers.

Modern federated and streaming learning systems often release intermediate models, so privacy must hold for the full trajectory under adaptive interaction. Motivated by participation privacy, we study single-edit neighboring user streams, where one insertion/deletion shifts all subsequent updates and defeats standard Hamming-neighbor continual-release analyses. We give an auditable modular recipe. A randomized buffering wrapper emits bins of size $[U,2U]$, reducing single-edit streams to a Hamming-style per-bin update stream with explicit backlog/delay guarantees, where $U$ is calibrated by the privacy parameters $(\varepsilon,δ)$. We then prove a certification theorem identifying when a non-adaptive Hamming-neighbor DP proof for a continual primitive lifts to adaptive inputs: the primitive must use fresh per-round randomness and have a stable one-round privacy profile under common adaptive context. Together, these ingredients yield trajectory-level $(\varepsilon,δ)$-DP for single-edit streams using standard primitives (e.g., tree prefix sums), with an explicit privacy--latency link via $U$.

We analyze how four forces restructure the AI industry over 2026-2030: the DRAM/HBM price surge, frontier-capable open-weight models (GLM-5.2), rapid inference-efficiency gains (near-Shannon-limit KV-cache compression, lightweight local runtimes), and the entry of Meta and xAI into compute resale on fleets bought before the memory repricing. Formulating inference economics in dollars per petabyte of bandwidth delivered (\$/PB) -- model-agnostic for bandwidth-bound decode -- we show the entrant-incumbent cost gap never closes: a depreciation conveyor delivers newly amortized fleets to incumbents faster than hardware prices normalize (3.2x in 2026, 1.9x in 2027, re-widening to 3-4x by 2029-30). Training bifurcates into a luxury tier (\$18-38B per frontier run by 2030) and a mass tier (previous-frontier parity via RL/distillation falling toward \$5M). Solvency of the announced buildout is confined to a corridor requiring roughly 2x annual token-demand growth for four years with sticky premium pricing; a measurement critique shows public token trackers overstate monetizable demand, and all pre-Q2-2026 projections predate the industry's shift from token maximization to token minimization. A vintage-breakeven analysis finds 2026 and 2028-29 capacity each fatally exposed to one pricing regime, with only the 2027 vintage robust. A greenfield custom-silicon entrant removes the merchant margin but not the memory premium (central outcome: 25% success/34% mediocre/41% loss, improvable via staged go/no-go gates). China's LineShine LX2 -- domestic HBM on a standard ISA -- decouples its cost curve from the memory crisis. Scenario probabilities: Rotating Landlord Oligopoly 25%, Commoditization Crash 25%, Jevons Absorption 20%, System-Layer Re-differentiation 18%, Geopolitical Bifurcation 12%. Solvency now depends on monetized bandwidth demand, premium stickiness, and vintage ownership.

Adaptive optimizers mix several mechanisms: a metric or preconditioner maps gradients to descent directions, while estimation, memory, step-size control, constraints, stochasticity, target modification, and discretization determine which directions are available and how they are used. We introduce geometric--nongeometric optimizer calculus, a modular language for auditing reachable gradient methods under explicit oracle, budget, state, and rule constraints. The geometric module is a positive cometric family that maps covectors to parameter-space directions; the nongeometric modules are information, memory, control, operator, noise, target, and discretization mechanisms. The main formal result is a direction-expressivity theorem: away from critical points, full positive-definite geometry expresses exactly the strict descent directions. We then define restricted direction residuals for admissible metric families, prove exact expressivity conditions for diagonal and block geometries, and separate this direction-level diagnostic from condition-number geometric complexity. The resulting design problem is a Pareto optimization over module budgets, not a single universal optimizer ordering. We also lift pointwise residuals to a trajectory-level residual complexity that couples direction mismatch with the variation of the explaining geometry. We include diagnostic prototypes only as evidence for the language: a high-information full-metric probe solves deterministic quadratic benchmarks to numerical precision, while a practical Muon-style PyTorch candidate gives small-scale evidence that matrix-operator updates can be audited by the calculus. The paper is a theory and benchmark-language manuscript; it does not claim large-scale optimizer state-of-the-art performance.

Optimization geometrodynamics views optimizer state as evolving geometry. Its full positive-definite quadratic benchmark gives the least affine-invariant deformation needed to reduce condition number when arbitrary metrics are allowed. This paper records that benchmark in the present notation and develops restricted dynamic geometric complexity: an intrinsic certificate distance for reaching a target condition-number class when the metric is restricted to a specified family. The main proved results are monotonicity and submanifold-distance principles, diagonal and block reachability as linear matrix inequality feasibility problems, an exact two-dimensional diagonal complexity formula, and affine-invariant Kronecker projection theorems with normal equations, computable mismatch certificates, Armijo solver convergence, auxiliary self-conditioned K-target bounds, and Hessian-relative candidate certificates through an exact Kronecker Loewner-sandwich reachability condition, including a Kronecker expression threshold and a fixed-basis exact subproblem. Low-rank spectral models, curvature-proxy inflation, stochastic restricted complexity, discrete geometric length, and expression--estimation--flow--discretization accounting are presented as diagnostic interfaces rather than full optimizer characterizations. The resulting language turns structural preconditioner questions into geometric distance, reachability, and certificate problems. The repository includes deterministic toy and synthetic workflows that check diagonal expression gaps, block primal/dual certificates, Kronecker spectral width, and Hessian-relative Kronecker candidate certificates on small quadratic instances, together with low-rank spectral monotonicity.

Achieving a genuine quantum advantage relies on two distinct non-classical resources that restrict efficient classical simulation: entanglement and magic (nonstabilizerness). We investigate the interplay between these resources by characterizing the Pareto frontiers of extreme magic at fixed entanglement for systems of two qutrits ($d=3$) and two ququints ($d=5$). Unlike the case of two qubits, the Schmidt spectrum for two qutrits features two independent entanglement parameters, resulting in two-dimensional Pareto surfaces. For the lower frontier, we recast the minimal magic as a compact function of concurrence and negativity, with a maximal value of $\ln 2$. For the upper frontier, we determine the maximal stabilizer Rényi entropy to be $M_2 = \ln(81/17) \approx 1.561$, which tightens the previous theoretical bound of $\ln 5\approx 1.609$ and improves on earlier numerical estimates. The maximum magic is achieved at eighteen distinct maxima categorized into three families of six permutation-equivalent spectra. We provide analytical expressions for the maximal magic in the neighborhood of each maximum and for the corresponding maximally magical states which turn out to be Weyl-Heisenberg-covariant fiducial states for mutually unbiased bases. Finally, numerical analysis of two ququints ($d=5$) reveals six permutation-inequivalent maxima with a peak magic value of $M_2 = \ln(625/49) \approx 2.546$. Based on these findings, we conjecture that the maximal magic for a bipartite system of two qudits with prime dimension $d$ is given by $\ln [ d^4 / (2d^2 - 1) ]$, which reproduces the previously known value for qubits, as well as the values derived here for qutrits and ququints.

Across robotics, World Models (WMs) are increasingly used to evaluate action policies by simulating the consequences of actions in an imagined world, and returning a success or safety verdict. Yet a verdict is only as trustworthy as the WM that produced it, and the WM itself needs to be certified. In video-generation WMs, fidelity metrics such as Fréchet Video Distance (FVD) reward visual realism, but ignore whether the world responds correctly to the policy's actions, including those unseen in training. Classical simulation-based validation assumes a trusted simulator evaluating an untrusted policy, whereas generative WMs are themselves unverified learned artifacts. Hence, we argue that any WM used as a test oracle must first be accredited before its verdicts can serve as evidence. Building on credibility practices from safety-critical simulation, including Verification, Validation & Accreditation (VV&A), Safety of the Intended Functionality (SOTIF), and scenario-based testing standards, we define an admissibility ladder (L0-L4) that a WM must climb before its closed-loop verdicts are accepted as assurance evidence. Our framework is embodiment-agnostic, and is instantiated in autonomous driving (AD), where assurance methods for traditional simulation are most mature. Applied to two driving WMs, the lower rungs reveal a reversal: the model that ranks higher on visual generation quality (L0) ranks lower on action-following (L1-L2), so visual fidelity does not predict the action-robustness a closed-loop verdict depends on.

Vision-language models (VLMs) and agentic AI have shown strong performance on semantic visual tasks, but it remains unclear whether they can handle the physics and inverse problems that underlie computational imaging. We present ImagingBench, a benchmark of 20 computational imaging tasks spanning five categories: ray and wave optics, image signal processing, inverse reconstruction, computational sensing, and calibration. ImagingBench evaluates three complementary settings: Expert, fixed expert-guided inverse reconstruction; Planner, planner-guided inverse reconstruction; and Forward, forward-system simulation for consistency checking. We benchmark leading proprietary and open-source image-centric multimodal systems, including Gemini, GPT, and Qwen, and compare them with representative task-specific non-agentic baselines. Across tasks, agentic models remain consistently weaker than specialized methods, especially on computational sensing problems such as lensless imaging, event-based reconstruction, time-of-flight imaging, and holography. Planner guidance provides only modest and inconsistent gains over the fixed-prompt Expert baseline. Although the models often generate visually plausible outputs, their reference-based fidelity remains poor, revealing a substantial gap between semantic visual competence and physically grounded imaging performance. ImagingBench provides a unified testbed for measuring this gap and tracking progress in agentic AI for computational imaging.

Pre-deployment safety evaluations aim to inform the downstream risks of releasing a new AI model. Yet most evaluations provide limited evidence about how often undesired model behavior will occur in deployment: they generally have insufficient coverage, are unrepresentative, and are generally recognizable as tests. To address these concerns, we study a simple way to simulate a model deployment: starting from de-identified conversations from a previous model deployment, we hold fixed the initial conversation prefix and regenerate the next response using a candidate model. The resulting responses can then both be audited for novel misalignments and used to estimate the prevalence of model misbehavior before deployment. We evaluate deployment simulation across four GPT-5-series deployments, using registered, outcome-blinded predictions for GPT-5.4 and retrospective analyses of three earlier releases. We find that deployment simulation produces informative estimates of post-deployment misbehavior rates and outperforms baselines based on adversarially selected production data; its evaluation-awareness point estimates were also much closer to production traffic than those from traditional evaluations. We also identify the realism of tool resampling as a central challenge for further improving predictions and share results suggesting that this challenge is surmountable even in complex tool-use settings. Finally, we show that deployment simulation can be seeded from public chat datasets and remain informative about production misbehavior rates, suggesting a path for external researchers to run deployment-grounded evaluations without access to private production logs. Overall, deployment simulation helps evaluators forecast how language models will behave in the real world and supports more quantitative assessment of deployment risk.

Document Visual Question Answering (DocVQA) presents a complex multimodal challenge, requiring models to exploit visual, textual, and layout information from documents. Although Vision-Language Models (VLMs) have shown remarkable performance in text-vision tasks, their robustness and transferability to different document domains remains underexplored. In this study, we present a comprehensive evaluation of 8 open-source pretrained VLMs on DocVQA in three different document domains: industrial documents of varying type, infographics, and presentation slides. We systematically assess model performance under zero-shot evaluations, fully supervised finetuning with inter- and intra-dataset evaluations, and few-shot learning evaluations of knowledge transfer between domains. Our findings demonstrate that while large pretrained VLMs possess strong zero-shot baselines for structured layouts, their performance strongly decreases on visually complex layouts of infographics and slides. Although parameter scaling is a dominant factor on performance, supervised finetuning yields higher relative gains in smaller architectures. Furthermore, our cross-domain and few-shot experiments show that visual understanding is the main bottleneck for DocVQA, not a lack of knowledge from the VLMs. Using 50 target domain samples, the models finetuned in DocVQA with datasets of different domains rapidly adapt to the target domain documents, even surpassing their fully supervised counterparts in some cases.

Recent breakthroughs of Reinforcement Learning (RL) have highlighted its potential for complex agentic Large Language Model (LLM) tasks. However, existing efforts largely focus on single-task settings, whereas real-world deployment necessitates a generalist agent capable of solving multiple tasks simultaneously. In this work, we identify a critical yet underexplored phenomenon in multi-task agentic RL: different tasks can exhibit exploration-exploitation pace mismatch. Specifically, easier tasks may converge early to low-entropy policies that hinder learning on harder tasks, while harder tasks can, in turn, push easier tasks back toward high-entropy exploration. This back-and-forth interaction creates inter-task entropy crossovers and frequent entropy spikes. Inspired by this observation, we introduce Entropy Pacing Policy Optimization (EPPO) for multi-task agentic LLMs, which coordinates entropy across tasks to stabilize multi-task optimization. At the core of EPPO is a task-wise dynamic clipping mechanism that replaces the fixed clipping threshold in Group Relative Policy Optimization (GRPO) with a task entropy-aware adaptive bound, tightening updates for over-confident tasks while relaxing them for under-explored ones. Experiments on the multi-task agentic benchmarks demonstrate that the proposed EPPO yields results superior to its counterparts.

Logic Equivalence Checking (LEC), a fundamental hardware verification task, is often bottlenecked by synthesis-induced structural perturbations and XOR-dense regions that degrade SAT solver performance. We contend that the modeling of the miter is as critical as the SAT solver itself. To this end, we introduce a miter-aware mapping framework that strategically formulates the problem before solving. By constructing a LUT-based miter -- instead of a traditional, flat netlist -- our approach preserves critical structural correspondence between the two designs while making high-level logic relations explicit. Our framework uniquely integrates three techniques: equivalence-preserving mapping to structurally align the two circuits, Gaussian-guided XOR modeling to algebraically simplify dense arithmetic, and solver-oriented LUT selection to generate a representation optimized for efficient SAT reasoning. Evaluated on comprehensive datasets, our method achieves up to a \textbf{92.1\%} reduction across state-of-the-art SAT solvers. This demonstrates that a solver-aware modeling paradigm, which unifies structural mapping with SAT reasoning, can fundamentally enhance LEC efficiency.

In partially exploratory factor analysis (PEFA), the loading structure and factor numbers are weakly specified. The regularized variational approximation for partially confirmatory factor analysis (PCFA VA) recovers this structure via Bayesian variable selection, using spike and slab priors to assign inclusion probabilities to unspecified loadings. This research introduces a post selection assessment framework for this approach. We convert converged solutions into covariance models using either hard selection (thresholding probabilities into a sparse pattern) or soft selection (retaining them as weights for effective parameter counts). We derive the resulting degrees of freedom, absolute fit diagnostics (RMSEA, SRMR, CFI, TLI), and relative criteria (AIC, BIC, ELBO). To determine factor numbers, we propose a scale free gain rule with a sustained drop guard. Simulations show absolute indices successfully track loading recovery and flag under factoring. While raw criteria over factor, our gain rule accurately recovers true dimensionality, with the ELBO variant proving most robust. Finally, a 100 item PID 5 example demonstrates that our model fits better than a confirmatory 25 facet model and concordantly recovers major structures across disjoint specifications.

Different optimizers have different update biases, but these biases are usually implicit. Existing studies mainly analyze or control such biases from the geometry of the final solution. However, how optimizer bias forms during training still lacks a clear internal mechanism. This paper proposes an information allocation dynamics perspective. It interprets optimizer implicit bias as the relative allocation of training signals between weight-like and bias-like parameter pathways. This allocation can be described and adjusted by a continuous preconditioning exponent \(p\). To characterize this mechanism, we first analyze the update contributions of weight and bias to the same residual signal in a minimal linear model. The weight correction term preserves input-dependent residual signals, while the bias correction term preserves the residual mean direction. They therefore correspond to different projection pathways of the residual signal. After substituting the preconditioned update into the residual update equation, the optimizer can change the relative strength of the weight correction term and the bias correction term through different preconditioning factors. Therefore, optimizer implicit bias is not only reflected in the final solution or the global training trajectory. It is also reflected in the relative write-in ratio of training signals across different parameter pathways. Overall, this paper moves the analysis of optimizer implicit bias from solution-space geometry to update dynamics during training. It reveals that the relative update allocation between weight and bias-like parameters is an important dynamical mechanism that affects parameter trajectories and generalization behavior.

The Internal Waves Service screens the Sentinel-1 Wave-mode archive for internal solitary waves, routing detections to experts whose adjudication time is the resource the effort exists to conserve. Because attention is the cost of error, precision leads. Its classifier was trained and reported at a one-to-one class balance, fixed before the operational rate could be known. That rate has since emerged at roughly one scene in twenty, and a balanced-test score badly overstates the precision a validator meets. A model that scores 0.794 balanced-test precision scores 0.192 in real operation: the gap is a systematic artefact of reporting at the wrong prior, invisible to the metric most work quotes. We show the mismatch to be an evaluation problem in the costume of a training one at a fixed recall, prior correction and calibration cannot move precision, and answer it with a prior-matched reporting method based on three figures: balanced-test, operational-prior, and real post-deployment, whose contrast is the honest measure. A precision-first, leakage-controlled development cycle then improves the classifier lever by lever, each promoted only against a pre-registered margin; added capacity not clearing it, calibration inert, feature aggregation the one real lift, so the honest negatives are as much a result as the gain. Holding recall at a floor of 0.80 and certifying against a sealed, single-read lockbox, the promoted model reports 0.927 precision at the operational prior; an out-of-time check confirms discrimination transfers to unseen periods while a fixed operating point does not. Prior-matched reporting, begin balanced, then move to the prior as the stream reveals it, transfers to any operational Earth-observation service bootstrapping a rare-event detector under a prior it has yet to discover.

The key-value (KV) cache dominates the memory cost of long-context autoregressive inference, and a growing body of work compresses it through quantization, eviction, or offloading. We study a complementary question: once a position's KV state has been quantized to codebook indices, how should the resulting symbol stream be stored, and can the storage layer do more than store? A family of contractive iterated-map codes that serialize a symbol sequence into a sequence of low-dimensional real vectors is revisited, and it is shown that they form a natural archive format for a quantized KV cache with the following features. The method provides exactly the access pattern a growing cache requires. It is lossless, it runs in linear time, and supports O(1) random access and O(1) amortized append. A controlled study of the quantizer feeding this archive is conducted on GPT-2 with 1024-token contexts. Keeping a small exact window (4 attention sinks + 32 recent tokens) and archiving the rest, per-head residual vector quantization reduces the archived cache by 36-54x relative to an fp16 cache at a perplexity cost of 11-15%, and we quantify a sharp key/value asymmetry -- quantizing keys is roughly 4x more damaging than quantizing values, consistent with prior low-bit KV work -- and use it to allocate bits in a hybrid scheme. Finally, we show the archive is simultaneously a search index: approximate substring queries execute directly on the stored vectors, and matched context is decoded from the matched vector without ever materializing the surrounding text. We release all code; every number reproduces from a single command on a laptop CPU.

Newly developed items must ordinarily be field tested before their psychometric properties are known, creating a cold start problem for item calibration. Predicting item parameters from features is a long standing measurement problem dating back to the Linear Logistic Test Model; modern text embeddings now automate the design matrices traditionally specified by hand. We propose an evaluation framework combining regularized regression on item text embeddings, repeated cross validated R squared reported with its resampling standard deviation, and two performance upper bounds: a reliability ceiling derived from parameter standard errors, and a design ceiling derived from simulation based power calibration. Applying this framework to a mathematics item bank (EEDI) and a medical licensure benchmark (BEA 2024), we find that item difficulty is highly predictable from text (repeated cross validated R squared = 0.53, or about 57% of its reliability ceiling), whereas discrimination and pseudo guessing appear less predictable. However, evaluating these results against our ceilings reveals that this apparent hierarchy stems from target reliability rather than text signal strength: text uniformly recovers 57 to 63% of the reliable variance across difficulty targets, whereas the 3PL pseudo guessing parameter has a reliability ceiling near zero, making it an unviable target at current precision. On BEA, embedding based regression matches leaderboard RMSE despite explaining almost no variance, highlighting the critical need for scale free metrics and explicit ceilings in benchmarking. Finally, we show that a single train and test split can inflate apparent accuracy by 0.1 to 0.15 in R squared, underscoring the necessity of repeated cross validation for calibration support applications and future benchmark construction.

Language models perform a wide range of tasks at varying levels of abstraction with the capacity to flexibly infer tasks from context, execute multiple tasks simultaneously, and select among competing tasks. To study the role of model components in task behaviour, their causal influence can be investigated through interventions. Prior work on model steering has largely focused on interventions along global directions in activation space, modeling task representations as approximately linear and additive. By studying interventions at the neuron level, we find substantial, neuron-specific nonlinear effects on model outputs that are not captured by current steering approaches. We introduce Distributed Sparse Interventions (DSI), an intervention approach that considers nonlinearities and interactions between neurons across layers to identify sparse sets of neurons that elicit task-relevant computations. Across a range of tasks, we demonstrate that DSI can activate task behaviour in instruction-tuned language models by localising and intervening on as few as 0.01% of neurons, highlighting the effectiveness of sparse, distributed interventions in the neuron basis. Additionally, adopting a set-based perspective enables computations over the identified neuron sets, offering insights into the roles of individual neurons by analysing their effects across tasks. Through sparse interventions, DSI enables fine-grained control over model behaviour, localisation of task-relevant neuron sets, and furthers our understanding of task composition.

Lattice field theory is the workhorse of non-perturbative physics, used to simulate phenomena from the strong nuclear force to critical phenomena in materials. Its Boltzmann distributions are parametrized analytically by coupling constants, but these bare parameters are weak predictors of observables -- extracting physics typically requires extensive simulation. While normalizing flows have emerged as effective samplers at fixed couplings, it remains difficult to interpret what these networks have learned. This raises a natural question: can the physics be read off directly from the flow network parameters themselves, and can those parameters be generated for unseen theories? We propose lattice field theory as a testbed for neural network interpretability: because the target physics is qualitatively well-understood and smoothly varying, it provides ideal synthetic data with known ground truth. To this end, we introduce JEPAWG, a Joint-Embedding Predictive Architecture-based Weight Generator that maps couplings directly to flow weights via a learned latent space. On a scalar theory at lattices of size $6^2$ to $11^2$, the JEPAWG latent space recovers the correct intrinsic dimension of the underlying manifold, locates the phase transition, and encodes a finite-size shift aligned with the 2D Ising exponent $ν\approx 1$, allowing us to uncover physical structure by studying the network weights alone. This suggests the fascinating idea of treating the network weights as a new type of physical observable. As a generator, JEPAWG also interpolates and extrapolates to unseen couplings effectively and remains robust to weight-space discontinuities introduced by multi-seed training data, outperforming PCA, AE, and VAE baselines.

In text-to-image in-context learning (T2I-ICL), a model has to infer a latent compositional pattern from fewshot demonstrations for generating a query image. Recent studies show that state-of-the-art multimodal large language models struggle with this setting, particularly due to limited compositional reasoning and sensitivity to prompt construction. In this work, we propose a Tree-of-Thoughts (ToT) reasoning framework for T2I-ICL that introduces a multi-stage reasoning and selection layer that generates, evaluates, and selects among multiple candidate hypotheses before constructing the final prompt for image synthesis. By exploring alternative reasoning branches and selecting a coherent interpretation, the proposed approach mitigates prompt ambiguity and compositional errors. We implement the proposed approach in a complete ToT-T2IICL inference pipeline and evaluate it on the CoBSAT benchmark. Both qualitative and quantitative results show that structured multi-branch reasoning leads to more consistent and semantically aligned image generation compared to baseline and Chain-of-Thought prompting strategies, without any additional training or fine-tuning.

Safe autonomous driving requires both rapid responses to common high-risk events and deeper reasoning over rare, extreme long-tail scenarios in traffic safety. These scenarios are severely under-represented in naturalistic driving data, and existing trajectory and language-augmented datasets seldom provide high-risk event labels, semantic annotations, and verifiable safety signals. Here we present K-Risk, a knowledge-augmented dataset that combines structured driving trajectories with large language model generated semantic annotations for safety-critical driving scenarios. K-Risk integrates 20 human-driven and autonomous-vehicle trajectory datasets from Europe, China, and the United States, covering highways, urban freeways, intersections, and roundabouts. Using a unified risk-centric extraction pipeline, K-Risk curates 31,398 high-risk events, together with a 1,036-event extreme subset of near-collision cases. Each event is released as a synchronized trajectory, metadata, and language triplet containing structured scenario descriptions, abnormal-behavior notifications, and, for a representative subset, causal risk analyses and action recommendations validated through a closed-loop simulator with iterative reflection. By combining multi-dimensional risk annotations, interpretable language supervision, and verifiable decisions, K-Risk bridges structured traffic trajectories, semantic reasoning, and decision supervision, providing a standardized foundation for developing and evaluating next-generation risk-aware autonomous driving agents.

Proprioception is fundamental to robotic manipulation, yet standard fusion methods often treat it as an isolated vector lacking explicit alignment with visual tokens. Without a direct correspondence between 3D kinematics and 2D feature maps, manipulation policies struggle to ground the robot's state within the scene, frequently underperforming even vision-only baselines. To address this, we introduce GeoProp, a lightweight, plug-and-play adapter that aligns proprioception with vision through explicit geometric grounding and spatial feature sampling. GeoProp projects the robot state onto the image plane to sample localized visual features, constructing a grounded state token. It then injects state-derived spatial priors into the corresponding visual features via FiLM modulation. To capture motion intent, GeoProp further samples features at a short-horizon predicted coordinate derived from recent kinematics, providing look-ahead visual context. Across 67 tasks, GeoProp improves Diffusion Policy by 8.7% on 63 simulation tasks and pi_0 by 4.0% on the RoboTwin subset, and yields a 10.6% average gain across both policy families in the real world, while adding only 2-3% to the parameter count. These results demonstrate that GeoProp is a simple yet high-impact inductive bias for generalist embodied policies. Project page: https://alibaba-damo-academy.github.io/GeoProp/.

Safety evaluations of multi-agent LLM systems often compare a direct prompt with a planner-executor pipeline and report the difference as a single "pipeline effect." We argue that this aggregate is difficult to interpret because it conflates three mechanisms: harmful intent may be reframed as plausible operational work, the planner may refuse or transform the request, and the executor may act under delegation prompts implying prior approval. To separate these factors, we introduce a five-condition controlled contrast design, evaluated on 30 synthetic harmful scenarios and an exploratory external validation set from four agent-safety benchmarks using LLM-judged compliance. Our results show that aggregate pipeline safety is not a stable architectural property. Operational reframing is the most portable risk signal, increasing compliance for GPT, Gemini, and DeepSeek across both scenario sets, while Claude is comparatively resistant. Planner behavior can offset this risk mainly through refusal; however, when the planner produces executable steps, the executor may become more compliant than under the direct operational baseline. Approval-framed delegation is sensitive to prompt design, model pairing, and scenario source, and a skeptical executor prompt sharply reduces compliance. Raw-direct model rankings can also mispredict deployed planner-executor behavior. Gemini is safest under raw direct prompts in the primary set yet shows the largest amplification with a Claude planner, rising from 8.9 percent to 38.9 percent compliance. GPTs near-zero aggregate pipeline effect instead hides a reframing increase canceled by planner refusal. These findings suggest that multi-agent safety evaluations should report reframing, planner behavior, delegation framing, and model pairing separately before attributing failures to architecture itself.

Chiplet-based DNN accelerators provide a scalable path to balance performance and yield for modern AI workloads. However, such systems face critical challenges in area and thermal constraints. Design space optimization should jointly consider fine-grained task modeling, chiplet granularity, core granularity, and critical physical constraints. To the best of our knowledge, this is the first framework that involves all these factors. In this work, we propose ThermoDSE, a thermal-aware and comprehensive design space exploration framework for chiplet-based DNN accelerators. ThermoDSE integrates existing fine-grained modeling techniques into a unified simulation and optimization framework that jointly considers architecture design, task orchestration, and inter-chiplet communication under strict thermal and area constraints. Experimental results show that ThermoDSE achieves up to 3.5x improvement in Energy-Delay-Inverse-Yield, defined as E times D times inverse Y, compared with state-of-the-art Simba and other baselines. Furthermore, relative to simulated annealing and reinforcement learning-based methods, ThermoDSE converges to better design points with 3.7x and 29.4x runtime speedups, respectively.

In longitudinal Alzheimer's disease (AD) diagnosis support, clinical and imaging information is often collected at irregular visits. Integrating these multimodal observations may improve diagnostic assessment, but naive fusion can degrade performance when MRI is noisy or intermittently unavailable. We propose AT-Attn, a temporal-aware multimodal framework that combines Change-and-Time encoding, time-biased asymmetric cross-attention, and gated fusion to integrate MRI with longitudinal clinical information. We evaluate AT-Attn on an MRI-retained ADNI cohort of 1,520 patients using structural MRI, six cognitive-scale trajectories, and seven static clinical variables under patient-level five-fold cross-validation. The main asymmetric AT-Attn model achieves accuracy 0.719+/-0.024, macro F1 0.721+/-0.023, ROC-AUC 0.873+/-0.013, and PR-AUC 0.783+/-0.018, outperforming unimodal and naive multimodal fusion baselines while remaining competitive with strong tabular baselines. These results suggest that a temporal-aware and constrained fusion strategy can help structural MRI contribute clinically relevant complementary information for patient-level AD diagnosis support.

Relational Deep Learning (RDL) has become a standard methodology for machine learning on relational databases: the database is encoded as a heterogeneous temporal graph in which tuples become nodes and primary-key to foreign-key (PK-FK) dependencies become typed edges, over which a graph neural network is trained for downstream prediction. We study the adversarial robustness of this pipeline. We consider a white-box attacker who knows how the graph is built and the model is trained, reasons about perturbations on the graph, but can only act on the upstream database, by rewiring foreign-key references while preserving the integrity constraints of the schema (foreign-key validity, the degree-one FK constraint, and functional dependencies). This restricts the attacker to a constrained, combinatorial set of admissible edits under a global perturbation budget, which is intractable to explore exhaustively and made non-additive by GNN message passing. We investigate seven attack heuristics - two random sampling baselines and five gradient-guided variants that exploit differentiable edge masks - and evaluate them on the RelBench rel-f1 benchmark. Gradient-based attacks consistently outperform random baselines on regression tasks, whereas gains on classification are smaller, which we attribute to low label-flip rates and greater local stability of classification outputs.

The Adaptive Data Analysis (ADA) problem formalizes the challenge of preventing false discovery and overfitting when a dataset is repeatedly reused. Formally, our input is a dataset containing $n$ i.i.d. samples from an unknown distribution $\mathcal{P}$ over a domain $\mathcal{X}$, and our goal is to answer a sequence of $k$ adaptively chosen statistical queries with respect to $\mathcal{P}$. The main question is how many queries we can support (i.e., how large $k$ can be), primarily as a function of the number of samples $n$. This question has been intensively studied and is relatively well-understood for randomized mechanisms: there are computationally efficient mechanisms that support $k \approx n^2$ queries, and no computationally efficient mechanism can answer $k \gg n^2$ queries. In this paper, we address a fundamental question: is randomness necessary for ADA? Despite a decade of work on ADA, this question remains open. A folklore observation dating back to the initial works on ADA is that randomness is not necessary when the analyst is computationally bounded. Yet, the necessity of randomness against computationally unbounded analysts has remained elusive. Our main contribution resolves this gap in the information-theoretic Random Oracle model. Perhaps surprisingly, we show that randomness is strictly necessary to answer a non-trivial number of adaptive queries: when the analyst is unbounded, any deterministic mechanism can be forced to fail after just $k = \tilde{O} (n)$ queries.

Subsampling significantly reduces the number of measurements, thereby streamlining data processing and transfer overhead, and shortening acquisition time across diverse real-world applications. The recently introduced Active Deep Probabilistic Subsampling (A-DPS) approach jointly optimizes both the subsampling pattern and the downstream task model, enabling instance- and subject-specific sampling trajectories and effective adaptation to new data at inference time. However, this approach does not fully leverage valuable dataset priors and relies on top-1 sampling, which can impede the optimization process. Herein, we enhance A-DPS by integrating a deterministic (fixed) prior-informed sampling pattern derived from the training dataset, along with group-based sampling via top-k sampling, to achieve more robust optimization, method we call Prior-aware and context-guided Group-based Active DPS (PGA-DPS). We also provide a theoretical analysis supporting improved optimization via group sampling, and validate this with empirical results. We evaluated PGA-DPS on three tasks: classification, image reconstruction, and segmentation, using the MNIST, CIFAR-10, fastMRI knee, and hyperspectral AeroRIT datasets, respectively. In every case, PGA-DPS outperformed A-DPS, DPS, and all other sampling methods.

Functional brain networks exhibit a hierarchical organization across ROI, community, and whole-brain levels, supporting local processing, inter-community coordination, and global integration. Recent studies have demonstrated that brain community-aware modeling is beneficial for both diagnosis and biomarker identification of brain networks. However, existing brain graph modeling methods often struggle to model ROI-community interactions, thereby failing to fully exploit the hierarchy across ROI, community, and whole-brain network levels. To address this issue, inspired by deep hyperbolic learning in modeling hierarchical structures, we propose a novel framework, termed Hyperbolic Learning on Brain Graphs (HLBG), for brain network analysis. The core idea of HLBG is to exploit the inherent hierarchical geometry of hyperbolic space to model the hierarchical relationships among ROIs, functional communities, and the whole-brain network, thereby learning hierarchy-aware and highly discriminative representations for brain network data. Specifically, HLBG first projects representations from ROIs, communities, and the whole-brain network into Lorentzian hyperbolic space. Then, the multi-level hierarchy is imposed via two geometric entailment constraints. In addition, we introduce a new Graph-aware Mamba (GaMamba) model, which incorporates topology-derived structural prompts into Mamba to capture long-range dependencies while preserving graph topological information. Experiments on ABIDE-I and REST-MDD demonstrate that HLBG outperforms state-of-the-art methods and identifies disorder-relevant functional biomarkers.

Quantum diffusion models provide a physics-consistent route to generative learning by formulating noising and denoising directly on quantum states. However, applying such models to classical high-dimensional data is constrained by the qubit cost of state encoding and the computational burden of simulating large density operators. We propose a scalable hybrid generative pipeline that combines a classical autoencoder for dimensionality reduction with a mixed-state quantum denoising diffusion probabilistic model (MSQuDDPM) operating in the learned latent space. The autoencoder compresses data into compact latent codes that can be embedded into a small-qubit Hilbert space, after which the quantum diffusion model learns a generative distribution over latent density operators and decodes samples back to the original domain. Algorithmically, we simplify the reverse dynamics by predicting an estimate of the clean state $ρ_0$ at timestep $t$ and computing the one-step reverse update via an analytic backward propagation rule, rather than learning an explicit predictor for $ρ_{t-1}$. We demonstrate the proposed approach on MNIST image generation and discuss how mixed-state quantum diffusion can serve as a practical backbone for hybrid quantum--classical generative modeling under realistic qubit budgets.

Transformers have demonstrated a remarkable ability to learn algorithmic reasoning, yet mechanistic analyses have mostly focused on globally invertible operations such as cyclic addition and group composition. In this work, we investigate how small transformers learn modular integer multiplication over composite moduli, a fundamentally non-invertible operation due to the presence of zero-divisors. We propose the monoid extension: a localized generalization of Group Composition via Representation (GCR) that suggests the learned computation does not rely on a single global representation space. Instead, the model partitions the input space into local hierarchical algebraic regions, where group-like structure survives and Fourier mechanisms can be applied. In transformers trained on square-free modular multiplication, we find that embeddings organize around these regions, attention exhibits class-sensitive routing and low-rank write directions, and local character features explain a large fraction of the model's output logits. Our results suggest that representation-theoretic mechanisms previously identified for group operations can extend beyond groups to more general structures.

Monero is a privacy-focused cryptocurrency that deploys the Dandelion++ protocol and incorporates anonymity networks (such as Tor and I2P) to prevent malicious attackers from linking transactions with their source IPs. In this paper, we demonstrate that Monero's integration of the Tor network introduces a fundamental vulnerability: a Monero Tor node's originated transactions are exclusively forwarded to two outgoing Tor hidden service nodes (proxy nodes) prior to clearnet propagation, enabling an adversary to capture originated transactions by occupying the target node's outgoing connections. Based on this observation, we propose \textit{ProxyMark}, a three-stage deanonymization framework for the Monero Tor network, comprising node role identification, originated transaction identification, and node location deanonymization. Through experiments on the live Tor network, Monero mainnet, and testnet, we empirically demonstrate the effectiveness of \textit{ProxyMark} in successfully deanonymizing transactions originating from Monero nodes over Tor.

Inherently interpretable classifiers for tabular data typically rely on sparse features, rules, or patterns that users can inspect directly. The marginal feature-screening step common to these methods can discard variables whose predictive value emerges only through joint configurations with other variables. We present Interaction Aware Interpretable Machine Learning (IAIML), a framework that addresses this limitation through three coordinated mechanisms: adaptive per-feature discretization, finite-grid pairwise interaction scoring, and a partitioned explanation budget. Detected interactions are routed through one of two strategies: relaxing the screening filter so that interaction-supported variables enter the pattern search, or constructing explicit pair terms for a sparse downstream classifier. On a 40-dataset panel comprising 24 real-world tabular benchmarks and 16 synthetic interaction stress tests, evaluated under nested cross-validation, IAIML achieves mean AUC within 1.4 points of tuned gradient-boosted ensembles while requiring roughly 14--28 times fewer fitted explanation components. On datasets with strong pairwise interaction structure and low marginal signal, IAIML outperforms all baselines. Among compact interpretable methods, IAIML is comparable to RuleFit in AUC and component count and is less expensive to tune. EBM obtains a small but significant AUC advantage across the full panel, with a substantially larger lookup-table footprint. Performance degrades on datasets requiring higher-order interactions beyond the pairwise scope. Component-isolated ablations confirm that adaptive discretization and interaction-aware admission each contribute incrementally. These results support IAIML as a compact, interaction-aware framework appropriate for settings where bounded explanation size and controlled treatment of feature interactions are design requirements.

AI agents deployed for IT operations are typically permanent cost centers because every execution requires full LLM inference, even for previously solved problems. This paper introduces progressive crystallization, a lifecycle that treats agent exploration as a discovery mechanism rather than a permanent execution model. It defines a three-stage execution taxonomy, from fully agent-orchestrated to hybrid to fully deterministic workflows, together with an evidence-based promotion mechanism that converts repeatedly validated agent behaviors into cheaper and more reproducible deterministic workflows, while automatically demoting workflows that regress. Evaluated on a production cloud networking AIOps system processing tens of thousands of incidents per month, the approach increased deterministic execution from 0% to 45% over eight months, reduced per-incident agent costs by more than 70% despite doubling incident volume, and improved safety through greater reproducibility and auditability. The paper also presents the execution taxonomy, promotion and demotion criteria, trace extraction methodology, economic model, safety considerations, and discusses limitations and threats to validity.

Conversational image editing requires preserving not only visible content, but also content that temporarily disappears across turns. When newly added or modified content occludes a previously visible region, that region should reappear if it was never semantically changed. However, existing systems often fail to recover such occluded-but-unchanged content, producing inconsistent or hallucinated results. We introduce OCCUR-Bench, a diagnostic benchmark for temporal preservation in conversational image editing. OCCUR-Bench provides diverse occlusion-and-revelation scenarios with historical restoration references, enabling evaluation of faithful restoration rather than plausible regeneration. We also propose ReSpec, a training-free framework that makes implicit preservation explicit by pairing restoration-aware instructions with historical visual references. Given an editing history, ReSpec identifies what should persist, selects the historical image state that provides missing visual evidence, and conditions an in-context editor on the resulting instruction and reference image. Experiments show that ReSpec improves restoration fidelity and temporal consistency on OCCUR-Bench, highlighting the need to ground preservation in editing history rather than only the current image.

Agentic language models must learn when to call tools, when to consume tool responses, and when to answer directly. This makes multi-teacher on-policy distillation a natural training strategy: one teacher can specialize in tool calls, another in direct responses, and the student can learn from both on its own generated distribution. We show that this strategy can induce a behavior shift that is invisible from aggregate losses alone. In a two-teacher tool-use setting, vanilla generalized knowledge distillation improves tool-call recall but also moves the model toward over-calling, where it calls tools on examples that should be answered directly. Aggregate explanations are insufficient: tool-call samples do not receive more token exposure, and full-sequence per-token divergence is not larger for the tool-call teacher. We instead analyze behavior leverage imbalance: local token-level signals at mode- entry and structural positions, such as <tool_call> and function names, can have disproportionate control over the global generation mode. We propose Soft Clamp, a per-token divergence calibration method that dynamically compresses extreme token-level Jensen-Shannon divergence while preserving nonzero gradients. On APIGen-MT, Soft Clamp reduces over-calling from 13.7% to 9.0% relative to vanilla GKD while matching its decision accuracy. In a BFCL multi-turn diagnostic, it also lowers tool-call loops and repeated calls among GKD variants. These results suggest that multi-teacher OPD should monitor where teacher signals act, not only how large they are in aggregate.

Understanding the geometric structure of pre-trained language model embeddings matters for interpretability and safety. We ask whether sentence-level classification signal lives in the Riemannian geometry of contextual token embeddings, and probe it by extracting per-token pullback metrics from a learned encoder's analytical Jacobian and aggregating them with the Fréchet mean on the symmetric positive definite (SPD) manifold; we call this procedure Riemannian Mean Pooling (RMP). Across three datasets with non-trivial linguistic structure (CoLA, CREAK, RTE), RMP outperforms Euclidean mean pooling, while on FEVER-Symmetric, a benchmark constructed to remove annotation-driven lexical artifacts, the method correctly stays at chance. Ablations show that a randomly initialised encoder combined with Fréchet aggregation already beats Euclidean pooling on two of the three signal-bearing datasets, localising the source of the gain to the geometric aggregation rather than to learned manifold structure; the trained encoder contributes additional signal specifically on CREAK, the most knowledge-heavy of the three signal-bearing datasets.

Large language models (LLMs) are widely used in intelligent services due to their remarkable capability in generative tasks. Typically, LLM-based services process the inference requests of the users in a centralized data center. Unfortunately, such centralized execution has limitations for end-users, such as increased response latency with communication overhead and privacy leakage risk. To alleviate the aforementioned limitations, there have been increasing pushes to execute LLM inference locally on user-end devices. However, the limited resources of a single edge device impose restrictions on achievable accuracy of LLMs. To overcome the issue, we first propose to leverage multiple user-end devices available at the edge for LLM inference, enabling the execution of larger models. Specifically, we propose Voltron, a novel on-device LLM inference framework that elastically utilizes multiple user-end devices for LLM inference execution while adapting to diverse real-world edge environments. In our evaluation, Voltron achieves up to 16.5% higher accuracy than state-of-the-art LLMs that can be executed on a single edge device, satisfying user QoS requirements.

How can we measure intelligence beyond human capability? Human-authored benchmarks saturate, and above human capability, examiners may not know which tasks are both hard and verifiable. We argue that this difficulty is inherent to absolute-scale evaluation and propose a new paradigm based on relative measurement in which models generate public challenges that separate other systems. Aggregating these outcomes yields an adversarial psychometric rating system that can scale with the systems being measured. We describe practical protocols that reduce incentives for private-information attacks, support judge-free adjudication, and naturally scale with agent capabilities. We instantiate the framework across verifiable and open-ended, non-verifiable domains, illustrating how model-generated evaluation can continue to measure systems beyond the human frontier.

The architecture of deep feedforward neural networks is ubiquitous in deep learning, either as a whole system or as a subnetwork of other architectures, and thus its mechanism is a key ingredient of the black box of neural networks. On the basis of the simplest two-layer ReLU network, this paper systematically studies the mechanism of deep feedforward ReLU networks with multiple hidden layers and successfully explains the training solution obtained by the back-propagation algorithm. The concept of a path, especially in terms of the relationships between paths, plays a central role in uncovering the mystery of the black box. It is shown that a unit of a deep ReLU network can form a piecewise linear manifold to divide the input space, instead of a hyperplane of the two-layer case. How to efficiently use the hidden-layer units to produce both linear functions and partitions of the input space is also a central problem. The principles of a two-layer ReLU network can be generalized to the deeper case to a large extent, such as multiple strict partial orders and continuity restriction. The combination of the basic and simple principles proposed can yield complicated instantiations including the training solutions, and in this sense the black box of deep feedforward ReLU networks is revealed.

We introduce Intrinsic Green's Learning (IGL), a framework that models a target function on a manifold as the solution to a linear PDE whose source term is learned from data. Rather than approximating the target directly, IGL learns a source and integrates it against a Green's kernel. An encoder discovers a low-dimensional coordinate chart on the manifold where both the source and the kernel decompose as low-rank tensors, collapsing a high-dimensional integral into independent one-dimensional integrals with cost linear in the intrinsic dimension. A two-stage algorithm separates coordinate discovery from source fitting, a near-convex linear solve, preventing the dimensional collapse of joint training. Learnable gates on each coordinate automatically discover the intrinsic dimension of the manifold. We validate IGL on synthetic manifolds and on MNIST, where it simultaneously achieves near-optimal classification and automatic recovery of the intrinsic dimension.

Large vision-language models incur substantial inference costs because high-resolution inputs introduce thousands of visual tokens, many of which are redundant for a given query. Existing pruning methods often combine query relevance and token diversity, yet these objectives can conflict under aggressive compression: relevance-driven selection may overconcentrate the budget on correlated local evidence, while diversity-driven selection may suppress indispensable tokens or retain distinct but uninformative regions. We introduce AnchorPrune, a training-free framework that first constructs a protected relevance anchor and then expands it with complementary visual context. AnchorPrune adaptively determines the anchor size from the novelty profile of relevance-ranked tokens, preserving a compact set of query-critical evidence, and allocates the remaining budget through importance-weighted novelty to recover informative, non-redundant context relative to the anchor. This ordered design prevents contextual expansion from displacing indispensable query cues while improving overall visual coverage. AnchorPrune is lightweight, architecture-aware, and requires neither retraining nor model modification. Across image and video vision-language models and benchmarks, it consistently improves the accuracy-efficiency trade-off over training-free baselines, particularly under severe compression. On LLaVA-NeXT-7B, AnchorPrune preserves 97.6% of full-token performance using only 160 of 2,880 visual tokens. These results establish relevance-anchored contextual expansion as an effective principle for efficient multimodal inference. Code is available at https://github.com/MULTI-cau/AnchorPrune.

Spectral positional encodings (PEs) for \emph{directed} graphs face two obstacles: magnetic Laplacians require an $O(n^3)$ Hermitian eigendecomposition per potential, and their complex eigenvectors are defined only up to unitary gauge, which prior work handles with basis-invariant architectures. We propose learnable spectral PEs of the form $h_θ(A_q)\,R$, where $A_q$ is a normalized magnetic operator, $h_θ$ a learnable scalar spectral response, and $R$ a block of random probes. Because the PE is a \emph{matrix function} of the operator, it is gauge-invariant by construction. We compute it in a Hermitian block Krylov subspace from sparse matrix--vector products only, prove that $k = O(\log(1/\varepsilon))$ block steps suffice uniformly over heat--resolvent response families, and give a covering-number argument for why low-dimensional structured families generalize where free per-eigenvalue weights overfit. On a directed SBM whose symmetrization is uninformative by construction, direction-blind PEs stay at chance while magnetic Krylov PEs converge to the exact-eigendecomposition oracle as the depth grows. The same probes yield gauge-invariant pairwise features with $1/\sqrt{s}$ Monte-Carlo error, and the undirected $q{=}0$ case improves heterophilous benchmarks over no-PE and polynomial baselines.

Reinforcement learning (RL) policies can be unsafe and vulnerable to attacks. Ensuring their reliability is often a pain point as existing automated testing methods target only selected environments, testing scenarios, and RL algorithms. To address this, we propose a comprehensive framework for testing single- and multi-agent RL policies under varying conditions. Our implementation of this framework, Gimitest, is an open-source tool that supports various gym frameworks and allows for modifications of their integrated components. This article describes the framework and details Gimitest's functionality and architecture. It showcases its effectiveness in testing multiple RL policies in environments such as the official Farama Gymnasium and PettingZoo.

While generative models enable encoding of complex neuroimaging data for feature generation and reconstruction, developing optimal architectural frameworks with appropriate encoding and latent space processes is crucial for studying structural and functional properties of the brain. We design a multimodal generative framework for structural and functional magnetic resonance imaging (MRI) features through systematic evaluation of encoding strategies, latent multimodal fusion, and generative model selection. Using structural gray matter volume (GMV) and static functional network connectivity (sFNC) features from a large neuroimaging dataset, we analyze generative frameworks involving variational autoencoders (VAEs), transformers, generative adversarial networks (GANs), and diffusion models. Architectures that employ modality-aware graph encoding of functional connectivity into a lower-dimensional latent space outperform vectorized encoders or direct data space approaches. The proposed multimodal graph VAE (gMMVAE) surpasses alternative generative variants across multiple metrics for generation fidelity, reconstruction quality, efficiency, and latent space discriminability, highlighting its potential for robust multimodal neuroimaging analysis.

Constrained decoding is essential for serving LLMs, ensuring that generated outputs follow specific structures such as JSON schema-formatted function calls. Existing systems are designed for autoregressive models and assume left-to-right generation, masking out invalid next tokens at each step. Diffusion language models, however, break this assumption: they sample multiple positions simultaneously from a fully-factorized mean-field distribution at each denoising step. In this paper, we present an exact and tractable algorithm for sampling from the constrained mean-field posterior under any constraint expressible as a finite automaton. Viewing finite automata as graphical models, we obtain tractable representations of the constrained distribution that enable efficient inference. The approach guarantees constraint satisfaction by construction, supports both greedy and sampling-based decoding, and is compatible with parallel and block-wise decoding under arbitrary remasking schedules. Applying depth-reduction techniques from arithmetic circuit theory, we further reduce sampling depth from linear to logarithmic in the sequence length. Empirical evaluations on Dream-7B and LLaDA-8B show substantial accuracy gains across various tasks including function calling (xLAM, BFCL), planning (Sudoku, Countdown), text-to-SQL (Spider), and math reasoning (GSM-Symbolic), with little inference overhead relative to unconstrained decoding. For example, on BFCL-Live, our approach improves Dream-7B's greedy decoding accuracy from 63.9% to 71.5%, and stochastic sampling accuracy from 22.3% to 69.0%, where the unconstrained baseline collapses, with under 5% wall-clock overhead.

Supervised fine-tuning (SFT) is often treated as a capability-adaptation step, while alignment is attributed to later preference optimization or reinforcement learning. This separation is incomplete: when examples are scored and kept online during fine-tuning, the choice of which data to train on already changes the model's behavioral preferences. We study online data selection as an implicit alignment mechanism. Given the same base model, optimizer, and selected-token budget, we compare random, loss-based, quality-based, and diversity-based online selectors and measure the behavioral drift they induce without any preference optimization. The proposed evaluation tracks helpfulness, refusal rate, verbosity, truthfulness, sycophancy, calibration, and jailbreak robustness, together with diagnostics for which behavioral modes are over-represented in the selected data. We formalize online selection as a reweighted SFT objective whose weights define an implicit preference over response styles and safety postures, so that an online scorer plays the role usually assigned to a reward model. This view predicts that high-scoring data can systematically favor longer, more assertive, more compliant, or more refusal-prone behaviors depending on how the online score is defined. Empirically, selectors that are statistically indistinguishable in task accuracy diverge sharply in refusal rate, verbosity, and sycophancy, and we show that the direction of the shift is predictable from the attribute mixture of the selected data. We introduce Alignment Drift Auditing (ADA), a controlled protocol for quantifying selection-induced behavioral movement, and Alignment-Aware Selection (AAS), a diagnostic online selector that retains data efficiency while constraining drift along safety and style axes.

Humans continuously coordinate with others in dynamic interactions, often through implicit, hard-to-quantify social norms that act as shared tacit expectations among interacting agents. As AI agents, including large language models (LLMs), become embedded in daily life, they increasingly participate in such interactions and reshape social interaction structures. Yet they often fail to coordinate with humans in an effective, considerate, and natural manner. We hypothesize that this gap arises because existing approaches align model behavior with human demonstrations without explicitly quantifying the underlying norms that generate such behavior. We selected pedestrian-vehicle interaction as a representative dynamic interaction and developed a simplified experimental platform that captures its key interactive features. From 3,456 dynamic human interactions collected via this platform, we identified three principles underlying human social norms: outcome predictability, value alignment, and advantage awareness. Incorporating these principles into AI agents significantly improves human-AI coordination. In the closed-loop interaction task with humans, the social-norm-informed LLM achieved a nearly fourfold higher total score than the baseline strategy and outperformed human-human interactions by 43%. These findings indicate that formalizing tacit social norms into explicit, quantifiable principles can enable AI agents to achieve mutually beneficial coordination in dynamic interactions, supporting their more natural integration into human society.

Accurate forecasting of cellular network traffic is essential for network planning, resource allocation, and quality-of-service assurance in modern mobile communication systems. Real-world traffic often exhibits bursty endogenous dynamics and disturbances triggered by external urban events, which makes reliable prediction highly challenging. Most existing spatiotemporal traffic forecasting methods primarily focus on intrinsic traffic patterns or structural relationships within a single modality, and rarely model burst behavior together with exogenous contextual signals. To address this issue, we propose \textbf{MSPF-Net}, a multimodal cellular traffic forecasting framework that integrates external contextual information. Specifically, MSPF-Net consists of a Spatiotemporal-Frequency Traffic Encoder for capturing temporal, spatial, and spectral traffic patterns, a Peak Enhancement Module for extracting burst-aware representations of sudden spikes, a News Context Representation Module for encoding urban news streams into exogenous contextual embeddings, and a Dynamic Fusion Prediction Module for adaptively integrating these heterogeneous signals to generate forecasts. Experiments on the Milano, Trento, and LTE traffic datasets demonstrate that jointly modeling traffic dynamics, burst patterns, and news contextual signals can effectively improve forecasting performance.

A common method for the representation and analysis of time-series data is the hidden Markov model (HMM), where each observation is associated with a hidden state that evolves over time. However, many real-world systems are influenced by multiple independent factors, which are more naturally represented by factorial hidden Markov models (fHMM), where several hidden Markov chains jointly generate the observed data. Although an fHMM provides a richer and more realistic representation of many real-world systems, it can be reformulated as an equivalent HMM, but with a significantly larger state-space, leading to a severe increase in computational cost. In particular, the forward filtering algorithm, which is central to evaluation, decoding, and estimation tasks, becomes prohibitively expensive even for small systems. This work focuses on developing scalable methods for time-series analysis using tensor algebra to exploit the multidimensional structure of fHMM directly, without constructing intermediate HMM representations. Our novel filtering approach significantly improves computational performance and enables the efficient analysis of large systems and datasets, extending the scope of fHMM and providing a practical framework for data intensive applications.

High-quality fuzz harnesses are essential for effective gray-box fuzzing. While Large Language Models (LLMs) offer promise for automating this task, existing one-turn generation methods suffer from hallucinations and inadequate coverage due to coarse-grained function targeting and misaligned generation workflows. We present SynapseFlow, an automatic harness generator that addresses these limitations through two key innovations: dataflow-aware function aggregation and a staged, rollback-enabled generation workflow decomposition. SynapseFlow first analyzes source code to construct Structural Flow Graphs and extract coherent Function Triplets. It then synthesizes harnesses via a decomposed four-stage process governed by a staged rollback algorithm to ensure correctness. We evaluated SynapseFlow on 25 real-world open-source software projects. The experimental results indicate that SynapseFlow outperforms state-of-the-art tools (OSS-Fuzz-Gen, CKGFuzzer, PromeFuzz), achieving 3.07$\times$, 1.71$\times$, and 4.26$\times$ higher branch coverage, and 1.77$\times$, 1.51$\times$, and 1.36$\times$ higher bug detection rates, respectively. Most importantly, SynapseFlow discovered 7 previously unreported bugs (5 assigned CVEs), demonstrating its practical effectiveness in real-world bug discovery.

Large Language Models (LLMs) frequently exhibit sycophancy, where they agree with a user's statement even when incorrect. While sycophancy is often treated as a single defined behavior, it can manifest in substantially distinct ways and circumstances, raising the question of whether this multi-faceted nature is reflected in its internal mechanisms. To address this gap, we dissociate the representations of sycophancy into factual and opinion subtypes -- motivated by the distinction between verifiable claims and subjective beliefs. We train linear probes and construct steering vectors on activations of one subtype and evaluate their transfer to the other subtype to measure to what extent they share representations. We find evidence that different LLMs represent these subtypes differently, with either more unified or more distinct and causally interfering representations. This method of dissociation offers a promising framework for studying the representational structure of complex model behaviors.

This paper presents a physics-guided machine learning (PGML) framework for fuel density prediction, integrating physics constraints and domain knowledge into deep learning models to enhance model accuracy and stability. We explore three deep learning architectures -- ConvLSTM, Adaptive Fourier Neural Operator (AFNONet), and Video Vision Transformer (ViViT) -- to model the spatiotemporal evolution of fuel density. Our approach incorporates differentiable physics-informed terms in the loss function, including a mass-conserving fuel transport term and a rate-of-spread estimation. Experimental results, averaged across multiple independent trials, demonstrate that the proposed PGML framework outperforms purely data-driven baselines without physics constraints in both accuracy and stability. This framework enables computationally efficient, physically plausible fire forecasting to support adaptive prescribed burn management.

Customer behavior modeling underpins recommendation, marketing, and decision support, yet existing approaches either optimize predictive accuracy without explaining decisions or simulate users without grounding them in real behavioral data. We present the Large Behavioral Model (LBM) that learns customer decision making directly from large-scale retail transactions through a unified Person-Environment formulation. Customer state is represented by a behavioral profile derived from historical purchases, while product context is incorporated through retrieval-augmented generation. The model is trained using continued pre-training on verbalized behavioral data, supervised fine-tuning for decision generation, and reinforcement learning with verifiable rewards for evidence-based calibration. We evaluate the proposed framework on purchase prediction, hard-negative discrimination, basket completion, promotion response, and cross-domain voucher redemption. The model consistently outperforms frontier general-purpose language models on in-domain retail tasks while demonstrating strong zero-shot and fine-tuned transfer across retailers and decision domains. Ablation studies show that continued pre-training is the primary driver of behavioral generalization, retrieval is most effective when applied during both training and inference, and reinforcement learning improves reliance on explicit behavioral evidence over generic language-model priors. These results demonstrate that behavioral knowledge encoded in transaction histories can be effectively learned by language models, providing a scalable foundation for customer digital twins and behavior simulation.

Steering robot foundation models (RFMs) toward new task variants or user-preferred behaviors remains challenging, often requiring additional robot demonstrations, task-specific fine-tuning, or long-context conditioning. We present WAM-TTT, a test-time training framework for steering world action models from raw human videos. Rather than treating human videos as trajectories to imitate, WAM-TTT absorbs them into a lightweight adaptive memory inside a frozen WAM through self-supervised video prediction. To make this memory useful for control, we introduce a meta-training stage that aligns human demonstrations with robot behaviors using paired human-robot data and a key--value memory reconstruction objective. At test time, only unlabeled human videos are required to adapt the memory, while the pretrained WAM remains frozen. This enables efficient and reusable steering without robot actions, human-side annotations, or task-specific fine-tuning, while preserving the generalization ability of the foundation model. Extensive experiments show that WAM-TTT consistently outperforms in-context human-video conditioning baselines across diverse manipulation tasks and generalization settings.

Reinforcement learning (RL) has become the standard paradigm for enhancing the complex reasoning capabilities of large language models (LLMs). To achieve sample efficiency, modern RL frameworks rely on importance sampling (IS). However, these algorithms suffer from an exploration-stability dilemma. Pure IS often leads to catastrophic training instability, while standard clipping mechanisms used to mitigate this instability strictly constrain the policy update budget. By formalizing the concept of Probability Capacity (Cap), we reveal that conservative clipping structurally stifles exploration by prematurely truncating the update budget for correct but low-confidence reasoning paths. To break free from these constraints, we propose Unbounded Positive Asymmetric Optimization (UP), a universal and plug-and-play objective. UP theoretically restructures the optimization process by anchoring the policy to its current state via the stop-gradient operator. This asymmetric design unleashes unclipped, stable gradients for positive advantages to maximize exploration, while maintaining standard clipping safeguards for negative advantages to prevent training instability. Furthermore, our formulation readily extends across different optimization granularities, including token-level (GRPO, DAPO) and sequence-level (GSPO) frameworks. Extensive experiments demonstrate that UP enhances exploration capacity and achieves superior reasoning accuracy across diverse RL algorithms (DAPO, GSPO, and GRPO), model architectures (Dense, MoE, and vision-language), and training modalities (language and multimodal), validating UP as a truly universal plug-and-play enhancement for RL-based training.

Convolutional neural networks (CNNs) have demonstrated encouraging results in image classification tasks. However, the prohibitive computational cost of CNNs hinders the deployment of CNNs onto resource-constrained embedded devices. To address this issue, we propose EdgeCompress, a comprehensive compression framework to reduce the computational overhead of CNNs. In EdgeCompress, we first introduce dynamic image cropping (DIC), where we design a lightweight foreground predictor to accurately crop the most informative foreground object of input images for inference, which avoids redundant computation on background regions. Subsequently, we present compound shrinking (CS) to collaboratively compress the three dimensions (depth, width, and resolution) of CNNs according to their contribution to accuracy and model computation. DIC and CS together constitute a multidimensional CNN compression framework, which is able to comprehensively reduce the computational redundancy in both input images and neural network architectures, thereby improving the inference efficiency of CNNs. Further, we present a dynamic inference framework to efficiently process input images with different recognition difficulties, where we cascade multiple models with different complexities from our compression framework and dynamically adopt different models for different input images, which further compresses the computational redundancy and improves the inference efficiency of CNNs, facilitating the deployment of advanced CNNs onto embedded hardware. Experiments on ImageNet-1K demonstrate that EdgeCompress reduces the computation of ResNet-50 by 48.8% while improving the top-1 accuracy by 0.8%. Meanwhile, we improve the accuracy by 4.1% with similar computation compared to HRank, the state-of-the-art compression framework. The source code and models are available at https://github.com/ntuliuteam/edge-compress

Federated Learning (FL) enables training shared models on private, on-device data, but production deployments remain constrained to slow, multi-day refresh cycles due to the complexity of coordinating massive client populations. For applications such as feed ranking, ad targeting, and personalized recommendation, model freshness: the ability to rapidly adapt to new user-local data is critical for maximizing objectives like click-through rate. This lag leaves models stale and unresponsive to volatile data distributions driven by viral trends and shifting user intent. Bridging this gap requires addressing three challenges overlooked by existing FL systems: transient client availability, dynamic data heterogeneity, and delays between model predictions and observable outcomes. We present FeLiX, an FL orchestration framework that minimizes wall-clock time-to-target accuracy on live interaction streams. FeLiX introduces three primitives: (i) streaming-aware availability tiers that leverage lightweight telemetry to identify ready clients at scale; (ii) fresh-utility selection, a dual-tier mechanism that prioritizes statistically valuable updates from devices able to meet tight refresh deadlines; and (iii) informativeness-aware, delay-robust aggregation that incorporates late, high-value updates containing ground-truth outcomes without biasing the global model toward stale distributions. Unlike prior systems that rely on unrealistic oracular knowledge of client availability, FeLiX achieves near-oracular performance in real-world settings. Across CIFAR-10, Google Speech, and realistic low-availability traces, FeLiX reduces wall-clock time-to-target accuracy by up to 2.37X while reducing communication bandwidth by 1.30X compared to state-of-the-art synchronous and asynchronous FL baselines.

In this paper, we propose an efficient hybrid least squares/gradient descent (LSGD) method for MIONets to accelerate training. This method generalizes the LSGD method for DeepONets. Since MIONet is the sum of the entrywise product of multiple branch networks and a trunk network, it can be viewed as a multilinear function with respect to the last layer parameters of each branch network. These sets of parameters can be optimized using the alternating least squares method, where we solve the LS system for a single branch network in turn. To handle the large-sized system matrix, we introduce Kronecker and Khatri-Rao products and tensor permutation matrices to factor the large matrix into small ones. Our method is compatible with a general type of $L^2$ loss with regularization terms for the last layer parameters of each branch, where linear operators can be applied to the MIONet output in each loss term.

Large language models (LLMs) increasingly improve their reasoning at test time via additional computation, yet most existing works treat each problem in isolation. When problems arrive sequentially, accumulating reusable experience across them can further improve performance. Existing memory-based methods either store whole-solution templates that generalize poorly to novel problems or use heuristic step-level selection that is not optimized for final-answer correctness. Learning selection policies requires large-scale training data and fixed action spaces, making such approaches unsuitable for test-time settings where memory expands incrementally and only limited supervision is available. We propose MILES (Modular Instruction Memory with LEarnable Selection for self-improving LLM reasoning), a framework that dynamically expands step-wise memory and applies correctness-optimized memory composition under realistic test-time constraints. MILES maintains modular memory units consisting of asymmetric pairs of sub-goal embeddings and sub-instructions, each associated with a learnable selection head. This memory structure enables a coarse-to-fine retrieval mechanism: The coarse level enables memory expansion and collects supervision for training selection heads from confident samples, while the fine stage applies learned selection heads to rerank coarse-level candidates and guide reasoning for uncertain samples. MILES consistently matches or outperforms prior methods while achieving superior accuracy-efficiency tradeoffs. Extensive experiments demonstrate its effectiveness, robustness, and transferability.

We introduce a new context-enriched, multimodal time series forecasting benchmark, TimesX. TimesX contains a wide selection of high-quality real-world time series with diverse domains and textual contexts obtained from an automated data generation pipeline, which helps address three main issues of existing multimodal forecasting benchmarks: (1) poor generalization due to the small scale and synthetic nature of benchmark data, (2) very limited types of textual contexts in the benchmarks, and (3) an inability to mitigate data leakage in evaluation. We conduct a thorough empirical study of zero-shot multimodal forecasting approaches on TimesX. Our results suggest that many approaches that perform well on existing benchmarks may fail on TimesX. In contrast, simple ensemble methods that leverage rich textual context accompanying time-series can outperform strong baselines on TimesX.

Predicting the structure of short peptides in protein binding pockets remains difficult because this regime requires physics-based conformational search, yet existing methods do not provide a practical way to carry out that search on current hardware. We present QSAD, a quantum-classical framework that reformulates peptide structure prediction as amino-acid-level Hamiltonian sampling and replaces iterative optimization with non-iterative Hamiltonian evolution. Executed entirely on IBM Heron R2 across 101 binding-pocket peptides (5-18 residues), QSAD improves prediction accuracy by 27-71% over all evaluated AI and quantum baselines while maintaining the lowest variance across tested lengths. QSAD also tolerates noise levels 3-5x beyond typical hardware error rates, where iterative methods fail, and reduces mean quantum execution time by 27x relative to VQE. The sampled ensemble further supports approximate reconstruction of protein energy landscapes. These results establish coarse-grained quantum sampling as a practical computational path for structure prediction in regimes where data-driven methods lack sufficient signal.

Bridging the gap between human pilot intent and autonomous flight operation is critical for real-world electric vertical takeoff and landing (eVTOL) aircraft deployment. Flight planning traditionally relies on classic algorithms that struggle to incorporate flexible human preferences. We present FRAMe, an End-to-End Large Language Model (LLM) Flight Planning tool with RAG-based Memory and Multi-modal Coach Agent. Our system integrates a planner LLM with a multi-modal coach agent and retrieval augmented generation (RAG)-based memory to generate flight plans that satisfy mission constraints while aligning with human flight operator preferences. We demonstrate the system in a range of real-world-inspired scenarios of varying difficulty levels. Across four LLMs, the full FRAMe system (RAG and coach) yields the highest validity for every planner (up to 93.8% aggregate, 99% on Easy scenarios for the strongest planner) and shifts preference-relevant metrics in the operator-favored direction where the metric has headroom. FRAMe signifies how advanced LLMs can be deployed for human-centric mission planning, translating natural language instructions into safe, efficient, and flexible flight routes. The code is available at: github.com/amin-tabrizian/FlightPlanningLLMs

Large Language Models (LLMs) and generative AI (GenAI) systems, such as ChatGPT, Claude, Gemini, LLaMA, Copilot, Stable Diffusion by OpenAI, Anthropic, Google, Meta, Microsoft, Stability AI, respectively, are revolutionizing cybersecurity, enabling both automated defense and sophisticated attacks. These technologies power real-time threat detection, phishing defense, secure code generation, and vulnerability exploitation at unprecedented scales. Following a rapid surge where LLM-generated malware grew to account for an estimated 50% of detected threats by 2025, up from just 2% in 2021, navigating this highly automated threat landscape in 2026 demands next-generation security frameworks. This paper presents a comprehensive survey of the beneficial and malicious applications of LLMs in cybersecurity, including zero-day detection, DevSecOps, federated learning, synthetic content analysis, and explainable AI (XAI). Drawing on a review of over 70 academic papers, industry reports, and technical documents, this work synthesizes insights from real-world case studies across platforms like Google Play Protect, Microsoft Defender, Amazon Web Services (AWS), Apple App Store, OpenAI Plugin Stores, Hugging Face Spaces, and GitHub, alongside emerging initiatives like the SAFE Framework and AI-driven anomaly detection. We conclude with practical recommendations for responsible and transparent LLM deployment and trustworthy AI, including model watermarking, adversarial defense, and cross-industry collaboration, setting a new benchmark for rigorous, holistic cybersecurity research at the intersection of AI and threat defense, and offering a roadmap for secure, scalable LLM systems that serves as a critical reference for researchers, engineers, and security leaders navigating the complex challenges of AI-driven cybersecurity.

Realistic and diverse traffic simulation is essential to autonomous driving development. Yet prevailing benchmarks predominantly reward realism, and recent methods have optimized accordingly, leaving diversity underexplored. We introduce \textbf{Flow-ERD}, a multi-agent simulator that pursues realism and diversity jointly. Its backbone, \textbf{Agent-Type Aware Flow Matching} (AFM), couples flow matching's multi-modal expressiveness with type-specific kinematic execution. It preserves fine-grained diversity while keeping motions consistent with each agent type. A second stage, \textbf{Entropy-Regularized Distillation} (ERD), fine-tunes the closed-loop rollout distribution with an entropy-regularized reverse-KL objective. This mitigates covariate shift while explicitly preventing collapse onto high-density modes. We evaluate Flow-ERD with a log-free diversity metric alongside standard realism scores. Flow-ERD ranks first on the WOSAC test benchmark and dominates the realism--diversity Pareto front among reproducible baselines. Our project page is available \href{https://seulbinhwang.github.io/flow-erd-project-page/}{here}.

Wearable and mobile sensing technologies have demonstrated strong potential for health inference; however, most sensor models are designed for specific disease types, limiting their transferability across different health risks. Wearable foundation models offer a more generalizable approach in diverse health risk types. Nevertheless, most existing methods rely on high-frequency raw sensor data, raising concerns about privacy, computational overhead, and scalability across devices and populations. In this paper, we propose StepFM, a foundation model built solely on step counter data for broad-spectrum health prediction. Leveraging the ubiquity and low-dimensional nature of step data, StepFM provides a practical, privacy-preserving, and computation-efficient alternative to traditional sensor-based models. We design a scalable pre-training framework that captures temporal dynamics and behavioral patterns from large-scale step sequences, enabling transfer across more than 20 health risk prediction tasks spanning diverse devices, new regions, and novel disease types. Extensive experiments demonstrate that StepFM achieves strong performance compared to existing methods while maintaining robustness across heterogeneous settings. Furthermore, our analysis reveals interpretable and generalizable relationships between physical activity patterns and various health risks, offering new insights into activity-based health modeling. Our work establishes step-based sensing as a viable foundation for scalable and real-world health monitoring.

We propose a quantum model for finding a resetting input sequence (RS) which can take a finite state transition system (FA), to particular state independent of its current state. The complexity of finding such sequences for various types of FA can be NP-Hard or even PSPACE-Complete. To this end, we represent the FA states, inputs, and transition function in quantum space. Accordingly, we propose a model to represent the execution of an input sequence of a particular length $l$ starting form an initial FA state. The model is extended considering the application in superposition of all input sequences of length $l$ to an initial state of the FA. The model is further extended considering the application of all input sequences to all initial states of the FA capturing for every input sequence the collection (ordered list) of states reached by applying the sequence to all states of the FA. The amplitude amplification algorithm is then used as it combines similar collections of reached states while preserving all input sequences that reach these collections. A Grover search for a reached collection where its elements correspond to the same FA state provides a RS for the FA. Our approach offers a quadratic gain over the exponential complexity of traditional brute-force method, which is the only method that can be applied to a general FA class. As a proof of concept we provide results of several simulated FAs on a quantum simulator.

The accuracy of existing leaf-wood segmentation methods for tree point clouds varies across forest types and sites. Self-supervised learning (SSL) on point clouds has improved the generalization of deep learning models for forestry point cloud tasks, including biomass regression and individual tree segmentation, but its applicability to leaf-wood segmentation remains untested. In this study, we pretrained Point-M2AE, a widely used SSL architecture for point clouds, on ShapeNet-55 augmented with 2,400 individual tree point clouds. For fine-tuning and inference, we used recursive voxel subdivision to handle the wide variation in point density across inputs, allowing the same model to operate at both individual-tree and plot scales without architecture change. Compared to the model without pretraining, the pretrained model improved wood IoU from 60.5% to 70.0% for needleleaf and from 69.7% to 76.3% for broadleaf trees. On a benchmark spanning four countries across three climatic zones, the pretrained model achieved the smallest cross-site variation and highest overall performance among compared methods (LeWos, CWLS, and PointTransformer). Plot-level segmentation maintained accuracy comparable to individual-tree performance, with mIoU of 84.7% for broadleaf and 77.7% for needleleaf plots, showing that the model generalizes across scales without additional finetuning. As a downstream test in tropical forests, where dense canopies make segmentation challenging, we applied our model and a quantitative structure model to estimate wood volume for 28 trees from Guyana, Indonesia, and Peru to assess whether the segmentation improvements from SSL pretraining translate into improved downstream performance. The resulting volume estimates achieved the lowest error among all methods tested (MAE = 2.40 m$^3$), less than half that of algorithmic baselines (LeWos: 5.94 m$^3$; CWLS: 5.27 m$^3$).

The remarkable performance of large language models (LLMs) in linguistic tasks underscores an urgent need for comprehensive evaluation of their response quality. Prevailing methods, often confined to singular dimensions, fall short of capturing the full spectrum of model capabilities. This study introduces a multifactor scoring paradigm, integrating accuracy, conciseness, factual consistency, readability, and coherence, complemented by a graphical user interface (GUI) for visualizing outcomes. Evaluations on the TruthfulQA dataset unveil mainstream LLMs' strengths in reasoning tasks (peaking at a composite score of 0.6104) alongside pervasive limitations in navigating complex facts and ambiguities. Transcending the narrow lens of traditional metrics, this framework offers a transparent, adaptable avenue to illuminate model potential and deficiencies. Though presently focused on English tasks, its horizons beckon toward multilingual domains. This work carves a novel path for knowledge engineering and model refinement.

Reinforcement learning (RL) is increasingly grounded in tools from probability, optimization, and operator theory. This survey organizes the mathematical structures that underpin the design and analysis of modern algorithms in RL. We begin from Markov decision processes (MDPs) and the Bellman operators, emphasizing contraction mappings, monotonicity, and fixed-point theory that yield convergence guarantees and rates for value and policy iteration, and temporal-difference schemes. We then develop the optimization perspective: stochastic approximation and martingale methods, convex duality and the role of regularization linking mirror/proximal methods. Function approximation is treated through linear and non-linear settings, covering stabilization, error decomposition, and sample-complexity via concentration inequalities for dependent data and mixing processes. We further cover off-policy evaluation/learning, constrained RL and constrained MDPs (CMDPs). Throughout we unify algorithmic templates under common operator and variational lenses, highlighting both finite-sample bounds and asymptotic results. Our presentation is intended to provide a unified mathematical entry point for researchers in probability, optimization, and statistics interested in reinforcement learning.

Missing data is prevalent in practical applications, making effective imputation an essential preprocessing step for downstream analysis. Real-world datasets often exhibit complex latent structures composed of multiple subgroups with distinct distributions. However, existing methods often overlook such population heterogeneity. Without explicit structural guidance, these methods tend to produce generic estimates that blur subgroup boundaries and lack instance-level fidelity. While incorporating subgroup information offers a remedy, it faces a circular dependency: reliable subgroup identification requires complete data, while data completion is the imputation objective itself. To resolve this, we propose CAGI (Cluster-Aware Generative Imputation), a framework that reformulates clustering and imputation as a mutually reinforcing co-optimization process. CAGI employs a ``Partition-Guide-Restore'' strategy where dynamic cluster assignments act as local priors to condition a Generative Adversarial Network. An iterative feedback loop is established to progressively refine both cluster structures and imputed values toward faithful subgroup distributions. To ensure distributional stability, CAGI further employs a multi-level optimization objective combining instance-level reconstruction with distribution-level regularization. Extensive experiments on 14 benchmark datasets with 15 representative baselines demonstrate the superiority of CAGI. The source code is available at: https://github.com/supercocachii/CAGI

Music aesthetic assessment is a challenging yet underexplored problem, requiring models to capture fine-grained, multi-dimensional human perceptual judgments. Progress in this area has been limited by the lack of large-scale datasets with structured aesthetic annotations. We introduce MADB, a large-scale dataset and benchmark comprising 9,999 tracks annotated by 30 trained annotators. Each track is rated by around 10 annotators across 10 perceptual dimensions and one overall score, with additional textual comments for multimodal analysis. We establish a unified evaluation framework over multiple pretrained models. Results reveal substantial gaps between model predictions and human judgments, exposing key limitations of current approaches. MADB provides a new benchmark for human-aligned music understanding. Project page: https://github.com/knownree/madb

Compact world models that condition on a language goal promise to ground relations such as ``put the red block left of the blue block'' using a sparse set of explicit \emph{reference anchors}. We ask when such references actually ground a relation, and identify a trap: a goal-conditioned predictor reaches a striking $0.90$ relation-readout accuracy, yet this is \emph{instruction transcription}, not perception. Withholding the goal collapses it to chance ($0.90\!\to\!0.27$, three seeds) and a counterfactual instruction makes the predicted anchors follow the \emph{false} instruction $94.5\%$ of the time (true scene $2.3\%$; $N{=}256$). Tested across three settings and a within-task ablation, our central claim characterizes the confound: \textbf{instruction leakage occurs when the scored quantity is transcribable from the instruction (when the instruction names the answer) and is essentially independent of how predictive the non-instruction inputs are.} Our tabletop and the external BabyAI benchmark leak, whereas a Language-Table forward-dynamics world model whose instruction names \emph{referents} does not, until the instruction is augmented to name the direction; and degrading the action never increases leakage, the opposite of what predictor-competition predicts. The diagnosis prescribes the fix: keep the goal out of the dynamics (it belongs to the planner's cost) and supervise the \emph{read} path, recovering genuine, instruction-independent grounding ($0.88$, identical with and without the goal). The detection protocol and remedy apply to any goal-conditioned world model whose instruction names the scored quantity.

On analog neuromorphic hardware, intrinsic device noise is normally an accuracy tax. We ask whether it can instead consolidate memories. We cast per-synapse consolidation as a Doob h-transform: condition each weight's stochastic dynamics on never crossing a memory-critical barrier around its consolidated value. The conditioned diffusion gains an extra drift sigma^2 d/dw log h, a restoring force amplified by the noise variance itself that diverges at the barrier. We are explicit about novelty: the anchored drift -s(w-mu) our rule also contains is not ours (the limit of OUA, MESU, and EWC), and we surrender it. We claim only the conjunction of (a) the Doob barrier-conditioning as a synaptic rule, to our knowledge unclaimed (every h-transform use we found is generative modeling, none synaptic), and (b) a falsifiable prediction: increasing intrinsic noise non-monotonically improves sequential-task retention, an inverted-U that anchored-drift methods cannot produce. We pre-registered this as a go/no-go gate; it passes. On single-head Split-MNIST (8 seeds) the rule lifts retention 10.9 points at an interior optimum (paired Wilcoxon p=0.004), while matched OU/EWC/MESU anchors are monotone. Ablating the conditioning removes the effect; the optimum tracks the barrier; the inverted-U survives a second task stream and the realization where noise enters the forward pass. We then measure the intrinsic noise on real BrainScaleS-2 silicon (additive, trial-to-trial independent, tunable via on-chip averaging) and run the rule on the chip with its noise in the training loop: barrier-conditioning retains a prior task 15.6 points better than the matched control at matched average accuracy, a stability-plasticity shift, not a net-accuracy win (single seed; retention measured, energy modelled). Intrinsic analog noise thus becomes a consolidation dividend a digital accelerator must spend energy to generate.

Deep learning applications have been widely adopted on edge devices, to mitigate the privacy and latency issues of accessing cloud servers. Deciding the number of neurons during the design of a deep neural network to maximize performance is not intuitive. Particularly, many application scenarios are real-time and have a strict latency constraint, while conventional neural network optimization methods do not directly change the temporal cost of model inference for latency-critical edge systems. In this work, we propose a latency-oriented neural network learning method to optimize models for high accuracy while fulfilling the latency constraint. For efficiency, we also introduce a universal hardware-customized latency predictor to optimize this procedure to learn a model that satisfies the latency constraint by only a one-shot training process. The experiment results reveal that, compared to state-of-the-art methods, our approach can well-fit the 'hard' latency constraint and achieve high accuracy. Under the same training settings as the original model and satisfying a 34 ms latency constraint on the ImageNet-100 dataset, we reduce GoogLeNet's latency from 40.32 ms to 34 ms with a 0.14% accuracy reduction on the NVIDIA Jetson Nano. When coupled with quantization, our method can be further improved to only 0.04% drop for GoogLeNet. On the NVIDIA Jetson TX2, we compress VGG-19 from 119.98 ms to 34 ms and even improve its accuracy by 0.5%, and we scale GoogLeNet up from 20.27 ms to 34 ms and achieve higher accuracy by 0.78%. We also open source this framework at https://github.com/ntuliuteam/ZeroBN

We describe the methodology used to alias the free-text author/committer identities of the entire World of Code (WoC) collection (version V2604, ~107M distinct author strings over ~6B commits) into canonical persons, extending the fingerprint-based anti-aliasing of ALFAA and the 38M-identity resolution of Fry et al. by an order of magnitude. At this scale the central problem is over-merge, not missed merges: a few bridge identities (bots, role accounts, placeholder emails, multi-author commit fields) transitively weld unrelated clusters through the global union step into million-member "mega-clusters." We report the full experimental record (more than twenty experiments, including unsuccessful ones) behind the deployed design. Node-level gates (information score, project spread, degree) preserve recall but cannot dissolve the mega-cluster; per-value blocklists of high-quality-but-shared attributes are recall-safe but cannot break a redundant mesh; the working composition is a betweenness cut over the exact union graph plus a per-edge classifier trained on 2.6M labels mined from GitHub no-reply identifiers. That classifier, filtering dormant cross-project shingle groups and joined by GitHub's own account assertions, then recovers the recall the precision work had foregone. Against human-adjudicated pairs the per-edge model transfers at AUC 0.99; end to end the largest cluster falls from 170,431 (and a predecessor's 3.0M) to under 7,000 w hile gold recall rises from 0.44 to 0.70 at increasing precision, and on an independent 21M-alias GitHub ground truth the final map outscores its predecessors and the published state of the art among global, privacy-preserving resolvers. The record doubles as a catalog of scale lessons: structural cuts do not transfer to edge sets they never saw, and recall-only and precision-only benchmarks invert verdicts unless read together.

Artificial intelligence (AI) analysis of micro-ultrasound ($μ$US) has shown promise for prostate cancer (PCa) detection. However, most existing AI methods focus on the analysis of single $μ$US images in isolation. By contrast, expert $μ$US readers typically assess a full recorded video study, which provides three-dimensional context, to improve PCa detection compared to single-frame analysis. Inspired by this clinical workflow, we propose Compass, a novel AI methodology which models a $μ$US study as a stream of 2D images. Compass jointly integrates rotational sweep videos of the prostate with $μ$US frames acquired at the moment of biopsy, and performs evidence aggregation across the study using a transformer conditioned on the probe's rotational angle. Finally, a decoder head predicts frame-level and study-level risk scores for the patient. The model is trained and evaluated using a multi-center clinical trial dataset of $μ$US studies, including continuous rotational scans of the prostate and videos captured during biopsy acquisition. We compare the proposed method to baseline AI methods from the literature and to risk scores provided by clinical experts. Our framework shows strong performance, highlighting the value of multi-view context for $μ$US PCa detection, and providing a potentially powerful tool to complement human expertise in $μ$US-based PCa diagnosis. Our code is available at: https://github.com/mharmanani/Compass.

Pre-trained Vision Foundation Models (VFMs) provide strong visual representations for diverse downstream tasks. The key challenge of VFM adaptation stems from the prohibitive costs of full fine-tuning and catastrophic forgetting. To address this, Low-Rank Adaptation (LoRA) has emerged as the prevailing paradigm for Parameter-Efficient Fine-Tuning (PEFT). However, LoRA is typically designed for transformer self-attention layers parameterized by 2D matrices. Since convolutional kernels inherently couple spatial and channel information within a 4D tensor, forcing them into a monolithic 2D matrix disrupts the inherent spatial topology. In this paper, we propose Low-Rank Convolutional Adaptation (LoCA), a convolution-aware PEFT framework that addresses spatial-channel entanglement by decoupling channel and spatial adaptation. LoCA introduces a low-rank channel adaptation for dense cross-channel mixing and refines spatial bases extracted from pre-trained kernels via Singular Value Decomposition (SVD). Experimental results show that LoCA preserves pre-trained spatial priors and achieves competitive or state-of-the-art performance across fine-grained classification, domain-generalized semantic segmentation, and generative benchmarks.

Scaling down the resolution of input images can greatly reduce the computational overhead of convolutional neural networks (CNNs), which is promising for edge AI. However, as an image usually contains much spatial redundancy, e.g., background pixels, directly shrinking the whole image will lose important features of the foreground object and lead to severe accuracy degradation. In this paper, we propose a dynamic image cropping framework to reduce the spatial redundancy by accurately cropping the foreground object from images. To achieve the instance-aware fine cropping, we introduce a lightweight foreground predictor to efficiently localize and crop the foreground of an image. The finely cropped images can be correctly recognized even at a small resolution. Meanwhile, computational redundancy also exists in CNN architectures. To pursue higher execution efficiency on resource-constrained embedded devices, we also propose a compound shrinking strategy to coordinately compress the three dimensions (depth, width, resolution) of CNNs. Eventually, we seamlessly combine the proposed dynamic image cropping and compound shrinking into a unified compression framework, Smart Scissor, which is expected to significantly reduce the computational overhead of CNNs while still maintaining high accuracy. Experiments on ImageNet-1K demonstrate that our method reduces the computational cost of ResNet50 by 41.5% while improving the top-1 accuracy by 0.3%. Moreover, compared to HRank, the state-of-the-art CNN compression framework, our method achieves 4.1% higher top-1 accuracy at the same computational cost. The codes and data are available at https://github.com/ntuliuteam/smart-scissor

Agentic AI development today runs on token maxing: buying capability with tokens -- longer reasoning traces, more turns, wider tool payloads, bigger replayed contexts -- so tokens per task grow faster than task value. Falling per-token prices mask the pattern; total spend rises anyway. We argue the decisive lever against token maxing is the harness: the orchestration layer that assembles context, exposes tools, sequences turns, delegates work, and carries enterprise observability and governance. We isolate it with a controlled swap: 22 locked evaluation tasks, six foundation models (Claude Sonnet 4.6, Gemini 3.1, Gemini Flash 3.5, Qwen 3.6, GLM 5.1, Palmyra X6), changing only the orchestration layer -- a frozen conventional production loop versus the Writer Agent Harness. Holding models constant, the harness cuts blended cost per task 41% ($0.21->$0.12), median wall-clock 44% (48s->27s), and tokens per task 38% (14.2k->8.8k), with task-completion quality at parity (0.78->0.81, directional at this sample size). Efficiency is model-invariant -- every model gets cheaper (33-61%) -- while quality gains are capability-dependent: a model's gain correlates almost perfectly with its baseline strength (r=0.99, n=6), a phenomenon we term harness leverage. Quality per dollar rises 82%; task-completions per million tokens rise from 54.9 to 92.0. On this workload the orchestration layer moved cost per task more than the full spread of the model menu did. We formalize token economics at the orchestration layer (including effective input price under prompt caching), detail the six mechanism families behind the effect -- cache-shape discipline to failure-spend governance -- compare six widely used agent systems on the same axes, and argue the harness is the one component whose efficiency multiplies across every model an organization runs -- present and future.

The Stochastic-Oracle Turing Machine (SOTM) framework models AI-augmented computation as the interaction of a probabilistic Turing machine with an oracle whose responses are drawn from context-dependent distributions. This paper studies what an SOTM can achieve under two oracle-response schemes: in a cached-response oracle, each distinct query receives one response that is reused on later calls to the same query, while in a fresh-response oracle, each call returns an independent response. In both schemes, the SOTM first computes from its input and internal random source to generate its first query, then proceeds adaptively, computing from its query-response transcript (the record of queries issued and responses received) to generate each subsequent query or produce a final output. Cached responses impose two transcript-based ceilings on achievable performance: a correct-identification ceiling governed by the total variation distance between the transcript distributions induced by the hidden states of the oracle, and an output quality ceiling equal to the expected score of the best output the SOTM can compute from the transcript. Fresh responses can raise these ceilings by allowing repeated calls to accumulate independent evidence toward correct or high-quality outputs. In the binary single-informative-query case, the error probability decreases exponentially in the number of calls to the same query at the Chernoff rate. For output quality, query-count bounds characterize threshold stopping when the score function is incorporated as part of the SOTM, and majority-based amplification bounds characterize the binary candidate-output model when it is not. Together, the results identify how response reuse, transcript information, and access to the score function determine what an SOTM can compute and at what token cost.

Deep learning image classifiers achieve strong predictive performance yet remain opaque in how decisions are formed. A model may predict correctly while relying on irrelevant cues, shortcut associations, peripheral structures, or device level artifacts instead of task relevant regions. On large scale datasets this opacity is especially problematic, since inspecting heatmaps one sample at a time cannot scale to thousands of predictions. We propose Relevance Based Model Decision Explainability (ReMoDEx), a framework for systematic, dataset scale assessment of model decision behaviour in image classification. ReMoDEx defines a stepwise pipeline: model inference, target class selection, relevance map generation, heatmap standardisation, similarity based grouping of patterns, cluster level interpretation, and spatial relevance assessment. Local methods GradCAM++, Integrated Gradients, Occlusion Sensitivity, and Layerwise Relevance Propagation are each combined independently with a single global module that summarises an entire set of relevance maps into a few decision strategy clusters, replacing sample by sample inspection with an automatic, scalable summary. To demonstrate ReMoDEx, we applied it to a VGG16 based classifier distinguishing COVID-19, Normal, Lung Opacity, and Viral Pneumonia. The classifier showed stable performance (86.27% test accuracy, 0.9624 test AUC). However, each explainer combined with the global module consistently produced two recurring strategies: central thoracic region decisions and border/corner sensitive decisions, indicating possible shortcut learning that conventional metrics could not reveal. Masked image validation confirmed that model confidence and predicted class changed when central or peripheral regions were occluded. ReMoDEx thus provides a scalable relevance based decision assessment framework and an essential complement to accuracy based evaluation.

Most self-supervised image clustering models, actually almost all deep learning approaches, are based on gradient descent: In order to calculate the loss, every optimization step requires a clearly defined target, whether a contrastive split, a masked patch or entity, an EMA-teacher output, a pseudo-label, or a differentiable information-theoretic functional. We propose a self-supervised framework that drops this requirement for image clustering. Without any prior knowledge, we have to assume that each pixel is i.i.d. according to the Principle of Maximum Entropy. Taking this as our null hypothesis H0, we define a "surprise score" that measures how unlikely the model's output representation would be under H0. Maximizing the surprise score forces the deep learning model to reject H0 - equivalently, to discover non-random feature from data. Also, here is our fundamental assumption: a surprise score cannot, in general, be reduced to a per-step loss. Hence, we propose the "converge-to-surprise" scheme to optimize our model: an evolution-strategy (ES) outer loop, which directly maximizes the surprise score without needing its gradient, paired with a periodic gradient-descent inner loop, which uses the surprising clusters already discovered by ES as surrogate targets. On standard image benchmarks, our framework achieves new state-of-the-art results in non-parametric self-supervised image clustering - the strictest deep-clustering setting, in which the number of ground-truth classes is not given to the model.

We consider the problem of finding stationary points for stochastic convex optimization problems. Rather than surrogates to stationarity, such as a proximity-to-stationarity guarantee or small gradient of the Moreau envelope, we ask for a stronger notion: that the subdifferential of the objective actually contains a small element. This criterion is non-trivial, because subdifferentials of convex functions fail to converge uniformly, even in arbitrarily small neighborhoods of the optimum. Our convergence guarantees rely on dimension theory to decompose the graph of the subdifferential of a convex function, showing how stochastic sampling preserves "pieces" of these graphs, and allowing effective application of proximal-point-like methods.

Visual navigation policies built on large pretrained models have so far followed a common recipe: a dedicated visual encoder, a bespoke action head, and training on thousands of hours of cross-embodiment datasets. We ask whether this recipe is necessary. In this paper, we introduce GemNav, a visual robot navigation policy that adapts a frozen Multimodal Large Language Model (MLLM) for short-to-medium horizon waypoint navigation using Low-Rank Adaptation (LoRA) on the language tower alone, with no auxiliary visual encoder and no continuous regression head. Waypoints and categorical navigation signals share a single discrete token vocabulary generated by the language-model head, and a soft-decoded auxiliary loss recovers the metric structure that pure cross-entropy training discards. On a single 8.7-hour open corpus, roughly three orders of magnitude smaller than competing training sets, the policy transfers zero-shot to four physically distinct unseen environments and stops within 0.25-0.42m of the goal across 20 real-world trials covering an open carpark, an obstacle carpark, a long outdoor chemical yard, and an indoor warehouse. Conditioning on short image histories improves offline metrics but yields no robot benefit, pointing to a ceiling on what temporal context adds once pretrained vision features are in place. These results indicate that discrete-token adaptation of frozen MLLMs can provide a data-efficient, deployable alternative for foundation model robot navigation.

A multiple double is an unevaluated sum of doubles. An NVIDIA tensor core is a specialized high performance compute core for matrix multiplication. The Ampere A100, released in 2020, introduced tensor cores capable of 64-bit floating-point arithmetic. Every multiple double arithmetical operation requires renormalization, which involves branching, for which tensor cores are unsuited. To solve this problem caused by renormalization, we apply a solution similar to the Ozaki scheme [Ozaki et al, Numerical Algorithms, 2012]. Our software is available under the GPU GPL license on github.

Best-arm identification is a canonical model for data-driven decision-making, but in many applications each reward observation is costly. Motivated by the growing availability of cheap predictions from machine learning and large language models, we study fixed-confidence best-arm identification in which each costly reward pull is paired with a cheap but correlated proxy score. The marginal mean of the proxy can be estimated offline and is treated as known, whereas its correlation $ρ$ with the reward, which governs how much the proxy helps, is unknown and must be learned online in pair with real rewards. We show that a control-variate adjustment turns this model into a heteroscedastic identification problem whose oracle sample complexity improves by residual variance $1-ρ^2$. The central difficulty is that the correlation must be learned from the same costly samples that identification consumes online, and that a plug-in estimate of the residual variance is anti-conservative and can compromise correctness. We propose PROBE (PRoxy OLS for Best-arm Exploration), a phase-elimination algorithm that directly maintains an upper certificate on the residual variance with an ordinary least squares fit, whose exact chi-square law keeps the certificate valid regardless of the unknown correlation. We prove that PROBE is $δ$-PAC and attains the known-correlation oracle sample complexity up to a constant multiplicative factor and a constant additive calibration cost. The guarantee extends to the $(ε,δ)$-PAC setting under minimal changes to the algorithm. Numerical experiments on synthetic instances and on an auto-loan pricing replay with large language model and tabular proxies confirm that the sample savings of PROBE scale with the strength of the reward-proxy correlation, exactly as the theory predicts.

Understanding and forecasting audience reactions to video content are crucial for improving content creation, recommendation systems, and media analysis. To enable audience reaction prediction and other content engagement applications, we introduce $\textbf{Video2Reaction}$, a multimodal dataset that maps short movie segments to a distribution of $\textit{induced emotions}$ of viewers in the wild, as expressed through social media. $\textbf{Video2Reaction}$ spans more than 10,000 videos and serves as a reliable benchmark as well as a training resource for audience reaction prediction. To enable cost-effective continuous annotations as reactions may change over time, we develop a two-stage multi-agent pipeline using only open-source LLMs, achieving 86% correctness under blind human verification despite the inherently noisy and subjective nature of the task. We establish the first benchmark for video-to-reaction-distribution prediction in the wild and show that pretrained foundation video models fail in zero-shot settings, while finetuning transforms them into state-of-the-art predictors capable of modeling both full reaction distributions and dominant responses from video alone. However, the task remains challenging: even the strongest methods achieve only 77% Top-3 F1 in dominant reaction prediction (LLaVA-Next), highlighting a substantial gap in modeling collective audience reaction. \modification{Dataset and code are available at our project page: https://information-fusion-lab-umass.github.io/video2reaction-bench.github.io

Conversational LLM agents can cause real-world harm when their internal workflows fail, such as completing a transaction without confirmation. Testing these state-dependent failures is difficult because critical boundaries, such as identity checks and confirmation gates, are hidden behind multi-turn conversational prerequisites, rendering them inaccessible to standard tests. We present AgentEval, a black-box testing framework that discovers and stresses these stateful boundaries. AgentEval interacts with an agent to mine a \emph{conversational workflow graph}, a model of its behavior. Instead of prompting blindly, AgentEval uses this graph's structure to enumerate specific guards and prerequisites as test targets, replaying the conversational path to a boundary before applying a perturbation. AgentEval then executes each test, determining whether it passes or fails using only the conversation turns. We benchmark AgentEval against a privileged, white-box auditor with access to the agent's underlying source code, which AgentEval never sees. On four $τ^3$-bench agents, AgentEval successfully generates tests covering $23$--$38$ distinct boundaries per agent; ablation studies attribute the gain to the graph's structure: $23$ distinct boundaries versus $12$ with a prompt-only baseline, at lower duplicate and false-alarm rates.

Container image pulling accounts for the majority of pod startup time in Kubernetes environments. Standard pull downloads the entire image before the container can start, even when the application accesses only a fraction of the image content at startup. We present SOCI (Seekable OCI), a lazy-loading architecture that enables containers to start without downloading the full image. SOCI builds an external index over standard OCI images, mapping files to byte ranges within compressed layers. At runtime, a FUSE filesystem intercepts file accesses and serves them via HTTP range requests. Unlike prior approaches that require image format conversion, SOCI works with unmodified images and standard registries. The index is stored as an OCI referrer artifact, requiring no changes to images, registries, or deployment tooling. On a 1.3 GB Python web service image, SOCI reduces cold-start pull time from 20 seconds to approximately 2.8 seconds (7.4x speedup), with pull time independent of image size. Larger images see larger speedups (9.3x on a 2.5 GB image) because SOCI pull time is constant while standard pull scales linearly. We measure a crossover at 80% access density: below this, lazy loading wins; above, parallel full pull is faster. SOCI lazy loading is deployed in production on Amazon EKS and Amazon ECS Fargate (which launched 18.4 million tasks per day during Prime Day 2025), and has been serving lazy-load requests since 2023. EKS Auto Mode uses SOCI's parallel pull mode for GPU instances.

Prior work suggests that diffusion representations capture low-level geometry but struggle with high-level semantics. We demonstrate that state-of-the-art video diffusion models overcome this limitation. By systematically probing their intermediate activations using recent mutual-kNN alignment metrics, we reveal a highly structured latent space where visual representations evolve across both network depth and noise levels. We show that while moderate noise levels yield linearly separable global semantics, fine-grained details persist at lower noise levels but become spatially scattered, requiring attention mechanisms to decode. Building on these insights, we introduce Gen4U (Generation for Understanding), a framework that repurposes these generative representations with a single forward pass. Our experiments establish that frozen, large-scale video diffusion models function as highly competitive video encoders across a wide spectrum of tasks, spanning semantic and non-semantic objectives (video classification, depth estimation, camera pose estimation, image and video captioning). Bypassing fine-tuning, Gen4U unifies the generation and understanding paradigms, achieving strong perception performance while fully preserving the model's ability to generate high-quality video.

On-policy distillation is a practical post-training recipe for large language models, supplying dense teacher supervision on the student's own trajectories. In privileged-context self-distillation, teacher and student are the same model conditioned on the same prefix, but the teacher also sees a hint or the full solution trace. This makes supervision abundant but harder to trust: the teacher can be confident about continuations its privileged view makes obvious but the student cannot yet justify. The distillation pull is strongest where teacher and student disagree most, and over many updates it accumulates into drift that degrades out-of-distribution (OOD) reasoning. We introduce GeoSD, a geometric self-distillation objective that treats this drift as movement in the student's predictive behavior and counters it in two complementary ways. A Hellinger loss scales each teacher preference by the overlap the student already shares with it, attenuating the pull on tokens the student cannot yet support. Since these pulls still compound over training, a proximal term penalizes how far the student's predictions drift from a recent checkpoint, measured as a Fisher-Rao distance. Both are distances in the same geometry of next-token distributions, and a natural-gradient update takes its steps in that geometry rather than in parameter space. Across mathematical reasoning benchmarks and three model families, GeoSD preserves the in-distribution gains of self-distillation while improving average OOD accuracy by 5.7-8.6 points over the base model, with gains holding across model scales from 1.7B to 32B. Analyzing why standard matching fails out of distribution, we find it wins agreement with the teacher by draining mass from alternatives at high-entropy states, resulting in confident agreement on wrong answers, whereas GeoSD keeps those alternatives in reach.

Reinforcement learning agents for imperfect-information card games are only as strong as the opponents they train against, and they are hard to grade, since they beat a random opponent over 99 percent of the time and only tie copies of themselves. So we build a strong, fixed, rule-based expert for Gin Rummy and use it only as a yardstick, never for training. It beats every agent we trained 70 to 99 percent of the time. Across more than a hundred runs, we isolate what makes a lightweight agent stronger. Trust region updates, a well-aimed reward, a curriculum of tougher opponents, warm starting, and keeping the best checkpoint all help, and stacking them lifts a self-play champion from about 30 to 36 percent against the expert. Several ideas did not pay off. Short-term and longer-term reward shaping, learned state embeddings, imitation and DAgger, and a live large language model opponent were each unhelpful, too slow, or too heavy to train at scale. Comparing MLP, convolutional, set-based, attention, and recurrent encoders shows that extra capacity does little to break the ceiling, suggesting the limit is information rather than network size. We add standard baselines (neural fictitious self-play and information set Monte Carlo search) and confirm the approach carries over to Leduc Hold'em, where the optimum is computable. The result is a lightweight, game-agnostic recipe that trains competitive agents without training on the expert, for any game a small model can handle, reported with robust statistics and released as a reusable package.

African American English (AAE), a rule-governed dialect spoken by over 30 million people, is routinely misinterpreted and "corrected" by large language models (LLMs). Across six instruction-tuned LLMs (14B to 70B), we show that state-of-the-art models systematically prefer Standard American English (SAE) continuations even when the preceding context is in AAE, effectively rewriting AAE into SAE. We present an end-to-end framework to audit and mitigate this bias. For auditing, we introduce conditional Dialect Group Invariance (cDGI), which isolates true model bias from translator-induced artifacts, and a feature-level localization analysis that identifies which AAE markers most strongly trigger bias; we find that syntactic constructions, especially negative concord (e.g., "ain't nobody"), are universal triggers across all models. For mitigation, we introduce, to our knowledge, the first application of activation steering to dialect bias: a training-free, test-time method that extracts dialect directions via causal tracing and injects them into bias-relevant layers. Activation steering reduces bias 5 to 20 times more than prompting while preserving SAE fluency. To enable this work, we release REAL-AAE , the largest real-AAE parallel corpus to date: 17,479 AAE/SAE/ AAE_back triplets from natural tweets (2 to 6 times larger than prior real-AAE resources), validated automatically (BERTScore F1 = 0.95) and by three native AAE speakers (83.0% semantic agreement).

Diffusion models offer a powerful framework for sampling from complex probability densities by learning to reverse a noising process. A common approach involves solving for the time-reversed stochastic differential equation (SDE), which requires the score function of the evolving sample distribution. The logarithm of this distribution's density is governed by a Hamilton-Jacobi-Bellman (HJB) type partial differential equation (PDE). However, current methods for solving this PDE, such as PINNs or trajectory-based techniques, often suffer from long training times and significant sensitivity to hyperparameter tuning. In this work, we introduce a novel and efficient solver for the underlying HJB equation based on the functional tensor train (FTT) format. The FTT representation leverages latent low-rank structures to efficiently approximate high-dimensional functions, enabling both model compression and rapid computation. By integrating this efficient representation with a backward-in-time iterative scheme derived from backward stochastic differential equations (BSDEs), we develop a fast, robust and accurate sampling method. Our approach overcomes primary bottlenecks of existing techniques, enabling high-fidelity sampling from challenging target distributions with improved efficiency.

Existing NAS benchmarks (e.g., NAS-Bench, NATS-Bench) cover only narrow, task-specific regions of the architectural design space and lack cross-domain or deployment-aware evaluation. LEMUR 2 introduces a large-scale, extensible framework unifying generative, evaluative, and deployment pipelines to unlock neural-network diversity. It comprises over 14,000 distinct architectures and more than 750,000 structured training records documenting model performance, hyperparameters, and task outcomes. These models were produced through AST-based code mutation, genetic and reinforcement-learning evolution, generation of fractal architectures, and synthesis guided by a Large Language Model (LLM). This includes deep models generated with the retrieval-augmented system NN-RAG, which derived and used architectural motifs from over 900 PyTorch modules extracted from public repositories. LEMUR 2 further employs NN-VR and NN-Lite pipelines for automated deployment and latency benchmarking on heterogeneous mobile and Unity-based VR platforms, providing real-device performance metadata. It spans multimodal tasks, image captioning, text-to-image synthesis, and language modeling, supporting cross-domain analysis of architectural transferability. By linking diverse architectures, tasks, and deployment data, LEMUR 2 provides the data foundation for LLM fine-tuning and coupling diverse architectural origins with large-scale, cross-platform empirical validation. This dataset defines a new basis for reproducible and data-driven AI design, advancing the emerging paradigm of LLM-driven AutoML and architectural generalization across modalities and hardware.

Sampling stochastic signals supported on a graph underlies many graph machine learning tasks, including recommender systems, forecasting in financial markets, and wireless network optimization. In these settings, the target signals are realizations of unknown conditional distributions. However, prevailing approaches rely mostly on intricate, application-tailored designs that often regress to a conditional mean instead of sampling from the conditional law. This paper unifies such problems as conditional graph signal generative modeling and tackles them with a single denoising diffusion framework. We learn a reverse diffusion process, parametrized by graph neural networks (GNNs), that draws graph signals conditioned directly on the graph topology and on node-feature side information. The reverse process is realized by a novel architecture, the U-Graph Neural Network (U-GNN), which generalizes the image-convolutional U-Net to graph-structured signals. The U-GNN performs multi-resolution encoder--decoder processing in which pooling and unpooling reduce to a learned node selection, expressed by nested selection matrices, and a zero-padded lifting of coarse signals back to the full node set. The graph convolutions are carried out on the original graph, with a stride that sets their hop reach, so the U-GNN bypasses explicit graph coarsening at every resolution. We demonstrate our method on two generative tasks: stock price forecasting and optimal wireless resource allocation, with extensive numerical results in both domains.

Speech-to-text alignment means finding the temporal boundaries of each word in the audio. Some models provide such an alignment directly and others do not. Connectionist temporal classification (CTC) and transducer models have an alignment by construction, whereas attention-based encoder-decoders (AED) and speech large language models (LLMs) do not, and their word timings are usually read off the attention weights instead. All of these signals live on the encoder frame grid, which bounds their temporal precision. We study a generic gradient-based alignment that applies to any differentiable ASR model. We take the gradient of each teacher-forced token log probability with respect to the input, reduce it to a per-frame saliency, and decode the resulting matrix into word boundaries with a single dynamic-programming pass. The method needs no training, no model modification and no alignment heads, works across all model families including the speech LLMs, and aligns on the input grid rather than on the coarser encoder grid. We evaluate it on sixteen models from four families, on read (TIMIT) and spontaneous (Buckeye) speech, each against the model's own native or attention-based alignment. We find that the gradient yields a usable alignment for every model, that it is usually somewhat behind a strong native aligner but better where the native alignment is weak, as for the streaming models, and that its main disadvantage is the cost of one backward pass per token.

Physics-informed learning promises data-efficient and stable dynamics prediction, yet its strongest geometric guarantees have largely remained confined to closed conservative systems. This excludes many robotic systems of practical interest, where actuation, dissipation, and constraints continuously exchange energy and momentum with the environment. We introduce CaLiSym, a lightweight framework that extends exact symplectic learning to such systems by changing where the geometric prior is imposed. Rather than enforcing symplecticity on the measured physical state, CaLiSym embeds the state and its physical ports into a structured lifted canonical phase space, where the learned dynamics evolve through an exactly symplectic map. The lift is explicit and algebraic, requiring neither recurrent latent states, transformer decoders, implicit optimization, nor inference-time ODE integration. We instantiate the framework with generalized-ridge SympNet predictors and introduce GRB-SympNet, a B-spline variant that combines local approximation with exact symplectic structure. Experiments on a controlled dissipative double pendulum, a real-world quadrotor, and a contact-rich quadruped demonstrate consistent improvements in out-of-distribution autoregressive prediction while using parameter-efficient models. At the same time, the learned lifted dynamics preserve the symplectic form to numerical precision. These results show that symplectic learning can be extended beyond conservative mechanics through structured canonical lifts, enabling geometry-preserving dynamics models for real-world robotic systems.

Recent advances in AI for Mathematics have focused largely on autoformalization and theorem proving, leaving the role of Computer Algebra Systems (CAS) in agentic LLM workflows underexplored. We propose a ReAct-style agentic setup that combines LLM reasoning with verifiable feedback from SageMath, together with Context7 for the up-to-date documentation. We evaluate this agentic setup across frontier models for solving research-level mathematical problems from the RealMath benchmark in a setting that emulates a computational-mathematics research loop. We also propose a refinement to the RealMath benchmark by introducing a multi-step post-processing procedure and a multi-stage validation pipeline, both of which improve the quality and reliability of the extracted problem set. Our experiments reveal substantial performance gains from SageMath access across all evaluated models on +9.7~pp on average, the gains range from 1.5~pp to 27.8~pp and narrow the gap between open-weight and closed models. Qwen~3.7-Max benefits from SageMath the most, while GPT-5.5 achieves the highest solve rate of $75.2\%$ and the lowest token usage among tool-enabled configurations. Our findings suggest that CAS-augmented agents represent a promising direction for assisting mathematicians in computational exploration, and we believe that this work is a step towards automated conjecture discovery. The project repository is available online.

For any E-commerce website it is a nontrivial problem to build enduring advertisements that attract shoppers. It is hard to pass the creative quality bar of the website, especially at a large scale. We thus propose a programmatic solution to generate product advertising headlines using retail content. We propose a state of the art application of Reinforcement Learning (RL) Policy gradient methods on Transformer based Masked Language Models. Our method creates the advertising headline by jointly conditioning on multiple products that a seller wishes to advertise. We demonstrate that our method outperforms existing Transformer and LSTM + RL methods in overlap metrics and quality audits. We also show that our model-generated headlines outperform human submitted headlines in terms of both grammar and creative quality as determined by audits.

While enabling effective collaboration on complex tasks, LLM-based Multi-Agent Systems (MAS) face critical security challenges due to vulnerabilities at the agent and interaction levels. Most existing MAS security defenses are built upon two core assumptions: semantically-explicit malicious attacks and explicit graph-based modeling of the MAS topology and agent-level interactions. In practice, real-world attacks are becoming more semantically stealthy, while MAS execution is typically asynchronous without the temporal alignment assumed by graph-based propagation models. To address these limitations, we propose AcMAS, an activation-based framework for malicious-behavior detection in MAS. By analyzing internal reasoning states in the activation space of local agents, AcMAS detects even stealthy attacks in a synchronization-robust fashion, without relying on explicit interaction graphs. Moreover, our activation analysis provides critical signals to guide AcMAS in restoring the functionality of compromised agents, rather than the disruptive agent isolation commonly used by the state-of-the-art methods. Comprehensive evaluation demonstrates that AcMAS significantly outperforms graph-based baselines against stealthy attacks, by +0.22 F1 in synchronous settings (0.94 vs. 0.72) and by +0.55 F1 in asynchronous settings (0.93 vs. 0.38), with generalization across diverse open-source LLM backbones, attack intensity, and MAS scale.

We present ECO/CPO-DAG, a domain-specific accountability protocol for adversarial supply chains that formalizes contradiction detection as a supplemental validation layer rather than a consensus or truth-establishing mechanism. Participants publish signed Event Claim Objects (ECOs) into a causally ordered, append-only directed acyclic graph (DAG) whose edges encode happened-before relations. When two claims about the same subject violate a domain constraint, any observer can compile a Contradiction Proof Object (CPO), a self-verifying object binding the two signed claims and the violated rule, which, on public verification, triggers economic slashing of a determinately blamed party. We map constraints to GS1 EPCIS 2.0 event semantics (spatial uniqueness, temporal monotonicity, quantity conservation, quality monotonicity, regulatory validity), so detection targets inconsistencies that are meaningful in practice. Selective disclosure via commitment schemes and, optionally, zero-knowledge contradiction proofs lets parties withhold claim contents until a challenge forces the minimal opening. We give an analytical treatment: an independent-observer detection model $1-(1-p_{\min})^h$, a deterrence condition $S>g(1-p)/(kp)$ under $k$-party collusion, and a storage estimate of order 1 GB per participant per year under stated assumptions. The protocol's boundary is explicit: it detects provable contradictions, not consistent lies; a party that never contradicts itself is invisible to it, so the layer complements, and does not replace, source verification and oracle aggregation. A single-machine reference implementation corroborates the detection model, with the predicted coverage band overlapping the measured 95% confidence interval at every observer count, and records zero false accusations; the fully zero-knowledge CPO, multi-party propagation, and adaptive-adversary evasion remain analytical.

Open physiological corpora are heterogeneous: they use different sensors, labels, sampling rates, recording settings, and clinical endpoints. They can support detector design, but they do not directly specify which detector rules should be built for a new contactless monitoring platform. We report a controlled four-analyst large-language-model (LLM) workflow for converting 68 public physiological corpora, screened for commercial-use compatibility, into an auditable library of candidate rule shapes for prospective validation. Four independent commercial LLM families read the corpus documentation under a controlled prompt and produced 695 candidate rule markers (top-markers). Deduplication retained 649 rule records; a threshold-bounds audit then flagged 51 sanity violations for clamping or curator review. Cross-corpus consolidation produced 436 unique rule shapes. Gate-tagging against two hard invariants, native target-hardware channel availability and no multi-night per-patient personalization, identified 94 build-now detector components across four detector-family buckets. The pipeline does not produce a validated clinical detector. It produces an auditable engineering cascade in which analyst disagreement, threshold checks, curator review, and automated continuous-integration (CI) checks route literature-derived rules toward prospective hardware validation.

Evaluating uncertainty in AI-generated SQL queries requires estimating whether a query is correct, where correct means it executes to the same result as a human-written reference. We study which signals predict correctness on hard multi-table text-to-SQL, using AUROC to measure how well each ranks correct queries above incorrect ones. On BIRD and Spider, black-box signals such as string, structural, and execution self-consistency, a schema-relevance score, and query executability all fall between about 0.61 and 0.68 AUROC, with string self-consistency strongest at 0.675; white-box log-probability is similar (0.67). The signals that move past this ceiling are verification-based: an LLM judge scores from 0.72 (GPT-4o-mini) to 0.78 (Claude). Judges from different providers make different errors, so a two-provider ensemble reaches 0.82 AUROC with a well-calibrated probability (expected calibration error 0.03) and supports useful abstention frontiers (for example, answering 27% of questions at 24% selective risk) where self-consistency offers no valid low-risk subset. The pattern holds across two benchmarks, two generators, and two judge providers. We also ask whether a verifier can be trained. Fine-tuned verifiers, both encoder and generative, reach about 0.77 to 0.79 AUROC in-distribution but fall to about 0.66 on unseen schemas; scaling to 7B, adding schema diversity, distilling a strong judge's rationales, and cross-benchmark training all fail to close that gap. Cross-schema transfer appears to track model scale and reasoning rather than fine-tuning. In practice, correctness uncertainty for text-to-SQL lives in reasoning-based signals: a fine-tuned verifier is a good in-domain tool, but a verifier that generalizes across schemas currently means a large frozen reasoning model.

Deep learning has achieved remarkable success in various domains including time series analysis, computer vision and natural language processing. However, high computational and memory demands of state-of-the-art architectures pose challenges for deployment in resource-limited environments. Knowledge Distillation (KD) addresses this by transferring knowledge from a large teacher model to a smaller, more efficient student model while maintaining competitive performance. In this work, we investigate the effectiveness of KD for Time Series Classification (TSC) across three architectures: the classical Fully Convolutional Network (FCN), the convolutional Inception model and the transformer-based ConvTran model. We evaluate our approach on UCR Archive, the largest benchmark repository of time series datasets, by modifying architectural components such as convolutional filters, Inception modules and attention heads across the three architectures. Our results consistently show that KD most effectively benefits student models of intermediate complexity across all three architectures, with the distilled FCN student reducing parameters by a factor of 38, the distilled Inception student achieving nearly the same performance as the teacher with 42% fewer parameters and the distilled ConvTran student with 2 attention heads showing the most significant improvement through distillation. To encourage further research and reproducibility, we provide our implementation at https://github.com/MSD-IRIMAS/KD-4-TSC.

Standards bodies, including TM Forum, 3GPP, and ETSI, are converging on Agentic AI as the foundation for next-generation network management, where Large AI Model (LAM)-based agents autonomously interpret intent, coordinate resources, and adapt operational behaviors at runtime. However, achieving this vision at the scale and complexity of 6G networks requires management systems that can generate and evolve their own automation software during operation. We introduce Autogenic network management, a reference architecture that extends agentic capabilities with self-programming, self reflection, self-orienting, and self-architecting capabilities. The architecture supports practical staged deployment beginning with human-supervised LAM-based agents and progressing toward autonomous operation as confidence builds. We demonstrate the approach through high-priority operator scenarios drawn from TM Forum's autonomous network use cases, showing how autogenic management addresses real operational challenges. We conclude with a research roadmap outlining the technical advances needed to make autogenic network management realistic in future 6G networks.

In this work, we investigate the fixed-architecture neural network approximation with explicit parameter bounds and elementary activations. While prior work demonstrated super-expressive approximation using fixed-size networks, they lack quantitative and non-asymptotic characterizations of parameter magnitude with respect to the approximation error. We resolve this issue by introducing the Chinese Remainder Theorem as a constructive encoding mechanism. For Lipschitz continuous functions on $[0,1]^D$, we construct a width-$\max\{D,4\}$, depth-$5$ network with explicit parameter-error trade-offs. For Hölder-smooth functions in $C^{r,γ}_A\left([0,1]^D\right)$, our fixed network of width $\max\{2D,\ D+5N+1\}$ and depth $r + 9$ achieves the parameter magnitude $\mathcal{P}$ bounded by $\log_2 \mathcal{P}=\mathcal{O}\bigl(\varepsilon^{-2D/(r+γ)}\log(1/\varepsilon)\bigr)$. This is the dual result compared to those in the parameter-bounded and architecture-unbounded paradigm.

We introduce an efficient Bayesian deep ensemble method for predictive regression designed to enhance interpretability while maintaining competitive predictive performance and computational efficiency. Our method combines the statistical rigor of Bayesian inference with the scalability of deep ensembles, providing calibrated uncertainty estimates that enable its use not only for standalone prediction but also as a component within broader learning systems. To achieve these goals, our work relies on three key design components: (i) low-dimensional ensemble representation: predictions are expressed as a combination of a small number of trained neural predictors, enabling scalable inference whose cost depends on ensemble size rather than dataset size; (ii) closed-form Bayesian aggregation: ensemble predictions are combined using Bayesian linear regression, yielding interpretable posterior weights and calibrated uncertainty without approximate inference; and (iii) Independent ensemble training: multiple neural networks are trained separately, producing diverse predictive representations that improve robustness and uncertainty calibration. Empirical results on standard regression benchmarks demonstrate that the proposed approach achieves competitive predictive performance while maintaining reliable uncertainty estimates across settings.

Learned optimization aims to improve upon hand-designed optimizers (e.g., Adam and Muon) by meta-learning small neural network optimizers over a distribution of tasks. While recent work has greatly advanced the architectural design and inductive biases of learned optimizers (LOs), current meta-training approaches still suffer from two main difficulties: (1) they cannot efficiently scale meta-training to long-horizon inner problems and (2) they often fail to surpass comparable hand-designed optimizers. To address these limitations, we propose Efficient Long-hOrizon (ELO) learning, an efficient meta-training algorithm that (1) reallocates redundant meta-training compute to longer failure regimes, achieving efficient long-horizon learning, and (2) enforces decoupled progressive expert supervision, providing stable meta-learning signals that additionally improve the generalization of LOs. Our empirical study evaluates ELO for meta-training both element-wise and matrix-based LOs. Across downstream language modeling (GPT-2-124M/350M on FineWeb) and image classification (ViT-B/16, ResNet-50 on ImageNet-1K) tasks, ELO substantially improves the long-unroll performance and out-of-distribution generalization of the base LOs. In particular, ELO-Celo2 consistently outperforms well-tuned AdamW across all evaluated tasks, while remaining competitive with Muon on language modeling. \textit{Notably, all ELO baselines require less than 7 H100 GPU-hours for meta-training.}

This paper investigates over-the-air federated learning (AirFL) in wireless systems where the access point is equipped with a multi-waveguide pinching antenna system (PASS). We adopt the widely studied learning-oriented AirFL formulation, which seeks to maximize the number of selected devices while keeping the aggregation distortion below a prescribed threshold. The resulting joint optimization of device selection, receive beamforming, and pinching-antenna placement is highly nonconvex due to the intricate coupling among these system variables. To address this challenge, we develop AirPASS, an alternating optimization framework with two main components: a homotopy-Riemannian margin-consolidation method for device selection and receive beamforming under fixed PASS configuration, and a homotopy-assisted geometry optimization method for updating the pinching-antenna positions under fixed selected devices and beamformer. Experiments show that AirPASS consistently outperforms conventional co-located MIMO baselines, remains close to ideal FedAvg, and achieves an attractive performance-complexity tradeoff relative to SDR-DC and matching-pursuit scheduling alternatives.

At Amazon Prime Video, we face the critical operational challenge of managing code deployments during live events and rapid feature releases without causing service outages. Current change control approaches use blanket deployment freezes that block all changes regardless of risk, creating significant developer toil. While prior research has explored risky change predictors, these rely on developer-specific metadata or extensive historical data, raising privacy concerns and limiting applicability to new projects. We introduce a framework centered on diff-aware features, characteristics derived directly from code modifications. Our key contribution is the systematic identification of which quantitative metrics (code-level and change-level metrics) and qualitative indicators (coding style violations, change type classification) are necessary for risk prediction. We employ LLMs as multi-language feature extractors, demonstrating their effectiveness for code analysis beyond generation tasks and eliminating the need for language-specific tooling. We evaluated our framework on two datasets: Prime Video's production environment and the public ApacheJIT dataset. Our best-performing model achieves an average recall of 0.83 and F1 score of 0.81 across both datasets for detecting risky code changes. Notably, ablation analysis reveals that change-level volume metrics (e.g., lines added/deleted) are noisy predictors, while structural code complexity provides a substantially stronger risk signal. These results demonstrate that thoughtful feature curation enables effective change risk assessment across different programming languages and organizational contexts while avoiding privacy concerns.

Recent progress on ARC-AGI-1 from disclosed architectures has come broadly from two regimes: heavy test-time compute over frontier models (evolutionary search, exhaustive sampling, extended chain-of-thought), or benchmark-specific training in which small models are fine-tuned on ARC data, often with task-specialized architectures. We study a third regime: an open-weight model in non-thinking mode (DeepSeek V3.2) under a strict budget, with no ARC-specific fine-tuning. We study what is recoverable through architecture alone, building agentic harnesses that decompose pattern-discovery and program-synthesis stages explicitly. First, we introduce an Explorer-Definer Pipeline that separates pattern discovery from executable transformation synthesis, implemented as a two-stage agent pipeline. Next, we present the Reflective Orchestrator, which augments the pipeline with autonomous exploration of new transformations when previous hypotheses fail on training pairs. On the ARC-AGI-1 public 400-task evaluation set, the pipeline reaches 57.50% pass@2 at \$0.25 per task, and the orchestrator reaches 67.25% pass@2 at \$0.62 per task. Together these architectures lift a 15.50% one-shot baseline by ~52 points without benchmark-specific training or heavy test-time compute. Furthermore, the orchestrator-driven lift tests a falsifiable diagnostic the pipeline produces; unbiased pass@k analysis suggests the pipeline is generation-bound, not selection-bound (selection via training-pair accuracy captures ~95% of the candidate ceiling) and predicts that significant improvement requires broader generation, not better ranking. The orchestrator implements this prediction via adaptive re-exploration and confirms it (unbiased pass@1 lift +9.81 pp, matching selection-mediated pass@2 lift). An additional pipeline ablation identifies its think tool as a significant component, with removal reducing pass@2 by 5.75 pp.

Speculative decoding greatly increases the interactivity of autoregressive language models by trading off computation for extra tokens generated in a single forward pass. Factorized draft models are especially efficient because they predict future-token marginals in parallel, but their independence assumption causes acceptance rates to degrade sharply as the speculative budget grows. We analyze this limitation and introduce Weaver, a lightweight autoregressive adapter that constructs proposal trees from the top-K marginals of a factorized drafter. Weaver restores conditional dependencies between proposed tokens while avoiding a full-vocabulary projection. To support fast verification for models with Gated Delta Net layers, we derive a rollback-free tree-verification algorithm and implement optimized CUDA kernels in SGLang. By combining these model and systems contributions we achieve a 4.37-fold speedup over autoregressive decoding, and outperform a highly optimized DFlash baseline by 24.7%.

Autonomous systems under partial observability act on beliefs, not raw sensor events. QANTIS treats the quantum processor as a calibrated belief-update service in that loop: it receives a prior and an observation model, estimates the rare-event evidence term, and returns an ordinary posterior to a classical planner. This paper asks whether that service can be reused across a sequential Tiger POMDP horizon on present IBM Heron hardware without corrupting the planner-facing posterior. We answer with a controlled hardware case study rather than an end-to-end autonomy or wall-clock speedup claim. The study compares no amplification, guarded Grover amplification, and all-step fixed-point amplification on the same trajectory, then checks whether the returned posterior would change the downstream action. All-step FPAA preserves the Tiger posterior across the reported 8-step and 12-step primary runs, and the 20-step and 32-step controls remain inside the same operating band. In every reported decision check, the hardware posterior and the exact Bayes posterior select the same immediate action. Boundary-aware BIQAE stabilizes amplitude estimation near zero and near one, while a rare-event sweep maps the logical sample-complexity envelope for one-in-a-million evidence. The result is an operating envelope for a hardware-calibrated belief-update primitive, not a standalone hardware-advantage claim.

Agent-based modeling (ABM) has the capability to model millions of individuals and their interactions, which is useful for policy making. However, ABMs have traditionally relied on static prior, which prevents the models from adapting to real-time changes. Our research provides a novel approach to addressing this information gap. Large language models (LLMs) offer new opportunities to predict human decision-making. Here, we introduce a scalable Hybrid Agent-based and Language-driven Epidemic (HALE) modeling framework that leverages LLMs to predict human decision-making in an ABM simulation. As a proof-of-concept, we use HALE to simulate COVID-19 and its effects in Salt Lake County, UT.

Smart home automation platforms increasingly rely on user-authored YAML configuration files to define device behaviors, but these files are prone to syntax, formatting, and semantic logic errors that can cause automation failures and safety risks. Existing YAML validators, static analysis tools, and general-purpose large language models offer limited support for end-to-end diagnosis and repair because they lack domain-specific understanding and validated correction workflows. This paper presents SmartHomeSecure, a prototype for automated detection and repair of Home Assistant configuration errors using lightweight program analysis and constraint-guided large language model generation. SmartHomeSecure parses YAML files, detects syntactic and common semantic errors, normalizes error context, applies deterministic auto-fixes for routine defects, and constructs constrained prompts that guide LLMs toward minimal and structurally valid repairs. The system is implemented as a modular web application with four layers: UI Shell, Feature Orchestrator, Domain Engine, and Integration Layer. Its repair pipeline was evaluated on 100 real-world Home Assistant YAML files with manually injected errors across five categories: syntax/parsing, indentation, mapping, sequence, and scalar quoting errors. Four models were tested: gpt-oss-20b, gpt-oss-120b, llama-3.1-8b, and llama-3.3-70b. Results show that three models achieved 100% error detection accuracy, with repair success rates ranging from 87% to 93%. Manual verification found no hallucinated or incorrect repairs among successful outputs. These findings suggest that combining domain-aware program analysis with constrained generative AI is a feasible approach for improving the reliability and usability of smart home configuration repair.

Soft robots have attracted significant attention in applications such as medical intervention, rehabilitation, and robotic manipulation due to their inherent compliance, flexibility, and high degrees of freedom. Modular soft robots (MSRs), composed of multiple interconnected segments, represent an emerging class of robotic systems with highly deformable and reconfigurable structures capable of performing complex tasks. However, designing controllers for MSRs remains challenging due to their nonlinear dynamics, modeling complexity, and hyper-redundant nature. Existing approaches typically require controllers to be retrained from scratch whenever the robot morphology changes. In this work, we address these challenges through a continual learning inspired control framework capable of incrementally adapting to changes in robot morphology while preserving previously acquired knowledge. Specifically, the proposed framework enables the controller to sequentially learn new MSR configurations without forgetting previously learned ones. In addition, for MSRs with fixed configurations, the same framework can be employed in a distributed manner to learn module-specific dynamics, enabling localized control and improved precision. The proposed approach is validated through closed-loop trajectory tracking experiments in simulation using a tendon-driven soft robot, as well as on a real-world three-module pneumatic soft robotic arm. Furthermore, we demonstrate the adaptive capabilities of the framework through a reaching experiment in which the controller selectively activates only the necessary modules to reach a virtual target position, thereby reducing computational overhead.

The Beagle framework is a GPU-based genetic programming framework that enables highly efficient genetic programming search using large population sizes by leveraging NVIDIA GPUs. This technical guide provides an introduction to the Beagle framework and provides detailed instructions for using the framework for symbolic regression problems.

Most gradient-based optimization methods move parameters through a fixed background geometry, even when their internal states implicitly define changing notions of length, curvature, and preconditioning. We introduce optimization geometrodynamics, a benchmark language in which optimization is a coupled evolution of a parameter trajectory, a transported distribution of particles, and a controlled time-varying Riemannian metric. The language separates invariant obstructions from improvable geometric mismatch: positive metrics preserve critical points and Morse indices, and cannot remove global geodesic-convexity obstructions, but can alter conditioning, distributional transport, and flux away from exact critical points. We introduce dynamic geometric complexity, the minimum geometric cost required to reduce an optimization difficulty observable. In the oracle benchmark model of strongly convex quadratic objectives with full positive-definite metric control, this complexity is exactly the affine-invariant distance from the relative log-spectrum to a low-condition-number set. We also analyze Hessian-matching flows, spectral Onsager relaxation, discrete exponential projection updates, gauge-invariant observables, and fixed-time local Morse-saddle flux. The paper is theory-only: its claims are formal statements with proofs, intended to provide invariants and benchmark costs against which implementable adaptive optimizers can be compared once their admissible metric families, curvature estimates, and discretization errors are specified.

Training large language models (LLMs) with extended reasoning has enabled in-context search, in which models iteratively generate, critique, and revise solution attempts. We provide a theoretical analysis of in-context search by modeling it as approximate inference over reasoning traces, where the base model defines a prior and self-reflection provides feedback for posterior updates, and study the resulting inference-time sampling complexity - the number of sequential attempts needed to achieve high success probability. We show that when reflections reliably localize early mistakes, in-context search can yield exponential improvements over the base model, solving problems with exponentially small zero-shot pass rates using only a polynomial number of sequential attempts, whereas when this property fails, conditioning on past attempts offers no asymptotic benefit over parallel sampling. We further show that these gains are robust and learnable: approximate posterior updates suffice, and cross-entropy training on search rollouts recovers the required behavior with polynomial sample complexity. Finally, we show that under a stagewise abstraction of reinforcement learning with verifiable rewards, the optimal policy extension implements the same posterior reweighting rule. We validate key qualitative predictions of the theory on real large reasoning models.

This paper introduces a graph-theoretic approach for predicting market regimes in foreign exchange (FX) currency prices. Specifically, the proposed model incorporates exogenous macroeconomic variables to update localized node features via message-passing operations. Utilizing the Graph Tsetlin Machine (GraphTM) framework, we empirically demonstrate the efficacy of this approach in anticipating market regimes for the US Dollar and Japanese Yen currency pair (USD/JPY). By representing multivariate macroeconomic drivers and technical indicators as hypervectorized directed multigraphs, the GraphTM leverages structured message passing to construct deep, interpretable logical clauses capable of recognizing complex sub-graph patterns.

Large language models are rapidly moving towards closing the development cycle, transitioning from simple assistive companions to autonomous contributors deeply embedded into collaborative development environments. Despite their accelerated adoption, existing evaluation techniques are limited due to their fragmented nature and distorted projection of true model capabilities, often obtained from hypothetical syntactic scenarios. This research aims to bridge this gap by providing a comprehensive evaluation methodology for LLM-powered agents that is grounded in real-world software development practice. Our evaluation approach focuses on contamination-awareness, in-the-wild agentic behavior assessment, and trajectory-aware benchmarks and metrics capturing realistic coding contexts, human-aligned behavior, and model failure modes.

Vision Language Action (VLA) models unify visual perception, natural-language understanding, and action generation within a single foundation model, allowing a robot to follow instructions such as fold the towel or fly to the red building directly from camera images. Because VLAs inherit world knowledge from internet-scale pre-training, they have become the dominant framework for learning-based manipulation, with bimanual coordination serving as the most demanding testbed: two arms with 7 degrees of freedom each must move in concert to fold, assemble, and reorient objects. Unmanned aerial robotics faces a structurally similar challenge: a drone must coordinate thrust, attitude, and increasingly gripper commands from visual observations under strict latency and payload constraints. This review covers 183 contributions spanning 2017-2026 and organized along seven dimensions: VLA architectures, training recipes, action representations, bimanual coordination (2022-2026), unmanned aerial vehicle (UAV) navigation and control (2017-2026), language grounding, and cross-cutting concerns including memory and world models. We show that the coordination strategies, training recipes, and action representations developed for bimanual VLAs transfer to unmanned aerial systems and identify fourteen research directions across both domains.

Interactive simulators have become powerful tools for training embodied agents and generating synthetic visual data, but existing photorealistic simulators suffer from limited generality, programmability, and rendering speed. We address these limitations by introducing SPEAR: A Simulator for Photorealistic Embodied AI Research. At its core, SPEAR is a Python library that can connect to, and programmatically control, any Unreal Engine (UE) application via a modular plugin architecture. SPEAR exposes over 14K unique UE functions to Python, representing an order-of-magnitude increase in programmable functionality over existing UE-based simulators. Additionally, a single SPEAR instance can render 1920x1080 photorealistic beauty images directly into a user's NumPy array at 73 frames per second - an order of magnitude faster than existing UE plugins - while also providing ground truth image modalities that are not available in any existing UE-based simulator (e.g., a non-diffuse intrinsic image decomposition, material IDs, and physically based shading parameters). Finally, SPEAR introduces an expressive high-level programming model that enables users to specify complex graphs of UE work with arbitrary data dependencies among work items, and to execute these graphs deterministically within a single UE frame. We demonstrate the utility of SPEAR through a diverse collection of example applications: controlling multiple embodied agents with distinct action spaces (e.g., humans, cars, and robots) across several in-the-wild UE projects; rendering photorealistic city-scale environments; manipulating UE's procedural content generation systems; rendering synchronized multi-view images of detailed human faces; coordinating an interactive co-simulation with the MuJoCo physics simulator; and editing scenes with natural language via an AI coding assistant.

Weighted empirical measures on compact manifolds arise in importance sampling, particle approximations, posterior summaries, quadrature, and representation learning. Standard weight-only summaries, such as ordinary effective sample size, ignore the geometry of the support. We introduce heat-kernel entropy profiles, a multiscale summary that diffuses weighted atoms by intrinsic heat flow and tracks nonuniformity across scales. For order-two Rényi entropy, the profile is computable from pairwise heat-kernel overlaps and yields a geometric effective sample size that discounts nearby or duplicate particles while matching ordinary effective sample size for well-separated particles. We prove monotonicity, small- and large-scale asymptotics, deterministic-weight consistency, and a bounded-ratio self-normalized importance-sampling extension for compact manifolds without boundary. On spheres, the unlogged profile decomposes into spherical-harmonic energies that recover mean-direction, von Mises-Fisher-type, and Bingham-type summaries. Sphere-based experiments show that the profile reveals antipodal, girdle, multimodal, and duplicate-particle structure missed by weight-only and first-moment spherical summaries.

Finance, sensing, and demand streams violate the exchangeability that IID conformal prediction and the IID bootstrap assume, and existing libraries implement either a general resampling engine or conformal calibration without the other. tsbootstrap provides block, residual, sieve, and wild resampling, classical bootstrap confidence intervals, and adaptive conformal calibrators (EnbPI, ACI, NexCP, AgACI) through a single typed API in which a specification object selects each method. In a controlled coverage study the IID bootstrap undercovers sharply under dependence; dependence-aware methods reduce the coverage deficit, the sieve nearest to nominal under short-memory linear dependence. On the shared fixed-statistic path a compiled backend runs several times faster than arch, and a streaming reduce avoids materializing the $O(Bn)$ replicate tensor, limiting peak extra memory to $O(B)$ for the statistic array. The software is MIT licensed (v0.6.1).

Knowledge workers switch between applications thousands of times per day, spending nearly a tenth of the work year transitioning between digital applications in a process called digital fragmentation. Whether this fragmentation reflects who an employee is, where they work, or what kind of day they are having, has remained an open question. We analyzed 103 million application events recorded second-by-second from 1,017 employees across eight organizations that largely employ knowledge workers (e.g., law, financial services). Day-to-day variation in fragmentation within individual employees accounted for 44.6% of the variation in digital fragmentation, slightly exceeding stable individual differences between employees (35.8%), and far exceeding variation between organizations (19.6%). Fragmentation rose over the work week and reset after weekends and holidays. Higher-than-typical use of communication applications coincided with more fragmented work. Generative AI use also occurred on more fragmented days, but the period following AI use was marked by narrower, longer, and more predictable application use. These findings identify the workday as a key level for understanding and intervening on digital fragmentation and suggest that AI may help structure fragmented work rather than merely intensify it.

Unified 3D foundation models aspire to generate 3D assets and reason about them in language within a single backbone, but their text-3D interaction remains largely implicit. Existing methods concatenate text and 3D tokens into a flat sequence and rely on self-attention, collapsing coarse structural cues and fine geometric details into one undifferentiated representation. We introduce ELSA3D, a unified 3D model that addresses this with elastic semantic anchoring, structuring language and geometric reasoning jointly along matched abstraction scales. ELSA3D represents geometry with a scale-aware octree tokenizer and introduces Anchor Tokens, sparse cross-modal units that select semantic cues, route them to the most relevant 3D scale, retrieve scale-specific geometric evidence, and write the fused signal back into the unified representation, keeping interaction sparse yet precise. A lightweight per-block router makes both computation and reasoning elastic, choosing which text tokens instantiate anchors at which geometric scale so that cross-modal capacity concentrates where alignment is most needed. ELSA3D achieves state-of-the-art performance across image-to-3D generation, text-to-3D generation, and 3D captioning, outperforming the strongest unified baseline while roughly halving FLOPs and inference latency relative to the non-elastic version of the same model.

Machine learning models are often intended to augment rather than replace human decision makers, by providing information that is complementary to human judgement. Yet, in practice, human decision makers routinely fail to realize such complementary gains, even when models provide useful signal. In this work, we study how asymmetric information about the quality of information available to a human decision maker vs. an AI impacts the ability of a decision maker to extract complementary value from AI predictions. We show that a key factor is the error correlation structure between human and AI predictions. In particular, when the AI's prediction errors are \textit{negatively correlated} with those of the human, the decision maker can construct robust strategies which guarantee improvements in expected utility. We empirically investigate whether these conditions for complementarity arise in practice, using real-world forecasting benchmarks.

Denoising graphs is a fundamental problem in graph learning and the core operation of graph diffusion models. Attention-based architectures like graph transformers have recently shown promise in denoising graphs. However, our principled understanding of attention-based graph denoising remains limited, making it unclear whether standard attention is the right mechanism for this task. Here we show that, under a denoising objective, linear attention is suboptimal and can only learn an average spectral denoising filter over the training distribution. This creates a fundamental limitation as graphs often vary spectrally across the distribution. To overcome this limitation, we introduce Spectral Attention, which directly utilizes the input graph spectrum and provably outperforms linear attention by a margin governed by the spectral diversity of the distribution. We then derive Graph Convolutional Attention (GCA), a practical and permutation-equivariant realization of this idea that implements spectral denoising through graph-filtered queries and keys. For stochastic block models, GCA provably matches the idealized Spectral Attention mechanism. We further show that the softmax operation, that follows the attention, provides additional denoising by approximately projecting noisy eigenvectors onto the clean eigenspace. Empirically, replacing linear attention with GCA consistently improves graph denoising and diffusion on synthetic and real datasets, with gains strongly correlated with spectral diversity. In DiGress, GCA matches standard graph-transformer performance without computing expensive structural features, and when combined with the recently proposed PEARL positional encodings, avoids explicit eigendecomposition computations resulting in faster inference without degrading quality. The code can be found here: github.com/shervinkhalafi/graph_conv_att

As Artificial Intelligence (AI) makes inroads into different parts of the Indian subcontinent, there is significant interest in studying how AI impacts the linguistic and cultural foundations of this civilization. AI is seen as a ''double-edged sword'' where on the one hand, it can enable access and inclusion for a large population, on the other, it can homogenize worldviews and exclude underrepresented languages and worldviews. In this paper, we try to characterize this problem by addressing the extensive characteristic nature of Indian linguistics and the way they closely connect to cultural practices and worldview. We then perform a longitudinal survey of how Natural Language Processing (NLP) techniques have evolved in this space, tracing the historical development of Indic NLP, covering key milestones, methodological shifts, and resource creation efforts. In addition, the paper also examines the structural and sociolinguistic characteristics of Indian languages, such as rich morphology, complex scripts and grammar rules, diglossia, and large dialectal variation, and explains how these create unique challenges for building AI foundation models. We then discuss the growing role of Indic foundation models and analyze how these models address these long-standing resource and representation gaps. Finally, we propose a research direction called 'Culture Sensing', which re-imagines AI based on hermeneutic reasoning. Culture Sensing aims to address open problems such as ensuring equitable performance across low-resource languages and producing outputs that are culturally meaningful. By bringing together past work, current techniques, and emerging trends, this paper outlines research directions that can guide the next phase of Indic NLP and contribute to the development of more robust and inclusive Indic foundation models.

In this report, we present Pelican-VLA 0.5, a unified VLA model that integrates vision-language understanding, future-frame generation, and action prediction within a single architecture. Pelican-VLA 0.5 achieves attention-level generalization: without object annotations, segmentation masks, attention supervision, or task-specific fine-tuning, its action pathway already focuses on the instruction-relevant object and contact region. This behavior persists across unseen scenes and unseen robot embodiments, and is substantially stronger than in other open-source VLA baselines. We verify that this ability originates from the learnable Reasoning Slots inserted between perception and action: by routing task-relevant visual information through a compact bottleneck, the slot interface induces manipulation-centric attention during pre-training and remains effective across different policy structures, including a MoT-style architecture.

Dependency parsing consists of finding a tree representation for a sequence. Unsupervised dependency parsing aims to develop parsing methods without a gold standard during model training. In human languages, an unsupervised parser can be evaluated because some gold standard is usually available or can be created. For other species, a gold standard is unknown. Thus one may conclude that it is impossible to determine the accuracy of an unsupervised parser and, consequently, dependency parsing is unfeasible in other species. However, here we apply recent advances in network science to demonstrate that the proportion of correct edges retrieved by a parser must be high for the sequences of vocalizations or gestures that non-human primates produce due to the fast decay of the sequence length distribution. In contrast, human language sequences lack that property. Therefore, evaluation without a gold standard is feasible in non-human primates but a hard problem in humans.

Developing seamless, high-performance, native intelligent full-duplex Spoken Language Models (SLMs) remains a critical challenge and long-standing goal for the speech and NLP community. Despite notable progress, recent endeavors are fundamentally constrained by severe modality interference, which causes substantial knowledge degradation and compromises semantic integrity -- ultimately making full-duplex SLMs feel unnatural and unintelligent. In this paper, through an exhaustive fine-grained analysis of model optimization dynamics, we uncover the root cause of such performance degradation, revealing that modality interference arises from inherent gradient conflicts between acoustic and semantic modeling when the two modalities are forced to share a deep parameter space. Guided by this key insight, we introduce Lychee-FD, a native end-to-end full-duplex framework designed to mitigate modality interference. Importantly, we propose a hierarchical parameter separation strategy that decouples conflicting modalities in deep layers while preserving cross-modality coherence via a dedicated semantic alignment channel. Extensive experiments on multiple full-duplex benchmarks demonstrate that our method significantly advances the state of the art, yielding substantial improvements in both speech intelligence (+7.4% on Spoken QA) and full-duplex interaction fluidity (+28.5% on FullDuplexBench 1.5) without compromising inference efficiency. To the best of our knowledge, this work is the first to achieve two key advances: 1) uncovering and elucidating the root cause of modality interference in full-duplex SLMs, and 2) designing an elegant hierarchical model together with a practical solution for seamless, high-performance, native intelligent full-duplex SLMs.

Federated learning (FL) enables collaborative model training across institutions without sharing sensitive patient data. However, the inherent heterogeneity of electrocardiogram (ECG) data across healthcare providers presents significant technical challenges for robust classification. We propose FedDualAtt, a personalized federated learning approach that splits transformer attention heads into global and local branches. Global heads are aggregated via FedAvg to capture shared cross-site patterns, while local heads remain client-specific to adapt to institution-level recording characteristics. Experiments on FedCVD, an FL benchmark for cardiovascular disease detection, demonstrate that FedDualAtt outperforms existing FL and personalized FL methods in ECG classification tasks. Analysis of global-local head ratios reveals that different clients benefit from varying levels of architectural personalization.

Mixed-integer linear programming (MILP) instances used for solver development are hard to obtain when models come from private or application-specific pipelines. A generator must keep the structure that solvers and learned policies rely on. Existing general generators usually choose their generation unit from a formulation template, summary statistics, local graph edits, or blocks found after recombination. These units do not explicitly record how a local part of the MILP is coupled to the rest of the instance. We propose GraphBU, a graph-native generator whose basic unit is a local subproblem plus its interface. The method promotes coupling nodes into master constraints or boundary variables and uses the resulting block units for compatibility-checked replacement. The analysis focuses on the properties needed by this construction: promotion separates interfaces, replacement can preserve feasibility under an interface-slack condition, and the graph construction is invariant to row-column permutations. On MILP instances generation, this unit keeps graph statistics close to the source family, preserves feasibility on most datasets, and improves downstream Predict-and-Search training. Genrated by GraphBU, The average graph-statistical similarity was approximately 0.934, the average feasibility was approximately 96.7%, and the average increase in the main index of downstream PS was approximately 8.0%.

- Objective: Multimodal deep learning models in oncology are currently limited by monolithic designs that rigidly couple data ingestion, clinical routing, and artificial intelligence (AI) inference. To address this inflexibility, we propose the Large Cancer Assistant (LCA), a model-agnostic, post-hoc orchestration framework designed for scalable clinical decision support. - Methods: The LCA is mathematically formalized as a 7-tuple architecture grounded in the principle of Algorithmic Impermeability, ensuring the orchestration logic remains strictly independent of underlying black-box AI models. We introduce the Entry Theory, leveraging Geometric Deep Learning (GDL) to standardize multimodal patient data along distinct structural and medical axes. The system dynamically orchestrates data via a Cancer Switching Module and intentionally isolates the core AI execution from volatile hospital IT infrastructures by outputting a Standardized Intermediate Payload (SIP). - Results: A Proof of Concept (PoC) validated the orchestration logic across four technical scenarios. The framework executed a nominal flow with negligible orchestration overhead. It empirically demonstrated algorithmic impermeability by maintaining an invariant routing projection during AI model swaps, and it validated strict failure-safety by achieving a 100\% recall rate in generating targeted Supplementary Data Requests (SDR) under injected data anomalies. Multi-protocol execution capability was also successfully verified. - Conclusion: By structurally decoupling multimodal ingestion from feature inference, the LCA provides a highly adaptable and modular orchestration foundation. The SIP establishes a clear architectural boundary, natively setting the stage for downstream Electronic Medical Record (EMR) interoperability as an independent future paradigm.

Fine-scale socioeconomic information is often unavailable across rapidly ur-banizing regions of the developing world, like India, limiting the ability to delineate intra-urban variations in affluence and deprivation. This study pro-poses a scalable, grid-based urban delineation framework using building morphology derived from open-source satellite imagery. Urban areas across 59 Indian cities and towns are partitioned into high-resolution spatial grids and characterized using interpretable morphological indicators, which are combined into a transparent, rule-based scoring framework to delineate areas with contrasting levels of urban affluence. The resulting classifications are validated through ground-level Google Street View observations, revealing a sharp contrast between the grid classes which are consistent with the ex-pected effects of the lifestyle affluence indicators. We further investigate density-based clustering of building footprints in Mumbai to identify dense urban settlements, demonstrating that the resulting clusters exhibit substan-tial spatial overlap with known informal settlements across the city. Finally, we conduct an exploratory analysis mapping consumer loan delinquency across the derived affluence classes. By relying entirely on publicly available geospatial data, the proposed framework provides a scalable, interpretable, and cost-effective approach for granular urban affluence mapping across In-dian cities.

Multi-hop Question Answering over Knowledge Graphs faces a critical challenge: traditional retrieve-then-read pipelines break differentiability, preventing the retriever from learning to bridge the semantic gap where intermediate nodes lack lexical overlap with the query. To address this, we propose RSF-GLLM, a framework decoupling differentiable graph reasoning from answer generation. Our Recurrent Soft-Flow (RSF) module employs a GRU-guided query updater to propagate continuous relevance scores, utilizing a dynamic gating mechanism to traverse semantically dissimilar bridge nodes via structural cues. We introduce flow sparsity regularization to theoretically guarantee convergence from soft probabilities to discrete reasoning paths. These paths are extracted and textualized to fine-tune a Large Language Model (LLM), ensuring generation is grounded in factual topology. Experiments on WebQSP and CWQ demonstrate that RSF-GLLM achieves competitive performance with superior inference efficiency compared to LLM based computationally expensive approaches.

Long-context language model inference is increasingly limited by the memory bandwidth and capacity required to store key-value caches, yet existing compression methods often apply uniform budgets across layers or tokens and degrade retrieval when lexical cues and semantic states require different preservation. We introduce DepthWeave-KV, a token-adaptive cache compression method that factorizes key and value states across neighboring transformer layers using shared low-rank channel bases while retaining lightweight token-specific residuals where attention behavior is sensitive. DepthWeave-KV combines cross-depth residual factorization with a token-conditional depth router that allocates higher reconstruction rank to instruction-bearing and retrieval-critical tokens, and uses calibration-free online error tracking from attention-output probes to adapt compression during generation without retraining the base model. A fused CUDA implementation jointly performs basis lookup, residual dequantization, and attention projection to reduce decode-time memory traffic. Across LongBench, Needle-in-a-Haystack, L-Eval, and long-form QA and summarization benchmarks, DepthWeave-KV achieves near-full-cache task quality with substantially lower memory use, improving average score and retrieval accuracy over prior compressed caches while reaching 8.3x KV memory reduction and 72.8 tokens per second at 64K context.

Vision-language models (VLMs) struggle to generalize in interactive physical reasoning, particularly under unseen tasks and environments. Two key failure modes are prominent: hallucinated chain-of-thought (CoT) reasoning that contradicts physical reality, and misalignment between the model's reasoning and actions. We present VAORA (Visual Action Outcome Reasoning Alignment), a novel reward design that directly addresses both issues. VAORA introduces two complementary rewards: Visual Alignment Reward, which anchors VLM reasoning to the visual context independent of the agent action itself, and Visual-Action Alignment Reward, which grounds reasoning in the visual outcome induced by the model's action. Together, these rewards suppress hallucinated CoT and reduce the gap between reasoning and behavior. To improve training stability, we further employ smooth, dense rewards by estimating success probabilities using a pre-trained in-domain expert agent. Experiments on PHYRE and Virtual Tool support our performances across novel-task and unseen-environment settings, confirming that grounded and generalizable physical intelligence can be induced through VAORA.

Rough path signatures are a universal feature map for continuous paths and, via the expected signature, characterise path distributions. These guarantees do not directly extend to cadlag paths of Temporal Point Processes (TPPs), limiting the use of signature methods for event sequences. Furthermore, neural TPP models, including recent generative approaches, optimise per-event objectives with no global sequence-level loss, while evaluation of variable-length event sequences lacks distributional discrepancy measures. This paper proposes a common pathwise framework for addressing these limitations. We introduce the interarrival embedding, a stable, injective lift from jump paths to continuous paths of bounded variation, extending signature methods to discrete event sequences. Our theoretical contributions give rise to sigTPP, the first signature-based generative model for TPPs, trained using a path-level loss on complete trajectories. We further analyse the space of counting paths and derive three distributional discrepancies, providing mathematically justified tools for evaluating generative TPP models. Across synthetic and real-world datasets, sigTPP achieves the best average rank based on eight complementary metrics, outperforms or is within a standard error of the strongest baseline in 64% of the dataset-metric pairs, and according to a relative score, improves against every baseline by at least 19% on average.

Long-context LLM inference is increasingly limited by the memory and bandwidth cost of KV caches, yet aggressive compression can remove the layer-specific evidence needed for retrieval and multi-step reasoning. We introduce FreqDepthKV, an inference-time cache compression method that factorizes adjacent-layer KV states into shared low-frequency depth components and sparse high-frequency residuals. A lightweight online probe assigns attention heads to shared-depth, residual-depth, or exact cache modes according to their contribution to reconstruction-sensitive attention logits, allowing the compression policy to adapt to prompt structure without retraining. Across long-context question answering, needle retrieval, summarization, and code generation benchmarks, FreqDepthKV preserves task accuracy under substantially smaller cache budgets. With a 32k-token prefill window, FreqDepthKV reaches 58.3 Exact Match, 63.0 F1, 32.5 ROUGE-L, and 48.1 pass@1, closely matching full KV while outperforming prior compressed-cache methods. It also improves decoding throughput to 70.4 tokens/s, reduces TTFT to 2.06 seconds, and lowers peak KV memory to 6.2 GB, achieving a 3.9x effective compression ratio.

Federated learning (FL) over mobile and edge devices increasingly involves multimodal models in which clients differ in both sensing capability and computational capacity. Existing update compression schemes typically apply uniform policies across layers and devices, without accounting for modality-specific differences in spectral structure and compressibility. We propose MESH-FL, an entropy-guided matrix product state (MPS) update-compression framework for modality-heterogeneous FL on resource-constrained devices. MESH-FL estimates the spectral entropy of each layer-wise update via truncated singular value decomposition and allocates MPS compression ranks adaptively across layers, modalities, and devices under per-client payload budgets. We show that higher spectral entropy necessitates a higher reconstruction rank under the majorization order on singular-value energy distributions. Building on this result, we prove that the proposed entropy-guided allocation solves a convex surrogate rank-allocation problem, preserves monotonicity under the exact payload model, and achieves convergence with an explicit compression-dependent error term. Experiments on a 15-node heterogeneous Raspberry Pi~4/5 cluster with modality-heterogeneous clients show that MESH-FL achieves up to $56.8\times$ compression while surpassing the uncompressed FedAvg baseline in final accuracy by up to 2.01%, and reduces total transmitted data to reach convergence by up to $66\times$.

We present FootsiesGym, an open-source environment for learning in a non-trivial two-player, zero-sum, imperfect-information game. Built on HiFight's minimalist 2D fighting game Footsies, it isolates the cyclic, non-transitive strategic interactions of fighting game neutral play while remaining simple enough for efficient analysis. We provide a vectorized simulator that enables high-throughput training on standard hardware, making the environment accessible and reproducible. We describe the design of the environment, benchmark several reinforcement learning algorithms, and discuss open research directions it enables. The code is available at https://github.com/como-research/FootsiesGym.

Multi-hop retrieval-augmented generation (RAG) acquires evidence sequentially, with each new document potentially revealing missing facts, bridge entities, query defects, or sufficient support for answering. Existing methods provide useful operations such as iterative retrieval, query reformulation, evidence critique, and sufficiency judging, but typically organize them within method-specific pipelines or predefined control topologies. This leaves underexplored how to learn a shared state-conditioned policy that chooses among currently valid evidence operations. We introduce DynaKRAG, which formulates multi-hop evidence acquisition as state-conditioned control over atomic evidence operations. At each step, a validity layer constructs the executable action set, and a learned controller selects the next operation. The resulting transition updates the evidence state and may enable new operations at subsequent steps. With Qwen2.5-7B-Instruct, DynaKRAG achieves F1 scores of 0.5998 on HotpotQA, 0.5340 on 2Wiki, and 0.3061 on MuSiQue, outperforming the strongest controlled baseline on all three benchmarks. Replacing the learned controller with a uniform-valid policy reduces F1 by 3.96--5.78 points, while removing sufficiency feedback hurts all three datasets. Controlled retrieval-cap experiments further show that additional retrieval is not uniformly beneficial. Together, these results demonstrate the benefit of coordinating retrieval, diagnosis, and gap-directed acquisition under an evolving evidence state.

GitHub hosts hundreds of millions of public repositories, but the platform exposes no native mapping from repositories to standardized industry sectors. This gap limits empirical work on the geography of innovation, the industrial composition of open-source production, and the diffusion of new technologies across economic sectors. We present NAICS-GH, a publicly released corpus of 6,588 GitHub repositories drawn from source pools covering the United States, the European Union, and Australia, each labeled with a 2-digit sector from the North American Industry Classification System (NAICS 2022). Labels are produced by a retrieve-and-verify pipeline that combines BAAI/bge-large-en embeddings, FAISS retrieval, and GPT-4.1 rubric scoring. The pipeline narrows about 1.37 million source repositories to 31,178 candidate repository-sector pairs and retains 6,588 high-confidence labels with score at least 8. Re-running the retrieval pipeline end to end reproduces the candidate set to within 0.03 percent. On a 2,421-repository human-validated random sample, the released labels attain 96.98 percent precision, with Wilson 95 percent confidence interval [96.23, 97.59]. We benchmark six pretrained encoders on the released corpus; RoBERTa-large reaches 86.45 percent F1 and 86.35 percent accuracy on a held-out 20 percent test set. The dataset, Croissant metadata, pipeline code, prompts, and fine-tuned checkpoint are released under CC-BY-4.0 and MIT licenses.

Recent years have witnessed the emergence of multivariate modeling using time series foundation models (TSFMs), which achieve advanced zero-shot generalization. Modern multivariate TSFMs are predominantly pretrained on multivariate synthetic data, which is easier to scale but may fail to capture the complex temporal dynamics and cross-variable relationships present in real-world time series. This raises a key question: Whether and to what extent the leading TSFMs trained with the real-world corpus perform better than those trained with synthetic data? To answer this, we establish the RMISC corpus, a considerably large-scale, high-quality, openly accessible, real-world, and multivariate time series archive that contains around 200 datasets and 142 billion time points across diverse domains. Furthermore, we pretrain four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data and evaluate their zero-shot generalization capabilities on standard in-distribution and out-of-distribution benchmarks. Experimental results show that incorporating real-world multivariate data predominantly improves the generalization performance for both univariate and multivariate TSFMs. These results provide a deeper understanding of how real-world multivariate data contributes to the development of stronger TSFMs.

Large language model (LLM) agents solving multi-step tasks frequently commit to trajectories that are doomed to fail, yet continue to consume substantial inference compute before the failure becomes observable. We show that failure is predictable early from the agent's internal representations: lightweight per-round probes on hidden activations anticipate eventual episode failure as early as the first interaction round, where scorers reading only the agent's observable behavior are barely better than chance. We turn this signal into a practical abort cascade: one distribution-free calibrated gate per round, with per-round recall budgets jointly searched so that eventually-successful episodes survive all gates at a user-specified global rate; this episode-level guarantee is the one that matters in deployment, since false-abort risk accumulates across gates. Across two agent models on TextCraft, the cascade meets every recall target from 90% to 97% and, at the 90% target, saves 47.1% +/- 10.3% (Qwen-2.5-7B) and 37.2% +/- 8.8% (Llama-3.2-3B) of inference compute, 1.6--1.7x the best single-gate policy. An otherwise-identical cascade reading only behavior saves roughly half as much, and adding behavioral features to the probe yields no further gain: the hidden states capture what behavior reveals. Finally, we characterize the sample complexity of certifying high recall targets, telling practitioners which recall promises their data can, and provably cannot, back. The code will be released soon.

Live sports commentary is grounded generation under a deadline: statements concern real, named athletes, the grounding state changes every few seconds, and no reference text exists at generation time. We present Pitwall, a production system that generates natural-language Formula 1 strategy briefings in English, Spanish, and Portuguese, treating faithfulness as an architectural property rather than an aspiration: every published sentence is decomposed into typed factual claims (positions, gaps, tyres, pace, overtakes, race control) and each claim is verified against the probabilistic race state that prompted it. The same verifier gates the fine-tuning data: of 3,045 model-written targets, only the 81.9% whose every claim is state-supported are retained, the rest falling back to a provably faithful template, so the generator never sees an ungrounded target. Verification is meaningful because of the grounding substrate: a vectorized Monte Carlo engine (N=2,000 per-lap race continuations) calibrated on 126 races (2018-2024) and validated on fully held-out 2025-2026 seasons (winner-in-top-3 90.3% over 155 backtests; held-out Brier 0.0745). A recurring finding spans both halves of the system: virtues trade off and must be gated separately. In simulation, calibration-optimal is not decision-optimal; in generation, fine-tuning on richer targets buys vividness that collapses into hallucination when the grounding state is sparse -- a failure a four-base replication traces to base-model instruction adherence, not scale, and that sparse-context auditing removes from the production model. End-to-end operation -- live timing to verified trilingual briefings -- was confirmed at two consecutive live Grands Prix (Austria and Britain, 2026); at Silverstone a timestamped probability trace, committed to disk before the outcome was known, locked onto the eventual winner ten laps before the flag.

The dairy industry in Ireland has a large potential for the integration of renewable energy and the reduction of carbon emissions. However, researchers of distributed generation control are mainly focused on residential and commercial applications. To contribute to the effective integration of renewable energy in the dairy sector, this paper presents a multi-objective optimisation control system based on differential evolution and multi agent Deep Reinforcement Learning. The proposed control is organised in two layers: the upper layer uses dynamic pricing, and the lower layer is based on multi-agent reinforcement learning for battery management. This paper also simulates the electrical response of the proposed control system in a rural distribution circuit. The simulation results show that the proposed control framework can improve profits from energy arbitrage up to 18% compared to using Rule-based models, increase the use of distributed generation without significantly increasing cost, and comply with the Irish grid code in terms of voltage variation.

Vision-language models (VLMs) are increasingly deployed on infrared (IR) remote sensing imagery in security-critical settings, yet their adversarial robustness remains unexamined. We present AirflowAttack, to our knowledge the first adversarial attack for IR remote-sensing VLMs and the first to weaponize thermal-airflow turbulence as the perturbation prior. A lightweight generator synthesizes a single input-agnostic perturbation regularized toward physically plausible airflow patterns. Optimized on one surrogate CLIP model, it attains a mean zero-shot scene-classification attack success rate (ASR, the fraction of samples whose top-1 class changes) of 48.5% across five diverse CLIP backbones, far exceeding four IR-specific physical baselines (27.7--37.0%). Applied to six state-of-the-art VLMs, it cuts scene-classification accuracy by up to 38.2% relative, yet paradoxically makes some models more confident in their IR analysis, confabulating the perturbation as genuine thermal evidence such as temperature gradients and convection. Ablations show the airflow prior raises physical plausibility at no measurable cost to attack success. Together with a benchmark spanning eleven models and four tasks, these findings expose critical vulnerabilities in the rapidly expanding IR VLM ecosystem.

Poisoning attacks against public datasets lead to major concerns, such as (i) misclassification of perceived objects when the poisoned data is used for training and (ii) embedding of backdoors that may eventually be triggered later on, when specific conditions in the system apply over the learned models. Its impact over data augmentation models is unclear. While data augmentation reduces the likelihood of poisoning attack success, some valid questions remain. Is data augmentation affecting the impact of poisoning attacks? can it increase the number of poisoned samples or injected backdoors? We explore in this paper some of these questions. We assess the effects of augmenting poisoned 3D point cloud datasets and validate that poisoning is able to evade the sanitizing nature of augmentation techniques when using the concrete case of Generative Adversarial Network (GAN) techniques to exemplify the case of data augmentation processing. We also validate that poisoning propagates over the augmented datasets and perturbs the decision made by general-purpose classifiers, in the end. All the experimental material (including tools, datasets, and classifiers) is publicly available, to facilitate reproducibility and to foster further research in the topic.

Current benchmarks for evaluating Large Language Models (LLMs) in data analysis often fail to reflect real-world settings. They typically focus on fact retrieval from small tables and overlook the challenges of large multi-tabular datasets, external knowledge integration, and exploratory insight discovery. We introduce DataGovBench, a benchmark derived from governmental open data designed to evaluate LLMs in practical scenarios. The benchmark includes two tasks: Table QA that requires solving complex decomposable questions and producing textual answers or visualizations, and Table Insight that evaluates the ability of models to generate expert-level findings through exploratory data analysis. Comprehensive experiments with state-of-the-art LLMs, both with and without agentic frameworks, reveal significant performance gaps across both tasks. These results suggest that current LLM-based systems remain far from satisfying the demands of real-world data analytics. DataGovBench provides a challenging benchmark for advancing research on LLMs capable of both answering analytical queries and discovering insights from data. Code and sample data are available at https://github.com/SoHasegawa/datagovbench.

We present PACR-Video, a parameter-efficient framework for multi-shot long video extrapolation that preserves recurring entities, scene structure, visual style, and causal progression without full generator fine-tuning. PACR-Video keeps a text-to-video diffusion transformer frozen and augments it with low-rank temporal adapters conditioned by learned shot-role prompt tokens. To maintain long-horizon coherence, it builds a recursive prompt bank that stores compact entity, location, action, and style prompts from previous shots, then routes them through adapter gates according to predicted narrative dependencies. A Shot-Local/Story-Global tuning objective combines next-shot reconstruction, cross-shot identity contrast, and prompt sparsity regularization, while an adapter composition schedule balances early-shot visual consistency with later-shot event progression and viewpoint change. Across six multi-shot and long-video benchmarks, PACR-Video outperforms text-to-video, tuning-based, memory-augmented, streaming, and recursive-context baselines on distributional quality, semantic alignment, identity consistency, temporal smoothness, motion stability, transition coherence, and human preference. These results show that compact prompt routing and lightweight temporal adaptation provide sufficient controllable capacity for stable long video extrapolation.

Physics-informed neural networks (PINNs) provide a promising framework for solving partial differential equations while embedding the underlying physical laws directly into the learning process. This study presents a PINN-based framework for modeling transient elastodynamic wave propagation in bimaterial systems governed by the axisymmetric equations of linear elasticity. A steel-aluminum specimen representative of a Split Hopkinson Pressure Bar configuration is considered, and the governing elastodynamic equations, together with the corresponding initial, boundary, and interface conditions, are incorporated directly into the network through a physics-informed loss function. High-fidelity finite-element simulations performed using ANSYS Workbench Explicit Dynamics are used for validation and as supplementary data constraints during training. The proposed framework accurately predicts wave transmission and reflection across the bimaterial interface and reproduces axial and radial displacement histories, face-averaged responses, and the dominant stress and strain evolution with close agreement to the finite-element solutions. The trained network further demonstrates the ability to predict wave responses at previously unseen time instants and for modified material properties without requiring additional finite-element simulations, providing a continuous surrogate model for elastodynamic analysis. Mesh-sensitivity studies confirm numerical robustness, while additional material combinations demonstrate the generality of the proposed methodology. The results show that integrating physics-informed neural networks with explicit finite-element analysis provides an accurate and computationally efficient framework for elastodynamic wave propagation in heterogeneous solids, offering an effective surrogate modeling approach for high-rate solid mechanics and impact engineering applications.

Context: Domain-Driven Design (DDD) is a leading paradigm for managing software complexity, yet research remains largely theoretical; our prior work found nearly 39% of DDD studies lack rigorous empirical evaluation, leaving practical adoption largely unexamined at scale. Objective: We provide the first large-scale characterisation of the DDD landscape on GitHub, a data-driven baseline for how the paradigm is implemented and sustained in practice. Method: Using a Mining Software Repositories (MSR) approach with a hybrid strategy (topics and README keywords), we identified 11,742 candidate repositories. To address label noise, we built a novel semantic validation pipeline using GPT-4o with a triplicate majority-vote strategy, yielding 2,502 verified repositories. Validation against a manually labelled sample showed substantial agreement with human experts (kappa = 0.77). Results: DDD adoption accelerated sharply after a 2017 inflection point, and the resulting projects are notably long-lived: their median lifespan exceeds the typical GitHub project by over an order of magnitude, indicating sustained, professional-grade engineering rather than short-lived experiments. Layered and Clean Architecture dominate, while CQRS and Event Sourcing recur in distributed, data-intensive systems. Notably, the data challenge the Java-centric assumption of much academic work: C# and TypeScript, not Java, lead practical adoption. Conclusions: DDD has matured into a stable, professional-grade practice adopted across diverse languages and domains. However, a quarter of projects (25.3%) record no explicit business context, revealing a persistent gap between how domain intent is designed and how it is preserved in version control. We call for lightweight architectural traceability standards and offer guidance for teams reusing these repositories as reference implementations.

Multimodal Large Language Models (MLLMs) have demonstrated impressive performance on cross-modal tasks by jointly training on large-scale textual and visual data, where privacy-sensitive examples could be unintentionally encoded, raising concerns about privacy or copyright violation. To this end, Multi-modality Machine Unlearning (MMU) was proposed as a mitigation that can effectively force MLLMs to forget private information. However, the robustness of such unlearning methods is not fully exploited when the model is published and accessible to malicious users. In this paper, we propose a novel adversarial strategy, namely Prompt-Optimized Parameter Shaking (POPS), aiming to recover the supposedly unlearned multi-modality knowledge from the MLLMs. Our method elicits the victim MLLMs to generate potential private examples via prompt-suffix optimization, and then exploits these synthesized outputs to fine-tune the models so they disclose the true private information. The experiments on the different MMU benchmarks reveal substantial weaknesses in the existing MMU algorithms. Our POPS can even achieve a near-complete recovery of supposedly erased sensitive information on the unlearned MLLMs, exposing fundamental vulnerabilities that challenge the foundational robustness of representative MMU-based privacy protections.

While recent Large Language Model (LLM)-based Text-to-Speech (TTS) systems have achieved remarkable naturalness, they predominantly rely on implicit end-to-end generation paradigms, resulting in coarse-grained control. In scenarios demanding precise stylistic interventions and strict temporal alignment, such as audiobook narration and video dubbing, the inability to explicitly manipulate word-level acoustic attributes remains a critical bottleneck. This limitation is primarily amplified by the severe scarcity of fine-grained annotated datasets and the architectural challenge of integrating multi-dimensional control signals into discrete autoregressive generation. To address this, we propose a unified framework for highly precise word-level control. First, we construct WordVoice-5A, a massive 4.7k-hour bilingual dataset featuring five-dimensional word-level annotations (duration, boundary, energy, pitch and tone) developed through a rigorous linguistically-guided pipeline. Second, we introduce WordVoice to transform the implicit generation process into an explicit, highly controllable paradigm. Specifically, we introduce a bound-token mechanism within the LLM to formulate an explicit ``acoustic planning'' process, enabling adaptive multi-task prosodic planning and flexible manual intervention. Furthermore, we augment the token-to-waveform stage with a fine-grained acoustic modulation module, bridging the resolution gap to strictly align word-level attributes between highly compressed discrete tokens and continuous waveforms. Extensive experiments demonstrate that WordVoice achieves superior, decoupled control over multiple acoustic dimensions while maintaining competitive zero-shot synthesis stability. The code and audio samples are publicly available at https://xxh333.github.io/wordvoice-demo/.

Recent LLM-based mathematical reasoning agents have begun to tackle research-level problems and, in several cases, have contributed to the resolution of open problems. However, scaling and orchestrating such agents effectively remains challenging, due to the difficulty of coordinating parallel proof search while keeping intermediate claims organized and reliable. In this paper, we propose Danus, an orchestration system for research-level mathematical reasoning centered on a shared fact graph as a global memory-management mechanism. Danus consists of a main agent that performs planning and coordination, multiple worker agents that carry out proof search in parallel, and a stateless verifier that checks proposed mathematical claims before they are admitted into the fact graph. Each verified fact is stored together with its proof and logical dependencies, allowing the system to build long arguments incrementally while keeping the shared proof state organized. The main agent periodically summarizes the evolving proof state, redirects workers across promising directions, and supports interaction with human mathematicians through progress reports. We evaluate Danus through six research-level case studies in algebraic geometry, singularity theory, and combinatorics, illustrating how the fact-graph memory mechanism enables Danus to construct long, detailed mathematical proofs. Our results suggest that fact-graph-based orchestration provides an effective route toward scaling mathematical reasoning agents for long-horizon research problems. Danus is open source at https://github.com/frenzymath/Danus.

Latent reasoning methods perform multi-step inference entirely in the model's continuous hidden states, promising more compact and efficient reasoning. However, these opaque hidden states raise a question of faithfulness: whether these latent reasoning steps causally drive the final answer. Prior work investigates this question at converged checkpoints and reports several unfaithful behaviors, such as latent reasoning steps that can be replaced without changing the answer, but leaves how these behaviors form during training unexamined. We instead track how faithfulness evolves across saved checkpoints for different latent reasoning paradigms, applying a verifiable counterfactual edit on the input and a noise-ablation activation patch on the latent reasoning steps. We find that (i) at the output level, latent reasoning methods can look similarly unfaithful at convergence under counterfactual edits while following qualitatively divergent trajectories; (ii) at the activation level, the causal contribution of latent reasoning steps to the final answer decays across training for both paradigms, with the examples that flip on the output side in (i) also being the examples on which this contribution decays; and (iii) the activation-level trajectory diverges by answer format, decaying on binary choice and rising on open-ended decoding. These findings highlight that latent reasoning faithfulness depends on training stage and answer format.

Open Radio Access Networks (O-RAN) increasingly delegate near-real-time control to deep reinforcement learning (DRL) xApps obtained from third-party vendors, creating a new supply-chain attack surface. A backdoor policy behaves optimally until an adversary injects a covert trigger into the observed key performance indicator (KPI) telemetry, at which point it issues harmful control actions that degrade quality of service (QoS). We present ORAN-DEFEND, a retraining-free wrapper that sanitizes a frozen, potentially compromised xApp by projecting each KPI window onto a safe subspace estimated from a small number of trusted clean rollouts via singular value decomposition (SVD). We establish, both analytically and empirically, a precise recovery condition: the defense succeeds if the trigger energy concentrates in the orthogonal complement of the safe subspace, and we quantify this boundary through the trigger's $\Eperp$ energy fraction. On the Colosseum COLORAN dataset, we evaluate four structurally distinct DRL backdoor attacks, like TrojDRL, SleeperNets, BadRL, and Q-Incept, spanning inner-loop and outer-loop poisoning regimes and demonstrate $100\%$ return recovery and $\geq99.5\%$ defense success rate across all four when the subspace assumption holds. A geometry ablation reveals an intrinsic and previously uncharacterized limit of any linear projection defense: when the trigger collocates with the legitimate signal, the $\Eperp$ energy fraction governs recovery monotonically, and the linear residual detector collapses to chance even while a nonlinear classifier retains perfect separability.

This paper introduces Diffusion Semi-Relaxed Fused Gromov-Wasserstein (DsrFGW), a novel method for graph comparison that unifies node features and structural connectivity through optimal transport. While traditional Gromov-Wasserstein and semi-relaxed variants (srGW, srFGW) capture graph structure, they often struggle with sparse, noisy, or partially observed graphs. Inspired by Graph Diffusion Distance, which posits graphs are similar if they enable similar information transmission patterns, DsrFGW incorporates diffusion processes allowing information propagation across nodes, capturing local and global structural patterns while reducing sensitivity to noise or missing edges. An extensive evaluation on 36 synthetic pairwise graph matching tasks (easy, medium, hard) demonstrates consistent superiority over srFGW, achieving accuracy improvements of 0-20 percentage points and dramatic Adjusted Rand Index (ARI) gains: in medium-difficulty scenarios, srFGW often achieves negative ARI (worse than random) while DsrFGW offers better performance in terms of both internal and external clustering quality measures (i.e., Adjusted Rank Index and Accuracy with respect to the true underlying clusters, respectively). Even under severe noise, DsrFGW improves clustering quality in 92% of the synthetic tasks with optimal diffusion scales adapting to problem difficulty, establishing DsrFGW as a robust framework for graph comparison under structural uncertainty.

Determinantal point processes have recently emerged as a kernel-based alternative to standard independent sampling for constructing efficient minibatches, coresets, and other compact representations of large-scale datasets. In particular, sampling mechanisms based on DPPs are believed to demonstrate better approximation properties compared to classical i.i.d. samplers, even at the scale of the exponent. One of the key strengths of DPP based samplers is that they can be deployed over very general spaces, in contrast to more classical sampling methods beyond i.i.d. which tend to work in very well-structured settings, principally Euclidean spaces. In this work, we establish explicit rate guarantees for determinantal sampling in spaces that extend far beyond known Euclidean setups, focusing on spectral kernels obtained from eigenspaces of naturally associated Laplacian and other Markov diffusion operators. This includes, in particular, Riemannian manifolds and weighted networks. In determinantal sampling from compact Riemannian manifolds, we establish sampling rates that automatically pick up the intrinsic dimensionality $d_{\text{int}}$ of the underlying manifold. In the setting of networks, we investigate DPP-based samplers on the celebrated k-nearest neighbour graphs, as well as weighted random geometric graphs, and demonstrate a similar improved dependence on the intrinsic dimensionality of the data. Overall, our approach achieves guarantees of $\big(\text{sample size}\big)^{-\frac{1}{2}-\frac{1}{2d_{\text{int}}}}$ that match known rates on Euclidean spaces of comparable dimension. In terms of techniques, we connect to the celebrated Weyl's Law for manifold spectra, and leverage tools from the theory of Markov diffusions and Dirichlet forms as well as certain ingredients from the theory of pseudodifferential operators, which could be of independent interest in this area.

Backdoor attacks severely threaten large-scale AI models. When model owners delegate training to external compute providers within a decentralized training paradigm, adversaries can craft stealthy, low-frequency triggers to inject malicious behavior while evading standard audits. Traditionally, detecting these attacks requires a full re-computation of the training steps--a prohibitive overhead that directly contradicts the owner's resource constraints. To address this, we investigate the resilience of continuous optimization dynamics under Byzantine perturbations, where adversaries are forced to compete against a continuous influx of honest updates. Under a threat model where an adversary compromises f out of n total trainers, we quantify the minimum auditing overhead required by the model owner to probabilistically bound the attack success rate. We formalize this injection-absorption dynamic as a Discrete-Time Markov Chain (DTMC). Using this framework, we prove that the success probability of any bounded adversary asymptotically collapses to zero under a defense strategy combining natural absorption, a randomized scheduler, and lazy verification oracle. Empirical results demonstrate significant backdoor suppression with zero utility degradation even when invoking the verification oracle on merely 10% of the total training steps. This approach yields a provably sound and computationally efficient defense for safety-critical AI.

Active learning studies the fundamental question: what data should we choose to observe? The greedy algorithm in optimal experiment design is a common heuristic and also equivalent to myopic Bayesian active learning for linear regression, the common framework where long-term planning is replaced with the one-step optimal choice. In this work, we prove a first-of-its-kind approximation ratio for the greedy algorithm's risk that is tight up to an absolute constant. The approximation ratio is linear in the maximum initial leverage score (MILS), a newly identified quantity fundamental to the greedy algorithm's performance. Finally, we illustrate the results with simple numerical simulations.

Large language models (LLMs) achieve promising results on medical question answering benchmarks, yet their use in public health is constrained by hallucinations and the rapid evolution of official guidance. Retrieval-Augmented Generation (RAG) mitigates these risks by grounding responses in an explicitly maintained corpus, but end-to-end performance depends critically on retrieval configuration and on evaluation beyond multiple-choice formats. We extend PubHealthBench, a question answering (QA) benchmark of 7,929 questions derived from UK Government public health guidance, into a retrieval-augmented setting and systematically evaluate retrieval and generation choices. We compare dense, sparse, and hybrid retrieval across multiple embedding models and corpus variants, and show that hybrid retrieval consistently improves recall and ranking quality, with chunk length and topic interacting with ranking performance. Providing retrieved context substantially increases multiple-choice accuracy across a diverse set of LLMs, enabling smaller open-weight models to match or outperform larger models used without retrieval, with gains primarily driven by retrieval quality and careful context selection. To assess realistic free-form answering, we introduce a rubric-based LLM-as-a-judge covering faithfulness, completeness, clarity, and factual consistency, and validate it against dual human annotations. Judge-human agreement is strongest for faithfulness and completeness, while factual consistency and clarity are less reliably reproduced, motivating caution when interpreting those dimensions at scale. Overall, our results highlight retrieval as a primary lever for reliable public health QA and provide practical guidance for building and evaluating RAG systems grounded in official guidance.

A learned world model is usually judged by how faithfully it reconstructs its observations or predicts reward, as though quality were something the model simply has or lacks. But what a task actually needs from a model is narrower: the few predictive coordinates its queries depend on, which we call the closure. We show that how much of that closure a latent comes to represent is set not by the model's capacity or its observations but by the dimensionality of the objective it is trained against, and we measure this directly on a DreamerV3 stack in a controlled environment with known ground-truth closure. An aligned scalar value signal -- the objective at the heart of value equivalence -- installs only a one-dimensional projection of a closure that needs several dimensions: read through a single linear probe, the recoverable structure rises from R^2=0.10 to 0.76 as the scalar is replaced by the full objective. Sweeping the objective's dimensionality from one to four installs exactly that many predictive directions through an auxiliary head, and the same staircase appears -- at attenuated magnitude but the same rank -- through the model's own value head, so the dissociation is dimensional rather than an artifact of head form. Capacity-matched comparisons and in-situ pressure checks rule out the obvious alternatives. The law governs a regime, and we measure its boundary: on a companion closed-loop task whose structure is observable frame by frame, reconstruction installs that structure and the scalar objective suffices -- the objective decides what a latent represents exactly where cheaper training signals cannot already recover it. Value equivalence is thus not all-or-nothing but dimensional: the familiar single-reward objective is its rank-one corner, and a model installs as much of a task's structure as the objective it is asked to predict.

On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has already generalized. Reading effective rank at the grokking transition overstates the converged value by 3-5x on an MLP, and by 1.3-1.5x on a transformer trained to convergence; on the MLP it also erases which cells compress at all. Compression lags the accuracy transition by an amount on the order of the time-to-grok, at least 10,000 steps, rather than coinciding with it. A one-variable ablation shows what sets the lag size: adding LayerNorm to an otherwise identical transformer moves the fraction of compression done by the grok step from 0.87 to 0.25, and a pre-registered control rules out scale invariance as the mechanism. We package this as an audit that separates onset from compression, flags censoring, excludes boundary cells that never fully generalize, and checks that the reference floor has plateaued, with an adversarial suite that caught a false-confidence bug in our own branch. A secondary, MLP-specific depth law linking norm budget to converged floor fails a generality test on a transformer and flips sign under free weight decay. Code and the toolkit are released.

Self-paced learning (SPL) is an effective learning paradigm that simulates the human learning process by progressing from easy to difficult samples based on the value of the loss function during the learning process. It has shown great potential in improving model performance and training efficiency. However, the prediction results of samples with smaller loss values are not necessarily reliable, indicating that such samples are not always simple samples for the model. Hence, this article proposes an uncertainty-aware self-paced learning based on evidential neural networks, termed UASPL, which integrates predictive reliability into sample selection through a general loss function within the Subjective Logic framework. This loss function incorporates uncertainty estimation and can be extended to different variants of SPL. Moreover, this loss function couples a sample selection preference, thereby ensuring the interpretability of the sample selection process. Finally, the experimental results on multiple datasets show that UASPL outperforms other SPL methods in terms of classification performance, interpretability, and generality. The source code is available at: https://github.com/treelife979/UASPL.

In this work, we propose a unified approach for diagnosing misclassification and assessing the robustness of black-box classifiers. Central to our method is an optimization framework that modifies an instance so that the classifier predicts a specified target label, while ensuring that the modification remains easily explainable. The objective function contains two components: an explainability-aware $L_0$ (XA-$L_0$) penalty that promotes sparse and interpretable modifications, and a classifier loss objective that steers the perturbed instance toward the desired output. This integrated optimization formulation is used both to identify the underlying causes of misclassification and to evaluate robustness by determining how an instance can change within a tolerance region before being reassigned to another class. To quantify robustness, we introduce the Tolerance Region Confusion Matrix (TOR-Confusion Matrix), which measures a classifier's susceptibility by modeling the class-to-class transition probabilities induced by tolerance-bounded perturbations. We validate the proposed method on both image and tabular datasets, demonstrating its ability to jointly deliver interpretability and robustness assessment.

Large language models frequently generate code that appears correct on typical inputs yet fails on edge cases, invalid inputs, and other specification-defined corner conditions. A popular fix has the model write its own tests and repair until they pass, but the source of the gain is unclear: does it come from the tests merely existing, or from their grounding in a specification of what the code should do? We isolate this factor. Holding the tester, test budget, and repair loop fixed, we change a single prompt line that controls whether the tester receives the spec as a checklist of rules. The baseline is strong: it is already told to probe invalid inputs and edge cases. Grounding the tests in the spec produces correct code +38 percentage points more often than this baseline across three Claude tiers (Haiku 4.5, Sonnet 4.6, Opus 4.8), and +36 points on a held-out set. Grounding, not test quantity, is the primary driver: doubling the test budget barely helps, and combining eight independent ungrounded suites plateaus far below grounding. An ablation isolates the spec's content, not its format: given the spec as a plain paragraph the tester recovers 27 of 30 bugs, but asked to plan tests without the spec it recovers only 2 of 30. The effect survives stronger baselines: a property-based generator catches 28 of 30 bugs but invents out-of-spec requirements, and an AlphaCodium-style loop only matches the baseline. It replicates across vendors (GPT-5.3-codex +28, Gemini 3.5 Flash +19), with a task-level sign test over 18 tasks significant at p=0.002. Grounding improves both sensitivity and precision: it catches more real bugs and wrongly rejects far less correct code, cutting the false-alarm rate from 33% (68% against a Python standard-library oracle) to 0%. On well-specified algorithmic problems it neither helps nor hurts.

Compact networks built from Clifford algebra Cl(3,0) primitives are exactly SO(3)-equivariant and learn synthetic 3D vector laws from few samples. We ask whether the geometric algebra structure itself contributes anything beyond exact equivariance. We compare against a minimal scalarization baseline: invariant dot products fed to a small MLP that outputs coefficients on the equivariant basis {v_i, v_i x v_j}, which is also exactly equivariant. On single-stage laws (rotation by axis-angle, cross product, central force), scalarization matches or beats the Cl(3,0) network at a fraction of the training cost, so the geometric algebra adds nothing there. On compositional targets whose computation graph nests group operations (apply R2 R1 to a point; map a local force through an orientation, then take a torque), the Cl(3,0) network beats scalarization by an order of magnitude in the low-data regime, reaching with 100 samples what the baseline needs 3000 for, and the gap survives strengthening the baseline with the triple-product invariant and 17x more parameters, external Vector Neurons and e3nn baselines, and a multiplicative coefficient network. Ablations show the required network depth tracks the rotation chain length, and scalarization falls below the constant predictor on chains of four rotations. The advantage is not composition per se: on a rotation-free nested cross product, which flattens into polynomial invariant coefficients, scalarization wins by 24x. No tested model, equivariant or not, extrapolates invariant magnitudes: on radius and separation shifts every model is worse than a constant predictor once errors are normalized. We conclude that geometric algebra layers are not a general shortcut for low-data 3D learning, but become useful precisely when the target composes group elements in depth.

In this paper, we address the problem of multimodal federated learning with missing modality. Existing methods utilize an additional public dataset or perform naive feature synthesis that is based solely on the available modality. To address these limitations, we propose ProMoE-FL, a Prototype-conditioned Mixture-of-Experts framework for robust missing-modality feature synthesis in multimodal federated learning. ProMoE-FL builds a global client-aware prototype bank that captures clinically meaningful modality priors across institutions. Our Mixture of Experts is conditioned on these prototypes and modality indices to enable direction-aware expert routing for dynamically synthesizing missing features. We perform extensive quantitative and qualitative evaluations on four public chest X-ray datasets (MIMIC-CXR, NIH Open-I, PadChest, and CheXpert) and demonstrate that ProMoE-FL consistently outperforms state-of-the-art methods in both homogeneous as well as the more challenging heterogeneous settings.

Adversarial attacks are crafted data manipulations that aim to deteriorate the outcomes of prediction or decision-making algorithms. In the energy systems literature, adversarial attacks have been studied with a focus on problems regarding the electricity grid. Such problems include forecasting and grid state estimation, where adversarial attacks are also known as false data injection attacks. Only few studies have analyzed the potential impact that adversarial attacks have on the demand side. We analyze how manipulated price forecasts impact the decision-making in industrial demand response. To this end, we design adversarial attacks that aim to deteriorate the output of electricity price forecasting models and solve scheduling optimization problems of energy-intensive production processes using the distorted price forecasts. We make use of a generalized process model to investigate the vulnerability to adversarial attacks for a range of production scheduling problems with different levels of process flexibility. We find that adversarial attacks can erode the profits gained from demand response. However, when perturbations are limited in extent (so that they are hard to detect by the human user), demand response preserves about 90\% of its financial advantage compared to steady-state process operation. Further, we find that the impact of adversarial attacks on demand response does not only depend on the magnitude of the perturbations but rather on the orientation of the adversarial perturbations. Therefore, we argue that attack analyses should explicitly incorporate the sensitivities of scheduling optimization models into the attack design to enable more rigorous assessments of decision-making under adversarial attacks.

Video Diffusion Models (VDMs) have demonstrated superior generation quality but suffer from prohibitive computational costs. While recent few-step distillation techniques significantly accelerate inference, they typically enforce a static model architecture across all denoising stages, ignoring the varying computational demands inherent to different noise levels. In this work, we propose a novel post-training acceleration framework that exploits this redundancy by integrating dynamic structural sparsification directly into the distillation process. Unlike conventional post-hoc compression applied to a fixed diffusion pipeline, our approach jointly optimizes the denoising steps and structured model sparsity, transforming a pre-trained VDM into a compact, step-specific Mixture-of-Models (MoM). To address the training instability arising from this joint optimization, we introduce a Progressive Training Strategy coupled with an Output Rollout Mechanism, which ensures the coherent learning of structural decisions across timesteps. Furthermore, we develop a specialized inference engine to deploy the resulting MoM efficiently. Our method is orthogonal to existing acceleration techniques and highly effective: On Wan-14B, it removes 24% of the per-step FLOPs on top of 4-step distillation, adding a 1.2x wall-clock gain and reaching a 30x speedup over the 50-step teacher while preserving competitive generation quality.

Formal robustness certificates for embedded neural-interface models can pass while task accuracy collapses: at perturbation budget e=0.25, EEGNet classification accuracy drops by 25.7% under projected-gradient attack while the Lipschitz-style certificate remains valid for all 9 tested subjects. We argue that this gap between mathematical certification and operational safety is one instance of a broader alignment failure in neural interfaces, where training objectives diverge from user welfare. We propose a unified empirical audit framework organized around three such failures: verification insufficiency, in which certificates pass while task behavior degrades; proxy-fidelity divergence, in which task-optimized representations damage neural signal structure (a time-domain auxiliary objective reduces reconstruction MSE by 0.1132 while worsening spectral log-MSE); and latent information exfiltration, in which public-task embeddings retain private attributes (subject identity recoverable at 48.1% versus 6.7% chance). We instantiate the framework on BCI Competition IV 2a and SEED-IV using multiple deep and classical EEG decoders, official session-level validation, null controls, and paired statistical tests. The verification gap persists across EEGNet, CSP+LDA, and FBCSP+LDA, and is therefore architecture-independent. Our results establish that operational safety auditing, not certificate verification alone, is necessary for responsible neural-interface deployment.

The deployment of Mixture-of-Experts (MoE) models on production high-bandwidth superpods, such as NVIDIA's NVL72/576 and Huawei's CloudMatrix384, introduces critical challenges beyond raw interconnect bandwidth. While these systems provide unified global address spaces and high-bandwidth fabrics, their full potential for sparse MoE communication is hindered by three fundamental bottlenecks: (1) Strict execution serialization imposed by coarse-grained Bulk Synchronous Parallel (BSP) orchestration of interdependent communication phases; (2) Prohibitive synchronization overhead that fails to scale alongside high interconnect bandwidth; and (3) Severe load imbalance resulting from distance-agnostic scheduling of irregular token traffic. To eliminate these bottlenecks, we introduce UBEP (Unified-Bus Expert Parallelism), a production-ready communication library that rethinks MoE's All-to-All primitives for modern superpod architectures. Through large scale experiments, UBEP reduces All-to-All latency by up to 52.4% and MoE inference Time Per Output Token (TPOT) by up to 11.1%.

Brain age -- the age inferred from a physiological recording -- is an emerging biomarker whose deviation from chronological age tracks neurological and psychiatric burden, and EEG is an attractive substrate for it because it is cheap, portable, and temporally rich. Yet EEG brain-age models must contend with cross-site montage heterogeneity, small labelled cohorts, and dominant subject-level non-stationarity, and few EEG foundation models have been shown to deliver competitive age regression across the full pediatric-to-older-adult range in which such a biomarker would actually be deployed. We introduce STST-JEPA, a self-supervised transformer for resting-state and task EEG, pretrained on 47,703 sessions spanning ages 5-81 from the brain.space and Healthy Brain Network (HBN) corpora. The model combines a latent-prediction objective - predicting masked-token representations against an EMA-of-tokenizer target - with an auxiliary signal-reconstruction term, applied to 30-second multi-channel windows under spatiotemporal block masks. A lightweight attentive probe trained on frozen pretrained embeddings achieves a best held-out-validation mean absolute error of 3.06 years (r = 0.924) for age regression on 3,367 sessions, against a predict-the-mean baseline of approximately 10 years MAE. With light task-specific fine-tuning of the model's final layers, the same pretrained encoder achieves rank-1 placements - with the model's native 30-second windows - on the public NeuralBench x brain.space EEG leaderboard for sex classification (balanced accuracy 0.911), age prediction (r = 0.749), and psychopathology composite regression (r = 0.215). We further show that the model's age-prediction residual is negatively correlated with cognitive efficiency over several tasks we examined.

Which properties of a partially trained network are causally portable to a different, independently trained network? Single-trajectory interventions show necessity within one run, not portability across runs. We introduce cross-trajectory chimera interventions: given two runs from different seeds, we split each weight vector into a norm and a unit direction, recombine one run's norm with the other's direction, and continue training. On two modular-arithmetic tasks that grok, the components dissociate. Direction carries a transferable, donor-specific circuit identity: implanting a donor's direction at the recipient's norm drives the run to the donor's circuit in 40/40 cases, while an angle-matched random control yields no shift. The transfer is threshold-like, and its location is predicted by the recipient's norm, separating perfectly by norm class over all 20 pairs (joint permutation probability 1.9e-4). Norm carries only a modest, distributed delay effect and no identity signal. An adaptive bisection procedure localizes the threshold to +/-1/64. Direction indexes which solution a trajectory approaches; norm governs how susceptible that identity is to being overwritten.

Recent Vision-Language Models capture increasingly complex aspects of human cognition. Here we ask whether this alignment extends to reward valuation, which we assess in a mechanistic framework built on clinical tests that were developed to evaluate anhedonia and motivational deficits in major depressive disorder. In the brain, anhedonia is frequently linked to dysregulation in the Nucleus Accumbens (NAc) and the broader dopaminergic reward system. While neuroimaging has localized these deficits, establishing a causal link between NAc activity and specific behavioral symptoms remains a challenge. We use these ideas from neuroscience to functionally identify reward-anticipatory units in vision language models, and test their causal role via targeted perturbations. Perturbing NAc-selective units induces behavioral effects that mirror human anhedonia: the model shifts toward low-effort, low-reward options in effort-based decision-making tasks. Crucially, our results reflect a specific deficit in reward valuation and anticipation rather than a loss of task capability: the perturbed model maintains baseline performance when reward-based choice is removed. This induced vulnerability further aligns with clinical anhedonia and motivation scales, including DARS and MAP-SR. Taken together, these results reveal reward valuation circuits in AI models that parallel those in humans.

Process industries have accumulated validated specialist models, yet sensor drift, feedstock variation, and regime switching cause these models to degrade systematically in new scenarios. Collecting new labeled data and retraining is costly, while continuing with the original model incurs persistent bias. Existing adaptation methods require modifying model parameters with sufficient labeled data, making rapid response on deployed systems difficult. Using LLMs as direct predictors risks hallucinations and uncontrollable outputs. Such predictors also cannot incorporate unstructured scenario knowledge from the field. To address these limitations, this article proposes Reasoning-Driven Open Adaptation for Specialist Models (ROAM), a framework that uses LLM world knowledge and reasoning to adapt frozen specialist models to unseen scenarios without retraining. ROAM confines all corrections to a low-dimensional, semantically interpretable latent space. LLM-generated scenario judgments and online observations are fused under a unified probabilistic framework. A risk-constrained mechanism suppresses corrections under unreliable LLM evidence or abrupt scenario shifts and falls back to the original frozen model when evidence is insufficient. Experiments on a mineral thickening process and the public IndPenSim penicillin fermentation dataset show that ROAM reduces MAE by over 20\% in major shift settings such as hidden shifts with only 839 additional parameters and under 0.02\,ms per-step overhead. These results indicate that LLM reasoning can be turned into a conservative adaptation signal for industrial models already in service.

We present AgentLens, a production-assessed benchmark for interactive code agents. Most code-agent benchmarks reduce a run to a single bit -- did the task pass? -- but the people who actually use these agents experience the entire trajectory: how the agent follows instructions, uses its tools, verifies its own work, recovers from mistakes, and talks to them along the way. AgentLens evaluates that whole trajectory. It pairs formal verification, where an objective check exists, with LLM-written trajectory reviews and side-by-side comparisons, so that each run yields a readable explanation of why the score is what it is. This makes AgentLens useful for more than ranking models: we use it to diagnose model behavior, compare successive versions of our own agent, and catch product regressions in a nightly evaluation pipeline. We release the benchmark as open source at https://github.com/agent-lens/agent-lens-bench.

Process industries rely on time-series forecasting and soft sensing to estimate quality variables that are hard to measure online. Labeled data are scarce, operating regimes change frequently, and retraining models or rebuilding alignment pipelines for each scenario is costly. Such settings often provide variable tables and process documents that record variable names, units, physical meanings, and process roles. However, standard time-series backbones usually treat inputs as anonymous numerical columns. Existing text-enhanced methods also rarely make the semantic-logical relations between input variables and the prediction target available to the model within each numerical window. To address this problem, this article proposes Task-Semantic Field Factorization (TSF), a large language model (LLM)-guided framework. TSF builds a task-semantic field from task protocols and variable documents before training and uses the LLM only for offline semantic construction. Online training and inference remain with conventional time-series backbones. During training and inference, the current numerical window activates variable semantics, so semantic information participates in each prediction and supports adaptation to different prediction targets and operating shifts. On multiple complex industrial forecasting and soft-sensing tasks, TSF reduces MAE by 6.4\% on average in improved settings, with the largest reduction reaching 25.5\%. It adds only about 1.8--3.0k parameters, with less than 0.008 ms/step of additional online inference overhead. These results show that TSF turns existing process documents into measurable forecasting gains across backbones and semantic generators while remaining lightweight for deployment.

In this paper, we investigate whether a model-free RL agent can identify and exploit price manipulation opportunities more effectively than a traditional model-based approach that assumes correct specification of the data-generating process but relies on noisy parameter estimates. We consider a single-asset market in which prices evolve according to an Almgren-Chriss framework with non-linear permanent impact and linear temporary impact. We first establish the existence of price-manipulative strategies in discrete time and compute the optimal benchmark strategy using Sequential Least Squares Quadratic Programming under full information. We then compare two finite-sample learning approaches: a model-based procedure that estimates impact parameters from simulated execution data and an agnostic RL approach based on Deep Deterministic Policy Gradient, trained directly on the same amount of data. For intermediate volatility, the RL agent successfully discovers profitable manipulative strategies without explicit knowledge of the underlying model, even when training data are quite limited. More importantly, RL consistently outperforms the model-based approach when parameter estimates are affected by sampling error, despite the latter benefiting from the correct model specification. For large volatility, all methods are unable to identify manipulation opportunities, while for small volatility, the model based approach outperforms RL. These findings highlight both the effectiveness of RL in complex control problems and the risks associated with deploying learning algorithms in financial markets without appropriate safeguards.

Utilities increasingly rely on planning and operational tools to cope with the increased penetrations of distributed energy resources, yet the lack of realistic, openly available datasets remains a major barrier for benchmarking and comparison. Traditional test feeders, and recently proposed large-scale synthetic networks alleviate this issue but are typically based on heuristic rules and do not learn directly from data. This paper proposes a generative framework based on Generative Adversarial Networks (GANs) to create power distribution network layouts using image-based representations. The model is trained on rasterised views of distribution systems and can operate in two modes: an unconditional configuration that learns layout patterns from the training dataset, and conditional configurations that incorporate geographical context such as street maps and the spatial distribution of consumers. The methodology includes dataset preparation from Geographic Information System (GIS) sources, GAN architecture design, and the analysis of training stability and image resolution. Results from three representative cases show that the proposed approach can reproduce the topologies of low (LV), medium (MV) and high voltage (HV) feeders and align generated layouts with underlying geographical structures. At the same time, the study reveals limitations related to training stability, resolution-dependent artefacts and limits, and the absence of explicit electrical constraints. The proposed framework constitutes a data-driven complement to existing synthetic network generation methods, and could be applied to propose distribution network layouts for the electrification of new areas. This would require future extensions towards power flow, electrically validated models.

The pre-softmax score of an attention head is a bilinear form $score(i,j) = x_i^T M x_j$ in a learned operator $M = W_q^T W_k$. Because M is generally non-symmetric, hence non-normal, it has a complex eigenspectrum and non-orthogonal eigenvectors, the regime where non-Hermitian and random-matrix tools apply. We ask what this spectrum encodes, at three levels for previous-token and induction circuits. Statically, across seven pretrained models spanning three positional schemes, the strongest previous-token heads are spectrally rotational under RoPE and non-rotational, or content-like, where position enters outside QK (learned-absolute and ALiBi); the model-level separation is perfect at every top-k examined (exact permutation $p=0.029$), and zeroing the per-frequency RoPE phase $Im(M_t)$ eliminates induction on a pre-identified previous-token head in all three RoPE models. Dynamically, over public Pythia checkpoints every head originates at the random-matrix (Ginibre) null; the rotational signature emerges with the behavior, not before it, and the population-median suppression that yields the final profile follows circuit formation, so the profile is a consolidated fingerprint, not a precursor. Causally, and at toy scale, no spectral channel is necessary: constrained two-layer training reroutes around every ban with capability intact, albeit at a significant formation delay (four pre-registered contrasts, $q_BH <= 0.016$). The cost structure exposes each scheme's default: imposing symmetry slows learned-absolute models by a factor of 2.9, whereas a RoPE head with a fully symmetric static M still routes directionally via the phase channel, impossible under absolute positions. Within the settings examined, the positional scheme sets the default spectral algebra of an attention head's solution: a fingerprint sculpted after function, not a hard constraint upon it.

This study harnesses topological analysis in an attempt to reveal structure in the dynamics of a crowd. Topology and in particular persistent homology characterizes relational structures in data through the number of connected components and holes, that is, a loop of pairwise connection with no connections across it. We apply this universal data analysis method to a simulated time series of individual pedestrian positions of a crowd moving through a wide corridor -- either uni- or bidirectional. We consider two pedestrians to be connected, when they are sufficiently close. This approach leads to two matrices containing the persistence signatures for the whole time series, so-called CROCKERs. Despite the high level of data abstraction, the CROCKERs' first two principal components on time-delayed positional data show a clear separation of the different parameter configurations. This holds up to symmetry. Our results support our claim that persistent homology is a useful tool to characterize crowd dynamics without introducing any prior assumptions about the detectable spatio-temporal patterns.

Recent Multimodal Large Language Models (MLLMs) struggle to bridge the representational gap between 2D semantic understanding and 3D spatial geometry. Existing 3D-aware models either rely on costly 3D-specific data or utilize RGB-only inputs with heuristic sampling and monolithic, shallow fusion, which respectively disrupt essential spatiotemporal connectivity and induce modality contention across diverse spatial tasks. To overcome these bottlenecks, we introduce SpaR3D-MoE, an end-to-end framework that enables adaptive spatial reasoning by equipping MLLMs with geometry-aware capabilities from only sparse RGB inputs. First, we propose an adaptive spatiotemporal manifold sampling mechanism that constructs a geometry-aware spatiotemporal graph to extract informative keyframes, effectively mitigating sequence redundancy while preserving the scene's topological connectivity. Second, we introduce the heterogeneous geometry-inductive Mixture-of-Experts driven by an instruction-pose aware router, which adaptively routes multimodal tokens to specialized experts, resolving the cross-modal contention inherent in monolithic fusion. Extensive experiments on VSI-Bench, ScanQA, and SQA3D demonstrate that our method achieves state-of-the-art performance. Notably, SpaR3D-MoE achieves the highest average score of 63.5 on VSI-Bench, outperforming the strongest baseline by 7.8 absolute points, alongside relative improvements of 35.4% and 51.4% in Route Plan and Relative Direction tasks, respectively.

Modern processor verification struggles to reach deep architectural states due to the inefficiencies of traditional mutation-based fuzzing. We propose HiFuzz, a novel hierarchical reinforcement learning framework that replaces mutation with a structured, two-layer generation process: a Program Agent for global layout and a Basic Block Agent for precise instruction filling. To overcome reward sparsity, HiFuzz integrates an adaptive coverage reward mechanism and a semantic-aware basic block encoder providing intrinsic feedback. Extensive evaluations on three real-world RISC-V cores demonstrate that HiFuzz significantly outperforms state-of-the-art fuzzers in coverage and bug detection.

Following the CMIVQA, MMI-VQA, and M4IVQA challenges in NLPCC 2023--2025, we introduce the Difficulty-Aware Medical Instructional Video Question Answering (DA-MIVQA) shared task for NLPCC 2026. DA-MIVQA extends previous multilingual and multimodal medical video benchmarks by explicitly distinguishing questions according to the type and complexity of evidence required for answering. Specifically, simple questions can often be answered from subtitle-based textual cues, whereas complex questions require visual grounding, procedural understanding, and cross-modal evidence integration. The challenge contains three tracks: Difficulty-Aware Temporal Answer Grounding in Single Video (DA-TAGSV), Difficulty-Aware Video Corpus Retrieval (DA-VCR), and Difficulty-Aware Temporal Answer Grounding in Video Corpus (DA-TAGVC). The dataset is collected from public medical instructional channels, covers diverse scenarios such as first aid, emergency response, rehabilitation, nursing, and general medical education, and is manually verified with difficulty annotations. This paper presents the task motivation, dataset construction, evaluation protocol, participation overview, competition results, and representative systems of DA-MIVQA. DA-MIVQA provides a practical benchmark for evaluating medical instructional video question answering systems under varying textual, visual, temporal, and procedural reasoning requirements.

Wearable motion sensing provides a continuous and scalable window into human behavior and health, making it a natural fit for foundation models, yet its pretraining and scaling principles remain poorly understood. Prior work studies isolated design choices, such as sensor placement or sampling frequency, often under fixed settings and narrow downstream tasks that fail to capture real-world sensing diversity. We introduce Inertia-1, a fully open exploration of wearable motion foundation models. Using massive corpora of accelerometer data from global sources spanning more than 18.2M hours, we build a controlled framework for studying the full lifecycle of wearable motion foundation models, covering data choices such as sensor modality, device placement, sampling rate, window length; model choices such as architectures and model size; and training choices such as pretraining objective and data scale. Extensive evaluations across 15 datasets spanning human activity recognition, freezing-of-gait detection, and disease prediction reveal intriguing findings for building motion foundation models that generalize across tasks and sensing conditions. Collectively, Inertia-1 not only presents state-of-the-art recipes for diverse downstream tasks, but also serves as a comprehensive, practical, and open cookbook for wearable motion representation learning.

Label skew in federated learning (FL) causes client drift and degrades global accuracy. Synthetic data augmentation can reduce this imbalance; however, full class balancing requires substantial computation cost. We propose FedEAS, a policy that assigns each client an entropy-adaptive per-class generation budget computed from its local label distribution. The budget jointly decides \emph{how much} each client generates and \emph{WHERE} the samples go. Accordingly, the total generation budget follows from the per-client budgets rather than being fixed in advance. FedEAS recovers most of the accuracy gain of full class balancing while reducing the generation budget by 94.1\%. At the same total generation budget, it outperforms Uniform allocation by up to 18.82\% across CIFAR-10 and CIFAR-100.

Image forgery detection is a critical task in digital forensics, yet many deep-learning localization approaches are typically GPU-accelerated and computationally heavier than handcrafted screening methods. We propose a lightweight, interpretable feature engineering pipeline for image-level forgery screening using only CPU computation and gradient boosted trees. Our method introduces \emph{multi-scale Error Level Analysis} (ELA) computed at seven JPEG quality levels, combined with novel \emph{cross-quality ELA ratio} features that capture double-compression artifacts characteristic of spliced regions, augmented by spatial entropy, FFT energy bands, edge density, SRM residuals, and DCT blockiness, yielding a 405-dimensional feature vector. CASIA v2.0 contains a format confound (60\% of tampered images are TIFF while authentic images are JPEG/BMP and contain no TIFF samples), enabling a trivial \texttt{is\_tiff} classifier to reach 0.80 AUC. We address this through rigorous format-controlled evaluation: on the JPEG-only subset (9,501 images, eliminating the TIFF/JPEG container confound), our method achieves AUC~=~0.990 [95\% CI: 0.988--0.991] and F1~=~0.905 using 5-fold stratified cross-validation. Under a conservative source-aware group split (preventing related images from appearing in both train and test), AUC remains 0.976. An ablation study reveals that multi-scale ELA provides the dominant gain (+0.180 AUC over single-quality on the format-controlled subset), while cross-quality ratios provide complementary double-compression detection. These results support that the method detects compression-history inconsistencies rather than file-format shortcuts -- while offering feature-level interpretability, CPU-only deployment, and sub-second inference.

Accurate station-level demand forecasting is essential for the efficient operation of bike-sharing systems, yet it remains challenging due to complex spatio-temporal dependencies and the large scale of urban networks. This paper presents STAGformer, a Spatio-Temporal Agent Graph Transformer that achieves efficient global modeling with linear computational complexity. The model introduces a two-step agent attention mechanism, where a small set of learnable spatial and temporal agent tokens first aggregate global information and then broadcast it back to individual stations and time steps, effectively capturing long-range interactions while reducing the quadratic cost of standard self-attention to O(NT). STAGformer integrates four core modules: a spatio-temporal encoder that fuses dynamic node features with external contextual factors (weather, time, points of interest), a graph propagation module for spatial neighbor aggregation, a temporal convolution module for local pattern extraction, and the agent attention module for global dependency modeling. Extensive experiments on two real-world datasets -- NYC Citi-Bike and Chicago Divvy-Bike -- demonstrate that STAGformer consistently outperforms state-of-the-art baselines across multiple prediction horizons, achieving significant improvements in both RMSE and MAE. Ablation studies validate the contribution of each component, with the agent attention mechanism proving critical for modeling global spatio-temporal dependencies.

Generalist and code-focused Language Models (LMs) are increasingly applied to software engineering (SE), yet whether they are optimized for understanding SE textual artifacts (e.g., issues, commit messages, developer discussions) remains unclear, as most evidence comes from code-focused benchmarks. We study how to adapt encoder and decoder LMs to SE text, comparing continual pre-training (CPT) against pre-training from scratch (PTS) on a new SE corpus, and evaluating both domain adaptation (SELU) and general-language understanding (SuperGLUE). To keep the comparisons fair, we control pre-training under constant-token and compute-matched budgets. We find that across families and sizes, reusing an existing LM dominates training a domain-native one from scratch: CPT yields small and mostly inconclusive domain gains while leaving general-language understanding essentially unchanged, whereas PTS pays a large and usually decisive penalty on both axes and becomes competitive only for small LMs under a token-rich budget. We distill these results into practical guidance for adapting LMs to SE text and release our corpus and pre-trained LMs in our replication kit.

Federated Learning (FL) enables multiple clients to collaboratively train machine learning models while retaining data locality, thereby enhancing user privacy. However, traditional FL frameworks rely on a centralized aggregation server and assume honest-but-curious clients, making them susceptible to both server-side inference and client-side poisoning attacks. Although recent work has explored secure and Byzantine-resilient FL protocols, they face a fundamental trade-off among privacy, integrity, and verifiability, and incur substantial computational and communication overhead due to the heavy use of cryptographic primitives. In this work, we propose PRoVeFL-a novel, modular FL framework that is Privacy-preserving, Byzantine-Robust, and ensures Verifiable aggregation. PRoVeFL employs multiple servers leveraging multi-key fully homomorphic encryption. Each client encrypts its local model updates and distributes encrypted shares to all servers. This design enables a hybrid computation model in which ciphertext operations are carefully offloaded to the plaintext domain under strict privacy constraints to efficiently evaluate complex statistical aggregation rules. PRoVeFL is compatible with a wide range of state-of-the-art Byzantine-robust aggregation algorithms (e.g., Krum, Trimmed Mean, FLTrust, norm clipping, MESAS, and more) and further enhances them with verifiability mechanisms that require minimal trust in at least one honest server. We evaluate it across different settings and demonstrate its scalability with varying numbers of parameters and participants. PRoVeFL improves runtime over the prior works, Prio and ELSA, based on distributed trust with comparable security guarantees, up to 100x and 10x, respectively.

Automatically recognizing the sentiment, positive or negative, from speech is a challenging task, requiring both the analysis of vocal inflections and the interpretation of uttered words. Recent solutions rely on audio foundation models to solve the task, but it remains unclear if such models can take all aspects into account. To this end, we propose a multimodal solution that integrates audio and text information via cross-modal transformers, where text transcripts are automatically generated via an automatic speech recognition (ASR) tool. Moreover, we create multiple text modalities by automatically translating the transcripts into multiple languages via machine translation tools. Audio and multilingual text features are combined via a cascaded architecture comprising cross-modal transformer blocks that integrate modalities one by one. We further distill knowledge from the multimodal model, called teacher, into a unimodal (audio only) model, called student. We conduct experiments on a large-scale dataset, demonstrating that the automatically generated textual information can bring significant performance boosts in multimodal sentiment polarity classification. Our ablation study confirms that both automatic transcripts and automatic translations are helpful. Moreover, we show that the audio-only model can be enhanced via distillation, boosting performance without any computational overhead during inference. To reproduce the reported results, we publicly release our code at https://github.com/andreidurdun/cross-modal-audio-sentiment.

Portfolio optimization under uncertainty is inherently a multi-objective decision problem involving complex interactions among return, risk, market dynamics, and practical investment constraints. Existing reliability based portfolio optimization approaches primarily rely on static optimization frameworks and often fail to capture sequential decision making, tail risk, and market frictions such as transaction costs. To address these limitations, we propose a deep reinforcement learning framework for multi-objective reliability based portfolio optimization (MORP-DRL). The proposed framework jointly optimizes expected return and downside risk using three complementary risk measures: variance, Conditional Value-at-Risk (CVaR), and Entropic Value-at-Risk (EVaR). To model uncertainty and heavy-tailed market behavior, asset returns are represented using GARCH(1,1), Extreme Value Theory, and a t-copula dependence structure, while realistic scenarios are generated through quasi-Monte Carlo simulation. A Proximal Policy Optimization (PPO) based strategy is developed under practical constraints including transaction costs and portfolio bounds, and is benchmarked against NSGA-II. Experiments on ten global equity indices across pre-COVID, COVID, and post-COVID market regimes demonstrate that MORP-DRL achieves competitive risk-return performance, reduced downside risk during periods of market stress, and scalability to high-dimensional portfolio settings.

LLM serving optimization typically benchmarks many configurations and reaches for heavy profilers when latency targets are missed. We argue for the reverse discipline: estimation is the analytical layer of profiling -- without it, optimization degenerates to grid search. Floor First is a residual-driven triage workflow. Each decode step is modeled as a five-dimensional resource vector (HBM bytes, FLOPs, network bytes, network messages, KV capacity); summing within a resource and maximizing across resources gives an optimistic floor, the plain sum a pessimistic one. Where a measurement lands inside this [max, sum] interval reads out overlap quality before any profiler is opened, and profilers escalate only on residuals above a stated threshold. Deployment alternatives are compared by wall ordering -- which resource wall binds first as load grows -- rather than by point benchmarks. The account is compositional: new attention or state-space variants enter by declaring one module, and the workflow ships as a zero-dependency calculator plus an agent skill that enforces the discipline in agentic optimization loops. As a case study we analyze a DeepSeek-V3.2-style 671B MoE/MLA model on 16 NVIDIA H20 GPUs, whose ridge point of ~74 FLOP/byte (vs ~590 for H100) makes it an extreme decode-oriented part. The floors show TP16 decoding is KV-capacity-limited to ~70 concurrent 8K requests; sparse attention removes the KV-bandwidth term but not the capacity wall; an EP16+DP-attention layout accepts slightly worse same-batch weight traffic for an order-of-magnitude higher capacity wall (~644) -- while single-stream latency favors TP by 2.4x. The layout judgment is thus a computable function of the operating point, explaining why production deployments on identical hardware have shipped opposite attention layouts.

We propose D2PO (Dynamic Direct Preference Optimization), a principled framework for optimizing diffusion sampling policies with respect to timestep schedules and classifier-free guidance (CFG) weights. Our work is motivated by a fundamental limitation of existing student-teacher regression frameworks; low-NFE student samplers are trained to mimic high-NFEteachers, often sacrificing high-frequency texture fidelity while preserving coarse global structures, thereby misaligning the sampler with perceptual quality. D2PO addresses this challenge by reformulating sampler optimization as a preference-based alignment problem, leveraging the Direct Preference Optimization (DPO) framework. To make DPO applicable to diffusion samplers, we model the sampling policy as an energy-based model (EBM), transforming preference comparisons into tractable energy differences. We further introduce a novel energy formulation derived directly from the pretrained score network, enabling preference evaluation in perturbed spaces that jointly capture structural consistency and fine-grained details. Moreover, we introduce dynamic preferences, where the preferred samples used for alignment progressively improve as the sampling policies are learned. This self-improving mechanism replaces rigid static teacher supervision with an iterative, preference-guided refinement process, providing progressively stronger alignment signals. Extensive experiments demonstrate that D2PO aligns diffusion samplers with perceptual quality more faithfully, unlocking the full potential of high-quality teachers and consistently outperforming conventional regression-based schedulers under low-NFE constraints.

We present key challenges and future research directions in the security and privacy of agentic AI, based on a horizon-scanning exercise that brought together thirty leading international experts from academia, industry, and government to engage in focused discussions and collaborative exercises on the emerging risks associated with the growing agency of AI.

Accurate long-term forecasting in complex systems is frequently compromised by dataset-level distribution shifts, where diverse underlying behavioral modes and evolving system states drive the dynamic multivariate time-series. While existing methods predominantly focus on local temporal shifts, they fail to explicitly model the global structural challenge where datasets are composites of distinct operational regimes. In this paper, we propose NEST, a specialized framework designed to model and recompose these evolving structures through a two-phase dense MoE architecture. NEST first facilitates structural specialization by partitioning the dataset into distinct operational regimes through unsupervised clustering in a principled moment-entropy space. We introduce a regime-oriented router mechanism that generates initial expert weights based on temporal content, subsequently refined through geometric modulation to regime centroids. Crucially, rather than acting as monolithic predictors, individual experts function as specialized kernels that capture regime-specific dynamics by evolving unique variate-attention patterns. Extensive evaluations on diverse benchmarks, including heterogeneous network traffic and physical phenomena, demonstrate that NEST consistently achieves state-of-the-art performance. Our code and datasets are available at https://github.com/Aaralshin/NEST

For every loopless matroid $M$ and every Feichtner--Yuzvinsky building set $\mathcal{G}$ containing the top flat, we construct an integral tangent class $T_{M,\mathcal{G}}^{\mathbb{Z}}\in K_{\mathbb{Z}}(M,\mathcal{G})$; in the realizable case it specializes to the class of the tangent bundle of the corresponding wonderful compactification, it recovers the Hilbert series of the Chow ring through Hirzebruch--Riemann--Roch, and it satisfies the expected Chern-alpha lower bounds. This reproduces the tangent class and its key properties studied by the first author in arXiv:2606.22650. The main body of this paper was produced autonomously, without human mathematical guidance, by Danus, an AI mathematical reasoning agent. Danus solved the problem before arXiv:2606.22650 was publicly available, demonstrating the potential of AI agents in mathematical research. We reproduce its output faithfully, adding only editorial comments; the experiment is documented in Appendix B.

Conformal prediction is being adopted in drug discovery to put an honest number on model reliability: pick an error rate alpha, and the method returns prediction sets containing the true label with probability at least 1 - alpha. We show this guarantee can be dangerous on imbalanced datasets. Across four datasets, standard (marginal) conformal prediction hits its global 90% coverage target while leaving the minority class badly exposed: realized minority coverage falls to 64.8% on blood-brain-barrier penetration and to 4.2% on clinical-trial toxicity, where the rare class is nearly abandoned. The failure is not tied to one model: a random forest, a graph network, and a frozen chemical language model all reproduce it (p < 0.001 in every case), with severity tracking baseline calibration on rare labels rather than architecture. A conservation identity explains the effect: the minority's shortfall equals the majority's surplus amplified by the imbalance ratio, predicting the measured gap to within one point and ordering severity across datasets. The failure survives realistic scaffold splits and a second conformal score, while aggregate accuracy and overall coverage stay reassuringly high, which is exactly why it is easy to miss. Class-conditional (Mondrian) conformal prediction closes the gap on every dataset, restoring minority coverage to target for a modest increase in prediction-set size. We localize the failures to generic molecular scaffolds - plain benzene and pyridine cores occurring in both classes - propose a one-number diagnostic, and show with a cost model that abstaining on affected compounds flips a screening campaign from net-negative to net-positive utility. Our contribution is demonstrating on real chemistry how severe and invisible this known conformal-theory gap becomes under imbalance, and laying out a practical protocol restoring per-class reliability.

Real-time decoding is a major bottleneck in scaling quantum error correction (QEC) from noisy intermediate-scale quantum (NISQ) devices to fault-tolerant quantum computing. We present an adaptive confidence-gated decoding framework for the rotated surface code that treats decoding as a two-stage inference problem. A lightweight feed-forward neural network performs fast-path decoding for the majority of syndrome measurements, while only low-confidence predictions are escalated to a minimum-weight perfect matching (MWPM) refinement stage. We benchmark the framework on rotated surface codes with distances $d \in \{3,5,7,9,11\}$ under circuit-level depolarising noise using the Stim stabiliser simulator. The evaluation characterises logical accuracy, confidence-controlled accuracy-latency trade-offs, decoding throughput, per-shot latency, and decoding-graph resource scaling. Routing only 3.3%-6.2% of syndromes to the refinement stage improves logical accuracy from 99.21% for the neural-only baseline to 99.81% at a confidence threshold of 0.95 while incurring only a bounded increase in average decoding cost. Neural-decoder throughput saturates near $4.6 \times 10^{5}$ samples s$^{-1}$ at batch size 512 on commodity CPU hardware, indicating that the neural fast path is not the dominant throughput bottleneck beyond code distance $d=7$. We release the complete benchmarking pipeline, trained models, raw benchmark data, and source code, and explicitly distinguish the experimentally validated contributions from the broader hardware-aware QEC co-design roadmap, including hardware-constrained code discovery, GPU-accelerated inference, and multi-noise optimisation, which remain directions for future work.

We study repeated contextual procurement auctions in which producers have private costs and the platform must learn context-dependent product values from bandit feedback. The objective is welfare rather than revenue or a virtual-cost surrogate: regret is the total surplus loss relative to the full-information efficient procurement rule. We first show that the natural UCB allocation rule attains $\tilde O(\sqrt{ngT})$ welfare regret under truthful bids, but its adaptive bid-dependent learning path does not by itself give a truthfulness guarantee. To obtain exact incentives, we design a bid-independent explore-then-commit mechanism with empirical critical payments; it is dominant-strategy truthful and has $\tilde O((ng)^{1/3}T^{2/3})$ regret. We then introduce frozen-payment UCB, which estimates payments in an initial bid-independent exploration phase, freezes those payment estimates, and continues adaptive UCB allocation learning afterwards. Under a smoothed truthful-path margin condition, this mechanism gives a regret-incentive tradeoff: the near-UCB tuning attains $\tilde O(\sqrt{ngT})$ welfare regret, while the average per-round gain from any fixed deviation is at most $\tilde O(T^{-1/4})$ for fixed $n,g$. A matching lower bound shows that this frozen-payment frontier is unavoidable.

Computer-aided design (CAD) for industrial components requires long-horizon procedural modeling, robust feature dependencies, editable parametric geometry, and production-grade B-Rep execution. Existing text-to-CAD methods have made promising progress in generating CAD programs from natural-language descriptions, but they still struggle when user prompts are ambiguous, underspecified, or only describe high-level design intent. They also rarely exploit expert procedural knowledge naturally available in industrial workflows, such as CATIA operation recordings, macro logs, drawing notes, and engineering descriptions. We present ArtisanCAD, a skill-guided industrial CAD agent with expert-grounded knowledge distillation. The core of ArtisanCAD is CAD intermediate representation (CAD-IR), an executable procedural representation that encodes parameters, ordered operations, MCP tool bindings, dependencies, generated entities, and verification rules. CAD-IR plays two key roles: it first serves as the carrier for distilling expert CAD procedures into reusable parameterized skills; then it provides a procedural scaffold that turns vague or intermediate-level prompts into complete executable CAD operations. ArtisanCAD retrieves expert-derived skills, instantiates and revises CAD-IR, executes the resulting procedure through a dedicated CATIA-MCP backend, and uses multi-view visual feedback for iterative refinement, and finally generates production-ready B-Rep models. On the Text2CAD benchmark, CAD-IR improves generation from intermediate prompts by reducing mean Chamfer Distance from $14.83$ to $9.88$, showing its ability to bridge ambiguous textual intent and executable CAD construction. On four complex automotive components, CAD-IR enables expert CATIA recordings to be distilled into reusable skills, allowing ArtisanCAD to generate editable CATIA-native B-Rep models for new variant requests.

The notion of algorithmic fairness has been actively explored from various aspects of fairness, such as counterfactual fairness (CF) and group fairness (GF). However, the exact relationship between CF and GF remains to be unclear, especially in image classification tasks; the reason is because we often cannot collect counterfactual samples regarding a sensitive attribute, essential for evaluating CF, from the existing images (\eg, a photo of the same person but with different secondary sex characteristics). In this paper, we construct new image datasets for evaluating CF by using a high-quality image editing method and carefully labeling with human annotators. Our datasets, \oursceleb and \ourslfw, build upon the popular image GF benchmarks; hence, we can evaluate CF and GF simultaneously. We empirically observe that CF does not imply GF in image classification, whereas previous studies on tabular datasets observed the opposite. We theoretically show that it could be due to the existence of a latent attribute $G$ that is correlated with, but not caused by, the sensitive attribute (\eg, secondary sex characteristics are highly correlated with hair length). From this observation, we propose a simple baseline, Counterfactual Knowledge Distillation (CKD), to mitigate such correlation with the sensitive attributes. Extensive experimental results on \oursceleb and \ourslfw demonstrate that CF-achieving models satisfy GF if we successfully reduce the reliance on $G$ (\eg, using CKD).

Conditional computation can decouple language model quality from per-token inference cost, yet leading techniques act on a single axis in isolation: Mixture-of-Experts (MoE) sparsifies the FFN, Mixture-of-Depths (MoD) skips whole transformer blocks, and KV-cache quantization compresses attention memory. We argue these three decisions (attention resolution, expert selection, and cache bit-width) are strongly coupled and should be made jointly: a token rare enough to warrant full attention may also need high-precision caching regardless of which expert processes it. We introduce TriRoute, a single lightweight controller shared across all three axes that, for every token at every layer, emits a coordinated policy: (i) an attention mode (skip/local/full), (ii) a sparse set of FFN experts (with a null expert recovering MoD), and (iii) a KV-cache bit-width. The controller trains end-to-end via a heterogeneous relaxation (Gumbel-Softmax with straight-through estimation for categorical decisions and load-balanced top-k gating for experts) under a Lagrangian budget constraint that turns the average compute and memory cost into a controllable knob. We identify a cross-axis routing-collapse cascade in naive joint training, where collapse on one axis propagates to the others, and address it with per-axis normalization and a coupling-aware balancing loss. On decoder-only models from 160M to 1.3B parameters at compute-optimal token counts, TriRoute Pareto-dominates the best independent MoD+MoE+KV-quantization combination at matched inference FLOPs and memory, while better preserving tail-case robustness on rare entities, code, and arithmetic that pure perplexity optimization erodes. Post-hoc analysis reveals interpretable structure: the controller allocates full attention and high-precision cache to sentence-initial positions, rare subwords, and named entities, while cheaply routing function words.

Line segment detection is a key building block in visual SLAM, 3D reconstruction, and industrial inspection. Recent deep learning methods have greatly improved accuracy, yet even the smallest models require several megabytes of memory, exceeding low-cost MCU capacity. This work investigates the maximum achievable accuracy under a sub-megabyte budget. We propose MiLSD, a detector tailored for MCU-level constraints, and systematically compare three output representations within a compact fully-convolutional backbone. Our study shows that the proposed F-Clip center-with-length-and-angle formulation learns most effectively at small model sizes. We find that 8-bit quantization preserves full-precision performance, while 4-bit quantization causes significant degradation, particularly in angle regression, with quantization-aware training recovering only part of the loss. With a one-megabyte activation budget and inference enhancements including sub-pixel decoding, test-time augmentation, and a lightweight verifier, MiLSD improves sAP10 on ShanghaiTech Wireframe from 10.6 (25k parameters, 0.25 MB) to 24.1 within 1 MB. Rather than competing with GPU-scale parsers, we map the accuracy memory trade-off across representations, bit-widths, capacities, and post-processing strategies for embedded vision systems.

Heart rate measurement is one of the key requirements for real-time health monitoring, in particular for health caring of elderly people. Traditional heart rate measurement relies on contact sensing mechanisms such as some heart rate measurement devices at medical hospitals or some wearable devices with embedded sensors such as Apple Watch, etc. In this paper, we develop a system for non-contact, real-time, heart rate measurement using image processing with commodity cameras such as an embedded camera on a laptop, where we use an innovative algorithm to capture the relevant signals for the computation of heart rate in a time series in real life environments. The presented heart rate computation (HRC) process is composed with four major steps: (a) identify frames per second of the camera in use, i.e., 30 frames per second for a given camera, (b) face detection (FD) with shape predictor of 68 face landmarks using deep learning (DL) method, (c) time sliding window (TSW) algorithm to de-noise the signal by smoothing out the noise, and (d) compute heart rate based on identified signal periodicity. We test and analyze the developed prototypes against heart rate results by Apple Watch and check the difference range in multiple rounds and compute the mean of the difference for the measurement values of the heart rate of the same person at the same time. We will do further tuning and optimization of the present methods and deploy the system as a personal AI agent [6] for health monitoring as our future directions.

Public chest-radiograph (CXR) datasets are typically released with small, fixed label schemas such as CheXpert-14. However, the underlying free-text reports describe far more findings -- and which findings matter depends on the task, site, and reader. We release a pipeline that converts free-text reports into multi-label matrices and then reconfigures the label schema through dictionary edits rather than new inference passes, i.e., without relabeling the corpus. After this one-time pass, reconfiguring MIMIC-CXR (223K reports) from cached annotations takes 196 seconds with no API cost, compared to \$6.6K for an equivalent relabeling pass with Claude Opus 4.7. Using a 58-label taxonomy, we show that 43\% of CXR studies contain at least one finding outside CheXpert-14. Image probes trained on these labels match CheXpert-14 probes on shared targets while also reaching 0.78 AUROC on expert-reviewed long-tail labels that CheXpert-14 cannot represent. These results suggest a different unit of work for radiology labeling: once reports are structured, the label schema becomes a configuration to edit, not a corpus to relabel.

Trusted monitoring is a central defense in AI control: a cheaper trusted model scores an untrusted model's actions for sabotage, and the most suspicious are audited or deferred. Such monitors are evaluated against one or two untrusted models, and the accuracy is reported as a property of the monitor. We ask whether it is partly a property of the pairing. We make the untrusted policy family the controlled axis: we fit a monitor on family A's transcripts, apply it to family B, and decompose the cross-family AUROC into how obvious each family's sabotage is, how capable each monitor is, and the residual own-family advantage after both are removed: the interaction. On code-backdoor transcripts the interaction is positive and survives the dominant confounds: +0.172 (95% CI [+0.158, +0.185]) on four open-weight families on a strict leak-free basis, holding across an eight-family matrix spanning four frontier vendors, blatancy-matched controls, a trained transcript-reading monitor up to 34B parameters, and a second code benchmark. The gap is geometric: each lineage's sabotage occupies a different scoring direction, and rotating the monitor's scoring direction toward the target lineage recovers most of the lost detection while a matched random rotation recovers nothing. At a 1% audit budget under defer-to-trusted, an off-lineage monitor catches 19% of attack attempts where an in-lineage monitor catches 41% at the same audit cost, a +21.7-point gap. Control evaluations should report cross-family transfer matrices: a single-pairing accuracy overstates the safety a monitor delivers against a lineage it was not calibrated on. We give a four-step protocol that acts on the gap, with each step a measured result.

Personal AI agents powered by large language models can reason and act using available tools to access emails, manage calendars, and push code to remote repositories, all with minimal oversight. When augmented with long-term memory, an agent can recall specific details relevant to the current task, reducing the need for large context windows. Currently, long-term memory agents tend to fall into two distinct domains: conversational and action-planning agents. Personal assistant agents sit at the convergence of these two domains and handle sensitive information while interacting with untrusted information sources, creating previously unaccounted security vulnerabilities. In this work, we introduce the novel attack vector, GhostWriter, which exploits current memory subsystems in tool-using personal agents to poison their memory store. GhostWriter operates in two phases: injection, where an adversary sends a hidden attack payload to the target agent; and activation, in which the poisoned memory is retrieved. We show that GhostWriter achieves near-universal injection rates of approximately 98% and a high average activation rate of approximately 60% against state-of-the-art agents. This attack is possible due to the lack of security-focused memory governance. In response, we propose Agentic Memory Sentry (AM-Sentry), which leverages two mitigation techniques: a memory-saving policy and a memory-retrieval screen. Our experiments show that AM-Sentry dramatically reduces GhostWriter's success rate while preserving agent utility.

Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training. As models scale, post-training itself becomes a bottleneck. We study a weak-to-strong alternative: run RL on a smaller model where rollouts are cheaper, then reuse what that RL run learned to improve a stronger target model. Directly distilling the post-RL weak teacher is not enough, because the teacher's final policy mixes useful RL gains with the limitations of the smaller model. We propose Direct On-Policy Distillation (Direct-OPD), which transfers the teacher's RL-induced policy shift instead. Direct-OPD compares the post-RL teacher with its own pre-RL reference and treats their log-ratio as a dense implicit reward for the student. In plain terms, the checkpoint pair tells us which actions RL made the weak model more or less likely to take, and Direct-OPD applies that signal on the stronger student's own on-policy states. This directly reuses the weak model's RL supervision signal without running sparse-reward RL on the target model. Empirically, Direct-OPD consistently leverages weaker teachers to improve stronger target models; notably, it boosts Qwen3-1.7B from 48.3% to 58.3% on AIME 2024 in just 4 hours on 8 A100 GPUs. It outperforms step-matched direct RL and enables the sequential composition of multiple policy shifts. Our results show that RL outcomes can be reused across model scales as implicit reward signals, not merely as final models to imitate.

Scaling pre-training, post-training, and test-time compute have become the central paradigms for improving the capabilities of LLMs. In this work, we identify verification, the ability to determine the correctness of a solution, as a new scaling axis. To unlock this and demonstrate its effectiveness, we introduce LLM-as-a-Verifier, a general-purpose verification framework that provides fine-grained feedback for agentic tasks without requiring additional training. Unlike standard LM judges that prompt LLMs to produce discrete scores for candidate solutions, LLM-as-a-Verifier computes the expectation over the distribution of scoring token logits to generate continuous scores. This probabilistic formulation enables verification to scale along multiple dimensions: (1) score granularity, (2) repeated evaluation, and (3) criteria decomposition. In particular, we show that scaling the scoring granularity leads to better separation between positive and negative solutions, resulting in more calibrated comparisons. Moreover, scaling repeated evaluation and criteria decomposition consistently lead to additional gains in verification accuracy through variance and complexity reduction. We further introduce a cost-efficient ranking algorithm for selecting the best solution among candidates using the verifier's continuous scores. LLM-as-a-Verifier achieves state-of-the-art performance on Terminal-Bench V2 (86.5%), SWE-Bench Verified (78.2%), RoboRewardBench (87.4%), and MedAgentBench (73.3%). Beyond verification, the fine-grained signals from LLM-as-a-Verifier can also serve as a proxy for estimating task progress. We build an extension for Claude Code, enabling developers to monitor and improve their own agentic systems. Finally, we show that LLM-as-a-Verifier can provide dense feedback for RL, improving the sample efficiency of SAC and GRPO on robotics and mathematical reasoning benchmarks.

Visual generators excel at rendering, but they confidently fabricate what they do not know. User requests are unbounded, evolving, and deeply long-tailed: new characters, trending entities, post-cutoff events, and more. This world-knowledge bottleneck is structural: generators are trained on fixed corpora, but the visual world is open-ended. We construct SearchGen-20K and SearchGen-Bench, with 20,839 prompts spanning twelve failure categories and twenty-two domains, paired with a pre-executed multimodal SearchGen-Corpus-1M to support offline, reproducible research. On SearchGen-Bench, frontier open generators score only 21 to 28 out of 100, a 40-point collapse invisible to existing benchmarks. The natural remedy is to employ search tools, enabling agentic visual generation. However, we find that naive search fails: it retrieves indiscriminately, injecting noise into prompts the generator already handles. We trace the root cause to a generator-specific, evolving knowledge boundary: the divide between what a generator can internalize through training and what must remain in external context. Although this boundary is hard to specify in advance, we show that it is discoverable through a teach-then-search co-training framework. Even a minimal version of this co-training recipe produces monotonic improvement, laying the foundation for recursive self-improvement in visual generation that can meet world-knowledge-grounded requests. We release the full dataset, co-training corpus, and search corpus as a replayable harness for tool-augmented, world-knowledge-grounded visual generation.

As CMOS technology scales into the deep nanometer regime, digital circuit reliability is increasingly threatened by the combined stochastic effects of Bias Temperature Instability (BTI) and Process Variation (PV). Traditional reliability analysis methods, which rely on computationally intensive simulations or extensive lookup tables, fail to scale efficiently for large designs, creating a critical bottleneck in design space exploration. To address this, we propose SMART, a novel framework that integrates Machine Learning (ML) with Monte Carlo simulation to enable rapid, high-fidelity reliability analysis. SMART employs Random Forest regression to predict gate delay distributions directly, bypassing time-consuming atomic model parameter extractions. Crucially, the model utilizes Bayesian Optimization for automated hyperparameter tuning, ensuring maximum predictive robustness across diverse libraries. Experimental validation on ISCAS85 benchmark circuits demonstrates that SMART achieves a 94.54% reduction in analysis time compared to state-of-the-art methods, while maintaining a remarkable average accuracy error of just 1.63%. By shifting computational complexity to an offline training phase, the proposed framework offers a scalable, accurate solution for designing resilient, reliability-aware digital systems.

Understanding the physics of many-body complex dynamical systems may be a non-trivial task. High-dimensional analysis approaches are often deemed necessary to prevent losing important information. Typically, these use order parameters or descriptors capturing information related to, e.g., relative positions, symmetries, etc., of the units in the studied system. However, in many cases, gaining information related to the relative positions of the constitutive units (or their velocities) alone may be insufficient, and to reach a more complete physical knowledge, one should ideally learn and correlate with each other both structure and dynamics. Here we demonstrate how to achieve such a goal efficiently by building and navigating high-dimensional Time-Derivatives (TiDe) spaces. A TiDe space can be generated for virtually any type of system/phenomenon from the time-series data collected along its observation over time. Each TiDe's dimension corresponds to a growing-order time-derivative of the extracted data, thus containing information related to different physical phenomena/events, which can be easily extracted via unsupervised approaches. We demonstrate how, by definition, TiDes can be directly analyzed without a need for prior dimensionality reduction, providing results that are intrinsically intuitive to interpret. We show the potential of the method by analyzing two prototypical example datasets extracted from molecular dynamics simulations or experimental tracking of different types of complex dynamical systems. Our results demonstrate how efficiently one can navigate and learn in information-rich TiDe spaces, which provide a robust general framework for data analysis and for studying complex dynamical systems from the data collected along their observation over time.

Vibration-based damage identification in civil infrastructure is a challenging, ill-posed inverse problem due to measurement noise, sparse sensor arrays, and environmental variability. While deep learning is powerful for system identification, deterministic approaches lack reliable uncertainty quantification and can yield physically inconsistent results. This work proposes a robust probabilistic Scientific Machine Learning (SciML) framework: a physics-informed Gaussian copula variational autoencoder (PI-GCVAE) for structural health monitoring (SHM). First, we eliminate the need for data-driven surrogates by embedding a differentiable numerical eigenvalue solver directly into the VAE architecture. This ensures that latent space samples satisfy the governing equations of structural dynamics, reducing the trainable parameter space and improving generalization. Second, we replace the conventional independence assumption of latent variables with a Gaussian copula. This model captures complex, physics-dependent spatial cross-correlations between adjacent structural elements, defining feasible solutions while accounting for inherent system variability and measurement errors. Third, compared with alternatives such as Gaussian mixtures, our copula-based VAE provides an efficient distributional model for high-dimensional, strongly correlated latent spaces. We validate the approach using a synthetic dataset of a simply supported bridge subjected to various damage scenarios and corrupted with stochastic Gaussian noise. Synthetic data enables exhaustive validation against ground-truth stiffness values unavailable in practice. Results demonstrate that the PI-GCVAE accurately recovers the true posterior distribution, achieving 77.2% coverage. The proposed framework provides a reliable, scalable tool for early-stage damage diagnosis in operating bridges.

Cell-type deconvolution, the task of estimating the proportions of constituent cell types in a heterogeneous biological sample, is a core problem in computational biology. Methods that rely on epigenetic marks such as DNA methylation typically operate on aggregated methylation estimates, discarding the pattern-level information carried by individual DNA reads. Existing read-level approaches that exploit this information are scarce, and all remain restricted to few-class settings; scaling them further is an open problem because, at scale, non-discriminative reads dominate and hard labels conflict with the many-to-many mapping between methylation patterns and cell types, preventing classifier convergence. To overcome this, we propose data-driven soft labels that estimate the conditional cell-type distribution for each read, and integrate this scheme into Syto, a new modular framework for read-level classification-based deconvolution. On a whole-body atlas of 39 human cell types, Syto reduces MSE by 2.56$\times$ over SoTA, with gains transferring to an out-of-distribution dataset spanning 16 tissues. Syto lays the foundation for modeling increasingly large cell-type panels, with improved applications in biology and healthcare. The proposed soft-labeling scheme is further translatable to any setting with a many-to-many signal-to-label mapping.

Object detectors have many applications in safety-critical systems, but they are known to be sensitive to worst-case perturbations such as adversarial attacks, which limits their applicability in real-world scenarios. Compared with classification, adversarial robustness for object detection has received less attention, and existing methods are often tied to adversarial training, whose performance may not transfer across attacks, perturbation budgets, or architectures. In this work, we introduce Lipschitz-constrained variants of object detection architectures as robust-by-design alternatives to standard detectors. We validate this approach with LipSSD, a Lipschitz-constrained Single Shot MultiBox Detector (SSD), and provide a comprehensive study of its adversarial robustness using multiple white-box adversarial attacks and datasets. We first analyze the accuracyrobustness trade-off induced by Lipschitz constraints and show that it can be controlled through a single training hyperparameter. We then demonstrate that Lipschitzconstrained detectors are complementary to adversarial training: under the same training setup on the Pascal VOC dataset, adversarially trained LipSSD improves mAP@50 on unseen attacks by up to 15 points over classical adversarially trained SSD. Finally, we use more specific safety-critical datasets such as LARD and KITTI, and show that Lipschitz-constrained detectors can improve robustness while largely preserving clean performance. These results suggest that architectural Lipschitz control is a practical and attack-agnostic direction for improving the robustness of object detectors.

Preserving and revitalising traditional textiles such as Ulos, a cultural heritage of the Batak ethnic group in North Sumatra, Indonesia, requires balancing fidelity to tradition with innovative approaches that meet contemporary design demands. Traditional Ulos weaving faces two key limitations: a narrow range of motifs and a time-intensive design process. This study presents a generative AI framework that fine-tunes two pretrained latent diffusion models: Protogen v3.4 and Stable Diffusion v1.4, on a curated, annotated dataset of high-resolution Ulos motifs to generate culturally consistent yet novel designs. Model performance is evaluated quantitatively using Frechet Inception Distance (FID), Inception Score (IS), and qualitatively through assessments by traditional weavers and members of the public. Protogen v3.4 consistently outperforms Stable Diffusion v1.4, achieving substantially lower FID (~10.5x) and higher IS (2.0x), indicating superior visual fidelity, diversity, and closer alignment with the real Ulos motif distribution. We further examine the effects of strength and guidance scale on generation quality across both models. Lower strength values consistently yield higher fidelity (lower FID), while higher strength values increase generative diversity at the cost of realism, revealing a clear fidelity-diversity tradeoff for both models. Across all tested configurations, a guidance scale of 5-9 provides the most effective balance between fidelity and diversity, stabilising FID, KID, and IS, and is recommended as the operating range for high-quality, diverse Ulos motif generation. These findings demonstrate that carefully fine-tuned generative AI can support the creative renewal of intangible cultural heritage while preserving its stylistic and symbolic integrity.

Physical design quality-of-results~(QoR) optimization is hard and expensive. Choices made at one stage can help or hurt later stages. Each evaluation requires a costly EDA run through the full flow. While existing methods still treat optimization as flat parameter tuning or a LLM-based script generation task, we present AgenticPD, a stage-aware agentic framework for physical design QoR optimization. Instead of re-running the full flow after every trial, AgenticPD is organized around the stage boundaries of the physical design flow, where a Judge Agent navigates the search and stage-specialized agents make local decisions within their own stage using stage-local tools. Additionally, the agent harness in AgenticPD provides structured observations, execution history, and agent context management. As a result, the system can branch from prior intermediate states and reuse checkpoints to continue the optimization procedure, and every candidate is evaluated at the post-route signoff. Across these baselines, AgenticPD achieves strong post-route timing while remaining competitive in power and area.

Text embeddings for Portuguese have no dedicated benchmark: evaluation rests on translated corpora such as English MS MARCO or on thin multilingual coverage, with native tasks scattered and unconsolidated. We introduce MTEB-BR, a benchmark of 22 native Brazilian-Portuguese tasks across seven categories (classification, multilabel classification, pair classification, semantic textual similarity, clustering, retrieval, and reranking), admitting only data created or found in Portuguese and excluding translations by construction. We evaluate 93 models spanning 23M to 27B parameters: 73 open-weight and 20 closed commercial APIs. Alongside the leaderboard we report a statistical layer for every headline comparison: per-task bootstrap confidence intervals, paired-bootstrap significance, a task- and instance-level discrimination analysis (how sharply each task separates models) adapted from Item Response Theory, and a cross-leaderboard correlation. Three findings stand out. The benchmark cleanly separates about a dozen tiers of models, though the top six are statistically too close to order. An openly licensed, self-hostable model reaches that leading tier, so strong Portuguese embedding quality does not require a commercial API. And a model's rank on the global multilingual leaderboard predicts its Portuguese rank only moderately (Spearman rho = 0.75 over 55 shared models; one model ranks 3rd there and 49th here), so a native benchmark measures something the multilingual boards do not. We release every task, our code, and a public leaderboard, so practitioners can choose Portuguese embedding models on native evidence.

We present Wan-Streamer v0.2, a latency-preserving upgrade of the native-streaming, end-to-end audio-visual interaction model. v0.2 keeps the v0.1 modeling formulation, but raises the interactive output stream from 192x336 to 640x368 while preserving approximately 200 ms model-side signal-to-signal latency at 25 FPS. The higher-resolution stream supports scene-grounded mid-shot agents whose posture, gaze, hands, nearby objects, and local scene layout remain legible during real-time conversation. To support the larger visual stream without adding user-visible delay, v0.2 keeps the thinker as a single-GPU low-latency path for streaming perception, the short language/state Transformer pass that builds the generation cache, and final decoding. The performer becomes a multi-GPU Ulysses-style context-parallel group for the expensive next-unit latent generation. Each performer rank writes incoming K/V into a pre-sharded local cache. The long high-resolution latent video sequence is split across ranks for denoising and gathered through Ulysses communication, while the much shorter audio latent sequence is generated without sequence sharding. In this split, the thinker's language/state computation reaches the performer only as K/V conditioning, so no separate language sequence has to be communicated inside the performer group. This concentrates additional hardware on visual generation while preserving the compact thinker-performer boundary, keeping total remote interaction latency at approximately 550 ms when a 350 ms bidirectional network budget is included.

Generalist robot manipulation policies have advanced rapidly, yet existing benchmarks remain limited in systematically evaluating their capabilities. Many rely on simple, short-horizon, or skill-narrow tasks with limited capability coverage, and are often conducted only in simulation or only in the real world. Simulation enables scalable feedback but misses physical deployment challenges, while real-world evaluation is costly, time-consuming, and difficult to reproduce. We introduce RoboDojo, a unified sim-and-real benchmark for comprehensive evaluation of generalist robot manipulation policies. RoboDojo includes 42 simulation tasks and 18 real-world tasks covering diverse and complementary manipulation capabilities. The simulation benchmark evaluates five dimensions: generalization, memory, precision, long-horizon execution, and open-vocabulary instruction following, while the real-world benchmark exposes policies to challenging physical-world deployment conditions. RoboDojo supports scalable evaluation through heterogeneous parallel simulation in Isaac Sim and provides RoboDojo-RealEval, a reproducible real-world evaluation system with remote cloud access, standardized hardware, scene reset, evaluation protocol, and deployment interface. Together with XPolicyLab, policies can be integrated once and evaluated across simulation and real-world settings with minimal adaptation. We integrate 30 policies into XPolicyLab and evaluate them on RoboDojo, establishing a public leaderboard and systematic analysis of current policy performance. The website is available at http://robodojo-benchmark.com/.

Final-answer scores hide which agent transitions helped or harmed a trace. We introduce Agent Step Value (ASV), a replay framework that scores before/after states with a stateless LLM evaluator over a fixed candidate set. ASV reports entropy movement and Bayesian surprise measure belief movement, while offline gold-margin gain measures movement toward a reviewed target. It also quantifies evaluator-channel sensitivity by replaying the same frozen transitions under changed projection, rationale, prompt, or scoring rules. In a 100-question open-QA study with live PubMed retrieval and DeepSeek log-probability scoring, ASV evaluates 1,100 transitions. Entropy movement is 0.000 while mean Bayesian surprise is 2.693, exposing near-one-hot belief pivots. Under a 128-token rationale-conditioned protocol, mean gold-margin gain is -2.335 (95\% CI [-3.395, -1.272]); direct one-token scoring on the same traces gives +4.033. A 100-transition component audit traces the reversal to short generated rationales over full states. ASV turns prompt sensitivity into a measured channel effect and localizes the largest rationale-conditioned losses to extraction and audit.

Plant diseases, resulting from both biotic and abiotic stresses, cause an estimated 20-40% loss in global agricultural yield annually, resulting in economic damages exceeding USD 220 billion. Accurate and scalable stress quantification is essential for precision agriculture, yet traditional manual assessments are labour-intensive and subjective. This paper proposes a unified deep learning pipeline integrating semantic segmentation, regression-based severity estimation, and disease classification. Stress severity is categorised into four levels (Low to Very High) based on the proportion of infected leaf area. Experiments on the Apple Tree Leaf Disease Segmentation dataset (1,641 samples, six classes) evaluate four models: U-Net (MobileNetV2), SegFormer, FCN, and PSPNet. U-Net with MobileNetV2 achieves the best performance with 98.20% pixel accuracy, 0.70 mIoU, and 99.41% detection accuracy at 14.7 ms per image, making it suitable for real-time use. SegFormer performs competitively (mIoU 0.66), while FCN and PSPNet show lower spatial accuracy (approximately 0.49 mIoU). The computed severity index strongly correlates with expert annotations (r = 0.968, R^2 = 0.937), demonstrating the system's reliability for automated crop monitoring and decision support.

In 2017, Terry Rudolph introduced an elementary rewriting system that relies on a representation of quantum states as misty states to accurately describe the basics of quantum circuits and quantum computation to high-school and middle-school students. The accessibility and effectiveness of the system are remarkable: every calculation can be done to good-enough accuracy, and perhaps with a small overhead, using just a tiny, universal set of gates chosen to take advantage of a remarkable mathematical result by Yaoyun Shi, leveraging another powerful result by A. Y. Kitaev. The misty formalism greatly simplifies calculations and makes them accessible to first-time learners using only simple arithmetic, and without sacrificing accuracy; it, too, is universal, inasmuch as you can use it to do any quantum calculation with maybe just a small overhead. We don't advocate that we should recast all of quantum theory into this formalism. The misty state picture is a good way of getting people to the heart of some nontrivial quantum theory without having to first absorb a huge amount of (what might initially seem largely) irrelevant math. Our argument is that the misty formalism can effectively be used to facilitate a transition to the full, conventional quantum-mathematical apparatus. To this end, we start by reviewing the original proposal, consider its strengths and limitations, and show it in action via entanglement swapping. We then extend the formalism through a new category of (irreducible) misty states acting as fixed points, and present the GHZ game in this new, general setting and representational semantics.

DNA methylation (DNAm) serves as one of the most robust molecular biomarkers of biological aging. While conventional epigenetic clocks accurately predict chronological age from high-dimensional CpG profiles, they treat aging as a static regression task, meaning they can only output a single score rather than simulating how an entire profile continuously changes over time. To reconstruct these continuous dynamics, we frame lifelong human epigenetic aging as a trajectory inference problem across discrete age snapshots derived from widely available cross-sectional data. We introduce a two-stage computational pipeline: first, an age-regularized Variational Autoencoder (VAE) maps high-dimensional CpG profiles onto a chronologically ordered latent manifold while preserving a generative decoder bridge back to the original methylation space. Second, we model the continuous movement across this latent space via Regularized Unbalanced Optimal Transport (RUOT) that unifies deterministic drift, random diffusion, and non-conservative mass changes. By resolving this RUOT formulation using the DeepRUOT framework, our model fluidly accommodates population-level density shifts like survivorship bias and cellular attrition without requiring rigid biological priors. Evaluated on a large-scale, 80-year pan-tissue dataset, our model demonstrates robust distribution interpolation and uncovers a prominent late-life surge in the learned growth field that mathematically captures the variance expansion driven by stochastic epigenetic drift. Finally, by decoding continuous latent paths back to individual CpG sites, we reconstruct and empirically verify distinct biological aging archetypes, offering a rigorous, generative paradigm for simulating human molecular aging.

The accurate estimation of Above-Ground Biomass (AGB) in mature tropical forests remains a critical challenge in remote sensing, primarily due to the saturation of Synthetic Aperture Radar (SAR) signals in high-density areas and persistent cloud cover affecting optical imagery. To overcome these physical limitations, we propose the Trimodal Coherent Co-attention Transformer (TCCT), a physics-informed deep learning architecture. The TCCT natively fuses optical surface reflectance (Landsat-5) with complex-valued Polarimetric SAR Interferometry (PolInSAR) data from both P and L bands. Unlike traditional fusion methods, our architecture employs complex-valued encoders to preserve spatial phase coherence, coupled with a dynamic co-attention mechanism that acts as an adaptive gating module, reducing the weight of cloud-corrupted optical pixels and shifting reliance to microwave phase data. We also derived a localized spatial allometric calibration model via Levenberg-Marquardt optimization, tailored to the specific wood density of the Paracou region in the Amazon basin. Evaluated using a two-stage protocol, the TCCT first underwent a rigorous 5-fold cross-validation to establish robust global weights (achieving a global RMSE of 4.19 m). Subsequently, following a localized spatial fine-tuning phase over 200 epochs, the model attained an absolute RMSE of 3.78 m and an $R^2$ of 0.33 for Canopy Height Models (CHM), outperforming standard Random Forest, CNN, and Vision Transformer baselines. Our ablation study confirms that preserving phase coherence mitigates deep-canopy signal saturation. When converted to AGB, the fine-tuned TCCT map yielded a Relative RMSE (rRMSE) of 4.51% in dense forest areas above 50 Mg/ha. By meeting the European Space Agency (ESA) BIOMASS mission requirement of less than 20% error, the TCCT provides a robust framework for continuous carbon stock mapping in tropical biomes.

An isolated edge state observed in a finite open chain is usually expected to survive the thermodynamic limit (TDL), with a localization mechanism distinct from non-Hermitian skin accumulation, which localizes the \emph{entire} bulk continuum. We show that scale-sensitive non-Hermitian systems can generically admit a different fate: as we scale up the system size, a detached edge-localized eigenstate can remain sharply visible over a broad window until a critical scale is reached, where it forms an ephemeral bound state in the continuum (BIC) of the open-boundary bulk before being absorbed (entocytosed) at even larger system sizes. We call this phenomenon edge state endocytosis. Its mechanism is fundamentally traced to the Widom expansion of the open-chain characteristic determinant (energy dispersion equation) into contributions corresponding to admissible non-Bloch mode subsets. Each subset contribution factorizes into a boundary-projected Green's function (proj-GF) determinant, which encodes lattice truncation, and a subset-resolved bulk propagation factor, which encodes the system size dependence. We uncover the fundamental distinction: TDL edge states are zeros of the leading-subset proj-GF determinant, whereas endocytosed states are a hitherto-ignored class of hidden proj-GF zeros from subleading subsets that control the spectrum at finite sizes. Due to its fundamental mathematical origin, the endocytosis mechanism is completely platform-independent, occurring generically without fine-tuning when isolated edge states, topological or otherwise, are subject to non-Hermitian couplings that generate the requisite non-locality. Our new framework quantitatively predicts the endocytosis scale and sheds light on how its intricate competitive mechanism can be revealed through experimentally relevant Green's functions.

The coexistence of polar and nematic interactions, observed in a broad range of biological and synthetic active systems, gives rise to a rich phenomenology that continues to challenge our theoretical understanding of non-equilibrium collective behaviour. In this paper, we numerically investigate phase ordering and defect dynamics in a newly introduced minimal single-field model for dry nematopolar systems, where competing polar and nematic contributions enter the free energy, and activity is implemented through a self-advection contribution. At optimal balance between the two alignments, the system develops depolarization strings connecting half-integer defects and separating domains with opposite polarization, together with closed depolarization loops. We first characterize the elementary relaxation mechanisms of defect pairs and loops, showing that the interplay between polar and nematic alignment gives rise to non-monotonic string-mediated interactions, finite equilibrium separations and distinct loop-collapse pathways. Large-scale simulations from disordered states instead show dynamic scaling with a characteristic length growing as $\sim(t/\ln t)^{1/2}$, consistent with coarsening in systems with non-conserved order parameters and point-like defects. Upon introducing self-advection, sufficiently strong activity leads to the coexistence of positive integer and negative half-integer defects, which we term motility-induced charge symmetry breaking, and to saturation of the characteristic length scales, ultimately resulting in arrested coarsening. Overall, our results provide a simple unified framework for understanding the ordering and defect dynamics in biological and synthetic nematopolar systems.

We present an inhomogeneous fluctuation-induced asymmetric escape event of a Feller diffusion confined to a finite interval with two competing absorbing boundaries. The dynamics correspond to an overdamped Brownian motion in a shifted harmonic potential with state-dependent diffusivity. We set two exit points (equal potential energies), equidistant from the potential minima: an extinction site near the origin and an outbreak point. The multiplicative fluctuations are suppressed near the extinction boundary, biasing trajectories toward the outbreak state. The mean exit time exhibits a non-monotonic dependence on the initial condition and the drift-to-noise strength ratio, attaining a maximum when the particle is initialized toward the extinction (low-noise) site. The outbreak possibility is more likely even if the process started with a small initial bias towards the low-noise boundary. The spatial location of the lowest coefficient of variation (CV) is nicely corroborated by the maximum exit time, which is the hallmark of a stochastic escape from an interval. The fluctuation in escape time is found to dominate its mean as a non-trivial function of the drift-to-noise strength and the initial spatial bias. The observed asymmetries in escape also shape the speed-accuracy trade-off in stochastic decision-making events.

Despite being intrinsically athermal and strongly driven, granular materials can exhibit remarkably glass-like dynamics. Whether their rheology can be described by the same scaling concepts remains an open question. Here, we investigate the linear viscoelastic response of an air-fluidized granular bed using small-amplitude oscillatory shear over a broad range of fluidization states. We show that the frequency-dependent spectra collapse onto a single master curve when shifted by a state-dependent relaxation time, establishing a time-state superposition principle analogous to time-temperature superposition in molecular glasses. The master curve spans more than five decades in relaxation time and is quantitatively described by a Cole-Davidson relaxation spectrum. By comparison with continuous shear measurements, we identify tribocharging as the origin of history-dependent deviations from universal scaling. Our results demonstrate that fluidization primarily rescales a single structural relaxation time while preserving the underlying relaxation spectrum, establishing a direct connection between the rheology of driven granular matter and molecular glass-forming liquids.

We present a geometric interpretation of the sum photon blockade effect in multimode quantum optical systems, such as semiconductor microresonators. The blockade condition \(c^{(n)} \cdot v = 0\) reflects the orthogonality of the \(n\)-photon amplitude vector to a target mode vector in an \(N\)-dimensional Hilbert space, visualized as the confinement of the state to a hyperplane. A key result is the calculation of the maximum probability of the system remaining in the blockade subspace under the influence of decoherence processes (in particular, dephasing), which determines the practical feasibility and robustness of the effect. This approach extends to higher-order correlators \(g^{(2)}_Σ\) and cross-correlations, enabling the design of scalable quantum devices. We introduce the concept of "dark-state typicality": as the number of modes \(M\) increases, the dark subspace annihilated by the collective mode operator asymptotically occupies a unit fraction of the \(n\)-boson Hilbert space. This allows the transition from fragile, finely tuned mechanisms to macroscopically robust non-classical light in large multimode bosonic architectures. We consider continuum collective modes, hypotheses on correlation zeros and invariant manifolds, as well as the relationship between blockade and entanglement.

Localized interlayer excitons in moiré heterobilayers can form dipolar exciton ladders, yet their internal spin-valley structure remains unresolved. Here, we use helicity-resolved magneto-photoluminescence to identify the microscopic origin of the ladder in R-stacked WSe$_2$/WS$_2$ at charge neutrality and one-electron filling of the moiré lattice. At charge neutrality, the first two emission peaks correspond to a spin-triplet interlayer exciton and a triplet-triplet two-exciton state separated by 38 meV, reflecting the on-site dipolar interaction. The opposite Zeeman response of the apparent third rung of the ladder rules out its assignment as a spin-conserving three-exciton state and instead identifies it as a triplet-singlet two-exciton configuration with a 22 meV offset set by the WS$_2$ conduction-band spin splitting. At one-electron filling, the correlated electronic background gives rise to charged one- and two-exciton states and intervalley/intravalley two-exciton configurations, while reducing the effective exciton-exciton interaction. Our results establish a spin-valley-resolved picture of dipolar exciton ladders beyond simple occupation-number physics in moiré heterobilayers.

DNA loop formation is a key mechanism in gene regulation, and looping kinetics are sensitive to mechanical tension acting on the DNA. In both single-molecule experiments and biological settings, this tension is typically transmitted through DNA segments flanking the looping region, rather than acting directly at the looping sites. How this indirect force transmission affects the looping time has not been systematically investigated. Using molecular dynamics simulations of a wormlike chain, we show that such flanking segments significantly steepen the force dependence of the looping time, an effect that is insensitive to their length once it exceeds the persistence length, and vanishes when the junction to the looping region is made flexible. We develop an analytical framework that accounts for this effect through a force-dependent shift in the effective free energy landscape of the looping segment. In the limit of small forces, this shift reduces to a zero-force equilibrium average, after which the entire force dependence of the looping time follows analytically. Applying this framework using a coarse-grained DNA model that treats individual bases as rigid bodies, we obtain predictions in quantitative agreement with experimental looping data. Our results demonstrate that the geometry of force transmission has a significant and predictable effect on looping kinetics, with direct implications for the interpretation of tension-dependent looping in both single-molecule experiments and gene regulatory contexts.

Solid-state quantum technologies, including qubits and quantum metrology circuits, demand milli-Kelvin operation to preserve fragile quantum states from classical noise. While the negligible electron-phonon coupling is the major impediment, reaching 50 mK electron temperature is further suffered by the high electrical resistance and sub-micron-scale dimensions of typical devices, limiting conventional heat dissipation. Though the phonons are effectively frozen, thermoelectric techniques could offer a viable path for heat management.This work explores thermally driven electrical transport in a gated quantum dot (QD) on a GaAs-AlGaAs two-dimensional electron gas (2DEG), to control heat flow between the source and drain reservoirs.By exploiting the QD's discrete energy spectrum and tuneable tunnel rates, a precise control over the polarity and magnitude of the resulting thermoelectric current is demonstrated. A temperature difference of 650 mK is maintained across the QD, a separation of 400 nm, by tuning the tunnel-rates. An experimental gate pulsing method is also introduced to directly measure the electron temperature differences across the QD, bypassing the need for any theoretical fits. The results presented here show that tuneable tunnel barriers can be used for local heat control, and could lead to advanced quantum refrigerators that work efficiently in mesoscopic circuits.

We investigate the subsystem entanglement asymmetry in random quantum automaton ensembles, which are generated by permuting the basis states in the Hilbert space and applying global phase shifts. We compute the ensemble average of the $U(1)$ subsystem asymmetry in different connectivity geometries, showing that the late-time limit of the ensemble associated to a 2-local circuit geometry coincides with the all-to-all ensemble average. By focusing on different subsystem sizes, we demonstrate that, similarly to Haar-random circuits, the system locally symmetrizes. However, in sharp contrast to the Haar-random setting, the scale at which symmetrization happens depends on the initial state, a phenomenon we associate with the interplay of conservation of the participation entropy and the uniform exploration of charge sectors. Additionally, we connect the growth of the subsystem asymmetry to the subsystem coherence and show that their growth is characterized by the same symmetrization scale.

Isospectrally patterned lattices exhibit a composite band structure with a tunable ratio of localized versus delocalized eigenstates that is controlled by the underlying phase gradient. We show that the lattice Hamiltonian can be interpreted as that of a single spin exposed to a rotating magnetic field which is allowed to hop with a spin-flip across the lattice. In the low- and high-energy part of the band the localized states show an envelope of oscillatory character separated by quasi-nodes. Spin peaks occur at the locations of these quasi-nodes and provide a unique spin texture to the eigenstates which becomes increasingly complex with increasing degree of excitation. The crossover from localization to delocalization and vice versa leaves its fingerprints in the Fourier spectrum of the eigenstates: the original bimodal frequency distribution widens with increasing degree of excitation, moves across the spectral window and finally culminates in an extremely narrow frequency peak. In the course of this evolution the spin texture undergoes a rearrangement transition involving different characteristic (ir)regular patterns which we quantify by considering the total variation of the local spin fluctuations. Our results demonstrate the variety of the spectral properties of isospectrally patterned lattices which holds great prospect in particular when considering higher lattice or cell dimensions.

Hopfions, associated with higher-dimensional topology through the Hopf fibration, exhibit {exotic} features like {complex knot} and improved stability compared to skyrmions, enhancing their appeal for innovative applications. In this paper, we study the nonlinear response of magnetic Hopfion to microwave fields. {We observe the emergence of higher-order harmonics of the driving microwave field as it interacts with the Hopfion}. By carefully selecting the {driving frequency, the corresponding harmonic can efficiently excite localized magnon state of a Hopfion. Our results demonstrate the promising potential of hopfions in nonlinear magnonics.

We study non-Hermitian antiferromagnetic resonance in an antiferromagnetic insulator/nonmagnetic metal junction with sublattice-dependent damping and spin-orbit torque. By formulating the linearized Landau-Lifshitz-Gilbert (LLG) equation as a 2x2 non-Hermitian eigenvalue problem, we derive the complex resonance frequencies, the exceptional-point condition, and the stability-threshold condition. We show that the largest absorption response is maximized by approaching the stability threshold from the stable side. Near the threshold on the stable side, the absorption spectrum exhibits strong enhancement and linewidth narrowing, together with pronounced helicity dependence in the sub-THz regime that can be switched by reversing the spin-orbit torque. We also confirm by solving the full LLG dynamics that crossing the threshold leads to gain and self-oscillation. These results identify the stability threshold as the key control principle for enhancing absorption and achieving polarization selectivity in the sub-THz/THz regime.

Hyperuniform systems, defined by the anomalous suppression of large-scale density fluctuations, are a paradigm of non-equilibrium self-assembly. While mechanisms underlying the self-assembly of hyperuniform states have been widely studied, the energetics of this process remain unexplored. This raises a fundamental question: what is the energetic cost of self-assembling a hyperuniform system? Here, we address this question across several noisy particle systems drawn from soft matter and machine learning, in which hyperuniformity can be induced by tuning noise correlations. Despite their distinct microscopic dynamics, we uncover a universal behavior across all systems: hyperuniform states are maximally irreversible, as quantified by the entropy production rate. Further, we develop a path integral formulation of the entropy production rate directly from the microscopic dynamics, which explains our observations. Our work establishes a direct link between emergent long-range structure and time irreversibility and opens a new avenue of probing the energetic cost of hyperuniform self-assembly, ubiquitous across physics, biology, and materials science.

High-resolution atomic force microscopy has transformed molecular imaging by revealing intramolecular structure directly in real space. A major remaining challenge is to extend this capability from largely planar molecules to non-planar molecular systems, where the most important structural information may be distributed across different heights above the surface. Here we use probe-particle-model simulations to predict the constant-force contours expected above molecules with increasing structural complexity. By extracting force isosurfaces from simulated three-dimensional force fields, we compare the molecular information retained in constant-height and constant-force images. For tilted benzene and pyrrole, constant-force images preserve the molecular framework across a range of adsorption angles and allow the molecular orientation to be recovered quantitatively. For larger non-planar and three-dimensional systems, simulations identify characteristic force-isosurface contrast associated with adsorption geometry, lower-lying molecular structure and curved molecular surfaces. These results provide target contrasts for force isosurfaces that could be extracted from three-dimensional force-mapping experiments, evaluating the molecular information retained by ideal force-isosurface imaging across progressively non-planar systems.

Root-mean-square deviation (RMSD) is the standard metric of structural comparison in molecular dynamics (MD) simulations. In its conventional form, RMSD assigns equal weight to all atoms regardless of mobility. Hence, flexible loops and disordered regions can dominate a global RMSD, while the rigid functional core contributes negligibly to the overall metric. To address this issue, we introduce the Bayes-optimal RMSD (BRMSD), which optimizes per-atom weights jointly with structural averages by maximizing a Bayesian posterior. In a trade-off between low RMSD and weight uniformity, a position-fluctuation parameter $σ$ controls the transition from classical RMSD ($σ\to \infty$) to a progressive focus on a rigid core ($σ\to 0$). The BRMSD framework supports analysis modules for structural alignment, focused alignment onto a user-specified domain, trajectory smoothing, soft $K$-means conformational clustering, and rigid-domain identification. These modules are implemented in the open-source Python package BRMSD (https://github.com/bio-phys/BRMSD) and benchmarked on two MD systems, the endoplasmic reticulum translocon-associated protein SND3 and the phosphotransferase adenylate kinase.

We demonstrate that the balance between polar and positional order can be systematically tuned through molecular engineering, providing direct control over the emergence of polar and modulated liquid-crystalline phases, allowing for versatile strategy for the design of functional ferroelectric soft materials. We show that polar orthogonal smectic phases (SmAF and SmAAF), promoted by the self-segregation of aromatic cores and sufficiently long terminal chains, are readily destabilized by strong longitudinal dipolar interactions that energetically penalize parallel alignment of molecular dipoles within a smectic layer. In contrast, the tilted ferroelectric SmCF phase is remarkably robust across the entire homologous series, indicating that molecular tilt efficiently relieves dipolar frustration within the smectic layers. We further demonstrate that the interplay between microsegregation and electrostatic interactions stabilizes the new modulated SmCM phase, characterized by incommensurate electron-density waves, particularly for compounds with short terminal chains. For longer homologs controlling the spatial distribution of fluorinated molecular fragments and terminal-chain length enabled the targeted formation of broken-layer-type modulated polar phases (2D or 3D).

We experimentally investigate spin-polarized electron transport for the centrosymmetric altermagnet CrSb, which is known to reveal both altermagnetic and topological features. We demonstrate pronounced first-harmonic anomalous and second-harmonic non-linear Hall effects for a single-crystal CrSb flake with ferromagnetic nickel contacts, while both effects can not be seen for the reference samples with non-magnetic gold ones. For the anomalous Hall effect, we demonstrate bow-tie hysteresis loop in Hall voltage, which is usually ascribed to surface spin textures in magnetic materials. The slope of the Hall curve changes a sign for two orientations of the Hall-bar contact configuration for the same sample, i.e. for the same sign of the charge carriers. We interpret the observed sign inversion and bow-tie hysteresis as the joint effect of the alternating bulk spin splitting and spin-polarized topological surface states in CrSb. The pronounced non-linear Hall effect with hysteresis in magnetic field confirms finite Berry curvature dipole under injection of spin-polarized electrons, i.e. the topological features for the altermagnetic candidate CrSb.

We study a bulk-driven nonlinear variant of the Kipnis-Marchioro-Presutti model of stochastic energy diffusion in which local collisions are biased to induce a net energy flow, resembling the effect of an external field. Starting from the microscopic master equation, we derive the hydrodynamic description of the driven system via a local equilibrium approximation, obtaining explicit expressions for the energy current and the associated diffusivity and mobility transport coefficients, which are nonlinear functions of the local energy density. We test our findings in kinetic Monte Carlo simulations of the model and, as a proof of concept, we demonstrate the versatility of this driving mechanism to control nonlinear energy transport by inducing time-crystalline phases. In particular, we show that appropriately designed packing fields induce the spontaneous formation of traveling energy condensates, exhibiting robust long-range temporal order reminiscent of continuous time crystals. Our results provide a simple yet powerful framework to study bulk-driven nonlinear energy diffusion in stochastic many-body systems, offering a bridge between microscopic dynamics, macroscopic transport, and controlled spatiotemporal order.

Inverse design - specifying a desired functionality and letting a computational algorithm find the optimal structure - has emerged as a powerful paradigm for magnonic device engineering. In this article, we survey the rapidly growing field of inverse-design magnonics, organising it along two axes: the design variables (topology, material parameters, and magnetic field landscape) and the algorithmic toolbox (gradient-free, gradient-based, and neural-network-based methods) together with the differentiable micromagnetic solvers that enable them. We then identify open frontiers that we consider most promising for the next phase of the field: sensitivity analysis and robust design to bridge the gap between simulation and experiment; input shaping and transducer optimisation; the incorporation of nonlinear spin-wave effects as an explicit design resource; spatially structured amplification; self-adapting media and machine-learning-based design; and the long-term vision of a universal, reconfigurable magnonic platform. We argue that magnonics and artificial intelligence are converging from two directions - machine-learning tools for designing magnonic devices, and magnonic devices as hardware for neuromorphic computation - and propose the term AI magnonics to describe this emerging paradigm.

Based on public data, we analyze the distributions of energy and carbon emission over world countries on a scale of the last 40-50 years using their presentation via Lorenz and Pareto curves. These curves in rescaled format remain remarkably stable on this time period being characterized by high values of the Gini coefficient indicating a strong inequality of energy distribution. To explain these distributions, we introduce the ENergy Thermalization Hypothesis (ENTH) according to which these distributions result from the Rayleigh-Jeans (RJ) thermalization and condensation of agents representing different countries. We show that this hypothesis provides an excellent description of Lorenz and Pareto curves obtained from data on the above time period. It also gives natural grounds for inequality relating it to the RJ condensation at low energy states. We additionally trace parallels with the wealth inequality in the world.

The deconfined quantum critical point (DQCP) in square lattice S=1/2 quantum antiferromagnets has been extensively studied with a large body of evidence pointing to a weakly first-order transition scenario. Recent studies, which focused on entanglement at this nearly continuous DQCP in square lattice J-Q models, have observed conflicting bipartite entanglement entropy (EE) scaling behavior. One bipartition choice gave scaling coefficients in remarkable agreement with predictions from the unitary CFT corresponding to the putative DQCP. While another equally natural choice gave scaling coefficients in complete violation of unitary CFT that may be attributed to lack of scale invariance at the known weakly first-order behavior of the model. This motivates the exploration of DQCP behavior via entanglement measures in lattice models with distinct crystalline symmetries. Here we study a S=1/2 honeycomb model that hosts a nearly continuous transition between Néel and valence-bond-solid ground states relevant to probing DQCP. Using large-scale quantum Monte Carlo simulations, we compute the Rényi EE for a variety of bipartitions and test the CFT based description of the DQCP on the honeycomb lattice. For smooth bipartitions, we find no evidence of logarithmic corrections, in accordance with CFT, thereby essentially ruling out contributions from Goldstone modes. For subsystems with corners, CFT predicts universal logarithmic contributions, which we extract for corners with 60 and 120 degree angles and find close agreement with an emergent SO(5) CFT. While we observe scaling consistent with a critical system in the majority of cases, we also demonstrate an intriguing counterexample of the hexagon subsystem that exhibits a subtle period three oscillation. This results in three separate finite-size series, where the sign of the logarithmic term apparently changes depending on the series.

We predict a class of bound states in the continuum (BICs) in non-subwavelength periodic metasurfaces -- Rayleigh BICs -- that emerge precisely at Rayleigh anomalies, where diffraction channels open. In contrast to conventional symmetry-protected and accidental BICs, which are typically engineered in the subwavelength regime with only a few radiation channels, Rayleigh BICs remain nonradiating far beyond this limit, even when multiple diffraction channels are open. Their formation originates from the interplay between collective lattice resonances, Rayleigh anomalies, and anapole states of the constituent metaatoms. We show that Rayleigh BICs are characterized by an unusual cubic scaling of the quality factor in reciprocal space, ($Q \propto 1/k^3$), an asymptotic behavior that is not characteristic of either symmetry-protected or accidental BICs. Our results establish a constructive route for engineering nonradiating states in metasurfaces beyond the subwavelength regime.

A liquid drop falling into a deep pool can create a vortex ring at the right impact conditions. Such drop-formed vortex rings are of importance in nature and technology and the dynamics of rings created by single and synchronised double drop impacts have been extensively studied. In practice, two neighbouring drops rarely impact a liquid surface exactly at the same time, yet the interaction of two unsynchronised vortex rings have not been studied. Here, we have performed experiments with two water drops impacting a water pool at varying time differences $Δt$. By using particle image velocimetry, we have quantified the time-evolution of the resulting vortex rings. We find four distinct categories of vortex ring evolution depending on $Δt$. At $Δt<0.5$ ms, fully symmetric merging of the vortex rings occur. An unsynchronisation larger than this drastically influences the collision and merging, which either becomes asymmetric and incomplete ($0.5 < Δt<7$ ms) or does not happen at all ($Δt \geq 7$ ms). At $7 \leq Δt<80$ ms, the creation of the second vortex ring is impaired, whereas at $Δt\geq 80$ ms, the first impact no longer affects the formation of the second ring and eventually the two rings evolve without influencing each other. We show that these different regimes can be explained by the capillary waves created by the first droplet. Our results demonstrate the importance of the time difference between drop impacts in the creation and subsequent interaction of two adjacent drop-formed vortex rings, which is important for achieving uniform and controlled mixing in high-throughput applications.

We study the emergence of superconductivity in a quantum spin Hall insulator and identify a disorder-driven enhancement of pairing arising from quantum geometry. Using sign-problem-free quantum Monte Carlo simulations of the attractive Bernevig-Hughes-Zhang (BHZ) Hubbard model, we obtain a quantum phase transition as a function of interaction strength for different impurity densities. In the clean limit, the system develops bulk superconductivity for Hubbard interaction $\vert U \vert$ above a finite critical strength. Interestingly, strong impurities significantly reduce such $\vert U \vert$ required for the onset of superconductivity. Our calculations indicate that Cooper pairing first nucleates in subgap ring states surrounding the impurities and then evolves into a globally coherent superconducting phase. Our results demonstrate that impurity-generated bound states can promote superconductivity in systems with strong quantum geometry. This mechanism is expected to be relevant in nearly flat-band systems like moiré materials where quantum geometry plays a dominant role.

Understanding how nonequilibrium driving modifies phase-separation kinetics remains a fundamental challenge. Here we show that phase separation in a two-temperature system exhibits a striking density-induced reentrant coarsening behavior. Using Brownian dynamics simulations and a coarse-grained field-theoretic model, we find that the characteristic domain size grows as $L(t)\sim t^{1/z}$, displaying a reentrant sequence $(t^{1/3} \rightarrow t^{1/4}\rightarrow t^{1/3})$ with increasing density. While the low- and high-density regimes are governed by classical curvature-driven bulk diffusion, the intermediate-density regime exhibits anomalously slow growth. We show that this slowdown originates from a transport bottleneck arising from the interplay of particle diffusivity, particle availability, and attachment kinetics, which suppresses the effective mass flux between domains. Unlike equilibrium phase separation, where density primarily affects morphology and crossover scales, the two-temperature drive renders density a key control parameter for coarsening pathways. Our results uncover a nonequilibrium mechanism for anomalous domain growth in two-temperature systems.

Tissue surfaces exhibit complex curvature during embryogenesis and oncogenesis. Evidence shows that cells can actively sense curvature to regulate behavior and fate, yet the underlying mechanism remains unclear. Here, we develop a vertex model for arbitrary curved surfaces and uncover spontaneous cell size patterning on ellipsoidal surfaces: cells in high-curvature regions are consistently larger than those in low-curvature regions. This non-uniformity arises from a mechanical competition encoded in Riemannian geometry: positive Gaussian curvature reduces the perimeter-to-area ratio of polygonal cells, relaxing cell-edge tension in high-curvature regions, which is compensated by area expansion to maintain global force balance. This area pattern is robust against variations in model parameters and matches observations in biological systems. The perimeter pattern, in contrast, is governed by competition between the intrinsic geometric tendency and the deformation required by force balance, and undergoes reversal beyond a critical shape index. Together, these findings establish self-organized spatial variations in cell size as a potential physical mechanism for curvature sensing.

The substitutional alloy SrTi$_{1-x}$V$_x$O$_3$ interpolates between the band insulator SrTiO$_3$ and the correlated metal SrVO$_3$, exhibiting a composition-driven metal--insulator transition whose origin combines Mott physics with local chemical disorder. Previous first-principles studies relied on individual supercells, which cannot capture the thermally disordered solid solution, since configurations of identical composition can display very different electronic properties. Here we treat the alloy within a generalized quasi-chemical approximation, a thermodynamically consistent statistical framework in which every property is obtained as an ensemble average over all symmetry-inequivalent clusters, weighted by occurrence probabilities that minimize the Gibbs mixing free energy. This provides a well-defined procedure to average over different supercells, and places the structural and electronic descriptions on an equal footing. From the mixing thermodynamics we obtain a miscibility gap with a critical temperature of 1443 K, consistent with experimental evidence. Combining the cluster ensemble with dynamical mean-field theory, we track the density of states at the Fermi level across the full composition range: whereas density-functional theory alone predicts a metal for all $x>0$, the correlated spectral function reproduces the transition, evolving from insulating below $x\approx0.3$ to metallic near $x=1$. Finally, classifying clusters as metallic or insulating and performing site percolation on a simple cubic lattice yields a sharp onset of system-spanning conduction near $x\approx0.4$. These results establish a thermodynamically consistent, configuration--averaged framework applicable to the broader class of correlated materials.

We study invariant measures for soliton systems described by Mealy automata. Motivated by recently introduced soliton models associated with 2-letter, 3-state Mealy automata, we formulate the time evolution induced by Mealy automata on bi-infinite configuration spaces. We provide sufficient conditions for the invariance of Bernoulli product measures and derive a criterion for the invariance of two-sided space-homogeneous Markov distributions. We then apply these general results to three soliton models, which can be interpreted as variants of the box-ball system (BBS). For two of these models, BBS-S(2) and BBS-V(2), we prove that Bernoulli product measures are invariant. For the remaining model, BBS-C(2), we establish a more general result: the invariance of two-sided space-homogeneous Markov distributions, which include Bernoulli product measures as a special case. Furthermore, for all three models, we compute the phase shift associated with the interaction of two solitons, as well as the velocity of an isolated soliton. Although the latter has already been studied previously, both quantities constitute fundamental characteristics for understanding the generalized hydrodynamics of these systems. These results provide a foundation for the study of invariant measures, generalized Gibbs ensembles, and generalized hydrodynamic behavior in Mealy-automaton soliton systems.

Direct, real-space imaging of spin-wave propagation and wavefronts in magnetic materials is crucial for advancing both fundamental understanding of spin dynamics and the development of functional devices. This, however, remains a significant challenge, especially in materials with complex magnetic characteristics at the nanoscale. Here, we employ scanning nitrogen-vacancy center spectroscopy to achieve visualization of spin waves in two archetypical magnetic films: yttrium-iron-garnet and lanthanum strontium manganese oxide. We reveal a wavelength-dependent spin-wave filtering effect near point-like magnetic scatterers and a modified spin wavefront in antiferromagnetically coupled stripe domains. The spin-wave characteristics are explained using micromagnetic simulations and analytical calculations. These findings point to possible fine control of spin-wave propagation near complex magnetic structures and extend the scope of spin-wave imaging based on nitrogen-vacancy centers beyond uniform magnets.

Using a Green's-function formalism, we study the spin-resolved local spectral function of a resonant impurity coupled to a two-dimensional $d$-wave altermagnetic substrate. It is found that the interplay between the direct impurity-STM tip tunneling and altermagnet-mediated tunneling leads to dual Fano resonance in the absence of external magnetic field. Moreover, the anisotropic spin-dependent oscillation of the local density of states and/or Fano factor provide the information of the altermagnetic splitting strength from the local/global perspective. In addition, spin-selective tunneling can be achieved by tuning the Fermi energy and the tip position. In the presence of strong magnetic field with Landau level quantization, the dominant STS signature becomes a spin-dependent nodal structure of the real-space pattern: the nodal mismatch between opposite spin channels produces a large local spin contrast. These results establish resonant-impurity STM/STS as a phase-sensitive local probe of altermagnetic band anisotropy.

Microscopic disorder in a coherent conductor is encoded in the magnitude of the transmitted current and in the energy- and channel-resolved structure of the scattering response. We develop a conductance-balanced learning framework for identifying microscopic disorder mechanisms from coherent quantum-transport spectra. Using armchair graphene nanoribbons as controlled multichannel tight-binding waveguides, we generate ensembles in which distinct disorder classes occupy the same intervals of integrated transmission, removing the dominant shortcut associated with average conductance. Within this constrained setting, we compare scalar transmission spectra with transmission-eigenvalue spectra, and use supervised classification, scale-normalizing controls, plateau-resolved tests, and principal-component analysis to identify the spectral structures that remain informative. The learned distinctions persist after normalization of the conductance envelope and are strongest in multichannel energy windows, where mode mixing and channel redistribution shape the scattering response. Principal components provide interpretable transport coordinates whose loadings identify the energy and eigenchannel sectors responsible for the dominant spectral deformations. The results establish a controlled AI-assisted protocol for learning disorder signatures from nonlinear quantum-scattering data.

The localization transition of the sub-Ohmic spin-boson model generates boson bath correlations that are, at low frequencies, analytically inaccessible to standard coherent-state-polaron variational ansatzes. In this paper, we introduce a ``squeezed polaron'' wavefunction incorporating generic Gaussian boson-boson correlations induced by the spin impurity. This gives the correct critical power-law scaling of bath observables near the localization transition together with a systematically improved ground state energy and a more accurate determination of the critical coupling. Our wavefunction captures the correct mean-field nature of the transition in the deep sub-Ohmic region. It also displays non-mean-field critical exponents in the shallow sub-Ohmic regime. Using the squeezed polaron as a starting point, we derive scattering phase shifts for bosons, and show how they encode the emergent energy scale that vanishes at the localization transition.

Electrons trapped above the surface of superfluid helium are a disorder-free platform for investigating the formation and dynamics of low-dimensional Wigner crystals. A characteristic nonlinear transport feature of this electronic solid suspended above the helium surface is the Bragg-Cherenkov effect, in which the mobility of the smoothly moving crystal is limited by the coherent emission of helium surface waves (ripplons). The effect has been understood in the conventional Cherenkov setting in which the crystal moves at a constant speed. Here we report on transport measurements of electrons on helium confined in a microchannel geometry to investigate the non-equilibrium response of the Wigner solid when it is subjected to a high-frequency driving field. Surprisingly, the experiments reveal a strongly nonlinear transport response of the confined Wigner solid at frequencies nearly an order of magnitude larger than the ripplon frequencies contributing to the conventional Bragg-Cherenkov effect. We relate this observation to the coupling of the Wigner solid to ripplons with higher-order Bragg vectors, which gives rise to a dynamical friction that provides a mechanism for the observed high-frequency pinning.

We train a neural-network Maxwell's demon to extract work from a model of an underdamped micromechanical cantilever subject to thermal noise. The demon, which periodically adjusts the position of a harmonic trap, is trained to maximize the power extracted under steady-state operation. When the demon is given the cantilever position and trap position as inputs it learns a refined version of an existing hand-designed protocol, yielding a substantial improvement in performance. When the demon receives the oscillator velocity as input it discovers a qualitatively different strategy that extracts substantially more work, close to the theoretical power bound. Analysis of the protocol shows that it implements {\em cold damping}: the trap position is displaced approximately linearly with velocity, producing an effective increase of the oscillator's damping coefficient and a reduction of its effective temperature. Thus a neural-network Maxwell's demon rediscovers a well-known cooling strategy from optomechanics, revealing a simple physical mechanism underlying near-optimal work extraction from thermal fluctuations in an underdamped system.

The Microscopic Dynamical Entropy (MDE) introduced in Ref. [1] describes irreversible relaxation of selected variables x within a finite closed Hamiltonian system of fixed total energy E. Here we extend the framework to arbitrary total-energy distributions and time-dependent Hamiltonians. This allows work interactions with external agents or protocols, crucial in stochastic and classical thermodynamics, and driven systems more generally, to be addressed directly. The central step is to reduce the full microscopic description in terms of selected variables x and unmonitored variables y to one in terms of the selected variables and the instantaneous total energy (x,E). The unmonitored degrees of freedom y in Y enter through their conditional phase-space volume (CPV) $Ω_Y(x,E,t)$. This construction allows work and heat to be defined directly for finite Hamiltonian composites, including cases where Y is so small that its temperature is not well defined. We demonstrate monotonic MDE growth, corresponding to the macroscopic second law, under mixing and time-scale separation even for driven systems with arbitrary energy distributions. At trajectory level, we derive detailed and integral fluctuation relations for the MDE, generalising standard large-bath results. We verify these numerically in a driven few-particle Hamiltonian system, showing the emergence of stochastic thermodynamics for collective coordinates, such as the centre-of-mass position, even in closed systems with only ten or twenty degrees of freedom. Physically, these fluctuation relations connect irreversibility to the change in the number of unmonitored microstates compatible with initial and final data for the observed coordinates. Together, our results provide a finite-system Hamiltonian foundation for the emergence of classical and stochastic thermodynamics and extend their applicability to surprisingly small heat baths.

We present a comprehensive computational study of driven quantum dot arrays in a square lattice configuration, subject to an external magnetic field and coupled to a cylindrical far-infrared photon cavity. The driving is introduced through a harmonic modulation of the full electron-photon interaction, therefore including both paramagnetic and diamagnetic contributions. The electron-electron Coulomb interactions are treated within density functional theory, while the electron-photon coupling is modeled using a many-body configuration interaction approach at each iteration of the density functional. By exploiting the unique properties of the cylindrical TE$_{011}$ cavity mode, we demonstrate selective enhancement of diamagnetic two-photon transitions. Our results reveal that the effectiveness of harmonic modulation of the electron-photon interaction is strongly dependent on both the driving frequency and the electron occupation number per dot. When the driving frequency approaches twice the cavity photon frequency, the system exhibits resonant behavior characterized by efficient photon pumping, occupation of higher-order photon replicas, and activation of collective radial Coulomb breathing modes. These findings establish a controllable mechanism for manipulating photon states in coupled quantum dot-cavity systems and provide insights into the interplay among harmonic modulation, photonic excitations, magnetic confinement, and many-body electron correlations in dimensionally reduced nanostructures.

Drying of nanowire-laden elongated droplets is a ubiquitous process in printed electronics fabrication, where the resulting deposition pattern critically determines device performance by controlling nanowire alignment, connectivity, and percolating charge-transport pathways. However, the physical understanding of evaporation-driven deposition is still largely derived from studies of spherical droplets on homogeneous substrates. This gap limits the ability to predict and control deposit morphology in realistic printing scenarios. Here, we use mesoscale lattice Boltzmann simulations to investigate the drying of nanowire-laden elongated droplets on wettability-patterned substrates, focusing on the effects of droplet geometry, nanowire interactions, and nanowire length. The elongated droplet geometry is found to intrinsically induce distinct axial and transverse inhomogeneities in the final deposit. Increasing the effective attraction between nanowires, which mimics changes in surface chemistry or solvent conditions, can improve electrical connectivity but also promotes clustering and local ordering, reducing structural uniformity. In contrast, increasing nanowire length yields a dual benefit by improving long-range connectivity while simultaneously enhancing deposit homogeneity. Our findings provide design guidance for balancing electrical transport and structural uniformity in evaporation-driven printed electronics.

We introduce a Microscopic Dynamical Entropy (MDE) for Hamiltonian systems, defined with respect to a chosen partition of degrees of freedom into a system X and its environment Y. The construction is based on the conditional phase-space volume (CPV), or conditional Boltzmann entropy, associated with the unmonitored degrees of freedom Y. The MDE is a microscopically defined entropy functional of the marginal distribution $ρ_X(t)$, obtained by discarding conditional microscopic information associated with Y from the Gibbs entropy of the joint XY system, while retaining exact Hamiltonian dynamics. This construction clarifies the microscopic origin of thermal entropy. The dependence of MDE solely on $ρ_X(t)$ is consistent with the thermodynamic assumption that the entropy increment of a heat bath Y depends on its heat content and temperature, not on details of its probability distribution. Indeed, the MDE recovers dS = dQ/T connecting entropy increments to heat flow between system and environment. More generally, it provides a consistent description of irreversible relaxation under exact Hamiltonian dynamics, while permitting transient entropy decreases in small systems and in spin-echo type protocols. Under time-scale separation between X and Y, the MDE becomes strictly monotonic in time, recovering the familiar structure of irreversible thermodynamics. The MDE can foreshadow thermodynamics even in a small isolated Hamiltonian system, if X is a well chosen subset of its degrees of freedom. An example is the centre of mass of interacting particles confined to a box. Even for as few as N=10 particles, the MDE increases during relaxation towards a maximum at equilibrium, with increasing monotonicity at larger N. Taken together, our results show that the MDE offers a microscopic interpretation of nonequilibrium thermal entropy and its time dependence within exact Hamiltonian dynamics.

The low-frequency dielectric response of liquid water is commonly described by a dominant Debye relaxation together with additional faster contributions whose microscopic origin remains debated. Here we show that the dielectric function of water between 0.14 and 1.21 THz can be represented by a collective Debye relaxation plus a Drude-Smith term constrained to the zero-dc-conductivity limit. The Drude-Smith spectral weight increases upon heating pure H2O from 20 C to 50 C and decreases upon isotopic substitution (D2O at 20 C vs. H2O at 20 C). Molecular dynamics simulations including nuclear quantum effects show correlated changes in the population of water molecules with unequal numbers of donated and accepted hydrogen-bonds. Ab-initio-based spectra calculations further indicate that the ~0.1-1 THz response contains both nuclear-motion and explicit electronic-polarisation/charge-redistribution contributions. We therefore interpret the excess low-frequency THz response as a localised, mixed nuclear-electronic dielectric response correlated with transient donor-acceptor imbalance in the hydrogen-bond network.

The unavoidable dissipation of heat in electronic nanostructures is a crucial problem, specially when their operation requires low temperatures. It demands finding devices able to control and redirect the excess heat, ideally without perturbing the electrostatic environment. We propose three-terminal junctions working either as thermal routers or as thermal single-pole double-throw switches controlled by a single external knob. Two models are discussed based on resonant tunneling energy filters and different couplings to the heat source: (i) Phase-coherent contact via a scanning tip modulates the relative amount of the two output currents via position-dependent quantum interference; (ii) Coupling via a gate voltage tunable filter selectively switches one of the currents in the presence of dephasing. In the later case, we find that the heat flow using ideal filtering is bounded by fourth the open conductor current.

Twisted bilayer graphene reconstructs from weak breathing corrugation to large common bending near the magic angle, but the origin of this collective crossover has remained unclear. Here we show that the crossover is a soft-mode condensation of layer-symmetric $A_1$ moiré flexural phonons: these modes soften on the breathing branch, lose stiffness near the magic angle, and freeze into the bending morphology. We call this mechanism Moiré Phonon Condensation (MPC). At $θ=1.08^\circ$, it is extremely surprising that displacements of all 11164 atoms in the moiré supercell, with a maximum atomic position shift of 2.30 Angstrom, is captured by only two $A_1$ phonon modes at more than $99.5\%$ spectral weight. A first-harmonic continuum theory identifies a dimensionless control parameter of the phenomenon, showing that as the twist approaches the magic angle, the growing moiré length scale amplifies a smooth stress-bending competition until the flexural stiffness changes sign. Mode-resolved tight-binding calculations further show that the condensed phonon coordinates are electronically active. This work identifies MPC as a twist-controlled structural order parameter for moiré reconstruction.

We investigate how magnetic fluctuations influence spin-orbit torque in ultrathin-film magnetic heterostructures whose Curie temperature $T_C$ is suppressed by confinement. Above $T_C$, the damping-like contribution to spin-orbit field is significantly enhanced while the field-like contribution is suppressed, with the two contributions exhibiting opposite field dependencies. We show that these behaviors are consistent with fluctuation driven mixing between the longitudinal and transverse interfacial spin conductances, which enhances absorption of transversely polarized spin current by the ferromagnet. This mechanism can be activated below $T_C$ by engineering the microscopic magnetic state and by harnessing spin current-generated short-wavelength magnons, suggesting a spintronic analog of heat-assisted magnetic recording.

We consider how the presence of conserved charges affects memory in a classical stochastic process, the symmetric exclusion process, with an observer constantly measuring a single site. We find that the observer's measurement record becomes Markovian (i.e., loses memory) on a timescale that depends on their knowledge of the global charge, namely the total particle number. In particular, when the global charge is unknown a priori, the observer's time series Markovianizes on a timescale constrained by their ability to learn it from their measurement record. Augmenting the observer's record with bulk measurements drives a charge-learnability transition between charge-fuzzy and -sharp phases. We show that the memory timescale tracks the learnability timescale, diverging in the fuzzy phase and remaining finite in the sharp phase.

Quantum circuits under sufficiently weak monitoring purify on a timescale $T_P$ exponentially long in the system size. This slowness underlies a universal purification dynamics, whose quantitative description has so far required the replica trick, with a delicate analytic continuation. We show that monitored Clifford circuits on $L$ qudits of prime dimension $q$ bypass this construction entirely: in the scaling limit at fixed $x = t/T_P(L)$, purification reduces to the Markovian decay of the density-matrix rank, an exactly solvable death process descending from infinity. We compute the full scaling functions in compact form: all Rényi entropies collapse onto a universal curve $\langle S(x) \rangle$. Exact stabilizer simulations at $q=2,3,5$ confirm the predictions, with no fitting parameter for the global model and $T_P$ as the only fitted scale for local brick-wall circuits. Also, the replica problem amounts to a tilted version of the same Markov process, in agreement with exact computations from the Clifford commutant. Finally, the quantization of the rank leaves two hallmarks that distinguish Clifford dynamics from generic monitored circuits: the entropy fluctuations saturate at short scaled times $x\to0$ to an $O(1)$ variance, instead of vanishing, and observables develop a temporal modulation periodic in $\log_q x$, which cannot be captured by the replica approach.

Accurate knowledge of the on-chip temperature is essential for understanding and optimizing the performance of superconducting qubits, yet direct thermometry at millikelvin temperatures remains challenging. While qubits themselves are sensitive to the temperature of their environment, other factors may affect the qubits` effective temperature, and using them as thermometers with any accuracy requires specialized measurement protocols and qubit designs, limiting their practicality for routine diagnostics and adding complex infrastructure to any hardware testing apparatus. Here we demonstrate a complementary on-chip thermometry method based on superconducting Dayem bridges that are integrated on the same chip as transmon qubits. By extracting the critical current of the Dayem bridge from I-V measurements, we obtain a local, quantitative measure of the chip temperature without the need for microwave calibration or qubit-specific control sequences. To demonstrate the utility of the Dayem bridges as thermometers, we fabricate them in-situ with qubits on the same chip, calibrate the Dayem bridge critical current as a function of temperature, and characterize its resolution and stability at cryogenic temperatures. We additionally perform simultaneous measurements of the Dayem bridge thermometer and qubit excited-state population, and show agreement over the relevant temperature range, validating the method against established qubit thermometry. Furthermore, we correlate the independently measured chip temperature with qubit energy relaxation and dephasing times, demonstrating the utility of this approach for diagnosing temperature-dependent decoherence mechanisms. These results establish integrated Dayem bridges as a simple, non-invasive, and scalable tool for cryogenic hardware development, and on chip thermometry in superconducting quantum circuits.

We evaluate the utility of the spin-zero manifold of an exchange-coupled array of $N$ spins for tasks in quantum computation and quantum simulation. Since pairs of electrons can be readily initialized into a product state of singlets in semiconducting quantum dot arrays, the full spin-zero manifold is available with exchange-only control, providing a Hilbert space of approximate dimension $2^N/(N/2)^{3/2}$, asymptotically close to the $2^N$ dimension of the full spin Hilbert space. Leveraging the spin-zero manifold enables larger computational space in a given array compared to traditional exchange-only control, in which spin arrays are organized into modular units of $n$ spins comprising $N/n$ encoded qubits, limiting to the exponentially smaller Hilbert dimension $2^{N/n}$. Here we focus on benchmarking metrics for this resource utilization by generalizing cross-entropy benchmarking, mirror benchmarking, and out-of-time-ordered correlators to this system. We show that operating in the spin-zero manifold can accelerate the realization of computational quantum advantage applications in semiconductor-based spin qubits.

A parallel in-plane magnetic field could, in principle, distinguish between two competing physical scenarios for the experimentally observed density-tuned 2D metal-insulator transition (where decreasing the carrier density leads to a crossover from an effective metal to an effective insulator): Wigner crystallization or Anderson localization. Since the main scattering mechanism in 2D doped semiconductors arises from screened random charged impurities and screening in turn depends on the electronic density of states, the in-plane magnetic field could distinguish between the two by decreasing screening through spin polarization and this enhances the effective critical density for Anderson localization compared with Wigner crystallization. We give the general theory and provide results for the quantitative magnitudes of the spin polarization effect on the transition density by focusing on two recent experiments [Z. Ge, et al, arXiv:2510.12009, T. Han, et al, arXiv:2604.00113], noting that the critical density may actually decrease if the dominant scattering is by short-ranged defects instead of long-ranged charged impurities. The difference between the two cases arises from whether spin polarization dominates screening (enhanced critical density) or the Fermi surface (suppressed critical density).

The origin of the fractional state at $ν$ = 1/2 observed in double-layer quantum Hall systems has been under debate for decades. Because of the variation of bilayer charge distribution and interlayer tunneling strength, the half-filling state can be attributed to a two-component(2C) or a one-component(1C) origin, which corresponds to Halperin state and Pffafian state, respectively. Here we report the magnetotransport measurement in decoupled twisted double bilayer graphene(TDBG), which has been proved to be a promising platform for double quantum Hall system. Fractional quantum hall states in both odd and even denominator fillings are observed. We also found that the half-filling state occurs at zero displacement field at $ν_{tot}$ = -1/2, which is theoretically consistent with two-component Halperin-Laughlin (Ψ331) state. Moreover, we report the transition from two-component state at zero D field to one-component non-Abelian state by tunning displacement field. Our observation of the half filling state and its transition from 2C to 1C state provides the tunability of decoupled twisted double bilayer graphene and shed light on the understanding of the ground states at half-filling factor in the double quantum Hall system.

We revisit the problem of two static nonmagnetic vacancies in the transverse-field Ising chain with first- and second-neighbor couplings $J_1$ and $J_2$, now on the critical line, using density-matrix renormalization-group (DMRG) calculations in open chains of up to $N=300$ sites. In contrast to the gapped regime studied previously, where the vacancy-vacancy interaction decays exponentially, along the entire quantum critical line the interaction becomes algebraic, $|Δ_b(r)|\sim r^{-α}$, with $α$ close to the universal Casimir value of unity and a weak but systematic dependence on the second-neighbor coupling, $α_\infty \simeq 1.070 + 0.091\, J_2/J_1)$ across $J_2/J_1\in[0.1,1.0]$. The transmission ratio of the spin correlator across a vacancy approaches a $J_2$-dependent plateau $T_\infty(J_2)$ that grows from $0.11$ to $0.33$ over the same range, and the Affleck-Ludwig boundary entropy is small and approximately constant, $\log g_\infty \approx -0.073$, well above the Ising fixed-BC value $-\ln\sqrt{2}$ and close to the free-boundary value. The three observables vary smoothly and monotonically with $J_2$, consistent with a one-parameter family of partially transmissive conformal defects controlled by $J_2$. Throughout, the critical line is located using the bulk spin-correlator exponent $η=1/4$, the order-parameter exponent of the Ising universality class, which provides a robust criterion in this open geometry.

Spherical colloids with harsh repulsive forces have long been used as experimental analogs of the hard sphere model, with demonstrated good agreement with computer simulations for bulk and structural properties of the fluid, glass and crystal phases. However, an enigmatic discrepancy remains for the crystal-melt stiffness coefficient. Here we perform computer simulations of colloidal hard spheres under tunable buoyant mass and show that the long-standing discrepancy can be traced to a hitherto unrecognized gravity dependent contribution of the stiffness coefficient. This effect is one practical realization of a more general result for the external field dependence of stiffness coefficients of arbitrary interfaces.

Interaction between different two dimensional materials can give rise to many exotic physical phenomena which are rarely observed in intrinsic materials. Recently, several theoretical and experimental works have revealed that magnetic proximity effect between pristine graphene and magnetic substrates can lead to the emergence of quantum anomalous Hall states and quantum spin Hall states. However, interplay between correlated states in graphene-based systems and magnetic materials has seldom been studied. Here we perform the transport measurement at ultrahigh magnetic field of twisted double bilayer graphene (TDBG) on CrOCl (COC) substrate, which is an antiferromagnetic material. Instead of a magnetic-exchange effect on graphene, we observe an enhanced correlated insulating state at half-filling factor of TDBG as a result of the charge-transfer process between TDBG and COC. The temperature and magnetic field dependence of this enhanced state are further studied. Our results demonstrate the influence of charge-related effect at the interface, and shed a light on a new route for manipulating the correlated states in graphene-based moiré systems using interfacial engineering.

Understanding phonon-mediated heat transport in structurally complex materials remains a central challenge for next-generation electronic and nanomechanical devices, where grain boundaries and interfacial disorder strongly limit thermal dissipation. Although classical interatomic potentials enable large-scale simulations, their limited transferability can lead to inaccuracies in vibrational properties and interfacial phonon scattering. In this work, we develop a machine-learning based framework for modeling thermal transport in bulk and nanocrystalline silicon by combining Gaussian approximation potential (GAP) and multi-atomic cluster expansion (MACE) models with lattice-dynamical calculations and non-equilibrium molecular dynamics (NEMD). Harmonic and anharmonic force constants derived from machine-learning interatomic potentials (MLIPs) are used within a unified Phonopy/Phono3py workflow to compute phonon dispersions, lifetimes, and lattice thermal conductivity, providing an internally consistent description of vibrational properties. In nanocrystalline silicon, NEMD simulations directly quantify the thermal boundary resistance associated with grain boundaries and reveal its sensitivity to interfacial roughness and the underlying interatomic description. Compared with the Stillinger-Weber and Tersoff potentials, the MLIPs provide a more accurate and internally consistent description of bulk and interfacial phonon transport, enabling more predictive modeling of nanoscale thermal transport in low-dimensional materials.

Random exploration is usually quantified by how fast new space is found, from the range of a single walker to the territory collectively covered by many walkers. In competitive exploration, first arrival secures an exclusive resource, as when foragers compete for food items or agents capture distributed targets. It is then no longer enough to know which sites have been discovered: one must determine, for each discovered site, which searcher reached it first. We introduce the discovery share $X_n$, the fraction of the first $n$ collective discoveries secured by a tagged searcher. For two identical competitors, exchange symmetry fixes $\langle X_n\rangle=1/2$, but the central question is whether this equal split emerges in each long exploration history or only on average, i.e. whether early competitive advantages are erased or persist. Here we show that the answer is controlled by the spectral dimension $d_s$, defined by the large-time decay of the probability that a single searcher is at its starting point after $t$ steps, $p_0(t)\sim t^{-d_s/2}$. Across ordinary diffusion, long-range superdiffusion and subdiffusion induced by crowding or memory, $d_s$ separates persistent randomness in recurrent exploration $(d_s<2)$, anomalously slow non-Gaussian concentration for $2\le d_s<3$, and Gaussian concentration, logarithmically corrected at $d_s=3$, for $d_s\ge3$. For $d_s\ge2$, we derive exact asymptotic variances, including prefactors, and the discovery scale on which competitive imbalances are erased. Two-point correlations of first-discovery labels identify the memory mechanism behind these regimes. The same phase structure persists under changes in geometry, competitor heterogeneity, number of competitors and memory, revealing a general fluctuation theory of first-arrival inequalities.

Modeling higher-order interactions (HOIs) in nonlinear networks with static topologies is often physically restrictive. We demonstrate that standard 3-body Kuramoto couplings are mathematically equivalent to pairwise connections modulated by latent variables of transmission channels. While standard HOI topologies emerge in the adiabatic limit of these variables, relaxing this constraint reveals that latent channel timescales dictate collective macroscopic states. Specifically, transmission inertia drives bistability for symmetric interaction tensors and anti-phase cluster synchronization for antisymmetric ones. Furthermore, dynamically induced clustering in global topologies emerges as a finite-size effect of the dynamics of the local channels. Ultimately, we show that relying exclusively on static topologies restricts interaction modeling. Integrating latent variables captures the transient inertia and fundamental asymmetry of physical networks, bridging the analytical utility of higher-order functions with the reality of the underlying transmission medium.

Wave turbulence is a regime of interacting nonlinear waves occurring in most physical systems. Kelvin waves are helical distortions that propagate along vortex filaments and are believed to play a central role in quantum turbulence up to atmospheric vortices. Yet, Kelvin wave turbulence has remained inaccessible to direct experimental observation. Here, we report the first direct experimental observation of Kelvin-wave turbulence along a single vortex filament in a classical fluid under controlled conditions. Using high-resolution spatiotemporal measurements, we resolve Kelvin-wave dynamics over a broad range of scales and obtain wave-amplitude spectra consistent with the predicted weak-turbulence cascade. We identify six-wave resonant interactions as the mechanism driving this energy transfer, providing direct experimental support for a long-standing prediction of weak-turbulence theory. These results establish an experimental platform for investigating energy transport along vortex filaments, with broader implications for both classical and quantum turbulent systems.

We investigated kink-antikink collisions in a $(1+1)$-dimensional $φ^8$ scalar field theory with multiple degenerate vacua. We presented explicit soliton solutions for different vacuum structures characterized by the ratio $n = p_2/p_1$. We focused on the cases $n=2$ and $n=3$ with four distinct vacua, and performed systematic numerical simulations across all topological sectors. We revealed that there is a complete annihilation regime in the $(1/2,\,1)$ sector, where kink-antikink pairs annihilate for all initial velocities. Up to our knowledge, such phenomenon is first reported in the kink antikink collisions. We revealed a comprehensive fractal cartography of multi-bounce resonance windows across sectors $(-1,\,-1/2)$, $(-1,\,-1/3)$, and $(-1/3,\,1/3)$. Given the kink antikink ($K\bar{K}$, or $\bar{K}K$) ordering, the effective potential and the spectrum of the Schrödinger like equation were presented. By collecting all the effective potential and collision phenomena, we proposed that the effective potential classification scheme provides a predictive framework for collision outcomes including escape, bion formation, sector change, and annihilation. Especially, it is identified that, when soliton pairs pass through each other, the abrupt changes of the potential type could explain both the sector-change and annihilation phenomena. Our work highlight novel dynamical features of the $φ^8$ model absent in lower-order field theories, and establish connections between topological structure, vibrational modes, and effective potentials.

The Hindmarsh-Rose (HR) neuronal network has recently been the subject of studies highlighting the influence of the electric field on the chimera states within it. In this study, we demonstrate the influence of the magnetic field on three categories of chimera states previously discovered in the same network: the traveling chimera state, the traveling multicluster chimera state, and the traveling multicluster chimera breather. The study is entirely numerical and proceeds in each case with three different applications of the magnetic field: first, the entire network is subjected to the field; then, half of the network is subjected to it; and finally, two symmetrical but distinct regions are also subjected to the field. Several phenomena emerge, the most notable of which are the multitraveling chimera state and the multialternating chimera state. This thus illustrates the ability of the magnetic field to transform areas of incoherence into areas of coherence, thus enriching the synchronization field and throwing more light on the field's influence on brain cells.

We investigate a (2+1)-dimensional nonlinear spin system containing an effective spin-current transport term. Based on its integrable structure, exact magnetic lump solutions are constructed on a rotating spin background, including both fundamental and higher-order configurations generated via the Darboux transformation. The obtained excitations are doubly localized in spatial directions, while their temporal evolution is characterized by intrinsic spin precession rather than translational motion of the localized envelope. It is shown that the effective spin-current contribution enters the localization coordinate and acts as a geometric control parameter for the spatial structure of the solutions. In particular, spin current induces anisotropic deformation of the localized profile, leading to a continuous transition toward a quasi-one-dimensional soliton-like state under specific parameter regimes. More importantly, this deformation mechanism is found to be universal across different hierarchical lump structures, including both fundamental and higher-order solutions, indicating that spin current governs a unified structural modulation law for the entire family of localized spin excitations. These results provide an analytically tractable example of spin-current-controlled anisotropic deformation and dimensional crossover in nonlinear spin systems, and further reveal a universal mechanism for geometric control of localized spin textures beyond individual solution types.

Pair generation in time-varying media is commonly attributed to time reflection at temporal boundaries or to amplification inside momentum k gaps. Here we show that these two processes are connected by the time domain Stokes phenomenon. A finite duration photonic time crystal (PTC) slab provides the necessary Stokes connection between the incident vacuum mode, transient k gap amplification, and time boundary scattering. With Kerr nonlinearity, the otherwise unbounded amplification is arrested, spawning Kerr stabilized k gap solitons. When these solitons cross the exit boundary of the time slab, Stokes induced mode conversion produces a secondary pair generation process, yielding four spatially separated and entangled pulse branches. Detection of a backward propagating light pulse therefore heralds its forward propagating partner. We further propose combined Hanbury Brown Twiss and Hong Ou Mandel measurements to test their nonclassical correlations. These results reveal a link between asymptotic Stokes physics and quantum temporal scattering in PTCs, and suggest a route toward ultrafast heralded quantum light sources.

Modular miniature robots could provide scalable function in constrained environments, but coordinating many imperfect modules remains difficult when computation, communication and reliability are limited. A central robotics challenge is to coordinate many actuator-sensor modules without assigning a privileged leader, prescribing a fixed gait template, or relying on dense communication. Here we introduce a programmable synchronization-graph framework for modular miniature robots in which each actuator-sensor pair is represented as a network node and locomotor coordination is encoded through graph coupling. Fixed intra-subgraph links synchronize heterogeneous actuator groups, whereas a small number of signed inter-subgraph links program phase relationships between groups. In physical robot collectives with up to nine modules, graph coupling drives the emergence of synchronization, signed links tune the phase difference from in-phase to out-of-phase motion, and floor experiments produce gallop-like and trot-like contact patterns in a five-module robot assembly. Replacing dense all-to-all coupling with sparse d-regular topologies preserves synchronization while reducing the coupling burden. The same graph representation also captures fault tolerance: increasing graph degree increases the number of module deactivations tolerated before desynchronization. Finally, an upper-confidence-bound edge-selection algorithm learns inter-subgraph links that drive the system toward target phase states. In a separate deactivation benchmark, the graph-based controller avoids the leader-specific failure mode observed in centralized leader-follower control and reduces worst-case phase error by about threefold. These results establish programmable network topology as a compact control layer for gait phase programming, online adaptation and robustness to unit loss in modular miniature robots.

Continuum descriptions of epithelial tissue mechanics can replace expensive individual-based simulations with tractable macroscopic models, yet the link between cell-scale forces and tissue-scale transport remains poorly understood. We show that dimensionality controls this link: long-time mechanical relaxation rates reveal generalized porous-media-type nonlinear transport phenomena, $D(ρ)\proptoρ^γ$. Exponents in nonlinear diffusivities are fixed by microscopic mechanics and dimensionality, providing a novel physical mechanism for emergent macroscopic transport phenomena.

Biochemical traveling waves transmit signals across cells and tissues, but the thermodynamic cost of reliable propagation remains unclear. We develop a stochastic thermodynamic framework for reaction--diffusion systems with stable traveling waves and show that diffusion of the wave position is bounded by the dissipation specifically associated with propagation. The bound follows by projecting noisy field dynamics onto the adjoint translational mode, which maps the wave position to an effective biased random walk. Its tightness is controlled by the non-self-adjoint part of the linearized dynamics, with finite wave speed and antisymmetric reaction dynamics generically producing deviations from equality. For excitable trigger waves in a FitzHugh--Nagumo model, we show that the slow inhibitor dominates the propagation cost, yielding a trade-off among wave speed, inhibitor amplitude, and dissipation. We test these predictions in stochastic simulations of a microscopic Belousov--Zhabotinsky reaction--diffusion system and find consistent signatures in mitotic trigger-wave experiments in \textit{Xenopus} egg extracts. The same relation further imposes an annihilation-limited bound on the reliable signaling rate of wave trains.

We collect rank two difference-differential Lax pairs for classical Painlevé equations in the literature and put each in $2\times 2$ matrix form with the coefficient matrix of the spectral equation a degree two matrix polynomial. We describe and apply a general method to obtain integrals of motion in characteristic $p$ from these Lax pairs. For every relevant Painlevé equation, this leads to a countable list of integrals of motion, with one entry for each prime $p$.

We demonstrate how the unified framework for binary-choice dynamics can be used to study the role of annealed and quenched disorders in homogeneous and heterogeneous systems. The framework defines the structure of interactions between agents without imposing their functional forms. Such a high level of generality allows us to connect many different models across disciplines and find universal rules that apply to all of them. Within this framework, agents update their states under the influence of two competing mechanisms chosen according to individual preferences. We review the literature to classify existing models as homogeneous or heterogeneous based on their preference distribution, and we discuss the role of annealed (changing) and quenched (fixed) disorders in modeling these preferences. Using the framework, we derive a constraint on the transition rates. When a model meets this condition, three major things happen: annealed and quenched dynamics become equivalent, any heterogeneous system can be mapped into a homogeneous one, and oscillations cannot emerge. We illustrate these consequences using models from statistical physics, opinion dynamics, and disease spreading. Finally, we discuss the framework limitations and its potential further developments.

Surface plasmon polaritons (SPPs) and magnetic polaritons (MPs) are fundamental resonance modes that are widely used to tailor the thermal radiation properties of micro/nanostructured metamaterials. Lumped circuit models (LCMs) are usually constructed empirically to describe the MP resonance conditions, and different LCMs have to be constructed for different orders of MPs, but these are difficult to be built for high-order MP modes due to the complex electromagnetic field distribution. This work proposes a new type of circuit model, distributed circuit model (DCM), to describe and predict multi-order MP resonances inside the structure based on the minimum total impedance condition. This allows both fundamental and high-order MP resonances to be predicted with a unified circuit, significantly simplifying the analysis of high-order MPs. More importantly, the DCM shares a similar and clear physical picture as the LCM for describing MPs. The MP resonance conditions for four typical structures are derived. Theoretical predictions based on DCMs are compared with and validated by rigorous numerical simulations. This study deepens the understanding and facilitates the design of MP-based thermal radiation metamaterials.

Single-cell molecular characterization remains a bottleneck in scalable biological analysis because of labeling requirements, limited multiplexing, and reagents that perturb physiology. Raman spectroscopy addresses these limits by providing chemically specific, label-free vibrational fingerprints, but long acquisition times and specialized instruments restrict high-throughput use. Here, we overcome this barrier by showing that spectral fingerprints can be reconstructed from brightfield microscopy using generative modeling. We introduce Pic2Spec, a framework that learns a shared latent biochemical representation linking image morphology to vibrational spectral structure, enabling virtual Raman spectroscopy without hardware. We validate Pic2Spec across mammalian and bacterial cells, generating high-fidelity spectra that reproduce measured Raman fingerprints with 98% cosine similarity and Pearson correlations of ~95%, while preserving biochemical peaks and population distributions. Beyond spectral similarity, Pic2Spec provides molecular-level resolution in bacterial systems: generated spectra discriminate mutation-driven transgenic states and predict GFP expression with accuracy approaching true Raman measurements, outperforming conventional image analysis by 20%. These findings establish Pic2Spec as a first demonstration of chemically informative virtual molecular fingerprinting from brightfield images, complementing slow, hardware-intensive spectroscopy with computational inference. By redefining microscopy as an inference-enabled molecular profiling platform, Pic2Spec democratizes label-free biochemical phenotyping and overcomes the hardware and time constraints that have confined spectroscopy to specialized laboratories. This enables high-throughput molecular analysis for clinical diagnostics, screening, and monitoring at the scale and accessibility of standard microscopy.

In semilocal density functionals the exchange potential decays too rapidly outside atoms and molecules and lacks the correct $-1/r$ tail; as a consequence, functionals from PBE to modern meta-GGAs such as SCAN support no bound Rydberg-like state of an atom. We revisit the gradient-corrected exchange functional that Gill and Pople (GP93) constructed to reproduce the exact exchange potential of the hydrogen atom, solve their equation as an inverse problem, and use the resulting enhancement factor -- the GP93 factor -- as a tail-generating ingredient of a switched semilocal functional. The GP93 factor is exact on the hydrogen $1s$ density and, by uniform coordinate scaling, on hydrogenic $1s$ ions, where it yields the eigenvalue $-Z^2/2$ exactly (hydrogenic exactness). Its large-gradient growth has the hydrogen-exact form $O[s(\ln s)^{2/3}]$, distinct from the $s\ln s$ form of Armiento and Kümmel. The factor is combined with a kinetic-energy-density indicator so that its divergent branch acts only in one-electron-like regions; the GP93 ingredient is nonempirical, while a few switching parameters are calibrated to balance tail recovery against self-consistent-field stability. The resulting functional produces Rydberg-like series of bound virtual Kohn--Sham states in systems whose outermost shell is a one-electron-like $s$ shell -- all-electron H and He, and Li, Na, and K with large-core pseudopotentials -- and moves the highest-occupied eigenvalue toward the experimental ionization energy. For $p$-shell atoms (Ne, Ar) the switch closes, so the domain of applicability is fixed by design. Fixed-density tests indicate that a reduced-Laplacian switch distinguishes atomic tails from covalent bond centers, and that a density-only kinetic-energy functional renders the entire switch orbital-independent, so that no generalized Kohn--Sham solver is required.

Most MLIP benchmarks reward static accuracy while ignoring inference efficiency and hardware scalability -- driving model bloat with unclear real-world value. We benchmark 23 mainstream open-source MLIPs on a low-cost NVIDIA DGX Spark (128 GB native memory, capped at 80 GB to mimic ordinary lab hardware), using a fixed 192-atom system under a unified ASE-based pipeline. We evaluate three dimensions: predictive accuracy, MD simulation throughput, and atomic scalability. Our results expose a sharp accuracy-efficiency trade-off: large SOTA models deliver only 3-5 meV/atom more accuracy than lightweight ones, but lose orders of magnitude in throughput -- in the worst case, becoming only marginally faster than DFT itself. Lightweight MLIPs, by contrast, sit on the Pareto frontier and run on modest hardware. The lesson is that single-dimensional benchmarks mislead the field, and that future MLIP development should value efficiency and scalability alongside accuracy.

Existing RLC circuit model leads to inaccurate predictions of the quality factor (Q-factor) of magnetic polariton (MP) resonances under imperfect absorption conditions due to the omission of radiation loss. Moreover, the lumped-parameter nature of RLC models also limits their applicability for predicting higher-order MP modes. In this letter, we propose a distributed circuit model (DCM) for predicting the Q-factor of MP resonances, which overcomes these limitations. By introducing a radiation resistance to characterize radiation loss and establishing a mapping between distributed and lumped parameters, we derive a unified analytical expression for the Q-factor of arbitrary-order MPs. Validation via rigorous coupled-wave analysis (RCWA) demonstrates that this model can accurately predict the Q-factors of MPs of all orders in various metal-insulator-metal (MIM) structures. This work provides a simple yet effective tool for designing metamaterial emitters/absorbers and advances the understanding of MP loss mechanisms.

Polarization measurements of a classical 1550-nm signal are collected and analyzed on 15-km hybrid aerial-inground fiber connections over 11 months. The spectral area and spectral moments9 of mHz-resolution Fast-Fourier-Transform (FFT) of these measurements are extracted, and related to temperature, humidity, wind speed, and time of day. Spectral area correlations show a strong11 diurnal structure: daytime maxima align with temperatures/wind speed peaks and humidity dips, with lower levels during the night. These diurnal patterns also show seasonality, with higher13 mean and variance in summer than winter. A random forest regressor is used to estimate FFT features from environmental measurements, informed by a theoretical model

Chiral patterns have been observed in various processes from swirling bacterial colonies to tissue morphogenesis and cytoskeletal organization, yet the physical mechanisms underlying chiral cell motion remain poorly understood. Motivated by experiments demonstrating directional bias in the circular motion of confined cells, we use the tools of dynamical systems analysis with computer simulations to identify minimal intrinsic and extrinsic mechanisms capable of generating persistent biased migration. The dynamical systems framework reveals a common organizing principle: directional bias emerges through changes in the stability and/or basins of attraction of the clockwise and counter-clockwise motility states. We find four distinct routes to such bias. First, intrinsic torque in a polarized cytoskeleton can be spatially integrated to produce biased circular motion. Second, anisotropic cell-substrate friction can generate directional preference when reduced friction along the polarity axis is coupled to a directional offset. Third, a chiral wall-alignment response can also produce a persistent directional preference. Finally, substrate patterns that break mirror symmetry, such as dextral or sinistral ridges and troughs, can likewise bias rotational direction. Together, these mechanisms yield distinct, testable predictions and suggest a unifying lens for experimental interrogation of cellular chirality and the design of synthetic systems with programmable chiral motion.

We present Hamiltonian Replica Exchange Transition Interface Sampling (HRETIS), a path sampling framework designed to efficiently sample rare events in systems with complex potential energy landscapes. HRETIS introduces a helper potential within a Hamiltonian replica exchange scheme, which enhances exploration of path space when the underlying potential is not well suited for conventional path sampling approaches. This is particularly advantageous for systems exhibiting multiple pathways separated by orthogonal barriers such as in drug (un)binding, where standard algorithms often show slow convergence since they become trapped within specific pathways. By exchanging Hamiltonians between the path ensembles, HRETIS overcomes these limitations and increases the decorrelation between subsequent paths in the Monte Carlo chain. We demonstrate that HRETIS provides robust and accurate kinetics in several systems, including coarse-grained simulations of amino acid permeation through a dipalmitoylphosphatidylcholine (DPPC) membrane. Moreover, HRETIS is found to improve sampling efficiency and convergence, illustrating its potential as a powerful tool for rare event sampling in complex molecular systems.

Computational on-chip spectrometers are emerging as a powerful platform for portable spectral analysis, combining photonic integration with advanced signal processing to enable a wide range of in-situ sensing applications. We propose a broadband reconstructive spectrometer based on wave chaos in a stadium microresonator with a nanostructured scattering layer for full-area speckle readout. Wavelength dependent interference within the chaotic microresonator encodes the spectral information into a spatial intensity pattern that can be computationally inverted to reconstruct the input spectra. The optimal fabrication parameters of the SU-8 polymer nanostructured layer yield a surface roughness of 176nm and a root mean square thickness of ~2um. We experimentally validate our spectrometer at visible and infrared wavelengths, with resolutions of 43pm at 630nm and 8.2pm at 1550nm. The spectral reconstruction is demonstrated for single and multiple narrowline sources as well as for a broadband (~1nm) pulsed laser source. The broad experimental validation and compact footprint (0.05mm2) establishes our chaotic microresonator-based speckle spectrometer as a robust and versatile platform for high-resolution, on-chip spectral sensing.

Rising temperatures create new challenges for local heat adaptation. Yet, it remains unclear how urban activity changes during hot periods and which urban environments people concentrate in as temperatures rise. Here, we perform a hyperlocal spatiotemporal analysis of urban activity across 10 German cities over a two-month period in 2024 with different levels of heat exposure. To monitor urban activity, we use fine-grained telecommunication data to map locations of people with high spatio-temporal resolution (i.e., hourly at 100m x 155m grid cells), yielding more than 100 million data points. We then link activity counts with hourly weather records and point-of-interest data. We find that sustained periods of hot weather, defined as at least three consecutive days with daily maximum temperatures $\geq 25$°C, are characterized by below-expected city-wide presence, with activity counts that are 1.5 percentage points below regular urban activity. During hot periods, urban activity concentrates more strongly around leisure- and culture-oriented amenities (e.g., cafés or swimming pools), with an increase of up to around 10 percentage points relative to cooler days, while public-service environments (e.g., educational and health facilities) show weaker or negative shifts. Our study provides policy-makers with fine-grained monitoring of which urban areas attract citizens during heat exposure, which can enable evidence-based, spatially-targeted urban heat adaptation plans.

Non-Hermitian frameworks extend conventional Hermitian physics, offering a powerful paradigm for describing open systems. Central to this field are various singularities within the complex parameter space, such as exceptional points (EPs) and scattering zeros, which dictate exotic physical behaviors. As research shifts from isolated singularities toward multi-singularity interactions, conventional planar metasurfaces remain constrained by limited tuning dimensions. Here, we propose a mirror-coupled design that maps a metasurface into a quasi-high-dimensional parameter space. By employing a metallic plane to generate image resonators, this scheme multiplies the system degrees of freedom without increasing the number of physical resonators. Its implementation on a reconfigurable platform integrated with PIN diodes yields the coexistence and manipulation of an EP and multiple reflection zeros. Through simulations and microwave experiments, we characterize the dynamic evolution of these singularities and exploit their synergistic effects for two distinct applications. First, for tunable absorption, multiple reflection zeros are spectrally coordinated to achieve a near-perfect absorption band exceeding $99.9\%$ across the X-band, thereby dynamically suppressing target scattering. Second, for enhanced sensing, a reflection zero couples with the EP to form a hybrid singularity. This hybrid state inherits the power-law sensitivity of the EP while substantially boosting robustness against fluctuations, resolving the conventional trade-off between sensitivity and stability and simplifying detection to direct peak tracking rather than complex multimode eigenvalue fitting. Our work provides a general methodology to circumvent parameter competition among non-Hermitian singularities, opening new avenues for multifunctional metadevices across the electromagnetic spectrum.

We study a driven selection mechanism on a fixed heavy-tailed network. At each step fresh mass is injected, its direction is recomputed from the current mass profile by a power-normalization rule, and the combined mass is transported by a primitive mixing matrix. The exponent $θ$ controls the feedback. Positive values give more weight to larger coordinates, while negative values favor smaller ones. When $θ=0$, the injected mass is distributed uniformly. After deterministic growth of the total mass is scaled out, the long-run injection profile is characterized by a nonlinear Perron-Frobenius fixed point on the simplex. Hilbert's projective metric gives a simple way to understand the stability of the system. The discounted network response brings positive profiles closer together, while the escort map scales their projective distance by $|θ|$. On heavy-tailed networks, this fixed point separates three effects that are often conflated: response or degree tilt, anomalous inverse-participation-ratio scaling, and genuine few-node localization. Positive feedback selects high-response nodes and, when response follows degree, a hub-directed branch. Negative feedback selects low-response nodes and typically produces a broad peripheral cloud unless the lower tail of the response field is itself thin. Numerical experiments on finite power-law networks support these results. They show convergence, illustrate when the forcing rate becomes unimportant because mixing is sufficiently fast, and confirm both the sign law and the crossover in the participation ratio. This mechanism is different from both conserved-mass condensation and graph growth. Instead, feedback selects a non-equilibrium profile on a fixed, heterogeneous network.

Biological fibers exhibit exceptional mechanical properties such as high stiffness, toughness, and elasticity. The stiffness of bio-fibers is governed by a hierarchical microstructure that is highly sensitive to the extraction method, post-processing, and environmental factors such as temperature. Commonly, the microstructure features an amorphous matrix comprising polypeptide chains that interact through weak intermolecular bonds and interconnect via crystalline domains. In this work, we develop a microscopically motivated energy-based model that sheds light on the underlying mechanisms governing the stiffness of bio-fibers. The initial deformation is driven by (1) the entropic extension of polypeptide chains, (2) the elastic stretching and rotation of rigid crystalline domains, and (3) the distortion of weak intermolecular interactions, leading to the relative sliding of polypeptide chains. The model captures the influence of key physical microstructural quantities such as chain alignment, chain stretch, intermolecular bond strength, and crystallite size on the overall stiffness. The merit of the model is demonstrated through a comparison to spider silk and cocoon silk fibers. We also employ the model to show that the reeling speed during the extraction of spider silk fibers governs the microstructure and, therefore, leads to different stiffnesses. We follow with a parametric analysis that sheds light on how different microstructural quantities affect the stiffness. Lastly, the framework is extended to account for thermally-induced microstructural changes and the model predictions are compared to experimental data on the stiffness of cocoon silk fibers as a function of temperature. The findings from this work delineate the role of microstructure on the overall stiffness and offer a pathway for the efficient design of tunable and optimized biomimetic fibers for target applications.

The implications of selecting different unit cells are often overlooked in both direct studies of microstructured materials and their homogenized equivalents. Investigating the effects of unit cell selection is crucial not only for understanding boundary phenomena but also for identifying which finite-sized metamaterial performs best for a given purpose (e.g., achieving zero or minimal transmissibility in the band-gap range). This study examines how the choice of a representative unit cell in periodic metamaterials influences boundary effects and, consequently, the transmissibility of finite-size samples, while providing a strategy based on eigenfrequency calculations that enables a priori optimization.

Light coupling into integrated microresonators is typically achieved using straight bus couplers that are optimized to achieve critical coupling at the pump wavelength but lack spectral engineering capabilities. Pulley couplers offer additional degrees of freedom and opportunities, but optimizing their spectral response requires intensive and time-consuming numerical simulations. To address this challenge, we developed a semi-analytical approach to quickly and efficiently calculate the coupling quality factor of pulley couplers. This enables the design of either broadband or wavelength-selective pulley couplers, depending on the target application. We illustrate the dispersion engineering capabilities offered by this new approach in the context of Kerr microcombs and experimentally validate it on the SiN photonic platform. We show that it is possible to achieve a non-dispersive broadband coupler, enabling the generation of frequency combs with similar spectral characteristics (span and intensity) irrespectively of the chosen pump wavelength.

Efficient optimization and control of photonic computing and communication systems increasingly rely on accurate surrogate models/digital twins. While data-driven models may achieve faster inference than traditional physics-based methods, they typically suffer from poor training data efficiency and limited generalizability. To address this trade-off, physics-informed data-driven modeling has emerged as a powerful hybrid paradigm. This paper presents a comparative analysis of these three modeling paradigms across three benchmark use cases: optical amplifiers, directly modulated lasers, and interferometer meshes. By evaluating model complexity, data efficiency, generalizability, and modularity, this work provides a detailed analysis of the respective trade-offs and highlights the advantages of combining physical insight with data-driven learning.

Electromagnetic waves are intrinsically vectorial and require description via polarization, unlike scalar fields such as acoustic pressure or electronic wavefunctions. In three dimensions, the transversality constraint further prevents any globally smooth transverse-polarization frame at the $Γ$ point, which would apparently rule out a simple scalar band structure for three-dimensional (3D) photonic crystals. We show here that site-adapted $p$-orbitals can realize scalar-wave dispersion: the induced band representation is isomorphic to the scalar elementary band representation up to a one-dimensional character twist, so the symmetry-enforced degeneracies and compatibility relations are the same. We demonstrate this mechanism experimentally in 3D photonic meta-crystals, where the local $p$-orbital axes adapt from site to site according to symmetry. In contrast to a fixed-polarization reduction (e.g., in 2D), our construction preserves site-polarization textures while simultaneously supporting a scalar network with one amplitude per site. Thus, it offers a pathway from vectorial photonic degrees of freedom to scalar band engineering, keeping polarization as an active design knob.

Molecular passivation has become central to reducing photovoltage losses in metal-halide perovskite solar cells, but its electronic action is still often inferred from device-level metrics rather than directly resolved at the nanoscale. Here, we use amplitude-modulated Kelvin probe force microscopy to examine how [3-(2-aminoethylamino)propyl]trimethoxysilane (AEAPTMS) modifies the surface potential and photovoltage dynamics of mixed-cation, mixed-halide perovskite thin films. AEAPTMS homogenises the dark contact potential difference (CPD), narrowing its distribution from ~45.7 to ~14.6 mV without obvious morphological changes. Under illumination, passivated films show a larger steady-state surface photovoltage (SPV) and faster stabilisation, with the SPV increasing from ~345 to ~417 mV and the stabilisation time constant decreasing from ~840 to ~470 s. Wavelength-dependent SPV further indicates reduced sub-bandgap electronic disorder. By separating grain-boundary and grain-interior contributions, we show that AEAPTMS suppresses grain-boundary potential barriers, linking amino-silane passivation to a more homogeneous and stable carrier landscape.

Turbulence plays a central role in shaping the structure and dynamics of the interstellar medium (ISM), governing the star formation rate (SFR) and the initial mass function (IMF). A key consequence of turbulence is the generation of density fluctuations, which regulate the amount of dense gas available for star formation. Accurate measurements of the three-dimensional (3D) turbulent density dispersion are therefore essential for understanding molecular-cloud structure and star formation. However, observations typically provide only two-dimensional (2D) column densities and are often affected by measurement/detector noise. The Brunt method estimates the 3D density dispersion from 2D column-density maps, but it does not account for finite signal-to-noise ratio (SNR). Here, we extend the method to recover the 3D turbulent density dispersion from noise-contaminated observations. Using numerical simulations spanning a range of density perturbation amplitudes and noise types, we identify a characteristic noise wavenumber, k_noise, corresponding to the intersection of the signal and noise spectra. Restricting the Brunt reconstruction to wavenumbers below k_noise yields a denoised density-dispersion estimate that closely reproduces the noise-free result. We provide a practical prescription to determine k_noise directly from the measurement SNR and image resolution. Alternatively, if the noise spectrum is known, it can be subtracted directly from the observed spectrum, eliminating the need to estimate k_noise. The proposed correction recovers the noise-free density dispersion with errors of <~5% for SNR>=3 and <~15% for SNR>=1, enabling substantially more reliable estimates of turbulent density fluctuations from noisy column-density data.

Magnesium ions are essential for RNA structure but difficult to model due to slow binding kinetics and experimental limitations. We present an enhanced-sampling strategy that accelerates Mg$^{2+}$ inner-shell binding by orders of magnitude, enabling quantitative exploration of ion-binding motifs in a large ribozyme. The method combines a barrier-flattening bias with Hamiltonian replica exchange to efficiently sample multiple equivalent binding sites, and builds on an approach that achieved top performance in the CASP16 blind assessment of RNA solvation structure. Using cryo-electron microscopy maps for validation, we introduce a local analysis framework that infers the population of individual binding motifs from their agreement with experimental density, enabling site-by-site validation. We find that insufficient sampling of inner-shell binding leads to significantly poorer agreement with experiment, whereas force fields predicting different inner/outer binding equilibria remain largely indistinguishable at the current experimental resolution. These results highlight the dominant role of sampling in modelling divalent ion binding and provide a general strategy for integrating simulations with experimental data in complex biomolecular systems.

Computationally efficient models for multimodal traffic flows with inter-modal interactions are foundational for coordinated traffic management. However, such models are lacking in the current literature. This study introduces a multimodal link transmission model (M-LTM) accommodating both continuous road traffic and discrete tramway traffic at the network level. M-LTM builds on a link model that captures the inter-modal interactions through the moving bottleneck theory. We further develop node models describing the flow transfer, tram operations, and mode interactions at tram stops and intersections. Specifically, the tram dwell process and dwell-induced congestion of road traffic are simulated at the stop node, and the intersection node model can reproduce queue spillback and tram priority. Case studies are conducted on two synthetic one-node networks, a synthetic arterial network, and the real network of Dresden to demonstrate the predictive power and the applicability of the proposed model in network traffic analysis and management. The GEH statistics of road traffic are below 4, and the average tram arrival time absolute errors are less than 10 seconds.

Compressible physical fields are widely present in the real physical world, but current artificial intelligence lacks an understanding mechanism for the non-differentiable features in compressible physical fields. Addressing the limitations of existing deep learning architectures in handling global non-differentiable features, we propose the Inverse Low-Dimensional Manifold reconstruction framework (ILDM). This framework couples the Non-differentiable Approximation Function (NAF) for capturing non-differentiable features in compressible flows with the Smooth Fluid Reconstruction (SFR) module tailored for smooth fluid regions. Extensive evaluations across 1D and 2D benchmarks, including Riemann problems and double Mach reflection, demonstrate that ILDM significantly outperforms cPINN and R-adaptive DeepONet. Specifically, ILDM achieves superior localization of non-differentiable interfaces and maintains robust super-resolution performance even with low-resolution inputs, establishing a physically consistent and scalable paradigm for data-driven fluid dynamics.

Reynolds-averaged Navier-Stokes (RANS) turbulence models are known to perform poorly in predicting the dynamics of Rayleigh-Taylor mixing when turbulence is not fully developed, particularly during the transition from an initially perturbed interface. In this work, we investigate the use of data-driven strategies to enhance a simple $k$-$\varepsilon$-$b$ model for this transitional regime. The turbulence model is first embedded within a surrogate physics-informed neural network (PINN), enabling the calibration of coefficients that account for parametric errors and the identification of corrective terms representing structural errors associated with missing physical processes. The learned corrections are then re-expressed onto the model state variables and relevant flow indicators, leading to explicit analytical modifications of the closure. The resulting fully interpretable corrected model is assessed against an extensive database of direct numerical simulations (DNS) of Rayleigh-Taylor flows. This framework enables improved predictions of the mixing-layer growth during the transition to turbulence.

Efficient terahertz (THz) generation with high field amplitude and pulse energy is studied through the interaction of a single-color long-wavelength infrared (LWIR) laser pulse with gaseous targets. Particle-In-Cell (PIC) simulations are performed to investigate the underlying mechanism and analyze the properties of the emitted THz radiation. The results reveal that THz pulses are generated via photon deceleration of the LWIR laser, driven by enhanced electron density accumulation at the pulse front in the self-modulated wakefield regime. The influence of key parameters, including target density, laser intensity, and propagation length, on nonlinear laser modulation and the resulting THz generation efficiency is analyzed. Important scalings governing the laser-to-THz energy conversion efficiency are identified from PIC simulations and validated through theoretical analysis. The study demonstrates a laser-to-THz energy conversion efficiency of approximately $4\%$, significantly exceeding previously reported values. The field amplitude of the emitted THz pulses is found to be on the order of 100 GV/m, with a pulse energy of approximately 50 mJ for the laser parameters considered in this work. The findings of this study provide valuable insights for the development of next-generation high-energy THz sources.

Molecular dynamics is used to study radiation cascades in solid C60 under ambient conditions. Simulations are performed for Primary Knock-On Atom (PKA) energies from 0.1 to 1 keV, and cascades are sampled over many PKA directions to collect statistics. Energies and forces are described using the Environment Dependent Interaction Potential for carbon paired with the Ziegler-Biersack-Littmark potential for short-range interactions, and cascade behaviour is characterized by tracking kinetic energy, hybridization and bond connectivity as a function of time. Compared to most materials, fullerite exhibits an unusual radiation response due to weak thermal transfer between C60 molecules leading to a thermalization phase lasting hundreds of picoseconds. The cascades damage the C60 molecules and link them together, and a linear relation is found between the number of cross-linked molecules and the number of new sp3 atoms. The threshold displacement energy computed is 18 eV, in agreement with experiments

Stimulated Brillouin scattering (SBS) microscopy enables label-free mapping of local viscoelastic properties, but frequency-domain implementations are often limited by uncertainty in the pump-probe frequency-difference axis. We demonstrate an RF-defined single-laser electro-optic-modulation SBS microscope in which the pump and probe are derived from the same optical carrier and their frequency difference is set by an electro-optically generated sideband. This architecture makes laser-frequency noise largely common mode and eliminates optical wavelength tuning during spectral scanning. It achieves Brillouin frequency shift and linewidth precisions of 0.07 MHz and 0.30 MHz, respectively. Comparison with a low-NA reference linewidth indicates a system-level spectral broadening of approximately 3.1 MHz, corresponding to an effective spectral resolution of approximately 3 MHz. Imaging of femtosecond-laser-modified chalcogenide glass resolves MHz-level Brillouin contrasts corresponding to 10^-4-level apparent longitudinal-modulus contrast. This work demonstrates the feasibility of transferring the frequency definition of SBS spectral scanning from optical wavelength tuning to RF-domain control, providing a new conceptual and technical basis for high-precision, high-spectral-fidelity Brillouin imaging.

In this work, a path integral Monte Carlo (PIMC) algorithm for fictitious identical particles (FIP) is proposed by introducing a $ξ$-ensemble and performing PIMC simulations on the resulting $ξ$-ensemble partition function. The PIMC algorithm with the $ξ$-ensemble allows us to obtain the thermodynamic properties of FIP for different $ξ$ values in a single simulation. Moreover, it also accelerates the simulation by improving the sampling efficiency, compared to the usual case where independent simulations are performed for each $ξ$ value, in the sense that the autocorrelation time of samples belonging to the same $ξ$ sector is decreased. Simulations of the uniform electron gas and uniform warm dense beryllium are performed to validate the improved algorithm and study the improvement in its sampling efficiency.

The problem of signal detection under an unknown background can be framed as one of inferring the weight of a mixture model with one misspecified component. Banerjee and Algeri (2026) show that, for this problem, the conservativeness of the inference is entirely determined by one single parameter, called the compensator. They demonstrate that, when the data are independent and identically distributed, an inferential approach based on the compensator circumvents the need to estimate the density of the misspecified component and the associated challenges. The main purpose of this manuscript is to broaden the scope of such an approach and extend it to the case in which, as is often encountered in modern experiments in physics and astronomy, the data consist of Poisson counts observed over a large number of bins.

Detecting terrestrial exoplanets in the habitable zones of nearby stars remains a critical challenge. Such planets can be \(10^8\) to \(10^{10}\) times fainter than their host stars and lie at diffraction-limited angular separations, where starlight strongly obscures the companion signal. Here we present an adaptive quantum measurement method for estimating the number, positions, and brightnesses of mutually incoherent point sources in the sub-Rayleigh, ultra-high-contrast regime, operating at contrasts down to \(10^{-8}\) -- five orders of magnitude beyond previous quantum imaging approaches to exoplanet detection. The method adopts a spatial-mode basis that is updated to maximize the quantum Fisher information per detected photon. Estimation is performed by maximum likelihood in log-brightness coordinates, and the source count is determined by Bayesian-information-criterion (BIC) model selection directly from photon-count statistics, without a tunable detection threshold. For point sources within sub-Rayleigh separations and with brightness ratios spanning eight orders of magnitude, the method reconstructs complete scenes with a mean success rate of \(72.5\%\). Furthermore, it is robust to misalignment, maintaining a \(71.3\%\) success rate under offsets of up to six pixels. These results demonstrate that terrestrial exoplanets can be detected below the Rayleigh limit, a regime previously inaccessible to direct imaging.

Hollow-core fiber (HCF) is often presented as a modestly improved transmission medium that can be inserted into networks originally designed for solid-core silica. We argue instead that recent progress -- most notably the reported attenuation below 0.1 dBkm$^{-1}$, together with a broad low-loss window, reduced propagation delay, and extremely low optical nonlinearity -- makes it timely to reconsider which long-standing design conventions are fundamental to optical communication and which are specific to silica fiber. By reviewing implications at the physical-layer, transceiver, and network-architecture levels, we suggest that the most durable benefits of HCF may arise not from its use as a drop-in replacement, but from cross-layer co-design. We also outline the studies and experimental demonstrations needed to determine where such advantages are genuinely achievable.

Purpose: To compare detuning performance and evaluate the power requirements of optical detuning methods, and to demonstrate the feasibility of an optically detuned four-channel receive array. Methods: Four optical detuning methods were compared in simulations, bench tests, and phantom measurements at 3T against conventional galvanic detuning. Passive detuning was also tested as an additional wireless detuning option. Optical power requirements for the detuning networks were investigated, and a flexible, optically detuned 4-channel shielded-loop resonator (SLR) array was constructed and tested in vivo. Results: A photodiode-PIN diode combination exhibited the highest unloaded Q (68.6) and Q ratio (1.9), with detuning performance and signal-to-noise ratio comparable to that of galvanic detuning at an optical power of 10 mW. Using this detuning strategy, in vivo images of the knee and brain were successfully acquired with a 4-channel flexible array. Conclusion: Optical detuning is a practical alternative to conventional galvanic detuning in flexible SLR arrays. With advances in optical signal and power transmission, optimizing optical detuning while meeting manageable power requirements is an important step toward fully optical receive-coil arrays. This study provides a baseline for the total optical power required for active detuning in such optical coil systems.

Biological and physical systems ranging from Fermi surfaces and skeletal structures to reaction--diffusion patterns and cosmological models may be viewed as binary mixtures in which a smooth interface separates two complementary phases. While the interface is often directly observable, the topology of one of the phases may remain hidden. To study such systems, we introduce tubular tilings, a geometric framework for discretizing binary mixtures on smooth manifolds of arbitrary dimension and topology. We prove that tubular tilings satisfy global Euler balance laws relating the topology of the ambient manifold, the discretized phases, and their interfaces. These balance laws provide a practical inference principle: topological information about a hidden phase can be recovered from the observable phase and the geometry of the separating interface. We further show that, in dimensions $d>2$, tubular tilings form a subclass of soft tilings, the recently discovered class of corner-free tessellations. Applications to Fermi surfaces and cosmological shell decompositions illustrate how the theory can be used to extract otherwise inaccessible topological information about complex geometric structures.

AstroForge seeks to mine platinum group metals (PGM) from asteroids. Asteroid reserves appear to be unlimited, and at current market price the gross margin of asteroid mining would be very high. It is natural to ask: when AstroForge successfully demonstrates economic space mining of PGM, will they cause the PGM market to collapse? We answer the question with a non-steady system dynamics model of the PGM market. We find that the market price for PGM will eventually drop towards the much lower cost of asteroid mining, but only after the entire supply has shifted off-world. In the meanwhile, huge fortunes will be made. And everybody on Earth will benefit from new applications of lower-price PGM.

All published models of ultraviolet (UV) inactivation of airborne viruses in saliva droplets have neglected UV light scattering. To the best of our knowledge, this work presents the first Computational Fluid Dynamics-Discrete Dipole Approximation (CFD-DDA) platform for investigating the physical mechanisms governing UV disinfection of virus-laden airborne saliva droplets. The DDA solver predicts UV light scattering by both spherical and irregularly shaped saliva droplets, while the CFD solver predicts droplet evaporation and transport in airflow. By coupling the DDA and CFD solvers, we demonstrate that infected saliva droplets, whether spherical or irregularly shaped due to evaporation, experience highly non-uniform UV light scattering that significantly affects virus inactivation and cannot be neglected. This phenomenon has not previously been investigated within a fully three-dimensional framework. The coupled Euler-Lagrange CFD-DDA model further quantifies the effects of (i) the initial droplet size distribution and concentration, (ii) airflow rate, and (iii) droplet interactions with the surrounding airflow and bounding walls on the total number of surviving coronavirus copies $N_s$, assuming a virion diameter of 100 nm, an air temperature of 21 $^{\circ}$C, and a relative humidity of 65%. Based on the DDA results, a new virus inactivation model, referred to as the Dbouk-Yurkin law, is proposed. This model extends the classical Chick-Watson law by explicitly accounting for UV light scattering in both spherical and non-spherical airborne saliva droplets. The proposed three-dimensional CFD-DDA platform provides a powerful framework for improving the understanding of UV-based airborne virus disinfection and for optimizing the design and performance of UV air purification systems.

The paper is devoted to the study of the connection between the numerical dispersion arising in FDTD modeling of electromagnetic signal propagation in nondispersive homogeneous media optically different from vacuum and the Courant number in the 2D case. The main results are formulated in the form of four statements, as well as a number of corollaries and remarks that determine the nature of the numerical dispersion, the optimal value of the Courant number and the limitations of the method. It is proved that the optimal choice of the Courant number eliminates the numerical dispersion and extends the capabilities of the developed numerical algorithm to media, which refractive index lesser than refractive index of vacuum, as well as media with negative refraction.

Surface characterization is essential for revealing the structural, chemical, and physical properties of materials. Yet high-resolution methods such as atomic force microscopy (AFM) require complex equipment and delicate skillsets, making them particularly challenging for applications involving soft and biological materials in liquids. Here, we propose and validate an innovative motion-enabled sensing scheme that uses untethered micromotors as robotic probes to interact with surfaces or objects, with their motion responses serving as sensing signals for characterization. This sensing concept is validated by employing 3D electrokinetic tweezers, which control micro/nanoparticles with up to 20 nm positioning precision in solution, to drive Au microsphere motors along designed scanning paths. When the motors encounter local chemical or structural variations, their locomotion changes, allowing motion itself to serve for the detection. This effort enables untethered motors, for the first time, to detect biomolecular patterns and lithographically defined microridge arrays in liquid environments. The work establishes robotic locomotion as a new sensing modality, opening a wireless, solution-compatible, and low-cost technical pathway to standard surface sensing.

A method is proposed for solving the two-dimensional problem of electromagnetic wave scattering by a cluster of an arbitrary number of parallel, infinitely long, homogeneous, non-overlapping right circular cylinders. The cylinders may have arbitrary radii and complex permittivities, and their axes, while remaining parallel, may occupy arbitrary positions in the transverse plane. The solution is constructed using an analytical expansion of the electromagnetic field in cylindrical harmonics. Multiple scattering is taken into account by Graf's addition theorem, which leads to a system of linear equations for the expansion coefficients. This system is solved numerically with condition number monitoring and, when necessary, extended-precision arithmetic, followed by a multistage verification of convergence. The method provides numerically verified solutions with controlled accuracy over a wide range of parameters, including densely packed subwavelength configurations. As an example, scattering of a normally incident, linearly polarized monochromatic plane wave by a subwavelength cluster of three identical aluminum nanocylinders (nanowires) is studied. The scattering, absorption, and extinction cross sections, as well as the scattering indicatrix, are computed and analyzed. Streamlines of the Poynting vector field are constructed, demonstrating redistribution of the energy flux between the cylinders of the cluster and the formation of localized regions of field enhancement near their surfaces.

Time-varying metamaterials involve a rapid temporal modulation of the permittivity, often at frequencies comparable to the oscillation frequency of light. However, pronounced physical effects at low modulation amplitudes are observed only when resonances sustained in the metamaterials are utilized. This requires an additional spatial structuring. Here, we demonstrate the first exploitation of bound states in the continuum (BICs) in such spatio-temporal metamaterials consisting of a multilayered metasurface. Leveraging Fabry-Perot BICs in a metasurface-based cavity, we realize polarization-insensitive scattering anomalies such as exceptional points (EPs), coherent perfect absorption (CPA), and lasing at extremely small modulation amplitudes. In a second example, by utilizing symmetry-protected BICs and breaking time-reversal symmetry of a multilayered metasurface, we obtain strong nonreciprocal behavior. Harnessing nonreciprocity, we further demonstrate a device capable of one-way monochromatic light transmission at perturbative modulation amplitudes. Our contribution establishes BIC-enabled spatio-temporal metamaterials as a scalable platform for low-power, tunable light-matter interactions, opening new pathways toward practical nonreciprocal photonic devices, dynamic wave control, and on-chip optical signal processing.

Ramkrishna, Kompala, and Tsao proposed the cybernetic model of microbial growth, in which cells allocate enzyme synthesis resources according to a matching rule that mimics rational decision-making. The matching rule was later shown to be optimal under general assumptions about the underlying return-on-investment structure, yet the specific objective the cell maximizes, and the constraints bounding that choice, were never written down as an explicit economic decision. Here we supply that missing decision, recasting cybernetic enzyme-synthesis control as a consumer choice problem from microeconomic theory: the cell allocates a limited proteome budget among competing catabolic enzymes as a linear program (LP), maximizing a linear growth utility subject to a linear proteome budget constraint. Because the utility is linear, the LP's solution is geometric: whenever the iso-utility line's slope differs from the budget constraint's, the optimum is a corner, and the entire proteome budget is allocated to the enzyme for the single most profitable substrate. Corner solutions correspond to diauxic growth, and sequential substrate consumption follows from the choice of corner rather than a distinct regulatory mechanism. Only when the two slopes coincide does the optimum spread across the entire budget line instead of concentrating at a single corner; this degenerate case underlies simultaneous substrate use. Using only parameters estimated independently from single-substrate experiments, the LP-derived cybernetic variables reproduced the diauxic and triauxic batch growth of Klebsiella oxytoca on glucose-xylose and glucose-xylose-lactose mixtures, achieving a fit comparable to the classical matching law. Thus, sequential substrate use is the generic outcome of growth-maximizing specialization under perfect substitutability, and co-utilization is the degenerate case of equal profitability.

Harmful pesticide effects exceeding specific protection goals (SPG) may go undetected in underpowered experimental designs. Regulatory honeybee field studies have consistently failed to reach the statistical power required under European Food Safety Authority (EFSA) guidance, which may have caused approval of high-risk substances. Therefore, EFSA advised a shift from testing the null hypothesis of 'no effect' to equivalence testing. Under this approach, a pesticide is classified as 'low risk' if the null hypothesis that its effect exceeds the SPG can be rejected. For honeybees, the recommended SPG is a colony size reduction below 10%. Critics have argued that this framework requires excessive site replication to demonstrate pesticide safety and proposed an alternative equivalence test defining treatment effects relative to the lower bound of the 90%-control-group confidence interval. Using simulations mimicking a regulatory honeybee field study, we show that although the two equivalence tests share the same trade-off between false 'low-risk' and false 'high-risk' classifications, only EFSA's original recommendation reliably identifies pesticides with effects > SPG at alpha = 0.2. Our results show that increasing site replication beyond the current practice is unavoidable for a reliable regulatory assessment. However, for pesticides with effect sizes of 5% or less, site requirements remain lower than those implied by the power requirement of the former EFSA guidance. Moreover, covariate adjustment through a model term or balanced colony allocation using anticlustering randomisation can reduce site requirements without losing power and thus save costs. Finally, we provide guidance and R functions for anticlustering randomisation and equivalence testing for pesticide risk assessment.

The Moran process with selection and recurrent mutation is a classical model in population genetics, yet how the placement of selection within the update rule shapes the stationary distribution has received little attention. We study a finite, well-mixed haploid population of constant size $n$ with $m$ labeled alleles, parent-independent mutation, and allele-specific fitnesses. Within this common framework we compare three Moran update kernels that differ only in the stage at which selection acts: during reproduction, when the offspring copies one of two sampled parents (Scheme~I); through fitness-biased mate choice, followed by neutral copying (Scheme~II); and at death, so that fitter individuals are less likely to be replaced (Scheme~III). Although all three favor fitter alleles, they define different Markov chains. For two alleles, each scheme reduces to a birth-death chain and admits an exact stationary law, but the three laws differ. For $m\ge 3$, the placement of selection becomes decisive: Schemes~I and~II are generally nonreversible when fitnesses are unequal, so no detailed-balance product form exists, whereas Scheme~III remains reversible for every $m$ and has a closed stationary distribution -- a Dirichlet-multinomial core modified by an explicit fitness factor. We further show that all three mechanisms can act simultaneously in the two-allele case without losing exact solvability, and we derive weak-selection expansions that make explicit how small fitness differences tilt the neutral beta-binomial and Dirichlet-multinomial benchmarks. Together, these results clarify when neutral stationary structure survives the introduction of selection and when multiallelic Moran dynamics become genuinely nonreversible

Phylogenetic trees are rooted trees with branch lengths that record genetic divergence or elapsed time, and quantifying differences between them is central to a wide range of evolutionary and epidemiological analyses. Graph-polynomial encodings of rooted trees provide an accurate, interpretable, and computationally efficient way to compare tree shapes, but existing polynomial encodings must be paired with auxiliary structures to study rooted trees with branch lengths. We introduce a bivariate polynomial encoding that incorporates branch lengths directly into a recursive computation from the leaf vertices to the root vertex of a tree. We prove that, for rooted trees with branch lengths and no vertices of degree two, which include all standard phylogenetic trees, two trees have the same polynomial if and only if their underlying unlabeled trees are isomorphic and the branch lengths of corresponding edges are equal. We apply the polynomial encoding to three published HIV-1 phylogenies sampled in different epidemiological settings and show that it accurately separates the three datasets based on their tree topologies and branch lengths, outperforming previous polynomial-based approaches for analyzing rooted trees with branch lengths.

Microbial keratitis requires rapid pathogen identification to guide treatment, but culture- and PCR-based diagnostics are slow and resource-intensive. We developed a triple-phase multimodal framework for bacterial-versus-fungal keratitis classification using slit-lamp photographs acquired under blue-light, sclerotic-scatter, and white-light illumination, together with clinical metadata. The model combines cross-modality contrastive learning, modality-specific fine-tuning, and feature-level multimodal ensemble learning for patient-level prediction. We evaluated the framework on a multicenter dataset of 1,645 patients and 17,158 images from India and the United States. The model achieved 85.84% accuracy, 84.46% average F1-score, and 0.885 AUC. Site-specific evaluation showed that pooled results were overly optimistic, whereas resampling- and balance-based re-evaluation provided a more realistic assessment of cross-site generalization. Under all settings, our framework remained the top-performing approach. The code is available at https://github.com/yqwang01/TPMKA and dataset access will be provided subject to University of Michigan data-sharing clearance.

Recent advances in multi-band wireless communication systems have driven the increasing need for dual-band bandpass filters. These types of filters are capable of isolating a small section of the frequency spectrum within a broader spectrum. Over the years, coplanar waveguide, microstrip, slotline, stripline, and other planar transmission line technologies have been widely employed in the design of microwave circuits and systems. This work employs a Folded-Arms Square Open-Loop Resonator (FASOLR) microstrip planar structure, designed and simulated using PathWave Advanced Design System (ADS). A third order (three-pole) single bandpass filter is transformed into a sixth order (six-pole) dual-band bandpass filter. The proposed six-pole dual-band bandpass filter is centred at 2.3 GHz, with a fractional bandwidth of 7% and produces two passbands centred at approximately 2.2 GHz and 2.4 GHz. The design is implemented on a commercially available Rogers RT/Duroid 6010LM substrate with a dielectric constant of 10.7, a loss tangent of 0.0023, a substrate thickness of 1.27 mm, and 35 micro meter copper cladding on both sides. The overall filter component has a compact footprint of 30.56 mm by 20.56 mm. The design is validated through comparison of circuit-level and electromagnetic (EM) simulation results, which show good agreement. The EM simulation responses indicate a return loss better than 10 dB on the first band and better than 19 dB on the second, and an insertion loss better than 1.5 dB, demonstrating its suitability for WiMAX applications.

Virtualized radio access networks (vRAN) run the compute-intensive multiple-input multiple-output (MIMO) baseband as software on shared servers, which makes energy efficiency (EE) a primary design objective. Distributed MIMO vRAN consumes power across virtualized distributed unit (vDU) baseband, fronthaul transport, and per-radio-unit operation. We build a power model that resolves these three components. We then develop a framework that jointly selects modulation, transmission rank, and per-subcarrier power to maximize system EE. Exponential effective SNR mapping induces a convex per-subcarrier power constraint, which yields a convex power minimization problem with a closed-form waterfilling-like solution. We show that radio frequency-only models underestimate the spectral efficiency range where single-input multiple-output (SIMO) transmission saves power, and our power model extends this range by 24%. We further extend the framework to a traffic-aware setting with realistic user trajectories from the multi-agent transport simulator. We propose a traffic-aware strategy that switches each radio unit among MIMO, SIMO, and sleep modes based on demand. Simulation results over 3GPP NR compliant fading channels show that, after a one-time offline calibration, the framework predicts link performance without further link-level simulation. The proposed framework achieves higher average EE than a traffic-agnostic always-on MIMO baseline, while maintaining comparable throughput at peak hours.

5G New Radio (NR) Non-Terrestrial Networks (NTNs) extend cellular connectivity through Low Earth Orbit (LEO) and Medium Earth Orbit (MEO) satellite constellations while enabling the reuse of downlink NR Positioning Reference Signals (PRS) to provide Positioning, Navigation, and Timing (PNT) services alongside broadband communications. However, the large inter-satellite differential propagation delays inherent to NTN geometry may cause PRS transmissions from non-serving satellites to overlap with the serving-satellite data stream. This paper analyzes this coexistence by deriving a statistical model for the slant-range distribution over the visible spherical cap and extending it to dual-shell constellations through a mixture formulation, yielding a closed-form cumulative distribution function (CDF) of the differential delay. The model is validated using a 10-day orbit simulation representative of a dual-shell European NTN constellation. Detection limits of non-serving satellite PRS under interference from the serving-satellite data stream are characterized in terms of the effective carrier-to-noise density ratio. The impact of periodic PRS transmissions on the uncoded bit error rate (BER) is also evaluated for standardized NR Frequency Range 1 (FR1) and Frequency Range 2 (FR2) configurations. Monte Carlo simulations show that the probability of simultaneous multi-PRS overlap remains below a few percent, depending on PRS duration and repetition period, while PRS detection remains feasible despite data interference. When the PRS is received about 25 dB below the data signal, its impact on uncoded BER is negligible over a wide range of repetition periods, whereas BER degradation increases with PRS duty cycle. These results demonstrate that NR-PRS-based PNT can coexist with broadband downlink in NTN under appropriate PRS periodicity design.

Semantic and terahertz (THz)-band communications are algorithmic and spectral enablers of future wireless networks. This work investigates deep learning-based semantic communication (DeepSC) over THz channels. We show that DeepSC models trained solely under additive white Gaussian noise generalize well to the tested THz block- and fast-fading channels when receiver-side compensation is applied. To enable fully data-driven reception, we propose a lightweight neural detector that does not require channel state information (CSI). At 0.3 THz, DeepSC outperforms a throughput-matched traditional coded communication system baseline over 0-12 dB signal-to-noise ratio (SNR), achieving more than 50 percentage-point higher Bilingual Evaluation Understudy unigram (BLEU-1) score. The proposed pilot-free detector outperforms minimum mean square error (MMSE) equalization with both perfect and imperfect CSI and remains robust to frequency offsets up to 50 MHz, highlighting the resilience of semantic communication to THz channel impairments.

Sparse vector transmission (SVT) has emerged as a promising technique for ultra-reliable low-latency short-packet communications. However, existing SVT schemes typically assume negligible phase noise (PN), an assumption that rarely holds in practical wireless systems. In this paper, a dual-mapping sparse vector coding (DM-SVC) scheme is proposed for short-packet communications subject to PN. In DM-SVC, pilot symbols are mapped onto multiple non-zero blocks and data symbols onto isolated non-zero elements within a single sparse vector, thereby enabling pilot-data separation through distinct sparsity patterns rather than explicit resource partitioning. Moreover, the indices of pilot blocks convey additional information bits, further improving spectral efficiency. A basis expansion model is adopted to represent the PN process, substantially reducing the number of parameters to be estimated. Furthermore, an iterative joint PN estimation and data decoding algorithm is developed, where pilot block indices are first detected exploiting block-sparse priors, after which PN estimation and data decoding proceed iteratively. Simulation results show that DM-SVC could achieve block error rate performance close to that of perfect PN compensation, while offering improved spectral efficiency and reduced codebook storage overhead compared to state-of-the-art SVT schemes.

Continuous fluid antenna systems (CFASs) represent an upper bound on the spatial diversity performance of fluid antenna systems (FASs), achieved when antennas may be positioned anywhere within a defined spatial region. This article examines the fundamental relationships governing CFAS performance. The focus is on the probability that the signal-to-noise ratio (SNR) exceeds a prescribed high threshold, termed the high SNR probability (HSP). This is among the few FAS performance metrics that admit the derivation of closed-form expressions. Following a survey of recent analytical advances in FAS performance limits, a dimensional scaling law derived for the HSP of a single-user, single-antenna CFAS is examined. This law is then applied to the per-user high signal-to-interference-plus-noise ratio (SINR) probability of a two-antenna, two-user CFAS employing minimum mean-squared error (MMSE) combining. For both scenarios, performance gains are shown to increase consistently with both dimensionality and region size. Remarkably, the scaling law remains accurate in the two-user case, showing that, in both scenarios, the influence of additional dimensions is dominated by the CFAS size and considered threshold. Moreover, the per-user high SINR probability of the two-user system exceeds the single-user HSP, despite the addition of inter-user interference.

UAV-mounted reconfigurable intelligent surfaces (RIS) are a promising enabler for 6G networks, offering dynamic control of wireless propagation for coverage enhancement, integrated sensing and communication (ISAC), and localization. By exploiting UAV mobility, RIS can maintain favorable line-of-sight links, improving channel quality in dynamic environments. However, UAV positioning uncertainties introduce channel distortions that degrade RIS phase alignment and coherent combining. This work develops a GUM-based uncertainty propagation framework for UAV-mounted RIS channels, mapping UAV position uncertainty through the geometric Tx-RIS-Rx model into the complex cascaded channel. We derive a closed-form stochastic propagation model capturing nonlinear phase uncertainty effects and quantify their impact on channel coherence. The results show that phase uncertainty induces exponential coherence loss, dominating performance degradation. To characterize this transition, we introduce a performance-driven coherence threshold (PCT) that defines the boundary where incoherent combining results in a predetermined performance loss. Results based on analytical scaling laws and Monte Carlo simulations confirm the tightness of the PCT in accurately capturing the coherence transition. This validated threshold is then leveraged to derive optimal UAV-mounted RIS placement, revealing that realistic positioning conditions significantly deviate from the conventional RIS intuition, which typically favors placement close to either the transmitter or receiver.

High-frequency communications strongly depend on the line-of-sight (LoS) path, and obstacle blockage can severely degrade the received signal power and achievable rate. Near-field Airy beams with curved trajectories can circumvent obstacles, offering a promising way to alleviate blockage. However, since an Airy beam carries most useful energy along a single curved trajectory, existing Airy beamforming methods are highly sensitive to estimation errors of transmitter-obstacle-receiver geometry. That is to say, even a small error in the estimated geometry may cause the mismatched Airy trajectory, leading to severe performance loss. To address this problem, we propose a multi-Airy beamforming scheme for blockage-robust near-field communications. Specifically, we first reveal and analyze the sensitivity mechanism of single-Airy beamforming. This mechanism motivates us to extend the single-Airy generation method to a coordinated multi-Airy generation method by deriving the phase offsets required to coherently combine multiple Airy beams at the target user. Based on this coordinated generation method, we partition the transmit array into multiple sub-arrays and configure a tailored Airy beam for each sub-array, so that the resulting Airy beams formed by multiple curved trajectories can be coherently combined at the target user. Simulation results verify the sensitivity of single-Airy beamforming and the robustness of multi-Airy beamforming under estimation errors of transmitter-obstacle-receiver geometry. Moreover, the proposed scheme achieves higher achievable rates than single-Airy beamforming in blocked scenarios without geometry estimation errors.

High-quality system measurement data is critical for power distribution system operation. As deep generative models (e.g., GAN, Diffusion, etc.) have been widely studied to solve the missing data restoration problem to enhance the data quality, their results may look "realistic" but not sufficiently "accurate" due to lacking physical guarantees. To address this limitation, a two-stage physics-guided framework, Flow-PIN, is proposed in this paper for system-wide multivariate profile inpainting. The first stage employs a conditional flow matching model, conditioned on topological and correlation graphs, to generate candidate values. A physical penalty is integrated into the loss function to constrain the generative vector field based on grid physical laws. The second stage introduces a topology-aware power-flow-guided refiner that utilizes Laplacian positional encoding to inject topology information into node embeddings. By coupling alternating current power flow equations with a differentiable correlation alignment mechanism, this refiner further corrects numerical deviations. Evaluations on an active distribution network dataset benchmark the proposed framework against ten representative baselines. The results show that Flow-PIN achieves high-fidelity profile inpainting across three dimensions: maximizing numerical accuracy, capturing temporal fluctuation, and preserving spatial topological correlations.

To achieve the desired coverage and capacity levels, future terahertz (THz) wireless systems are envisioned to utilize extremely large antenna arrays. At THz frequencies, the combination of short wavelengths and large array apertures often makes many of the conventional far-field assumptions invalid in practice. As a result, many UEs operate in the radiative near-field zone, where novel near-field beam synthesis methods become viable. This paper studies phase-only Bessel-like near-field beam configurations for downlink THz multiple-input multiple-output links under imperfect UE location knowledge. We first formulate a spectral efficiency maximization problem with respect to the "Bessel cone angle''. We then derive low-complexity closed-form approximations for the optimal Bessel beam configuration for: (i)deterministic UE location; (ii)Gaussian and (iii)uniform error in the UE location. Finally, through extensive simulations across multiple signal frequencies, UE locations, and array sizes, we show that our proposed simple closed-form approximations closely match (under 0.1% difference) the best performance achieved via exhaustive search, while simultaneously reducing the configuration complexity down to as low as O(1).

Fluid reconfigurable intelligent surfaces (FRIS) have recently emerged as a promising technology for enhancing wireless link reliability through spatial decorrelation. However, their performance analysis remains challenging due to the sum-product structure of the cascaded channel. This letter develops a rigorous analytical framework for FRIS-assisted wireless systems over arbitrarily correlated Nakagami-$m$ fading channels. Specifically, we introduce a physically consistent correlation model for Nakagami-$m$ fading and derive tractable statistical characterizations for the cascaded channel. These results lead to rigorous lower bounds for the outage probability (OP), with a simplified expression also obtained for the independent and identically distributed case. To the best of our knowledge, these are the first strict OP lower bounds reported for an FRIS-aided wireless system under arbitrarily correlated Nakagami-$m$ fading. CLT- and Gamma-based approximations are included as benchmark methods. Notably, the numerical results show that the proposed OP bound not only provides rigorous performance guarantees but also yields a noticeably tighter OP characterization than the CLT approximation in the high-SNR regime.

This paper proposes a novel cyclic-prefixed single-carrier transmission scheme that amalgamate a reconfigurable intelligent surface (RIS) with spatial modulation in the frequency-selective fading channel. The discrete-input continuous-output memoryless channel's~(DCMC) capacities of the proposed schemes are formulated, while their gradients with respect to the RIS phase shifts are derived in the closed form. Then, the gradients are used for iteratively and efficiently optimizing the proposed RIS-aided schemes with the aid of the gradient-ascent algorithm. Our performance results demonstrate that the proposed iterative algorithm enhances the DCMC capacity of the schemes, while outperforming the conventional RIS benchmarks. Moreover, the convergence behavior and sensitivity in the proposed iterative algorithm are analyzed.

This paper provides the first near-optimal lower bounds for one-bit compressed sensing of approximately sparse signals lying in a scaled $\ell_1$ ball, which is a commonly adopted relaxation of the exactly $k$-sparse assumption. In prior works, the best known upper bounds on uniform Euclidean error are of order $\widetilde{O}((k/m)^{1/3})$, where $m$ is the number of measurements. Under sub-Gaussian matrices, we establish nearly matching lower bounds for both the canonical one-bit compressed sensing model and the uniformly dithered model. Our argument is to first embed a small Euclidean ball into the signal set, which is straightforward for the dithered model but relies on a lifting map for the canonical model, and then construct two signals in this small ball that are separated in Euclidean distance by at least $(k/m)^{1/3}$ (up to logarithmic factor) but are indistinguishable from the binary measurements. Moreover, our argument extends to approximately sparse signals that live in a properly scaled $\ell_q$ ball $(q\in [0,1])$, yielding a lower bound $\widetildeΩ((k/m)^{\frac{2-q}{2+q}})$ that smoothly bridges the cases of exact sparsity ($q=0$) and $\ell_1$ sparsity ($q=1$). Finally, we discuss the extensions of our lower bounds to sub-Weibull matrices, adversarial bit flipping, matrix recovery, and characterize the transition to the non-sparse case.

We specialize Agarwal's multi-level collective spontaneous-emission formalism to the four-level case by formulating it in the fully symmetric \SU(4) representation of $N$ identical atoms. In the irreducible representation $(N,0,0)$, the occupation-number basis forms a tetrahedral weight lattice on which the six embedded $\mathfrak{su}(2)$ transition subalgebras act as ladder operators. From these algebraic factors we obtain a compact Pauli-type population-rate equation and a closed-form expression for the total emitted intensity that apply to any combination of open dipole channels. The formalism is then specialized to the seven dipole-allowed four-level topologies -- tripod, inverted tripod, Y, inverted Y, double-$Λ$, closed cascade, and diamond -- and the resulting rate equations are solved numerically for atom numbers up to $N=50$. In every case the emitted intensity develops a delayed cooperative burst whose peak height obeys a power law $I_{\mathrm{peak}}=aN^{p}$ with topology-dependent parameters $(a,p)$; the fitted exponents lie in the range $1.81\lesssim p\lesssim 1.92$, indicating a superlinear. The \SU(4) tetrahedral flow and the seven configuration-dependent transients together provide a unified geometric picture of multi-channel collective dissipation in four-level atomic ensembles.

The Baker--Campbell--Hausdorff (BCH) formula plays a critical role in many branches of mathematics and physics. It expresses the logarithm of the product of exponentials of non-commuting operators as an infinite series of nested commutators of the operators involved. The Zassenhaus formula is the dual of the BCH formula: the exponential of a sum of operators is written as an infinite product of exponentials involving the operators and their commutators. In practical computations, however, one typically has to truncate the expansions, and so understanding the error committed by the resulting approximations and eventually providing suitable bounds for this error is of paramount interest. In this work we present a general strategy to derive rigorous error bounds and explicit error constants for the BCH and Zassenhaus formulas when the operators involved are skew-adjoint, as is the case for quantum evolution problems.

The spectral condition number is a widely adopted measure of worst-case cost for quantum linear system solvers. Yet it can significantly overestimate the actual runtime for a typical problem instance. We present two quantum algorithms that produce the normalized solution $|x\rangle$ of linear system $Ax=| b \rangle$ to accuracy $ε$ with complexity independent of the condition number $κ=\lVert A^{-1}\rVert$. We focus on the standard input model where $A$ is accessed through a block encoding and $| b \rangle$ is prepared by a unitary. But we also introduce an affine dilation model that encodes $A$ and $| b \rangle$ jointly, allowing further refinements of the query complexity. Our truncation-based solver makes an optimal number of queries to $| b \rangle$ and $\operatorname{\mathbf{O}}\left(κ_{\mathrm{eff}}\operatorname{polylog}\left(\frac{κ_{\mathrm{eff}}}ε\right)\right)$ queries to $A$. We prove a family of upper bounds on the effective condition number, including $κ_{\mathrm{eff}}\leq\frac{\lVert(A^\dagger A)^{-t/2}|x\rangle\rVert^{1/t}}{ε^{1/t}}$ for positive even integer $t$ and $κ_{\mathrm{eff}}\leq\frac{\lVert A^{-1\dagger}(A^\dagger A)^{-(t-1)/2}|x\rangle\rVert^{1/t}}{ε^{1/t}}$ for positive odd $t$, overcoming the $κ$-barrier. Our filtering-based solver is extremely simple with a favorable runtime prefactor. In particular, the solver has query complexity $6\frac{\lVert A^{-1\dagger}|x\rangle\rVert}ε\ln\left(\frac{1}ε\right)$ to leading order when the solution norm is known. We then present a similarly simple solution norm estimator with the same asymptotic cost up to logarithmic factors. Our quantum linear system solvers thus substantially improve a recent algorithm of Li, enabling faster quantum linear system solving beyond the condition number.

We address the circular motion Unruh effect in 2+1 spacetime dimensions, as probed by a pointlike Unruh-DeWitt detector coupled to a massless scalar field. The effective temperature due to circular acceleration, operationally defined in terms of the detector's excitation and de-excitation probabilities, is known to be much smaller than the linear acceleration Unruh temperature when the detector's energy gap is small and the interaction lasts for a long time. It was shown by Parry et al. [Class. Quant. Grav. 42, 245012 (2025), arXiv:2508.19987] that a temperature of the order of the linear acceleration Unruh temperature can nevertheless be recovered in a simultaneous long-time-small-gap double limit, using suitable classes of detector-field couplings described by asymptotically scaled switching families (ASSFs). The successful constructions presented there required the coupling to change sign. Here we prove, within the ASSF framework and under certain technical boundedness and localisation conditions, that sign changes in the detector-field coupling are in fact *necessary* for obtaining a nonvanishing limiting effective temperature. Our analysis is motivated by current work towards an experimental verification of the circular motion Unruh effect in analogue spacetime experiments.

We introduce a family of simple $α$-attractor models that can interpolate between exponential and polynomial cosmological attractors. By varying the interpolation parameter $μ$ in these models, one can scan a wide range of values of the spectral index $n_{s}$ matching any combination of CMB and DESI data.

We investigate the phenomenology of black bounce spacetimes through a combined analytical and numerical study of relativistic observables associated with both time-like and null geodesics. Black bounce geometries provide a continuous interpolation between Schwarzschild black holes and traversable wormholes by introducing a regularization bounce parameter $α$, which removes the central singularity by replacing it with a finite-radius throat. Using the Hamilton--Jacobi and Lagrangian formalisms, we derive weak-field analytical expressions for the periastron advance of massive particles and the deflection angle of light, highlighting the leading corrections induced by the bounce parameter. These results are systematically compared with numerical integrations performed with the PyGRO code. We show that increasing $α$ enhances both the periastron precession and the light deflection with respect to the Schwarzschild case. Additionally, we analyze the relativistic redshift of bound orbits and identify cumulative temporal phase shifts associated with the modified geometry. For null geodesics, we also determine numerically the critical impact parameter separating scattered and captured (or transmitted through the throat) photon trajectories. Although this quantity remains identical to the Schwarzschild value in the regular black hole branch, it increases significantly in the traversable wormhole regime, providing a clear observational signature of the throat structure. Our results demonstrate that black bounce spacetimes can produce measurable deviations from standard black hole predictions, opening the possibility of constraining these geometries with current and future observations of strong gravitational fields.

We present the first numerical-relativity simulations of subsolar-mass binary strange star (SS) mergers and compare with binary neutron star (NS) mergers across equations of state, masses, and mass ratios. The self-bound nature of SSs makes them less deformed during the inspiral and keeps a sharp surface up to contact, driving strong shock heating and a large radial bounce that are far weaker in the NS. The more compact SS thus reaches a higher gravitational-wave cutoff frequency $f_\mathrm{cut}$ before contact but a lower post-merger peak frequency $f_2$. Within each class these frequencies follow quasi-universal relations with the tidal deformability, and their ratio $f_2/f_\mathrm{cut}$ cleanly separates the two classes. Both classes can eject $\sim10^{-2}\,M_\odot$ of material, neutron-rich for the NS and decompressed quark matter for the SS, a potential source of an electromagnetic counterpart whose observation could test the SS and NS hypotheses for subsolar-mass events.

The negatively charged boron vacancy center in hexagonal boron nitride is a premier candidate for quantum sensing, yet its performance is critically limited by longitudinal spin-lattice relaxation time ($T_1$). A microscopic understanding of spin relaxation in the high magnetic field regime remains elusive, as the relevant Zeeman transitions lie far below the optical phonon energies typically invoked to describe the relaxation process. Here, we apply an \textit{ab initio} acoustic mode spin-phonon relaxation theory to this problem and quantitatively reproduce the experimental magnetic field and temperature dependence of $T_1$ without empirical fitting parameters. We demonstrate that the relaxation dynamics are driven by a direct one-phonon emission and absorption process resonant with the Zeeman splitting. Furthermore, we identify the out-of-plane flexural phonon branch which is unique to two-dimensional hosts, as the primary source of decoherence, creating a distinct low-energy spectral function that facilitates spin relaxation. Our results provide a microscopic interpretation of the experimentally observed non-monotonic field and temperature dependence in two-dimensional quantum defect centers.

Multi-stage cryogenic readout chains with a wide bandwidth and added noise within a few quanta of the quantum limit are frequently constructed using traveling-wave parametric amplifiers (TWPAs) as the first stage, and a semiconductor amplifier as the second stage. Unfortunately for highly-scaled superconducting detector arrays, or quantum information systems, and space-based observatories, the power dissipation of the semiconductor amplifier becomes problematic from the perspective of available cryogenic cooling power at \mbox{3~K to 4~K}. Here we demonstrate a readout chain based on a two-stage kinetic inductance TWPA (KTWPA). This quantum-amplifier-based-readout-chain (QARC) provides sufficient gain that a cryogenic semiconductor follow-on amplifier can be eliminated without degradation of the system noise. In this way, the QARC dissipates approximately three orders of magnitude less power than readout chains containing semiconductor amplifiers while adding noise of less than 2~quanta over a 1~GHz bandwidth. In addition, by leveraging the high power handling of kinetic inductance technology, the QARC maintains an input compression point of -93~dBm, which exceeds that of many contemporary Josephson-junction-based parametric amplifiers.

Quantum error correction compatible with continuous symmetries is a fundamental problem in quantum information and a possible route to robust analog quantum simulation. Because the Eastin-Knill theorem forbids exact codes with continuous transversal symmetries, we construct explicit $SU(d)$-covariant approximate codes that exploit permutation symmetry to spread logical information uniformly across all physical subsystems. For one-, two-, and three-qudit erasures at known locations, we prove worst-case purified-distance scaling $Θ(1/N)$, matching approximate Eastin-Knill lower bounds up to constants, and we extend the reduced-state analysis to general flagged local noise. For single-qudit erasure, we construct an explicit near-optimal decoder from the Petz recovery map. We then use these codes as building blocks for encoded analog dynamics. Symmetry-preserving Hamiltonians generate block-structured dynamical Lie algebras implementable transversally, while controlled symmetry-breaking terms serve as non-transversal resources for universal dynamics. These results provide explicit non-Abelian covariant codes and a framework for robust analog quantum simulation.

The distribution of entangled photon pairs over standard optical fiber is a fundamental requirement for the realization of the quantum internet. However, real-world deployment is severely bottlenecked by the interplay of amplitude damping (photon loss) and phase noise (birefringence). In this paper, we numerically investigate the degradation of dual-rail polarization entanglement in telecom C-band fiber links. We demonstrate a critical disparity between the physical survival of quantum correlations and their practical utility in standard communication protocols. By evaluating the unconditional logarithmic negativity against the post-selected teleportation fidelity, we identify a distinct ``operational collapse region'' -- a distance window where the channel retains true quantum entanglement, yet standard coincidence-based detection architectures fail to provide any advantage over classical strategies. Furthermore, we reveal that the width of this inaccessible region exhibits a non-monotonic dependence on the phase noise rate, implying that simply minimizing fiber dephasing does not necessarily optimize the operational efficiency of the network. These findings provide vital guidelines for the design of practical quantum communication links.

We study the open quantum dynamics of the adiabatic curvature perturbation interacting with a massive entropic scalar environment during an inflationary scenario featuring a transient Ultra-Slow-Roll phase. Working within a Gaussian two-field effective Lagrangian, we employ the exact Transport Equations Method to track the full non-unitary, non-Markovian evolution of the system's covariance matrix across the SR-USR-SR transition. We find that the efficiency of decoherence is sensitive to the background kinematics at horizon crossing. Most importantly, the interaction with the environment leaves distinct observable imprints on the primordial scalar power spectrum: the characteristic interference dip preceding the USR-driven enhancement can be partially or completely erased, the growth slope modified, and oscillatory features induced near the peak. Propagated to second order, these distortions further imprint on the stochastic background of Scalar-Induced Gravitational Waves, breaking single-field predictions and yielding unique spectral signatures potentially accessible to LISA. Our results demonstrate that the quantum environment is not a passive spectator during inflation, but an active agent whose imprint on the primordial universe may be within reach of the next generation of cosmological observations.

Donor spin systems host a native Hilbert space whose dimension exceeds that of a qubit, meaning they can be used as qudits. Here we study a \ce{Si{:}P} donor spin system through leakage-aware randomized benchmarking (RB) of native ququart $\mathcal{C}_4$ and encoded two-qubit $\mathcal{C}_2^{\otimes 2}$ Clifford groups. We implement adiabatic ramps to operate electron dipole spin resonance (EDSR) pulses at the ionization point, where the electron is shared halfway between the donor and the interface, and to operate electron spin resonance (ESR) pulses near the interface, motivated by the sensitivity of the effective magnetic field to charge noise at the ionization point. By placing the electron near the ionization point only during EDSR control and using sufficiently long displacement ramp durations, leakage outside the computational basis is strongly suppressed, which is crucial for optimized qudit control. We find in our analysis based on leakage RB that $\mathcal{C}_4$ consistently achieves $\sim 40$--$50\%$ lower (lower-bound) error rates $\varepsilon^{\mathrm{LB}}_\mathrm{PT}$ with respect to $\mathcal{C}_2^{\otimes 2}$, due to its reduced circuit complexity. These results indicate that donor spin qudits benefit from genuine qudit operation as opposed to imposed encoded qubit operation.

We provide a thorough study of the properties of the Boulware vacuum in the spacetime of a spherical, static thin shell with a Minkowski interior. To this end, we calculate the renormalized vacuum polarization and stress-energy tensor of massless scalar fields via the extended-coordinate prescription, paying particular attention to their scaling as the shell approaches the black hole limit. Near the surface of the thin shell, we obtain the expected leading-order singular behavior of both quantities via two independent methods: a high-frequency approximation for the modes, and a weak-field approximation. At the center of the shell we find non-local, Casimir-like contributions that remain finite in the black hole limit, and whose backreaction effects we compute via the semiclassical Einstein equations. Away from these regions amenable to analytic treatment, we obtain numerical results for a wide range of shell compactnesses and field couplings. In the black hole limit, we show that the vacuum polarization and renormalized stress-energy tensor outside the shell quickly approach the ones generated by a Schwarzschild black hole, suggesting a possible universality in the vacuum outside highly compact horizonless objects. This work addresses the conceptual and technical aspects necessary for computing renormalized expectation values in matter configurations, laying the foundations for future explorations on the subject.

A fundamental question in PAC learning is determining the number of labeled examples required to learn a concept class to a desired accuracy and confidence. In classical learning theory, this quantity is characterized by the VC-dimension, while several quantum generalizations have established analogous results when examples are provided in quantum superposition. In this work, we study a distinct quantum PAC-learning model in which concepts are functions acting on quantum states. We demonstrate that the VC-dimension, although still relevant, fails to fully capture the sample complexity of this model. To further characterize this setting, we develop a new lower bound on the required number of samples and establish an upper bound when the states in the domain are linearly independent. Remarkably, this upper bound has a form similar to the classical PAC-learning bound. We further examine a setting in which the learner receives more informative data and show that the limitations of the VC-dimension persist in this extended model.

Quantum thermometry based on single-qubit sensor configurations enables the precise estimation of the temperature of a cosmological Anti-de Sitter (AdS) spacetime. In this work, we characterize the achievable estimation accuracy using the Quantum Fisher Information (QFI) and the associated quantum signal-to-noise ratio. For the first time, we introduce an ancillary Unruh-DeWitt detector between the sensor and the thermal bath, enhancing thermometric sensitivity by channeling temperature-dependent information into the probe qubit's coherence. We examine how detector acceleration in AdS space and the choice of boundary conditions modify the probe's thermal sensitivity. Despite the differing geometries, a unified phenomenology emerges: we characterize the scaling of the QFI with respect to temperature, detector energy gap, spacetime curvature, and interaction time. Finally, we identify optimal state preparation and measurement strategies that maximize the QFI, thereby establishing the fundamental limits of precision for non-Markovian sensing in curved spacetime.

The Born rule for computing probabilities of the outcomes of measurements is an indispensable ingredient of quantum mechanics. The standard textbook description of this rule gives the impression that it implies the unitarity of time evolution. This view relies on the argument that unless the dynamics is unitary, the probabilities of finding all possible outcomes of a measurement do not add up to 1, i.e., the total probability is not conserved. We show that this argument is flawed, and that the general expression for the Born rule ensures the conservation of total probabilities even when the dynamics of a quantum system is not unitary. This applies to the dynamics of ensembles of quantum systems in both pure and mixed states. We discuss the status of the local conservation of probabilities and the arguments against the plausibility of non-unitary time evolutions that are based on the identification of the Hamiltonian operator with the energy observable.

Designing fault-tolerant quantum circuits that are both algorithmically correct and hardware compatible remains a major bottleneck in the transition to scalable quantum computing. We introduce RubriQ, a scalable framework that formulates circuit synthesis as a large language model (LLM) code-generation task, optimized via group relative policy optimization (GRPO). Unlike conventional black-box neural critics, RubriQ employs a domain-grounded programmatic rubric as the reinforcement learning reward function, evaluating circuits for T-gate reduction, hardware topology compliance, and unitary fidelity. To support high-throughput training, RubriQ integrates GPU-accelerated CUDA-Q simulation directly into the reinforcement learning (RL) loop and is deployed on NERSC Perlmutter using DeepSpeed ZeRO2 across multinode NVIDIA A100 clusters. On benchmark tasks, RubriQ achieves a mean T-gate compression of 3.31x, significantly outperforming sparse-reward RL baselines (2.05x), converging 2-3x faster, and maintaining less than 1\% hardware-constraint violations. Validated on IBM and IonQ quantum processors, RubriQ establishes an automated, high-performance computing (HPC)-driven pipeline for generating hardware-ready, fault-tolerant quantum circuits at scale.

This study develops a hybrid quantum-classical framework for constrained vehicle routing problems, focusing on the pickup-and-delivery problem with time windows. Instead of casting the full routing problem as a stand-alone quantum optimization task, we embed shallow quantum samplers inside the repair phase of an Adaptive Large Neighbourhood Search (ALNS) heuristic. A Deep Q-Network controller decides whether each reduced repair subproblem should be handled by a classical repair heuristic or by a quantum sampler, using features that describe the local repair structure and predicted hardware reliability. IBM Heron experiments are used to calibrate an empirical noise-aware model for local quantum repair circuits. Across the tested instances, quantum repair is admissible in only about 16% of reduced repair states and is not superior on average. However, under selected matched repair budgets, quantum-enabled repair reduces the final gap relative to standard ALNS in 29 of 36 tested settings. These results suggest that near-term quantum sampling is most useful as a selective local repair mechanism rather than as a replacement for classical routing heuristics.

Accurately controlling entangling gates remains a major challenge for quantum technology applications with solid-state spin qubits. Here, we study a group-IV color-center with a strongly-coupled nuclear spin and approach the problem from a quantum control perspective. We show that there are three different types of entangling gates where the entanglement is mediated by the parallel hyperfine-coupling component, the orthogonal one or both. We derive the respective quantum speed limits (QSL) and show by means of dynamical decoupling, resonant driving of single- and double-quantum transitions, quantum optimal control and algebraic gate decomposition how these gates can be realized. We finally discuss the experimental applicability.

The performance of variational quantum algorithms depends in general on the structure of the parametrized quantum circuit, but the most common ansätze are typically based on local couplings. Motivated by the extended connectivity available with neutral atoms and trapped ions, we examine when structured long-range connectivity provides a useful resource, focusing on sparse power-of-two (PWR2) coupling graphs. Using dynamical Lie-algebra analysis, approximate unitary-design diagnostics, and finite-depth measures of expressibility and entanglement, we examine how these geometries enlarge the accessible operator space. This enlarged space alone is not sufficient to ensure trainability of the parameterized circuit for given target problems, and we explore performance across example problems with and without long-range coupling, identifying where sparse coupling graphs are or are not likely to provide an advantage. We also introduce a variational scheme that maps hierarchical long-range Hamiltonians to geometrically local ones that can be optimized with short-range circuits. Together, these results identify circuit geometry and qubit reconfigurability as task-dependent resources for variational algorithms, relevant to ongoing developments in quantum hardware with long-range connectivity.

Non-Markovian behavior in quantum systems is often studied in the context of bipartite systems consisting of a system of interest and an environment -- tracing over the environment results in non-Markovian behavior for the subsystem of interest. One may get a Markovian limit in certain regimes, which is studied using the Lindblad master equation, and corrections to this behavior can be obtained by techniques such as the Nakajima-Zwanzig formalism. In this paper, we obtain an exact non-Markovian equation for the dynamics of a simple model system that consists of a direct sum rather than a tensor product of two pieces, namely, a discrete state and an infinite ladder. This system, called the Bixon-Jortner model, was first developed in the quantum chemistry literature but has been utilized by the quantum optics community as a model system with interesting behavior, including a Wigner-Weisskopf limit of exponential decay. We attack the time evolution problem of this system directly in time-domain, and start with an integrodifferential equation describing the time evolution of the discrete state. Using tools from mathematical physics, we transform this equation to a delay differential equation, which makes the non-Markovianity completely transparent, and then we solve the delay equation using an intuitive ansatz. This allows us to obtain the analytic form of the dynamics directly in time domain, and demonstrate decay and revival behaviors coming from the aforementioned delay differential equation. We believe the explicit form of the time-domain non-Markovian equation we obtain and the accessibility the solution techniques we use make our results a useful case study of non-Markovianity in quantum systems.

We propose a novel approach to the minimax estimation of high-order functionals from the perspective of quantum computing. Specifically, for any real number $α\gg 1$, we present two estimators, one for the classical functional $\mathrm{F}_α(P) = \sum_{i=1}^S p_i^α$ of a discrete distribution $P$ and the other for the quantum functional $\mathrm{F}_α(ρ) = \operatorname{tr}(ρ^α)$ of a mixed state $ρ$. These functionals have close connections with the Rényi entropy and the Tsallis entropy. We show that both estimators achieve the minimax optimal $L_2$ rate $α\mathsf{n}^{-1}$ in the range $α\lesssim \mathsf{n} \lesssim α^{3-o(1)}$, where the support size $S$ of $P$ or the dimension of $ρ$ can be much larger than the number of samples $\mathsf{n}$. As a result, both estimators achieve the \textit{optimal} sample complexity $\mathsf{n} \asymp α$, improving upon the prior best upper bounds $O(α^2)$ established by Jiao, Venkat, Han, and Weissman (IEEE Trans. Inf. Theory 2017) for classical functionals and Chen and Wang (COLT 2025) for quantum functionals. Our estimators are constructed under a unified framework using quantum primitives and run in linear time on a quantum computer. This work reveals an unexpected path from quantum computing to statistics, suggesting a conceptually new methodology for functional estimation. It adds to the growing list of quantum proofs for classical theorems.

We consider a multiplexed quantum repeater that distributes entanglement between two end nodes. Multiplexing is achieved through optical integration of many quantum chips. Each chip hosts an optically addressable communication qubit and a separate memory qubit. The communication qubit serves as an entanglement generation interface between different quantum chips, and the memory qubit can be used to store entanglement. The quantum chips on the repeater are interconnected using a reconfigurable router, which makes it possible to dynamically assign quantum chips for entanglement generation with either of the two end nodes in every end-to-end communication cycle. We propose a dynamic multiplexing policy in which after an entangled link has been established with one of the end nodes, all remaining quantum chips are assigned to the opposite end node. We compare this dynamic policy to a policy in which the assignment of quantum chips to end nodes is fixed. We consider a parameter regime where on average less than one entangled link is generated per end-to-end communication cycle, which is the relevant regime for near-term quantum networks. We show that in this regime, the dynamic multiplexing policy can lead to a significant improvement in fidelity over a fixed policy, while marginally improving the rate. Moreover, even though the dynamic multiplexing policy requires a deeper, and hence, more lossy, router than the fixed policy, it can still achieve higher secret key rates in the parameter regime studied. This makes dynamic multiplexing with a many-quantum-chip repeater especially relevant for the development of near-term quantum networks.

With a single contested exception, every NISQ-era flagship demonstration of "quantum advantage" has been classically reproduced, or closed by a simulability theorem, within eighteen months of its announcement. Six theoretical results from 2024 through April 2026 explain the pattern: the regions of circuit-space NISQ hardware can run with sufficient fidelity coincide with the regions classical algorithms compress efficiently, because the features that admit one (low effective depth, strong algebraic structure, geometric locality) are the features that admit the other. This reading dates the NISQ programme from its 2018 articulation as an interim retreat from the unmet conditions of the 1996 threshold theorems, characterises the eight years that followed as a closed loop in which the demonstrations the hardware could run were drawn from the only regions classical methods could already attack, and locates the exit from the loop where the threshold theorems originally located it: in fault tolerance. The empirical pattern could in principle break with a demonstration that escapes the current simulability results. After eight years and more than thirty advantage-class announcements, the burden of producing such a demonstration falls to the defenders of NISQ.

We investigated the influence of a plasma environment on the optical and perturbative properties of a black hole with a topological defect, characterized by the parameter \(k\). We first established a straightforward correspondence between the real part of the quasinormal-mode (QNM) frequencies in the eikonal limit and the black-hole shadow radius. We then demonstrated that the Lyapunov exponent associated with the photon sphere exhibits only a weak dependence on the plasma frequency, while it monotonically decreases as the topological-defect parameter \(k\) increases. Subsequently, we analyzed massive scalar-field perturbations by deriving the associated effective potential and computing the QNM spectrum using the third- and sixth-order WKB approximations for both homogeneous and radially inhomogeneous plasma configurations, including the singular isothermal sphere (SIS) and non-singular isothermal sphere (NSIS) density profiles. Our results show that the presence of plasma induces shifts in both the oscillation frequencies and the damping rates of the modes, and that larger values of \(k\) systematically suppress the real part of the QNM frequencies. Among the plasma models considered, the NSIS profile generally yields slightly higher oscillation frequencies than both the SIS and homogeneous cases. Finally, we derived the dynamical equations governing electromagnetic perturbations in a cold, unmagnetized plasma and demonstrated that the axial and polar sectors decouple. In the axial sector, the plasma frequency enters as an effective mass term, thereby permitting the existence of quasi-bound states only in the case of a homogeneous plasma and only when the plasma frequency lies below a critical threshold that depends on the topological-defect parameter \(k\) and the multipole index \(l\).

The recent detection of a nanohertz gravitational wave (GW) background by pulsar timing arrays (PTA) has sparked extensive discussions regarding its origin-whether it arises from astrophysical supermassive black hole binaries (SMBHBs) or from primordial GWs generated by various early universe processes. Previous studies suggest that a key discriminant between these two origins is the non-Gaussianity of the GW background prior to the detection of any individual source. In this Letter, we introduce a hierarchical Bayesian inference framework for inferring population properties of GW sources. This approach enables not only the measurement of evidence for different GW origins using PTA data but also the inference of population properties of astrophysical SMBHBs, by optimally leveraging non-Gaussian information in individual bright sources and in power spectrum fluctuations of the GW background.

The common envelope phase plays a critical role in binary system evolution. In this study, we investigate the mass and momentum accretion rates during the interaction between a stellar mass black hole and the envelope of a red supergiant using simplified two dimensional simulations. We explore various Mach numbers and density gradients, finding that our simulations align with previous Bondi Hoyle Lyttleton accretion analyses. We observe the formation of a shock cone in the downstream flow, bow shocks in specific configurations, and subsonic regions within shocked flows. The shock cone is dragged when significant pressure gradients are present in the common envelope, with additional dragging near the black hole for highly rotating cases. We provide analytical fits for mass and momentum accretion rates, as well as bremsstrahlung luminosity, as functions of black hole spin, density gradients, and Mach number, offering first insights into general relativistic hydrodynamics modelling of the secular evolution of the common envelope phase.

We investigate pattern formation in a finite chain of bosonic modes whose dynamics is governed by a Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) master equation. The model combines local parametric driving and nonlinear damping with nonlocal dissipative couplings between modes that work on different discrete spatial scales. In the classical limit, these mechanisms generate a reaction-diffusion-like dynamics, allowing the emergence of Turing-type instabilities. The key aspect of the analysis is the coexistence and competition of different unstable spatial modes. Depending on the range of parameters, the system may select different stationary nonuniform configurations, oscillatory wave-like states, or regimes in which multiple modes interact before a dominant pattern is established, thus providing a mechanism for pattern selection. We compare the deterministic bifurcation scenario, generated by a reaction-diffusion-like system derived from semiclassical drift dynamics, with the quantum dynamics, derived via the GKSL master equation, using phase-space methods and reduced Wigner functions. The results show how Turing instabilities, mode competition, and pattern selection can be extended to multimode open quantum systems, providing a bridge between nonlinear dynamical systems, dissipative quantum mechanics, and spatial self-organization.

Light-electron interactions generate synthetic momentum-space dynamics that can be used to engineer free electron quantum states. Here we develop coherent control protocols in which the optical phase acts as the controllable hopping phase of a Floquet-Bloch momentum lattice. Pontryagin optimization designs phase-only waveforms that prepare selected momentum populations and coherent few-sideband superpositions with programmable relative phases. In a complementary Bragg regime protocol, dynamical phase matching selectively couples neighboring sidebands and enables deterministic sequential state synthesis. Full wave-packet simulations based on the minimal-coupling Hamiltonian identify the tolerance window set by phase noise, detuning, and finite momentum spread. The two protocols expose a speed-selectivity tradeoff between ultrafast multilevel interference control and slower resonant engineering, establishing programmable free electron sidebands as a platform for ultrafast quantum state synthesis.

In this paper, we study the deflection of massless particles due to a rotating, axially symmetric Kerr-Bertotti-Robinson (KBR) black hole via; material medium approach. We explored the effect of spacetime geometry on the trajectory of light rays in the presence of a uniform magnetic field. To derive an analytical expression for the deflection of light rays due to the Kerr-Bertotti-Robinson black hole, the frame dragging effect and refractive index were also studied in greater detail. From the analysis it is evident that the magnetic field actively adds to the black hole's gravity, making the bending of light stronger and permanently changing the space far away from the black hole, preventing it to act as a normal flat vacuum. From thermodynamical investigation, it is clear that entropy monotonically decreases with magnetic field strength and rotation parameter; whereas the Hawking temperature increases with a uniform magnetic field but decreases with spin parameter.

Engineered loss, gain, and non-reciprocity can sharply amplify a sensor's response, and exceptional points and the skin effect are widely proposed as routes to more precise quantum sensors. Some predict an exponentially large quantum Fisher information, the quantity that sets the best achievable precision, yet whether this amplified response is genuine measurement information has remained unsettled. We prove that it is not. For any open sensor whose parameter is imprinted coherently by a Hamiltonian, with loss, gain, non-reciprocity, and readout held fixed, the quantum Fisher information is capped by a single quantity, the spread between the largest and smallest energy the generator can impart, independent of the non-Hermitian dynamics. Amplification enlarges the response but not the information. The reported exponential gains arise only when that spread is unbounded, and any finite spread, such as a photon-number cutoff, restores the cap. The same limit governs dissipative probe preparation and spectral singularities. For photonic sensors the cap takes an operator form in which the information of any probe, classical or entangled, is set by the photon's dwell time at the perturbation, and it contains the recently derived scattering limits as special cases. Non-Hermiticity thus shapes a sensor's dynamics but does not source its precision, and the generator's spread and the attainable dwell time benchmark future non-Hermitian sensing proposals.

Finding rotating black hole solutions in higher-curvature theories of gravity is a problem of fundamental importance. Virtually every approach to reconcile gravity with quantum mechanics predicts corrections to the Einstein-Hilbert action, yet no systematic solution-generating method exists for the stationary sector. We close this gap with {\sc Akribeia}, a novel hybrid framework that pairs physics-informed neural networks with a pseudo-spectral refinement step, yielding certified neural-field rotating black hole solutions -- continuous, globally defined functions, parametric in the coupling constants -- whose residuals against the field equations are verified to extreme precision. We apply the method to theories quadratic and cubic in the curvature and construct, for the first time, families of rotating black holes featuring multiple non-vanishing angular momenta, parametric in the new coupling constants. After validating against previously known five-dimensional spacetimes, we present new solutions in scenarios leading to a highly non-linear/non-perturbative coupled system of ordinary differential equations. Our method can be systematically adapted to other setups involving partial differential equations as well.

It was previously shown that $H_0$ influences gravitational wave propagation beyond simple redshift, by modifying the effective wavenumber. While earlier studies focused on Pulsar Timing Arrays, we analyze the observability of this effect with LISA. Modeling the LISA arm geometry, we derive the timing residual and identify a clear angular enhancement at frequencies above 100 mHz. Using the stationary phase approximation and numerical methods, we show that the optimal incidence angle depends on $Z_A H_0$, providing a direct, redshift-independent measurement of the Hubble parameter. We estimate that LISA could determine $H_0$ to within a few percent using this method, independent of standard siren approaches.

Solving linear differential equations is a fundamental task in scientific computing and an important primitive for quantum computing. A recent one-ancilla quantum differential equation solver provides a hardware-friendly and locality-preserving approach with provable performance guarantees, making it highly suitable for the early fault-tolerant and near-term regimes. Its simple circuit structure comes with a natural trade-off: the maximum single-run circuit depth scales as $O (1/ε)$ in the target accuracy $ε$. In this work, we reduce this depth by combining the solver with classical step-size postprocessing. By running the one-ancilla solver at a logarithmic number of finite time step sizes and using classical post-processing to cancel leading discretization errors, we reduce the maximum single-run circuit depth to $O(\mathrm{polylog}(1/ε))$ without adding quantum ancillae or sacrificing locality. Technically, extending extrapolation ideas beyond Hamiltonian and Lindbladian dynamics requires regularity estimates for observable maps under nonunitary evolution, which we obtain through a holomorphic extension of the adjoint evolution. Numerical experiments on the Hatano-Nelson model (ODE) and the convection-diffusion equation (PDE) demonstrate the effectiveness of the approach.

Different diagnostics of black hole thermodynamic branch structure do not always identify the same local event. We examine this issue in charged AdS black hole examples where a thermodynamic topological phase transition is known to occur. In these examples, previously reported in [JHEP 06 (2024) 213}], the inverse temperature $τ_a$ at which the total topological number changes is separated from the inverse temperature $τ_b$ at which the fixed charge parameter branch response considered here becomes singular. Reanalyzing the same zero point branch geometry, we find that the two diagnostics track different local structures. The change at $τ_a$ is a boundary membership and index flow event of the positive radius physical branch set: a zero point with nonzero winding enters or leaves the counted domain. The singularity at $τ_b$ is instead caused by a turning point at finite radius, where a pair of branches annihilates. Since the two branches carry opposite windings, this event changes the branch count but preserves the total topological number. Thus, in the analyzed examples, the thermodynamic topological phase transition and the branch response singularity are distinct diagnostics of the same zero point branch geometry, with different local origins.

I formulate black-hole-horizon-induced decoherence of charged branch codes as the leading-soft QED restricted to an exterior algebra, formulated as an open quantum system. The fixed-history Feynman--Vernon identity ${\cal F}[J,J]=1$ remains exact. Decoherence enters through the unequal-history influence factor that survives exterior monitoring and belongs to the complementary horizon output. In the coherent eikonal regime, I derive the completely positive Schur channel $({\cal E}_H^{(0)}ρ)_{ab}=\langleΦ_b^{H,(0)}|Φ_a^{H,(0)}\rangle \, ρ_{ab}$. The leading soft input is the eikonal factor, projected onto the horizon radiative algebra. The channel yields Gram-positivity constraints, an exterior quantum-eraser bound, finite-time non-Markovianity tests, soft/hard scaling criteria, and a charged-qutrit interferometer measuring a leading-soft Bargmann holonomy. The holonomy phase is the rephasing-invariant symplectic area of a triangle in horizon soft phase space. I show that its orientation, common-mode, triangulation, and completely positive determinant identities render falsifiable tests beyond pairwise two-path visibility.

The study of non-Markovian quantum processes has attracted significant interest in recent decades, giving rise to several competing notions of quantum non-Markovianity. These notes serve as an introduction to the topic for graduate students familiar with quantum mechanics and probability theory. Owing to the vastness of the literature, we focus on two prominent characterizations of quantum Markovianity based on the divisibility of quantum channels and monotonically decreasing state distinguishability. The correspondence between classical concepts (stochastic matrices, Chapman-Kolmogorov equation) and their quantum analogs (dynamical maps, CP-divisibility) is emphasized throughout.

We consider the possibility of dark matter in a warped extra-dimensional theory in presence of a linear dilaton background, with a gapped continuum spectrum, in a brane-world cosmological scenario. Firstly, triggered by self-energy radiative corrections, we study the existence of an isolated resonance of massive gravitons, and its realization as a long-lived feebly interacting dark matter candidate, produced by the freeze-in mechanism. This massive graviton is proved to satisfy all theoretical and experimental constraints, in the sub-MeV mass range. We further consider the close relationship between the existence of this component of dark matter and the presence of an inflaton localized on the brane, with a mass around $10^{11}$ GeV and a sub-TeV reheating temperature, in a brane inflationary scenario that allows to reproduce the most recent cosmological observables. Secondly, the gapped continuum of gravitons, a particular five dimensional realization of the physics of unparticles, is identified as a holographic fluid which can play the role of holographic dark matter. The production of the holographic fluid goes by an ultra-violet freeze-in mechanism, with an abundance mainly depending on the reheating temperature. Depending on the values of the mass gap and the reheating temperature, one or both components of dark matter can be present.

The detection of transient gravitational wave signals relies on independent search algorithms that analyse detector data and assign significance measures to candidate events. However, varying performance complicates their interpretation. We use supervised machine learning combined with conformal prediction, a framework to quantify uncertainties, to merge multi-pipeline information into well-calibrated confidence scores. We demonstrate that this approach is robust across different classifier architectures and remains stable when trained on different simulated datasets. When applied to events across the GWTC catalogue up to and including the second part of the fourth observing run, the framework identifies several subthreshold candidates with elevated confidence, including the binary neutron star candidate GW200311_103121. We examine the reliability of these up-rankings, finding evidence that high-confidence predictions correspond to signal-like events. This framework enables simplified systematic candidate assessment for gravitational wave catalogues and real-time alerts by providing a single, well-calibrated confidence measure per candidate.

This work investigates the thermodynamic behavior of deformed AdS-Schwarzschild black holes by incorporating higher-order corrections within non-interacting spacetime models and extended entropy frameworks. To address the inadequacies of classical statistical mechanics in describing gravitational systems with non-local and long-range interactions, we employ non-extensive entropy formalisms, specifically Tsallis and Barrow entropies, which capture quantum-scale deviations and extended correlations. The resulting thermodynamic analysis reveals significant departures from conventional black hole behavior under strong entropy deformations. Notably, as the degree of non-extensivity decreases, the system asymptotically recovers classical features, indicating an emergent universality across statistical regimes. Furthermore, the Joule-Thomson (JT) expansion is examined to analyze the temperature-pressure response during adiabatic processes. Key thermodynamic quantities, including mass, temperature, heat capacity, Gibbs free energy, enthalpy, internal energy, and the JT coefficient, are computed under the influence of non-extensive entropy corrections. These results provide deeper insight into black hole thermodynamics in quantum-corrected spacetimes and offer new avenues for exploring gravitational systems beyond the traditional Bekenstein-Hawking (BH) framework.

The quantum boomerang effect (QBE) is a unique dynamical signature of Anderson localization, characterized by a launched wavepacket that initially drifts but ultimately returns to its initial position due to fundamental quantum interference. In this work, we theoretically establish and quantitatively characterize the QBE in a time-crystalline structure using a periodically driven quantum particle in a one-dimensional potential well. By constructing maximally localized Floquet-Wannier states and introducing temporal disorder, we rigorously map the continuous Floquet dynamics onto a discrete disordered tight-binding lattice. By positioning a detector at a fixed spatial coordinate, we monitor the temporal evolution of the wavepacket, to extract the mean temporal center of mass of the probability density in a time-crystalline structure. This mean temporal center of mass exhibits an initial ballistic expansion, followed by a pronounced U-turn, and ultimately returns to its initial temporal position after long-time evolution. These results confirm the existence of the complete QBE in the time domain. They also demonstrate that non-trivial dynamics can be explored within time-crystalline systems, even though these structures already possess an inherent temporal periodicity.

We study a quantum implementation of non-Abelian Thouless pumping in Lieb lattices using Rydberg synthetic dimensions. The lattice is encoded in twelve selected microwave-coupled Rydberg levels, forming a three-cell structure with six degenerate zero-energy states. These zero-energy states define the working subspace for cyclic modulation of the microwave couplings, while the remaining bright states provide the dominant leakage channels at finite evolution time. To choose the relative timing of the Gaussian pulses, we introduce a global adiabatic criterion (GAC), which evaluates the mean value and temporal fluctuation of a nonadiabatic factor obtained from a representative $Λ$-type transfer paradigm. With the resulting timing applied to the full twelve-level pumping dynamics, composing two elementary pumping cycles in opposite temporal orders produces distinct projected population maps. It is exactly consistent with noncommuting matrix-valued adiabatic operations in the zero-energy subspace. We numerically simulate the non-Abelian Thouless pumping using the Lindblad master equation with state-dependent Rydberg loss and representative perturbations. The results show that the GAC-selected timing within the same Gaussian pulse family gives higher target-state population than two literature-adapted Gaussian pulse schedules over the simulated parameter ranges. This quantum implementation of non-Abelian Thouless pumping, enabled by the GAC, marks a major milestone in finite-time geometric control and paves the way for transformative applications in holonomic quantum computing with Rydberg synthetic lattices.

This short report explores the (non exhaustive) current state of quantum technologies, their potential applications, and the challenges that must be addressed to harness their full potential. In particular we will focus on the quantum computer architecture and its ecosystem. Such architecture represents a complex, multi-layered system that integrates quantum and classical components to enable the execution of quantum algorithms. This manuscript is then organized as follows : first we will introduce the quantum processing unit, the lowest layer of a quantum computer. Then we will progress from the lowest to the higher layer of the system architectures : measurement, circuit control, error correction and mitigation system, the quantum compiler and finally the software stack, with particular emphasis on the interactions between these components

We investigate the stability of scalarized black holes in Einstein-scalar-Gauss-Bonnet-Ricci theory along their fundamental branches. We show that initially stable solutions first lose nonspherical stability in the eikonal regime, while lower multipoles remain stable. As the branch is continued, instability extends systematically toward lower multipoles, forming an ordered hierarchy of deformation instabilities extending down to the quadrupole mode, while the dipole sector remains stable. The instability thresholds obey a common scaling law and approach finite eikonal limits, defining the boundary of the angularly stable region. We demonstrate that the previously identified quadrupole and angular-Laplacian instabilities are connected by a continuous hierarchy of instability thresholds spanning the angular sectors of the theory. This hierarchy is distinct from radial stability, which changes only at branch turning points, and reveals a previously unexplored angular organization of instabilities in scalarized black holes.

We propose a framework connecting cosmic microwave background (CMB) observables with high-energy particle phenomenology, based on Starobinsky supergravity coupled to the Minimal Supersymmetric Standard Model (MSSM). Cosmic inflation and supersymmetry (SUSY) breaking occur within the hidden sector emerging from Starobinsky supergravity. The inflationary scale fixes the characteristic mass scale of the hidden sector, which determines the MSSM soft terms through gravitational mediation of SUSY breaking. The same hidden sector can also dynamically generate a high-scale $μ$ term. The resulting MSSM spectrum fixes the high-scale threshold corrections and the boundary conditions for the renormalisation-group (RG) evolution of the Higgs quartic coupling. Three-loop RG evolution of the Higgs quartic coupling gives a Higgs boson mass consistent with the measured value within the theoretical and experimental uncertainties, thereby linking the amplitude of primordial scalar perturbations to the Higgs boson mass. With conserved R-parity, the lightest supersymmetric particle is stable, making it a compelling dark matter candidate. The observed relic abundance selects a nearly pure thermal wino with a physical mass of about 3 TeV. Its loop-induced spin-independent wino-nucleon scattering cross section lies below the current sensitivity of the LUX-ZEPLIN experiment, but within the projected reach of next-generation multi-ton liquid-xenon detectors. Electroweak radiative corrections generate a small mass splitting between the charged and neutral wino states, leading to a long-lived charged wino and the characteristic disappearing-track signature. A future 100 TeV proton collider can discover the disappearing-track signal from a 3 TeV wino or fully exclude this thermal wino dark matter scenario.

Entanglement between two spatially separate matter particles can be generated via many means and often resides in the internal states of particles. Here, via Rydberg blockade in two spatially separate neutral atoms, we find that the photon recoil in Rydberg excitation can push one atom microns away provided the other atom exerts a state-dependent Rydberg-mediated blockade. When the atoms are recaptured by optical traps, a position-position entangled state between two spatially separate atoms can emerge. This realizes a Bell state of two atoms, where the entanglement exists in the position of each atom and the distance between the two possible locations of each atom can be in the hundred-micron regime.

Dynamical structure factors (DSF) measured with neutron-scattering experiments provide key insights into the structure of materials. Their computation requires both the preparation of an equilibrium state and the implementation of Hamiltonian dynamics. We demonstrate the feasibility of computing DSF on the Quantinuum Reimei trapped-ion quantum computer, comparing the DSF of 1D Heisenberg model on $20$ sites, and that of the copper sulfate crystal. To that end, we introduce a pumping approach for computing the DSF $S(q,ω)$ on quantum computers that enables targeting specific arbitrary values of frequencies $ω$. This method time-evolves the initial state using a time-dependent Hamiltonian perturbed by a source term oscillating at the target frequency $ω$. When targeting only a few frequency values, this approach provides a significant reduction in shot overhead compared to previous methods.

In this work, we investigate the formation of naked singularities for the $3+1$-dimensional Einstein-scalar field system without symmetry assumptions. We generalize the spherically symmetric and self-similar naked-singularity solution constructed by Christodoulou in [4] by prescribing non-spherically symmetric initial data along both incoming and outgoing initial null hypersurfaces. We then establish global existence for the resulting solutions and analyze the singular structure of the inner Cauchy horizon. Our construction is based on employing a notion of four-type differences and designing a system of scale-invariant weighted norms to control the corresponding geometry. We show that the constructed spacetimes retain a global naked-singularity structure, characterized by an incomplete future null infinity and a singular inner Cauchy horizon. Moreover, we derive detailed asymptotics near the inner Cauchy horizon and prove the desired $C^{1, \fracκ{1-κ}+}$ inextendibility of these solutions, where $κ\in (0,1/3)$ is the self-similar parameter. This indicates a connection between weak and strong cosmic censorship: for the class of non-spherically symmetric solutions constructed here, the failure of weak cosmic censorship in its strict formulation is accompanied by a quantitative inextendibility mechanism at the inner Cauchy horizon.

Exotic helium atoms act as unique atomic traps for heavy, negatively charged particles, protecting them from nuclear annihilation and nuclear capture on timescales long enough to enable high-precision laser spectroscopy. Such measurements serve as stringent tests of three-body quantum electrodynamics and offer a direct route to determining fundamental particle masses. Motivated by upcoming spectroscopic efforts targeting pionic ($π^{-\,4}\mathrm{He}^+$) and kaonic ($K^{-\,4}\mathrm{He}^+$) helium, we present a rigorous theoretical evaluation of the collisional and density effects governing these systems. Using an ab initio potential energy surface and coupled-channel quantum scattering calculations, we study the collisional stability of the candidate metastable states against inelastic quenching in a cryogenic helium buffer gas. Furthermore, we provide theoretical reference values for the pressure broadening and pressure shift coefficients of the targeted transitions. These results establish an essential benchmark for future experiments, paving the way for refined determinations of the pion and kaon masses.

We propose a method to create a room-temperature maser operating at approximately 2.9~GHz frequency using an ensemble of negatively charged nitrogen-vacancy electronic spins (NV) in diamond, without requiring population inversion. Our method considers a DC magnetic field of a few milli-Tesla (mT) applied along the perpendicular direction of an ensemble of NV spins aligned along a common axis. This perpendicular magnetic-field creates superposition states of $|m_{\mathrm{s}}=-1\rangle$ and $|m_{\mathrm{s}}=+1\rangle$ of the NV spin's ground state triplet levels and thereby makes it possible to drive all three transitions in the NV spin ground state. We model the system by including optical pumping of the NV spins, near-resonant driving of two transitions, and coupling the third transition to a near-resonant microwave resonator. Numerical estimates using experimentally realizable parameters show that inversionless masing can be achieved inside the microwave resonator using our method. As an application, we show that the output intensity of an inversionless maser ($1.1\times10^{14}$ spins) can be used for magnetic field sensing with a DC sensitivity on the order of a hundred pT/$\sqrt{\mathrm{Hz}}$. Our study opens a new direction in room-temperature diamond NV maser devices for quantum technological applications without the requirement of a strong bias magnetic field, as in conventional NV diamond masers.

Bose-Einstein-condensed (BEC) and superfluid dark-matter (SFDM) halos can contain coherent, wave-supported cores whose angular momentum is carried by quantized vortices. When vortices form a tangle, reconnections convert part of the vortex kinetic energy into dark phonons, density waves, Kelvin waves, and vortex loops, providing a microscopic channel by which SFDM vortex structure can affect halo-core evolution. We estimate the dynamical importance of this channel by combining the local Gross-Pitaevskii reconnection law with a halo-scale vortex-line density calibrated against Schrödinger-Poisson simulations. In the minimal model the released energy stays in the dark sector, and standard-model luminosity requires additional portal physics.

Efficient single-photon frequency conversion is investigated in a two-giant-atom waveguide-QED system, where a two-level giant atom and a $Λ$-type three-level giant atom couple to a common one-dimensional waveguide. While the $Λ$-type atom provides the inelastic channel, the two-level atom induces secondary coherent coupling, creating multi-path interference for the converted photon. Using the real-space approach and within the Markovian approximation, we derive analytical four-channel scattering amplitudes and reveal that the inelastic transmission spectrum, governed by three complex resonance poles, exhibits a multi-peak interference pattern. By introducing a local single-pole approximation, we reduce this complex spectrum to a local Fano lineshape, decomposing it into a coherent superposition of a local background term and a single-pole resonant term. The interplay between these two terms-controlled by the photon propagation phase between the giant atoms' coupling points-determines the conversion efficiency, with the background suppression condition leading to a Lorentzian reduction. Based on the single-pole resonance weight, we formulate a phase-selection criterion for highly efficient conversion. Compared with both the small-atom and single $Λ$-type giant-atom models, the two-giant-atom scheme achieves substantially enhanced inelastic transmission over a broader frequency-conversion range. This work reveals how phase-controlled local Fano resonance enables high-efficiency frequency conversion, establishing a general paradigm for engineering resonant light-matter interactions in structured quantum systems.

In a symmetry-restoration quantum Mpemba effect, an initial state with stronger local symmetry breaking can lose that memory faster than a state that starts closer to the symmetric manifold. We test whether this local ordering reversal is organized by chaotic thermalization in a clean U(1)-conserving spin chain, comparing spectral level statistics with crossings of the entanglement asymmetry for the same Hamiltonians. We find that Gaussian orthogonal ensemble (GOE)-like level statistics alone do not determine whether Mpemba crossings occur. Across field textures, GOE-like spectra can occur with or without entanglement-asymmetry crossings, and crossings can also appear away from the GOE reference. A near-staggered detuned control further shows that even an inversion of the total charge-sector coherence need not produce an entanglement-asymmetry crossing. Thus the crossing response is controlled not by spectral chaos alone, but by how local charge-sector coherence enters the reduced density matrix.

Gravitational-wave observations of compact binary mergers have enabled precision tests of gravity in the strong-field dynamical regime. Current approaches combine single-event results that assume deviations from General Relativity (GR) are uniformly distributed across events, limiting their flexibility and potentially biasing the inferred evidence. We introduce a simple mixture-model framework in which a fraction $ζ$ of events is consistent with GR, while a fraction $1 - ζ$ deviates from it, without imposing constraints on the population distribution of the deviation parameters. We apply this method to publicly available results from the LIGO-Virgo-KAGRA (LVK) collaboration on O1-O4a compact binary mergers obtained through FTI, pSEOBNR, and KerrPostMerger tests. We find that the data are consistent with all events satisfying GR. However, we obtain respective Bayes factors $B^{ζ=1}_{ζ\neq 1} \simeq 20 $, $10$ and $15$, which are much smaller than those inferred from existing LVK analyses, indicating that the data require greater flexibility in modelling possible deviations than standard approaches permit. In light of our results, we recommend using flexible mixture models to test GR across compact-merger catalogues, unless there are obvious physical motivations to impose more restrictive models, as in the case of graviton-mass estimates.

The twin paradox is a foundational thought experiment in the special theory of relativity where a returning twin ages less than the one who remains stationary. However, the intricacies of the twin paradox remain relatively underexplored in the curved spacetimes of general relativity. Here we explore the twin paradox in the vicinity of a rotating black hole, where the existence of multiple paths between two events creates significant complexity. We develop a numerical framework based on residual maps and optimisation to identify possible trajectories. We find a strong negative correlation between a traveller's experienced proper time and both the azimuthal distance travelled and the magnitude of acceleration. We apply numerical Jacobi field analysis to examine conjugate points along geodesics within the Kerr geometry, finding that only the geodesic with the minimal azimuthal distance contains no conjugate points. This provides beginner students of general relativity with a visual tool to understand general relativistic concepts, helping to correct flat-spacetime intuitions.

Recurrence Plot (RP) is a method employed to analyze the periodicity, chaoticity, and nonlinear characteristics of complex systems. Quantum Principal Component Analysis (QPCA), on the other hand, achieves dimensionality reduction of sample data using density matrices based on quantum circuits. We improve the distance threshold function of the recurrence plot algorithm using a density operator conceptually equivalent to the covariance matrix, integrate it with quantum circuits, and thereby develop a Quantum Recurrence Plot (QRP) algorithm. This algorithm achieves ultra-high efficiency in parallel computing, reduces computational costs, and simultaneously upgrades the traditional grayscale recurrence plot to colored heatmaps, enabling a better revelation of the system's dynamical characteristics.

We report an improved absolute frequency measurement of the $^{176}$Lu$^+\,(^{3}\mathrm{D}_1)$ optical frequency standard, evaluated via a remote link to the NRC-FCs2 caesium fountain primary frequency standard. Operating a single ion clock with 94.2% uptime over 10 days, and using an ambiguity-resolved precise point positioning (PPP-AR) link over the Global Positioning System (GPS), we determine an absolute frequency of $353\,638\,794\,073\,800.33(9)\,$Hz at a fractional uncertainty of $2.6 \times 10^{-16}$. This agrees with our previous result, which underpins the CIPM recommended frequency value, and reduces the uncertainty by a factor of 3.6.

We study lower bounds for Bayesian quantum parameter estimation, with a particular focus on qubit models. While several lower bounds on the Bayes risk have been proposed, including the Bayesian symmetric logarithmic derivative (B-SLD) type bound and the Bayesian Nagaoka-Hayashi (B-NH) bound, there is no definite proof available to show they are attainable except for special cases. Thus, identifying attainable bounds together with their corresponding optimal measurement strategies remains a central open problem in Bayesian quantum estimation. In this work, we introduce a new Bayesian lower bound, referred to as the Bayesian Gill-Massar (B-GM) bound, inspired by the logic of Gill-Massar bound in point estimation. We derive an analytical closed-form expression of the bound and show that it is attainable for any qubit model. In particular, we prove that the optimal Bayesian strategy can be realized by a projection-valued measure associated with a single effective direction determined by the weight matrix and the B-SLD-type Fisher information matrix. We provide numerical comparisons between the B-GM, B-NH, and B-SLD-type bounds in higher-dimensional models. Our results show that the B-GM bound has a limitation in high-dimensional models with few parameters, since it can be negative.

The Bumblebee gravity (BG) model, featuring spontaneous Lorentz symmetry breaking via a vector field non-minimally coupled to curvature, has been widely used to explore Lorentz-violating effects in cosmology. We investigate primordial black hole (PBH) formation within this framework, deriving the complete set of modified perturbation equations. We demonstrate that BG, sourced by a timelike vector field, introduces three distinct enhancements to PBH abundance--modified expansion history, suppressed collapse threshold, and amplified power spectrum--which together render PBHs viable dark matter candidates across the asteroid-mass window for modest Lorentz-violating couplings. However, a systematic analysis of the quadratic action reveals that these phenomenological consequences emerge from a theoretically pathological foundation. The vector sector exhibits an intrinsic ghost instability, while the requirement of a stable symmetry-breaking minimum simultaneously induces a tachyonic instability on timescales far below cosmological scales. The model thus suffers from a fundamental inconsistency: the conditions for cosmological viability and spontaneous symmetry breaking are mutually exclusive within the minimal Bumblebee framework. Our results illustrate both the notable power of Lorentz violation to influence early Universe observables and the necessity of a consistent theoretical foundation for such predictions.

We formulate the geometric structure of non-compact bosonic quantum systems in regimes where vacuum instability renders the relevant quantum states non-normalizable, causing conventional state-space quantum geometry -- described by the Berry connection, curvature, and quantum metric -- to become ill-defined. To overcome this breakdown, we develop a formulation of quantum geometry at the level of canonical operator frames, allowing for complexified Bogoliubov-Valatin transformations that lift the requirement that creation operators be Hermitian conjugates of annihilation operators. Canonical operator frames are defined as choices of bosonic creation and annihilation operators realizing the canonical commutation relations. A natural equivalence relation among such frames generalizes the phase ambiguity of quantum states and determines a parameter space that analytically extends the stable-regime parameter space. The space of canonical operator frames forms a principal bundle over parameter space -- the operator-frame bundle -- equipped with a natural Ehresmann connection that defines parallel transport while preserving the canonical commutation relations. In the stable regime, this construction reduces to the Berry connection, while more generally it yields a well-defined operator-space quantum geometric tensor (QGT) that remains valid across vacuum instabilities. Using the framework of rigged Hilbert spaces, we define a notion of quantum frame vacuum and obtain a consistent state-space QGT. Focusing on a single bosonic mode, we demonstrate analyticity of the QGT across the quantum frame-vacuum phase transition and present the corresponding phase diagram on the complexified squeezing-parameter plane. We further introduce a realistic physical setting that allows continuous paths connecting stable and unstable regimes, along which the QGT evolves smoothly across Stokes lines.

Macroscopic optomechanical systems offer a promising testbed for distinguishing whether gravity acts as a quantum interaction or as a classical field. Schrodinger-Newton (SN) theory is the nonrelativistic limit of semi-classical gravity where quantum matter couples to classical gravity. Based on SN theory, this work investigates how classical self-gravity affects continuous quantum state tomography of a macroscopic mechanical oscillator monitored by variable-angle homodyne detection. In the Schrodinger-Newton (SN) theory, the measurement record arises from a different conditional test mass dynamics from that in quantum-gravity (QG)/standard quantum mechanics, consequently, applying the QG-optimised reconstruction map introduces an additional state-dependent contribution. We show that this contribution makes the reconstructed covariance depend on the chosen set of tomography angles and can drive the SN covariance--after QG filtering--outside the standard Gaussian-covariance domain set by the Heisenberg uncertainty principle. We quantify the resulting QG-SN distinguishability via the Hellinger distance and analyse its dependence on measurement strength and temperature. We then formulate the same issue in the broader setting of nonlinear quantum mechanics: when the system's conditional dynamics during the readout process depends on the state being inferred, the tomographic map acquires nonlinear, model-dependent corrections to the usual Radon or Gaussian reconstruction map.

We obtain an experimental constraint on possible Born-rule violations at high-energy colliders. We model Born-rule violations with differential scattering cross sections $dσ/dΩ$ subject to an angular smearing by a narrow Gaussian of width $\varepsilon$ (with respect to $x=\cosθ$ for scattering angle $θ$). For large-angle Bhabha ($e^{+}e^{-} \rightarrow e^{+}e^{-}$) scattering, at a centre-of-mass energy $\sqrt{s}=29\, \mathrm{GeV}$, data from the PEP collider at SLAC allow us to set an upper bound of $\varepsilon<0.042$ at $95\%$ confidence. This corresponds to a Gaussian smearing over an angular range of twice the experimental bin width, and hence provides a physically meaningful limit on deviations from the Born rule. Future prospects for improving this limit are discussed.

We show that anyonic exchange statistics can activate non-Hermitian point-gap topology in models that are topologically trivial in its absence. The emergent topology oscillates more rapidly with the statistical phase as the anyon number increases, and exhibits a parity dependence on the particle number. A perturbative analysis reveals the mechanism: fractional statistics induces a mismatch between momentum terms that, combined with sublattice-dependent dissipation, produces particle-number-dependent non-reciprocity and complex spectral winding. As these effects rely on the formation and exchange of interaction-bound anyons, our results establish exchange statistics as a resource for enabling non-Hermitian topology under programmed dissipation.

We propose a component-selective scheme for improving two-photon sources in a cavity-coupled two-atom system, where a single cavity mode interacts with two two-level atoms driven by phase-controlled classical fields of the same frequency. By controlling the atomic detunings and driving phase, the system can be tailored toward optimized cavity-field two-photon blockade or strongly correlated fluorescence photon-pair emission. When the two-cavity-photon component is enhanced, the cavity field exhibits optimized two-photon blockade with simultaneous suppression of unwanted one- and three-photon components at a comparable two-photon population. In another parameter regime, strongly correlated fluorescence photon pairs can also be generated from the two atoms by selecting the double-atomic-excitation component in the same two-excitation manifold. This approach provides a route toward high-quality and versatile two-photon sources, with potential applications in few-photon quantum optics and quantum information processing.

We investigate rotational properties and universal relations of slowly rotating Bose-Einstein condensate dark stars in the context of General Relativity, both at the mean-field level and when the leading beyond-mean-field Lee-Huang-Yang correction is retained self-consistently. Adopting the polytropic $n=1$ equation of state appropriate to a dilute, self-interacting Bose gas, parameterised by the boson mass $m$ and the $s$-wave scattering length $a_s$, we integrate the Tolman-Oppenheimer-Volkoff equations together with Hartle's dipole equation for the frame-dragging angular velocity, and we compute the moment of inertia, the gravito-electric tidal Love number and the dimensionless tidal deformability. The resulting equilibrium sequences yield gravitational masses in the $1$--$2\,M_{\odot}$ range with radii of $10$--$20\,\mathrm{km}$, squarely within the window presently probed by NICER and the LIGO-Virgo-KAGRA network. We observe that the LHY term produces a measurable reduction of the dimensionless moment of inertia at fixed compactness, whilst the I-$Λ$ universal relation is preserved to within a few per cent. We supply polynomial fits for the I-$Λ$ and I-$C$ relations, and show that the LHY footprint is large enough to serve as a clean diagnostic of beyond-mean-field quantum physics in a putative dark star population, complementing existing dark matter constraints from pulsar masses and from the equation-of-state interpretation of the unusually light compact remnant HESS~J1731-347.

We study the atom-atom correlation length in an atomic array coupled to a waveguide under the Bragg condition with delayed non-Markovian interactions caused by a finite photon propagation time. Starting from a single excited atom, the excitation partially spreads among all atoms, reaching a steady state. The remaining excitation localizes near the initially excited atom, and the atom-atom correlation length decreases as a power law with the interaction delay. This localization phenomenon reveals how the delay-induced non-Markovian behavior affects the correlation transport in waveguide QED systems.

It is well known that the quantum value of an XOR nonlocal game, where the winning condition depends only on the XOR of the two players' output bits, may be approximated in polynomial time. We study a variant of the XOR game model, which we call tilted XOR games, where the winning condition can additionally depend on only one of the output bits. We show that this dramatically increases the expressive power: the computational complexity of the problem of approximating the quantum value of tilted XOR games to constant precision is RE-complete. Also, our result extends to succinct versions of tilted XOR games, where the questions can be polynomial-length binary strings, generated by a polynomial-time verifier. For classical strategies, the distinction between XOR games and tilted XOR games is inconsequential. Håstad (J. ACM, 2001) shows that they are both NP-complete to approximate, by using a reduction from linear systems to XOR games. Our approach is to show that this is also quantum-sound, but as a reduction from linear system games to tilted XOR games. Since titled XOR games are a special case of binary games (where each party outputs a single bit), our result implies that binary games are RE-hard to approximate.

We study the error-correcting properties of pure-state holographic geometries, in which mixed boundary subregions are replaced, via the surface/state correspondence, by the Ryu--Takayanagi (RT) geodesic bounding their entanglement wedges. In AdS$_3$/CFT$_2$ we derive a cross-ratio threshold relation $η'/η= e^{ΔH/2}$ for the connected/disconnected transition of the entanglement wedge when two holes are punched in such a geometry. The quantity $ΔH$ is sourced entirely by geodesics ending on RT boundaries. It shifts the standard two-interval threshold $η= 1/2$, and we classify when its sign is fixed by the pattern of hole endpoints. Turning to code properties, we show that the recursive hole-punching underlying uberholography cannot start within an RT-boundary, while an untouched asymptotic boundary can still fractalize, and we find numerically that in the configurations we study it does so with the universal fractal dimension $α\approx 0.786$. The resulting upper bounds on price and distance are nevertheless procedure dependent. In the configurations we study, punching holes on the asymptotic boundary while retaining the RT-boundary yields strictly tighter bounds than first tracing out the RT-boundary and then fractalizing.

We study three-dimensional Einstein gravity with negative cosmological constant on non-compact spatial boundaries within the Chern-Simons formulation. Using an exact fluid/gravity correspondence, we derive a closed radial flow equation for the quasi-local stress tensor and show that it realizes the holographic $T\bar T$ deformation at finite cutoff. We further develop the inverse-scattering description of the boundary dynamics, identifying the gravitational interpretation of the associated spectral data and analyzing the finite-cutoff deformation of soliton solutions. Although the boundary evolution is governed by an integrable bi-Hamiltonian hierarchy, we show that the radial flow itself is not Hamiltonian with respect to the canonical Poisson structure. Our results establish a unified framework connecting integrability, quasi-local gravitational observables, inverse scattering, and finite-cutoff holography on non-compact boundaries.

Chiral molecular structure can couple to electron spin, leading to unexpected spin polarization effects. This Chirality-Induced Spin Selectivity (CISS) was first reported for DNA molecules adsorbed on gold but its microscopic origin remains unclear. We demonstrated though simulation that the exchange arising from electron-electron Coulomb interactions within the self-consistent mean field (Hartree--Fock) approximation can yield significant ($\sim 2\%$) spin polarisation for $3$-methylcyclohexanone adsorbed on Cu(111) amplifying a much smaller ($\sim 0.0014\%$) initial bias, consistent with experiment. Symmetry considerations ensure the result is physically meaningful, while its ab-initio nature ensures all parameters are physically realistic. This amplification is connected to existing studies on spin-symmetry breaking in Hartree--Fock, providing a new pathway for understanding the magnitude of CISS as an emergent phenomenon of interacting electrons.

Quantum many-body states with sufficiently low energy variance can serve as approximations to thermal states, and they may be prepared by energy filtering simple pure states. In this work, we examine how narrow the filter width must be to guarantee thermal behavior. To this end, we analyze the problem in the Floquet regime, where filtered states are found to be equivalent to time averages. This equivalence allows us to reproduce the distinct Rényi-$α$ entropy scalings as reported in [Morettini et al., Physical Review Letters 133, 240401 (2024)]. Crucially, we show that under the Floquet eigenstate thermalization hypothesis, the trace distance between Floquet-filtered states and thermal states is bounded by the square root of filter width. We further demonstrate that these results extend naturally to the conventional Hamiltonian setting by mapping Hamiltonian-filtered states to their Floquet counterparts.

The production of large-scale entangled states is one of the main goals of next-generation quantum technologies, with an immediate potential for applications in the context of entanglement-assisted quantum sensing. A very promising platform to achieve this goal is offered by ultracold spinor gases, made of atoms with a large internal spin sensitive to magnetic fields. Here we show that the native spin-changing collisions in a spinor Bose-Einstein condensate, combined with an arbitrary quadratic Zeeman shift, can generate scalable spin squeezing in the collective spin of the ensemble, following the universal paradigm of the celebrated one-axis-twisting model. Squeezing dynamics is driven by the quadratic Zeeman shift when this shift is small; and by the spin-changing collisions for large shifts, in the form of stroboscopic squeezing. Turning off the Zeeman shift freezes out the collective-spin dynamics, so that the ensuing collective spin dynamics can be uniquely governed by an external field to be sensed. Our theoretical results pave the way for the use of spinor Bose gases with a large spin in fundamental studies of entanglement, as well as in advanced metrological applications.

Currently, remote energy transfer and immunity to dissipation are hot topics in quantum batteries (QBs). In this work, we propose a protocol to realize energy transfer between two remote atoms (a quantum charger and a quantum battery) each coupled to a separate optical cavity with the cavities connected by a fiber. The cavities and fiber are coupled to their individual baths. After optimizing inter-system couplings to achieve an efficient transfer, we uncover the effect of suppressing dissipation by introducing parity deformation of the cavities fields. We also prove that the charger-battery entanglement is a consumable resource for energy storage: it is initially stored until the charger and battery reach energy balance, and then subsequently consumed to maintain the increase in energy stored in the battery. The present scheme is the first execution of energy transfer to a distant battery assisted by entanglement, which may help better understand quantum thermodynamics and open new possibilities toward harnessing decoherence as a resource to improve the charging performance of QBs.

Pulsar timing array data analysis is computationally expensive, limiting the complexity of models which can be studied. As pulsar timing datasets and their respective models grow in size and sophistication, faster and scalable inference methods are essential. In this paper, we accelerate pulsar timing analyses by sampling in the space of Fourier coefficients instead of analytically marginalizing over them. Previous studies have shown the Fourier space induces a complex, high-dimensional posterior geometry, from which it is generally difficult to sample. We show that under an appropriate coordinate transformation the Fourier coefficients approximately follow a standard normal distribution, and may be efficiently sampled using a Hamiltonian Monte Carlo scheme. Under this coordinate transformation, for datasets of size and complexity comparable to the NANOGrav 15-year release, the new method produces converged posterior distributions for a range of models which include inter-pulsar correlations, stochastic, and deterministic signals in approximately 15 minutes on an NVIDIA GeForce RTX 3090 GPU. By comparison, the legacy pulsar timing analysis software \texttt{ENTERPRISE} would require months of computation on a CPU cluster to analyze comparable datasets under the same joint stochastic and deterministic models.

Continuous variable quantum key distribution protocols (CV-QKD) with discrete modulation have been intensively investigated to bridge the gap between ideal Gaussian modulation and modern coherent optical communication systems. To mitigate the penalty of discrete modulation, probabilistic constellation shaping (PCS) is applied to the modulation format and is typically performed by distribution matching (DM) algorithms. In this paper, we address the application of DM algorithms to perform PCS in CV-QKD protocols. We investigate the impact of approximating optimized Maxwell-Boltzman distributions with DM algorithms based on Huffman (HDM) and constant composition (CCDM) codes on the protocol's secret key rate (SKR) and tolerance to excess noise. Our results show that specifically symbol-by-symbol HDM degrades the SKR by at least 30\%, whereas CCDM matches the optimal SKR with code length of $10^3$ or more symbols. Furthermore, we also provide a statistical analysis of symbol dependence for both approaches, showing that CCDM must operate with blocks of at least $10^5$ symbols for the correlations become negligible. Finally, we propose an algorithm to generate independent symbols following near-optimal distributions.

Near-term physical implementations of quantum algorithms require efficient quantum error mitigation schemes to reduce quantum noise. In this letter we propose a new mitigation technique, by extending the applicability of our BBGKY-ISM scheme from quantum simulations of spin chains to arbitrary quantum circuits. We map executions of quantum circuits using Feynman's clock Hamiltonian to the Hamiltonian dynamics of a corresponding quantum system, whose time evolution obeys a BBGKY-like hierarchy of equations informing the BBGKY-ISM mitigation. We show that the method's classical and quantum overheads are polynomial in the circuit size and in the number of qubits. We apply our method to numerical simulations of tunable Bell state preparation circuits under state-of-the-art quantum noise, and numerically demonstrate its systematic and controllable quantum error reduction capability.

Majorana zero modes are the hallmark of topological superconductivity. In one-dimensional systems, these zero modes are usually introduced in the context of gapped, mean-field models that do not conserve particle number, such as the Kitaev chain. By non-locally encoding a conventional fermion across spatially separated Majorana zero modes, these systems become inherently immune to local decoherence. In this work, we show that signatures of Majorana edge physics persist in a number-conserving, gapless Luttinger liquid of spinless fermions with short-range attractive interactions. We identify the two-point correlator as a sharp diagnostic, revealing an edge-to-edge revival whose sign depends on the fermion-number parity. This revival is robust in the thermodynamic limit, and persists in the excited states of the system and at different fillings. A simple particle-hole ansatz for the ground state of the system with an odd number of fermions captures the physics of the system for a wide range of interaction strengths, interpolating between the free-fermion limit and the strongly interacting Majorana regime. Finally, we propose a concrete protocol to realize this model with ultracold dipolar molecules or atoms in an optical lattice, and to detect the revival via beam-splitter interferometry, opening an experimental route to Majorana physics beyond the conventional gapped-superconductor paradigm.

Einstein's equations of general relativity are one of the most complicated set of equations in all of physics and, for all but idealized physical settings, can only be solved by numerical methods on high-performance computing systems. Generating such solutions is a veritable Odyssey in its own right with adventures across the fields of mathematical theory, physical interpretation and computing challenges. These endeavors came to fruition in the mid 2000s when the two-body problem of general relativity was finally solved. And not too soon, as these results and their follow-up investigations came to play a key role in the Nobel-Prize winning discovery of gravitational waves by LIGO in 2015.

Grid state qubits offer a hardware-efficient approach to large-scale fault-tolerant quantum computing. They access the information redundancy required for quantum error correction by exploiting the large Hilbert space naturally available in harmonic oscillators. Superconducting architectures are particularly suitable to implement grid state qubits due to their fast and high-fidelity operations. Grid states in superconducting circuits enable quantum error correction (QEC) with performance beyond break-even. However, the state preparation and measurements (SPAM) errors of grid states has been a significant limitation to computational performances. In this work, we leverage high-performance QEC to enable repeat-until-success state preparation of both cardinal and magic states of the single-mode grid-state qubit. We combine this with an improved measurement protocol that corrects for both finite-energy envelope and auxiliary qubit readout errors, and increases robustness to photon loss. Our experiments, using both techniques, achieve a combined state-preparation and measurement error below $10^{-3}$. This represents two orders-of-magnitude improvement over the state of the art, bringing this platform on par with standard SPAM error levels measured in transmon qubits.

We herein study both the Joule-Thomson (JT) expansion process and the geodesic properties of a charged anti-de Sitter (AdS) black hole arising in modified gravity with nonlinear electrodynamic (NLED) sources. Our thermodynamic study reveals that the black hole charge has the most pronounced impact on the JT behaviour. The nonlinear electromagnetic sector together with the modified gravity parameters introduces further corrections to the inversion temperature and the associated cooling characteristics. At astrophysically relevant distances, the geometry closely reproduces expected outcomes.

Semi-device-independent quantum key distribution leaves the measurement devices uncharacterized while placing a trusted assumption on Alice's source. We formulate this source assumption operationally on Alice's four-preparation ensemble as a scalar bound on one of four physically motivated source tasks: full-label guessing, parity guessing, or their normalized composites with label exclusion. For the two-bit random-access code, we derive the exact classical frontier for each of the four source assumptions. Numerically, the BB84 strategy attains the maximal quantum deviation from all four frontiers, while the preparation-depolarized BB84 family and the direct-sum label-leakage family trace complementary branches of the arbitrary-dimensional quantum boundary for the two exclusion-assisted assumptions. Because all four task values are monotone under input-independent quantum channels, the same scalar source bound constrains every Bob--Eve extension compatible with the complete observed behavior. Using a three-setting extension that separates RAC testing from key generation, we obtain two dimension-independent security certificates over this feasible set: lower bounds on the conditional min-entropy and conditional von Neumann entropy, obtained respectively by direct optimization of Eve's key-guessing probability and by prepare-and-measure semidefinite relaxations based on the Brown--Fawzi--Fawzi variational bound. The exclusion-assisted assumptions certify positive key rates down to nearly vanishing preparation visibility, far beyond full-label or parity guessing alone. Under direct-sum label leakage, all four independently optimized rate bounds remain positive at every sampled incomplete-leakage point and vanish only at complete label revelation. These results show that robust semi-device-independent security depends not only on what Eve can identify, but also on what she can exclude.

While vacuum decay is traditionally described by Euclidean instanton methods, lattice simulations enable real-time modeling of dynamical observables relevant to cosmology and upcoming cold-atom analog experiments. We investigate the relationship between these approaches by extracting ensemble-averaged bubble profiles from zero-temperature simulations of a relativistic scalar field. Our observed profiles differ markedly from the bare Coleman bounce and classical thermal predictions. However, we find that instanton solutions in an appropriately renormalized potential reproduce both the measured profiles and their dependence on the UV cutoff, and predict decay rates consistent with simulations across the parameter range considered. The fact that a single renormalized Euclidean object captures these independent observables provides strong evidence that renormalization accounts for the discrepancy between the two formalisms, and establishes a quantitative link between instanton predictions, lattice simulations, and forthcoming empirical tests of vacuum decay.

Recent studies have identified a new component in black-hole ringdown from merging binaries, termed the \emph{direct wave}. This component was argued to be tied to the dynamical source evolution near the black-hole horizon, and thus to encode horizon information. Yet a firm theoretical foundation for the direct wave has been lacking. Here we fill this gap by deriving direct waves from first principles in Schwarzschild spacetime, using the causal structure of the Green's function. We show that the direct wave does not vanish and is governed by the near-horizon source dynamics. Our results establish a theoretical basis for direct waves as a probe of near-horizon dynamics, complementary to quasinormal modes.

We present \emph{spectral Born machines}, a class of quantum generative models that results from viewing and generalizing the class of IQP Born machines through the lens of group Fourier analysis. These quantum models exploit the quantum Fourier transform to create an inductive bias that make them naturally suited to learning integer-structured data, while remaining classically hard to sample from in general. Similar to IQP Born machines, spectral Born machines can be trained efficiently at scale on classical hardware via a maximum mean discrepancy loss based on graph spectral analysis, which we make available in a new \emph{tcdq} module of the PennyLane software platform. In numerical experiments, we show how the spectral bias of the model leads to significantly reduced parameter counts compared to unstructured approaches, and demonstrate the scalability of the software by training a 190-qubit model with over 1 million parameters to successfully learn a distribution of 93 nucleotide-long ribosomal RNA. Our results suggest that highly over-parameterized spectral Born machines may be immune to overfitting, even in strongly data-scarce regimes.

We construct Hilbert-space isomorphisms identifying the spaces of radiation fields on the event horizon and null infinity that are induced by the forward scattering map on the Schwarzschild exterior for fixed spherical harmonic mode solutions to the linear scalar wave equation arising from Cauchy data sets at $t=0$ which are regular at the event horizon, in the sense that the local energy at the event horizon is finite. We show that fixed spherical harmonic mode solutions with no radiation on the event horizon must decay exponentially in time, at a rate that is determined by the surface gravity. On the other hand, we construct examples of fixed spherical harmonic mode solutions with no radiation on null infinity and which have polynomial decay along the event horizon. Finally, we construct examples of polynomially decaying solutions to the linear scalar wave equation which are regular at the event horizon, have unbounded support in spherical harmonic modes, and induce no radiation on the event horizon.

We study gauge dependence in proper-time renormalisation group flows for asymptotically safe quantum gravity. Working in an essential scheme, we use field redefinitions to separate redundant off-shell contributions from the on-shell running of physical couplings. We consider two approximations: in the first, we work on general backgrounds and project onto all terms with up to four derivatives, neglecting boundary terms; in the second, we work on a maximally symmetric background and retain all orders in the Ricci scalar. Although the flow equation depends on the gauge parameters, this dependence can be cancelled order by order once the redundant terms are absorbed by field redefinitions. In the four-derivative approximation, the dependence on the retained gauge parameter drops out of the beta function for Newton's constant to third order in the coupling. On the sphere, the flow becomes independent of both parameters of the general gauge to all curvature orders, and all orders in the coupling. Crucially, this cancellation depends on the choice of regulator. We verify the mechanism explicitly at one loop, where the gauge-dependent fluctuation contributions are cancelled by the ghost sector on shell. The resulting essential flow displays a gauge-independent non-Gaussian fixed point, supporting the interpretation that universal information in quantum gravity is encoded in the on-shell essential sector.

Penning ion traps as quantum platforms have primarily utilized global control and symmetric Dicke states for quantum simulation and sensing experiments. The introduction of local control greatly increases the power of the platform as a quantum simulator but is technically challenging due to the rapid rotation of the ion crystals. Here we use an ultraviolet-compatible spatial light modulator (SLM) to imprint programmable AC Stark shift patterns with different azimuthal symmetries and gradients that co-rotate with the ion crystals, demonstrating localized coherent control of single plane crystals with greater than 100 ions. Comparisons of the measured ion qubit populations with calculations from independent measurements of the applied AC Stark shift patterns show good agreement, validating the technique and providing a path, with a higher format SLM, for parallelizable, coherent individual ion addressing in Penning traps.

We present the design and benchmarking of RP000, a quantum photonic processor capable of encoding a quantum system in the degrees of freedom of single photons, based on standard CMOS-compatible manufacturing processes, and working at room temperature. We benchmark it against machine learning tasks, evaluating three quantum-classical architectures of increasing complexity. Our experimental results and simulations show that RP000 achieves higher accuracy than classical networks of comparable size in multiple use cases. Compared to a superconducting quantum processor, RP000 exhibits superior noise tolerance. These findings demonstrate that RP000 can provide a scalable route toward efficient quantum applications.

This report presents a prototype VPN that uses QKD-derived keys delivered through the ETSI GS QKD 014 API. The VPN encrypts IP traffic with AES-256-GCM, transports ETSI key identifiers in-band, and retrieves matching keys from local KMEs. After validation with a controlled KME simulator, the system was tested on two Jetson Xavier NX devices connected to a LuxQuanta NOVA QKD platform. The experiment successfully transmitted bidirectional real-time audio and video traffic through the VPN for eight continuous hours, demonstrating the feasibility of integrating classical VPN applications with QKD infrastructure through a standardized key-delivery interface.

We present a Bayesian P-spline method for estimating the frequency-dependent cross-spectral density matrix of stationary multivariate time series. The inverse spectral matrix is parametrised through its frequency-varying Cholesky decomposition, which guarantees Hermitian positive definiteness at every frequency. Each real log-diagonal entry and each real and imaginary off-diagonal entry is given an independent penalised B-spline prior that controls smoothness. Inference uses a blocked, coarse-grained Whittle likelihood with safe-Bayes $η$-tempering to stabilise posterior calibration, sampled by the No-U-Turn Sampler from a variational initialisation. On synthetic VAR(2) benchmarks with known ground truth, the method recovers both diagonal and cross-spectral structure, attains near-nominal credible-interval coverage, and achieves a relative integrated squared (Frobenius) error (RISE) that decreases with sample size. We then apply the method to publicly released simulated LISA time-delay interferometry (TDI) data in two noise configurations. In the idealised symmetric case, the full multivariate model and a reduced model that assumes a diagonal AET noise covariance agree to within $\sim10^{-3}$ in RISE. Under realistic noise that is asymmetric across the six Movable Optical Sub-Assemblies (MOSAs), the AET-diagonal assumption fails by more than an order of magnitude in RISE ($\sim\!3.3\!\times\!10^{-2}$ versus $\sim\!10^{-3}$), whereas the full multivariate model recovers the cross-spectral structure.

We investigate the radial motion and tidal forces experienced by neutral test particles in multi-horizon black hole solutions arising from Einstein gravity coupled to nonlinear electrodynamics (NED). Focusing on the three- and four-horizon configurations, we examine how nonlinear electromagnetic corrections modify the causal structure, radial geodesic motion, and tidal-force profiles in comparison with the Schwarzschild and Reissner-Nordstrom (R-N) spacetimes. Our analysis shows that the NED field gives rise to multiple zero crossings in both the radial and angular tidal-force components, leading to successive transitions between stretching and compressive tidal regimes. More importantly, the radial equation of motion contains classically forbidden regions bounded by bounce-back points. For the class of trajectories considered in this work, these forbidden regions prevent particles from entering the spacetime domain where the tidal forces become divergent. In the super-extremal regime admitted by these solutions, the forbidden region may extend beyond the event horizon, preventing particles released from rest at sufficiently large distances from crossing the horizon. We further identify a systematic ordering of the critical charge values associated with the appearance of tidal-force zero crossings, additional horizons, and bounce-back points. For both the three- and four-horizon configurations, these critical values satisfy a hierarchical ordering, indicating that changes in the tidal-force structure precede the corresponding modifications of the horizon configuration. These results demonstrate that nonlinear electrodynamics can substantially modify the classical dynamics of neutral particles in multi-horizon black hole spacetimes through the combined effects of forbidden regions, multiple tidal transitions, and changes in the horizon structure.

With unitary coin operator that is a random matrix drawn from a uniform distribution with respect to the Haar measure, we construct a variant of a discrete quantum random walk using a d-regular undirected connected simple graph. With each step of the walk, the coin operator random matrix is drawn independently from the distribution. The expectation value over the distribution of random unitaries for the associated quantum channel gives a depolarization channel for the coin subspace and resembles the associated classical random walk where the direction a walker steps depends upon the outcome of a fair coin or balanced d-sided dice. Remarkably, despite the fact that the averaged channel depolarizes the coin subspace, measurements in the vertex subspace can be designed that would reveal information about the initial quantum state, even after many iterations of the channel. We illustrate with examples of quantum walks on Cayley graphs of Abelian groups, such as the cycle and hypercube graphs. For Cayley graphs of Abelian groups, the averaged channel is dephasing in the Fourier basis. These quantum walks are examples of bipartite strongly interacting systems, where one subsystem is strongly perturbed, yet information about the initial state in the other subsystem is potentially measurable forever. Due to its decoherence, a quantum random walk with a Haar-random coin would not be useful in search algorithms but could aid in understanding quantum systems with strongly perturbed subsystems.

Conical 2-designs are families of positive operators possessing crucial symmetry properties that allow them to efficiently characterize important quantum measures. However, little is known about their general structures and properties. We propose a construction method of conical 2-designs whose elements are not of equal traces, based on a generalization of equiangular weighted frames to non-projective operators. We provide a detailed analysis of their properties, relations to quantum measurements, and applications, together with examples that go beyond the known classes. This constitutes a major step toward a full characterization of informationally overcomplete quantum measurements.

Within the framework of generalised entropic cosmology and its equivalent $f(Q)$ gravity, we construct inflationary scenarios consistent with recent Atacama Cosmology Telescope (ACT) observations combined with Planck-BAO data. We work with $f(Q)$ gravity and use the reconstruction scheme from which one can obtain an arbitrary cosmological evolution consistent with desirable theoretical or observational considerations. Explicit inflationary cosmologies of the above theory, which pass ACT-Data Release 6 (DR6) data combined with Planck and BAO, are obtained. For $f(Q)$ gravity coupled with scalar, we obtain ACT inflation for large classes of $f(Q)$ via the choice of the scalar potentials. The entropic cosmology based on generalised entropy, which includes most of the known entropies like Tsallis, Rényi, Barrow, etc as a particular case, is investigated. The corresponding FRW equations, depending on the parameters of generalised entropy, are presented. Using the correspondence between the generalised entropy and $f(Q)$ gravity, we obtain generalised entropic inflation, which is consistent with the ACT-DR6. At the next step, the entropic gravity coupled with scalar is considered. It is demonstrated how one can obtain ACT-DR6-consistent inflation for such a theory based on any generalised entropy considered in this paper via the choice of scalar potentials. These findings may point towards the correct unifying entropy of the early universe.

We present the first calculation of single inclusive hadron and jet production at large transverse momentum in lepton-hadron scattering in a joint QCD+QED factorization approach. The scattering cross section is factorized into a convolution of infrared-safe hard coefficient functions with universal lepton distribution functions (LDFs) and parton distribution functions (PDFs) of the colliding lepton and hadron, respectively, together with fragmentation functions (FFs) of the observed hadron (or jet). With joint QCD+QED factorization, the DGLAP-type evolution equations for LDFs, PDFs, and FFs necessarily have evolution kernels calculated in both QCD and QED. We derive a default set of LDFs for our calculations and discuss a strategy to extract universal, non-perturbative LDFs from future data. We present our calculations for single inclusive hadron and/or jet production at the energies of Jefferson Lab and the future Electron-Ion Collider.

Hadronic vacuum polarization (hVP) enters simple atomic systems at a level that is small yet decisive for the precision spectroscopy now underway. We evaluate the hVP contributions to the Lamb shift and the hyperfine splitting (HFS) in ordinary and muonic hydrogen (H and $μ$H) and hydrogen-like helium-3 ions ($^3$He$^+$ and $μ^3$He$^+$), using the dispersive data-driven approach and state-of-the-art empirical parametrizations of the $R$ ratio. At the centre of the analysis is the interplay of recoil and finite-size effects: the recoil corrections that dominate the HFS in muonium (Mu), where both constituents are pointlike, are shown to be suppressed by the nuclear elastic form factors (FFs). Our results for the leading hVP contribution to the Lamb shift agree with the literature within uncertainties. Furthermore, we present a first evaluation of the subleading $O(Z^5α^6)$ hVP-finite-size correction, which is by no means negligible in $μ^3$He$^+$. Our results for the hVP contribution to the HFS deviate significantly from all previous evaluations. For the ground-state HFS, we obtain $2.153(11)~μ$eV in $μ$H and $-15.19(57)~μ$eV in $μ^3$He$^+$, as well as $0.0860(4)~$kHz and $-0.476(17)~$kHz in ordinary H and $^3$He$^+$, respectively. Notably, our result for $μ$H differs from previous evaluations by roughly ten times the experimental precision anticipated by the upcoming CREMA and FAMU measurements.

We establish a localization principle for odd-dimensional supergravity theories containing Chern--Simons interactions. Our construction relies on combining an equivariant extension of the gauge-invariant part of the action with an equivariant completion of the Chern--Simons anomaly form $Φ^{(\text{anom})}$ to form a relative equivariant cohomology class. This leads to a universal fixed point formula for the on-shell action of supersymmetric solutions that is expressed in terms of $Φ^{(\text{anom})}$ and an equivariant Euler class. We illustrate the formalism in $D=11$ and $D=5$ supergravity, with applications including the supersymmetric Casimir energy and superconformal index of M5-branes, as well as black ring and black lens solutions.

The ATLAS collaboration has recently performed a vast scan of the phenomenological Minimal Supersymmetric Standard Model (pMSSM) with a focus on the electroweak-ino sector, and analysed how their Run 2 searches for electroweak production of supersymmetric (SUSY) particles constrain this dataset. All the SLHA files from the scan as well as the constraints from the eight individual searches considered by ATLAS were made publicly available. We use this material to study how well the ATLAS constraints can be reproduced with SModelS v3.0. Moreover, we explore how the picture changes when also including CMS results, and what can be gained by the statistical combination of analyses. Finally, we discuss the part of parameter space with light electroweak-inos that remains valid despite the stringent LHC limits. Our results underscore the need of a broad, multifaceted approach for maximising sensitivity and closing loopholes in the extensive SUSY parameter space.

We consider a well-motivated class of scotogenic models (the "T1-2-A" variant), and carry out a comprehensive reassessment of its prospects regarding charged lepton flavour violating (cLFV) observables. Aiming only at explaining neutrino oscillation data and putting forward a viable dark matter candidate, a thorough exploration of the model's parameter space suggests that one can have sizeable rates for cLFV observables, especially in rare muon transitions. We have further considered the role of parity and time-reversal asymmetries for cLFV 3-body decays, $\ell_α^+ \to \ell_β^+ \ell_γ^+ \ell_δ^-$, which can be potentially studied in association with polarised muon and tau decays. The new set of observables offers further complementarity information on the scotogenic model under consideration, and possible means of testing it.

This thesis presents the first demonstration of a hybrid Cherenkov and scintillation optical detector at a beam dump facility for a proof-of-principle search for MeV-scale axion-like particles (ALPs). The work utilizes the Coherent CAPTAIN-Mills (CCM) experiment, a 10-ton liquid argon detector at Los Alamos National Laboratory instrumented with 200 PMTs. Coating 80% of the PMTs with wavelength-shifting material while leaving 20% uncoated enables enhanced Cherenkov sensitivity through wavelength discrimination and 2 ns timing resolution. This work includes detailed modeling of $^{22}$Na calibration data, providing the first comprehensive characterization of scintillation and Cherenkov light production and propagation in such a detector, including systematic uncertainties. Using the same source, it demonstrates the first event-by-event separation of Cherenkov radiation from sub-MeV electrons in a high light-yield scintillation detector. Additionally, this work develops machine learning-based position reconstruction with ~5 cm resolution and energy reconstruction with ~10% resolution at 1 MeV. Four observables exploiting Cherenkov timing, wavelength sensitivity, directionality, pulse shape, and event topology are combined into a likelihood ratio test statistic to suppress steady-state backgrounds for $\leq10$ MeV ALP-induced events. No significant excess is observed. Nevertheless, this improved background rejection excludes new regions of ALP mass-coupling parameter space at the 90% confidence level compared to a previous CCM analysis, despite a smaller exposure. Finally, this thesis explores liquid argon applications to supernova neutrino physics in DUNE and discusses next-generation large-scale hybrid optical detectors. These results demonstrate the potential of hybrid Cherenkov-scintillation detectors for future weakly interacting physics experiments.

Using a data sample corresponding to an integrated luminosity of 20.3 fb$^{-1}$ collected at center-of-mass energies from 4.18 to 4.95 GeV with the BESIII detector, we observe the process $e^+e^-\toπ^+π^-ηJ/ψ$ with a statistical significance of $6.0 σ$, including systematic uncertainties. The isoscalar partner of the $Z_c(3900)$, denoted $X(3900)$, is searched for in the $ηJ/ψ$ final state, and no significant signal is observed. The upper limits on the product of the Born cross section $σ^{\rm Born}[e^{+}e^{-}\toπ^{+}π^{-} X(3900)$] and the branching fraction $\mathcal{B}[X(3900)\toηJ/ψ]$ are given with various assumptions for the mass and width of the $X(3900)$.

We consider confining gauge theories with a non-vanishing topological angle $θ$, which induces CP-violating interactions among dark pions and dark baryons, with dark matter consisting of dark neutrons. The $θ$ term generates scalar pion--baryon couplings analogous to the CP-violating pion--nucleon interactions of QCD. These interactions give rise to an attractive long-range Yukawa potential mediated by the dark pions, whose strength is proportional to $θ$. Since the dark pions are naturally light pseudo-Goldstone bosons, the resulting force can lead to sizable dark matter self-interactions, providing a simple and theoretically motivated realization of the self-interacting dark matter paradigm. We also investigate the implications for direct detection in scenarios where the dark sector communicates with the Standard Model through a dark photon portal. The $θ$ term induces dark electric dipole moments proportional to $θ$, which couple directly to the electric fields of nuclei and can substantially enhance direct detection rates. We show how the underlying interactions shape the recoil spectra and discuss the possibility of identifying the CP-violating origin of the signal through its time dependence. Finally, we analyze the interplay between dark matter self-interactions and direct detection signals, showing that both are controlled by the same CP-violating parameter. Our results demonstrate that the topological angle $θ$ can play a central role in determining the phenomenology of composite dark matter.

Although the action of diffeomorphisms on pinor (or spinor) fields is geometrically unambiguous, the definition of a natural Lie derivative is more subtle and requires some understanding of the geometry of the space of metrics at a point. In this work we revisit the classical construction of Kosmann and Bourguignon-Gauduchon and develop the corresponding theory in generalised geometry.

The dynamics of Fierz-complete four-quark interactions and its influence on the QCD phase diagram have been investigated within the functional renormalization group approach to QCD at finite temperature and densities. It is found that in the vacuum the pion and sigma channels play the overwhelmingly dominant role, and all the other channels are negligible. However, when it is near the critical end point (CEP), the magnitude of four-quark couplings in other channels increases sizably and they become more and more important. In comparison to the single scalar-pseudoscalar channel of four-quark interactions, the dynamics of Fierz-complete four-quark interactions increases a bit the curvature of the phase boundary, and moves the CEP to location of larger baryon chemical potential and smaller temperature.

In this paper, we investigate five-dimensional $\mathcal{N} = 1$ supersymmetric $SO(2N)$ gauge theories coupled to hypermultiplets in spinor/conjugate spinor representation based on 5-brane web constructions with O5-planes. Based on the topological vertex formalism with O5-plane, we find new expressions for the partition functions for these theories, emphasizing the difference between the case with two spinors and that with one spinor and one conjugate spinor. Through the thermodynamic limit of these partition functions, we derive the Cameral and Seiberg-Witten curves. We show that the difference between the spinor and the conjugate spinor appears as a difference in the boundary conditions of the Seiberg-Witten curves at the orientifold planes.

We present new results for double quarkonium production in (un)polarized hadronic collisions at fixed-target experiments. Our approach incorporates the transverse momentum dependent factorization in combination with the Color-Singlet Model. We present new analytical expressions for the angular structure of the cross section for the $q\bar q$-induced channel, and provide predictions for the unpolarized cross section and transverse single-spin asymmetries for present and future fixed-target experiments at CERN and the LHC.

Given the latest JUNO implication for normal neutrino mass ordering, we parametrize three neutrino masses in a flavor cuboid: $m^{}_1 = m^{}_0 \sinξ$, $m^{}_2 = m^{}_0 \cosξ\sinζ$ and $m^{}_3 = m^{}_0 \cosξ\cosζ$. We find that this cuboid is able to accommodate both neutrino mass degeneracy and tribimaximal flavor mixing in its cubic limit with $ξ^{}_* = \arctan\left(1/\sqrt{2}\right) \simeq 35.26^\circ$ and $ζ^{}_* = 45^\circ$. Assuming $θ^{}_{12} = ξ$ and $θ^{}_{23} = ζ$ for the two large angles of neutrino oscillations and expanding them around $ξ^{}_*$ and $ζ^{}_*$, we propose a viable ansatz which predicts a normal but nearly degenerate neutrino mass spectrum and a nearly tribimaximal neutrino mixing pattern. Testing the achieved correlation among $ξ^{}_* - θ^{}_{12}$, $ζ^{}_* - θ^{}_{23}$ and $Δm^2_{21}/Δm^2_{31}$ will provide a smoking gun for the validity of this ansatz.

Understanding the internal structure of the pion-particularly the energy-momentum distributions of quarks and gluons and the internal mechanical properties encoded in its gravitational form factors-is a fundamental challenge in quantum chromodynamics (QCD). In this work, we study the gravitational form factors using light-front QCD (LFQCD), combined with the holographic QCD. Our main innovation is the introduction of an effective light-front wave function, with its five-dimensional component obtained from holographic QCD, which is then employed, within the light-front QCD framework, to calculate the pion's gravitational form factors $A(Q^2)$ and $D(Q^2)$ as well as its radius. Our computed pion gravitational form factors show good agreement with lattice QCD results, providing nontrivial support for the viability of our phenomenological model.

We reassess existing limits and derive new constraints on heavy axion-like particle (ALP) coupling to photons using gamma-ray bursts (GRBs). ALPs can be produced in the hot dense fireball plasma during the initial stage of GRB outflow, thus potentially disrupting the primary fireball and altering the GRB luminosity. We consider the ALP production rate for various GRB parameters in two different energy injection scenarios of GRB fireball formation, and point out that ALP production is less efficient than previously assumed unless a GRB event is exceptionally energetic. We update the existing energy loss bounds using more realistic GRB parameters. We also point out that in the region of parameter space previously constrained by GRB luminosity criterion, ALP production turns out to be still efficient enough to form a secondary fireball via ALP decay to two photons and their subsequent annihilation to electron-positron pair. This secondary fireball reprocesses the gamma-rays from heavy ALP decay into $X$-rays, emitted isotropically from its surface, thus allowing us to probe $\mathcal{O}(100~\mathrm{MeV})$-scale ALPs indirectly using $X$-ray (or future MeV gamma-ray) telescopes, not necessarily directed toward the GRB jet itself. We show that the future point-source sensitivity of $X$-ray and MeV gamma-ray telescopes may allow us to constrain new ALP parameter space.

The mass independent energy reconstruction of cosmic rays is crucial for understanding their origin, acceleration, and propagation. Precise measurement of the primary energy can also lead to better mass classification and could enable energy dependent anisotropy maps for individual elements. The GRAPES-3 experiment located in Ooty consisting of 400 scintillator detector array placed 8 m apart covering an area of 25000 m$^2$ with a dedicated muon detector made of 3712 proportional counters, is designed to do these kinds of measurements. Previously electron size calibration curves have been used to find primary energy in the GRAPES-3 data analysis framework however significantly better precision can be established using graph neural network. Thus, in this work we have implemented a modular and dynamic GNN based reconstruction algorithm that automates feature mapping. We demonstrate how the model is learning by studying its latent space and show that scaling the metric in the latent space can lead to further improvements in response resolution. Fine-tuned strategies are presented and a thorough comparison of the reconstructed energy and bias is done for different fine-tuned models along with studying the resolution variation for different mass groups and shower age.

The GRAPES-3 experiment is a ground-based extensive air shower array which consists of approximately $400$ closely packed plastic scintillator detectors and a large area muon telescope. Estimating the number of associated muons created in an air shower is crucial to understand the properties of primary cosmic rays. The GRAPES-3 muon telescope (G3MT) records these secondary muons, however, the punch-through hadrons can introduce background noise. This study aims to develop a machine learning pipeline to distinguish the tracks of secondary muons and hadrons at G3MT. We have used CORSIKA-simulated proton showers having energy in the range 100-158 TeV as an input for a Geant4-based detector simulation to analyze the signatures of both type of particles. Initially, single-particle classification was performed using decision trees, random forests, neural networks, and XGBoost, with XGBoost achieving the highest accuracy of 88.7\%. For multiparticle classification, we modelled Graph Neural Networks (GNNs) where each event was represented as a graph with detector hits as nodes. A GNN with edge convolution layers was developed to classify each node as a muon or hadron hit. Following this, a deep learning regression model using Dynamic Reduction Network was developed to estimate the number of particles and muons striking G3MT simultaneously. Details of the analysis and results of the multiparticle classification task will be presented.

We show that Airy function partition functions for M2-brane theories may be derived from relative equivariant localization of quantum M-theory. The eleven-dimensional Chern--Simons coupling gives the cubic term in the grand potential, while the $X_8$ correction gives the charge shift. Fixing the M2-brane charge turns the localized M-theory path integral into an Airy integral. In this way we derive the ABJM result, its toric Calabi--Yau generalizations, and gravitational blocks for black holes and other spacetimes, up to the prefactor and non-perturbative corrections.

Understanding the QCD phase structure and the possible existence of a critical point remains one of the central goals of the heavy-ion program at RHIC. In this proceeding, we present recent STAR results across multiple observables that probe different aspects of the hot and dense matter created in Au+Au collisions. These include two-particle transverse momentum correlations of mean transverse momentum, net-proton cumulants up to fourth order, identical-pion femtoscopy, and baryon-strangeness correlations. We also discuss femtoscopic measurements of baryon-baryon pairs, which provide insight into hyperon-nucleon and hyperon-hyperon interactions and the possible formation of strange dibaryon states. Together, these results provide complementary probes of the system's evolution across a wide collision-energy range (sqrt(sNN) = 3-200 GeV), offering new constraints on the QCD equation of state and the location of the QCD critical point.

For three-flavor neutrino oscillations in constant-density matter, the Cayley--Hamilton theorem forces the evolution operator into a quadratic polynomial in $\hat{H}$, with coefficients determined by the three real eigenvalues through a Vandermonde system we write out explicitly. The eigenvalues follow from Cardano's trigonometric formula, recovering the Zaglauer--Schwarzer expressions. The Denton--Minakata--Parke (DMP) approximation achieves fractional accuracy better than $10^{-4}$ because its $1$--$3$ rotation is a resummation: it removes the near-degeneracy that makes the naive expansion diverge at $\hat{A}\to 1$, replacing the unbounded $(1-\hat{A})^{-1}$ with an effective parameter $ε_0\lesssim 0.015$ bounded uniformly in energy. A density-matrix treatment with a Lindblad term handles open-system decoherence and wave-packet effects in the same language; matter-dressed coherence lengths satisfy $L/L^{ij}_{\rm coh}\sim 10^{-3}$--$10^{-2}$ for terrestrial baselines. The CP asymmetry $\mathcal{A}_{\rm CP}(ν_μ\toν_e)$ is split into genuine and matter-induced fake contributions. Closed-form Jacobians in the NuFIT~6.0 parameter basis feed Monte Carlo and linearized uncertainty-propagation schemes, the latter validated against a Feldman--Cousins profile-likelihood mapping near physical boundaries. The Denton--Parke NuFast-LBL algorithm [Phys.\ Rev.\ D {\bf 110}, 073005 (2024)] remains the tool of choice for production fits; the analytic expressions here supply what iterative solvers cannot -- parameter continuity, transparent limits, and Jacobians in closed form.

We calculate the full set of BSM contributions to the inclusive non-leptonic $B$-meson lifetime from charmless final states within the framework of the Weak Effective Theory up to next-to-leading order in QCD and to leading power in the heavy-quark expansion. We do so by computing cut diagrams and the corresponding phase space integrals. This involves calculating current-current, penguin-penguin, penguin-dipole and dipole-dipole diagrams with two-, three- and four-particle cuts. We describe the technical difficulties of the computation. We then discuss how our results can be used to constrain new physics scenarios related to the anomalies observed in semi-leptonic and charmless non-leptonic decays.

We investigate the holographic superfluid dynamics subjected to external rotation on a spherical geometry. Through a linear perturbation analysis, we identify several dynamically unstable phases in the phase diagram, each characterized by distinct unstable modes. Employing fully nonlinear numerical simulations, we further demonstrate that these unstable modes generically drive the system into vortex-antivortex configurations with definite winding numbers, determined by the symmetry of the corresponding unstable modes.

The $γ$-ray sky, as observed by the Fermi Large Area Telescope, contains a significant number of unassociated sources that may point to new astrophysical populations or more exotic phenomena. Machine-learning methods are widely used for source classification and searches for new physics, but most existing approaches rely on fully supervised training and therefore on explicit signal models. We explore weakly supervised classification as a less model-dependent strategy for analysing $γ$-ray source spectra. In a background-versus-mixture setup, classifiers are trained on samples with different signal admixtures rather than on fully labelled signal and background events. We study three representative scenarios: pulsar-active galactic nuclei separation as a controlled benchmark, the identification of dark-matter subhalos, and spectral irregularities induced by axion-photon oscillations. In each case we investigate the impact of signal fraction and sample composition on classification performance. Our results show that weak supervision can identify anomalous or signal-like subsets of data while reducing the reliance on detailed signal templates during training. In favourable cases, the method approaches the performance of fully supervised classifiers, while remaining applicable in situations where the signal model is uncertain or only partially specified. Weakly supervised learning therefore provides a complementary candidate-selection and anomaly-ranking strategy for $γ$-ray data analysis and searches for new phenomena.

We construct a simple equation of state of strongly interacting matter at zero chemical potentials that provides a unified description of lattice QCD thermodynamics in terms of hadronic and partonic degrees of freedom. The hadronic phase is described by the quantum van der Waals hadron resonance gas, extended by excluded-volume repulsion between mesons, while the quark-gluon plasma is modeled as an ideal gas of quarks and gluons supplemented with a phenomenological interaction term proportional to $T^3$. The two regimes are connected by a smooth crossover switching function. The three model parameters - the meson hard-core radius, the strength of the partonic interaction term, and the switching temperature - are determined from a fit to lattice QCD results for the trace anomaly. The resulting equation of state reproduces the lattice data on the pressure, entropy density, energy density, and speed of sound in the temperature range $T=100$-$500$ MeV. The fit yields a meson hard-core radius $r_M \simeq 0.2$ fm, a partonic interaction scale $A \simeq 600$ MeV, and a switching temperature $T_0 \simeq 216$ MeV, substantially exceeding both the pseudocritical temperature of the QCD chiral crossover and the chemical freeze-out temperature. This finding suggests that the transition from hadronic to partonic degrees of freedom is considerably more gradual than indicated by the chiral pseudocritical temperature alone, with hadronic states remaining an important component of strongly interacting matter up to temperatures of about 250 MeV, well above the QCD chiral crossover.

This paper presents diurnal measurements of the single-electron count rate in the PHELEX experiment using multi-cathode counters over the period from 2023 to 2026. The results obtained are interpreted as an indication of a possible effect from dark photons with masses ranging from 10 to 40 eV. The correlation of the measured rates in counters with Fe and Al cathodes suggests that the effect does not depend on the magnetic properties of the cathode material. This can be interpreted as an argument against many dark matter models related to magnetic moments. The result, if confirmed in an independent experiment, could narrow the range of possible candidates to those that interact primarily through an electric (or longitudinal) field, insensitive to spin. A dark photon with kinetic mixing fits this description ideally.

In this paper, we derive approximate expressions for the CP phase $δ$ in the CKM matrix by a perturbative singular value decomposition for hierarchical quark mass matrices $m_{q}$. The diagonalization is achieved through a seesaw-like procedure in which the heavier generations are successively integrated out, naturally leading to mixing matrices expressed in terms of the mass matrix and its inverse $m_{q}^{-1}$. As a result, in the basis where the up-type quark mass matrix is diagonal, $δ$ is reduced to the fourth-order invariant $δ\simeq \arg [ - m^{-1}_{d12} m_{d23}^{} / m^{-1}_{d11} m_{d13}^{} ]$, constructed from the down-type mass matrix and its inverse. Furthermore, in the Kobayashi--Maskawa parametrization, the CP phase is likewise expressed in the same basis as the invariant $δ_{\rm KM} \simeq \arg [ m^{-1}_{d13} m_{d33}^{} / m^{-1}_{d11} m_{d13}^{} ] \simeq π/2$.

These lecture notes are meant as a pedagogic guide to cosmological inflation and the early epochs thereafter. Inflation explains how the seeds for density perturbations, which evolved into the largest structures in our universe, could have formed during a period of exponential expansion. Apart from density perturbations, also tensor perturbations are generated, which may be observed as gravitational waves. The formalism is developed through explicit computations, paying attention to general-relativistic gauge invariance, and to thermalization (the mechanism that converts part of the energy density driving exponential expansion into the conventional hot big bang). For the steps best handled numerically or computer-algebraically, simple python scripts are provided. We aim at an unassuming style, hopefully accessible to students of theoretical high-energy physics.

The Dirac or Majorana nature of neutrinos remains one of the most fundamental open problems in particle physics. A natural intermediate scenario arises when small lepton-number-violating Majorana mass terms break the exact Dirac symmetry. These cause mass eigenstates to form nearly degenerate pairs, each separated by a small mass splitting. We study this quasi-Dirac scenario within a five-neutrino framework, extending the Standard Model by two right-handed neutrinos, which form a nearly degenerate sterile pair alongside the three active states. This extended mixing structure introduces additional phenomenological parameters, including new mixing angles, mass splittings, and CP phases. We use appearance and disappearance data from NO$ν$A and T2K to constrain the active-sterile mixing angles $θ_{sa}$ and the mass splitting within the sterile pair $Δm_{54}^2$ and forecast the sensitivity achievable at DUNE. We analyse how these parameters modify CP-violating observables relative to the standard three-neutrino case, comparing predicted event rates across both mass orderings and as a function of $δ_{CP}$.

Higgs Branches of 5d SCFTs are hyperkähler cones, symplectic varieties with interesting nested singularities governing the structure of possible Higgs branch RG flows. The main purpose of this note is to investigate 5d Higgs Branches via a geometric route: we study 5d SCFTs geometrically engineered from M-theory on non-compact Calabi-Yau threefolds, with non-isolated singularities. We match the (dynamical) complex structure deformations of the threefold with the symplectic leaves in the Higgs branch of the 5d SCFT, constructing its stratification purely relying on geometry. Consistency checks for our proposal are presented exploiting magnetic quivers.

We review existing and construct new two-dimensional field theories that exhibit BMS$_3$ symmetry, with and without central extensions. These include interacting electric, magnetic and canonical BMS$_3$ scalar theories, as well as couplings between the electric and magnetic sectors. We provide a careful analysis of boundary contributions at the corner points $u\rightarrow\pm\infty$, determine the counterterms and boundary conditions required by BMS$_3$ invariance, and study the corresponding variational principles. Furthermore, we introduce external sources and derive the associated flux-balance laws. Within the framework of flat-space holography, we show that a simple free electric BMS$_3$ model with the appropriate central charge reproduces the monodromy classification of three-dimensional Einstein gravity. Finally, we investigate the flat-space limits of both AdS$_3$ and dS$_3$, whose boundary dynamics are described by Carrollian limits of Liouville and Euclidean Liouville theory, respectively. Although the resulting BMS$_3$ theories possess isomorphic spectra, they differ in the signs of the supertranslation charges and the central charge. This suggests that the flat-space limits of AdS$_3$ and dS$_3$ provide complementary realizations of flat-space holography.

We study Collins azimuthal asymmetries for pion-in-jet production in polarized proton-proton and lepton-proton collisions. We adopt a hybrid transverse momentum dependent approach, with a collinear configuration for the initial state, and employ the transversity and Collins fragmentation functions extracted from semi-inclusive deep inelastic scattering and $e^+e^-$ annihilation data. After recalling the good description of the STAR data in $pp$ collisions, which supports the universality of the Collins function, we present predictions for Electron-Ion Collider kinematics, both at leading order and by including the quasireal photon exchange in the Weizsäcker-Williams approximation. This contribution is sizable but does not spoil the dominance of quark-initiated channels. This implies that $\ell p$ processes allow for a clearer access to the transversity distribution, including its sea-quark component.

A new field, \emph{Particle Cosmology}, emerged at the interface of elementary particle physics and cosmology in the last two decades of the 20th century. In this chapter of the \emph{Proceedings of the 4th International Symposium on the History of Particle Physics, the Standard Model and Beyond} I will review the development of Particle Cosmology in the period 1980-2000 centered on two events: a conference, \emph{Inner-Space/Outer-Space} held at Fermilab in May 1984 \cite{kolb1986innerspace} and the 1994 Snowmass Summer Study \emph{Particle and Nuclear Astrophysics in the Next Millennium} \cite{kolb1995particle}.

This document summarizes the contributions to the 5th DPHEP workshop March 5-6, 2026, CERN, and reflects the advancements since 2024, as well as future milestones and tendencies. Impressive progress in HEP data preservation is observed. Legacy data revival was showcased through successful reanalysis of archived data using contemporary methods, demonstrating the long-term scientific value of preservation. Sustainability challenges were noted, emphasizing the need for long-term funding and institutional support to maintain data preservation infrastructure, particularly for legacy experiments transitioning to archival modes. Innovative transverse projects display constant progress towards common technologies for a robust and transferrable DP. In particular, there is a clear shift toward automation, with increasing use of AI and machine learning for data curation, metadata extraction, and workflow optimization. Open science momentum is growing, with wider adoption of FAIR principles and open data policies, and experiments committing to public releases.

We present a technically natural microscopic realization of apparent phantom dark energy based on an axion coupled to $N$ copies of two-flavor dark QCD related by a $\mathbb{Z}_N$ exchange symmetry. The symmetry exponentially suppresses the axion vacuum potential, naturally generating the dark-energy scale, while reheating into a single dark sector induces a controlled breaking of $\mathbb{Z}_N$ that simultaneously restores the physical axion periodicity, sets the dark-matter abundance, and drives the late-time dark-energy dynamics. The selected sector contains dark-pion dark matter, whose finite density initially traps the axion away from its vacuum minimum. As the Universe expands and the dark-pion density redshifts away, the axion is released and rolls on the reheating-induced vacuum potential, generating an effective phantom crossing without tuned cancellations. We identify a viable parameter region that simultaneously reproduces the observed dark-matter relic abundance and dark-energy scale, satisfies cosmological and astrophysical constraints, and qualitatively reproduces the DESI preference for an evolving dark-energy equation of state.

We present an observable-driven study of leptophilic gauge interactions mediated by a neutral vector boson $Z_\ell$ at future $e^+e^-$ colliders. Focusing on the process $e^+e^- \to μ^+μ^-$, we investigate the sensitivity of angular observables to off-shell leptophilic interactions in regimes where interference effects dominate over resonant rate enhancements. The analysis is centered on the forward-backward asymmetry, which provides a robust probe of interference-induced angular deformations and retains sensitivity well beyond the resonance region. To characterize the relative angular response independently of the overall rate modification, we additionally introduce the interference-normalized quantity $\mathcal{I}_{\rm FB}$ as a diagnostic observable. Realistic collider effects, including initial-state radiation, beamstrahlung, and finite energy resolution, are consistently incorporated. Using benchmark configurations for FCC-ee, CEPC, ILC, and CLIC, we derive projected sensitivities in the $(M_{Z_\ell}, g_\ell)$ plane. The results demonstrate the complementarity between high-luminosity circular colliders and high-energy linear colliders, reflecting the interplay between precision measurements and the energy enhancement of interference effects. Overall, the present framework provides an experimentally grounded approach for probing leptophilic gauge interactions through interference-sensitive angular observables at future lepton colliders.

The Ding-Iohara-Miki (DIM) algebra (quantum toroidal algebra of $\widehat{gl_1}$) is related to a wide class of quantum many-particle integrable systems, a typical one being the Ruijsenaars trigonometric system with eigenfunctions that are a triad formed by the Noumi-Shiraishi power series, the Macdonald polynomials, and the Baker-Akhiezer multivariable function. Other integrable systems of this type are obtained from the Ruijsenaars system by twisting. At the same time, the Ruijsenaars Hamiltonians are directly related to the Hamiltonians of another quantum integrable system, the Cherednik DAHA Hamiltonians of type $A$ (and their twisted versions in the twisted case), due to the correspondence between the DIM algebra and the spherical DAHA. The eigenfunctions of the DAHA Hamiltonians are non-symmetric Macdonald polynomials. Similarly, there is a class of integrable DAHA Hamiltonians of type $C^\vee C$, the spherical version of which, in turn, allows one to generate integrable Koornwinder Hamiltonians. The eigenfunctions of these two integrable systems are, respectively, non-symmetric and symmetric Koornwinder polynomials, which are our main interest in this paper. Here we consider the cases of both type $A$ and type $C^\vee C$ systems, since they are sufficiently similar, and point out important distinctions between them.

We present a study of femtoscopic correlation functions for $K^{-}d$ and $K^{+}d$ pairs, and compare our results with recent measurements by the ALICE Collaboration in both Pb-Pb and high-multiplicity $pp$ collisions. The kaon-deuteron wave functions are derived from scattering amplitudes using a unitarized chiral effective theory model describing the elementary interactions of $K^{\pm}$ mesons with nucleons. We then evaluate the $K^{\pm}d$ strong scattering amplitudes by solving the Faddeev equations within two distinct frameworks: the Impulse Approximation and the Fixed Center Approximation, which accounts for multiple scatterings. We also incorporate the long-range Coulomb effects between the kaon and the deuteron. We show that the $K^{-}d$ correlation function exhibits large sensitivity to both the size of the emitting source and the relative momentum of the pair, being heavily influenced by rescattering processes. In contrast, the $K^{+}d$ correlation function is dominated by the weakly repulsive $K^{+}N$ interaction, showing deviations from purely Coulombic behavior only at small emission source sizes. Our predictions are in agreement with the ALICE experimental data, and also with the energy-shift and width of the $1s$ level of the kaonic deuterium preliminary results from the SIDDHARTA 2 Collaboration.

We investigate the Casimir response of a massive scalar field in a torsionless helicoidal spacetime with Levi-Civita connection. The background is ultrastatic and curved, with scalar curvature \(R=-2Ω^2\), and its off-diagonal metric component induces a geometric coupling between angular and axial quantum numbers. We show that individual modes exhibit a linear helicoidal splitting, whereas the linear contribution cancels in the nonchiral vacuum sum. The leading Casimir correction is therefore quadratic in the twist and defines a helicoidal vacuum susceptibility after local ultraviolet subtractions. For a cylindrical Dirichlet cavity, we compute this scheme-defined susceptibility and the associated correction to the radial Casimir force. The results identify torsionless helicoidal geometry as a controlled setting in which mode-level chirality produces a finite quadratic response of vacuum fluctuations.

Soft de Sitter Effective Theory is a well-motivated candidate for the correct effective late-time description of equal-time correlation functions in de Sitter space. In this work, we study its application to theories that enjoy classical conformal invariance, using the conformally-coupled $φ^4$-theory as a toy model. While quantum effects generate non-trivial late-time dynamics in such models, we argue that it is not described by the standard construction of the effective theory as discussed thus far in the literature. We show that the tree-level matching of the trispectrum onto the effective theory does not fit into the expected power-counting scheme, and we contrast it with the matching of the tree-level bispectrum in the conformally-coupled $φ^3$-theory, where it works consistently. We then propose a prescription to identify the leading superhorizon degrees of freedom in such theories, which should serve as the starting point for the construction of their late-time effective description. The interpretation of logarithms of the form $\log(-kη)$ in this context is briefly discussed.

Wilson loop operators in Chern-Simons theory have revealed profound links between quantum field theory, the fractional quantum Hall effect, topology, conformal field theory, and string theory. In Chern-Simons theories with charge conjugation symmetry, we construct a new class of observables: codimension-two monodromy defects around which fields return to themselves up to charge conjugation. Whereas Wilson loops are labeled by integrable representations of an untwisted affine Lie algebra, monodromy defects are labeled by those of the corresponding twisted affine algebra. The modular and fusion data of these two algebras determine the exact correlation functions of Wilson lines and monodromy defects, which together furnish a $\mathbb{Z}_2$-crossed braided tensor category. The spectrum of line defects in Chern-Simons theory thus gives a physical realization of every algebra in Kac's classification of affine Lie algebras: untwisted for Wilson loops, twisted for monodromy defects. We also determine the exact 't Hooft expansion of monodromy defects in $SU(N)_k$ Chern-Simons theory and identify their holographic duals in topological string theory. The insertion of the lightest monodromy defect has a striking effect: it replaces the resolved conifold background of the Gopakumar-Vafa duality by a specific orientifold of the resolved conifold, transmuting the dual theory of oriented strings into one of unoriented strings. Each excited monodromy defect is then realized as a collection of branes in the orientifold background, with the brane content determined by the representation of the twisted affine algebra that labels the defect.

Quantum anomalies provide a bridge between ultraviolet properties of a theory and its infrared sector. We study how this connection appears in axial-charge-flow observables. In the simplest example, an axial-charge detector probes the fermionic cut of the anomalous triangle and resolves its infrared content as an angular distribution. The massless limit does not commute with the angular integration: a contribution suppressed at fixed angle collapses onto the two beam-collinear directions while retaining the finite integrated sum rule fixed by the axial anomaly. We then replace the axial-charge detector by higher-spin helicity (zilch) detectors and study a family of axial-anomaly-controlled energy-weighted sum rules for the corresponding fluxes. We further show that the same singular localization mechanism and finite zilch-flux sum rules persist in the mixed axial-gravitational channel. We briefly comment on extensions to more general states and multipoint correlators.

We review recent progress in understanding certain aspects of condensed matter systems from a high energy theory perspective. We discuss effective field theories that describe collective bulk and localized excitations in a variety of solid and fluid systems. Particular emphasis is placed on the role played by spacetime symmetries and their spontaneous breaking. The resulting Goldstone dynamics can be seen as underlying a wide variety of phenomena. We attempt to bridge the language gap between subfields while underscoring the numerous conceptual similarities.

Five-dimensional $\mathcal{N}=1$ superconformal field theories admit a rich variety of generalised global symmetries, including higher-form symmetries, 2-group symmetries, and their 't$~$Hooft anomalies. We show that this data can be extracted directly from the blow-up equations that govern the instanton partition functions of such theories on the $Ω$-background. The central object is the classical prefactor $\exp(-V_n)$ that weights each magnetic flux on the blown-up geometry: evaluated on a background for the electric 1-form symmetry, the fractional parts of its exponents encode the cubic self-anomaly of the 1-form symmetry, as well as its mixed anomalies with the instanton, flavour, gravitational, and $\mathrm{SU}(2)_R$ symmetries. Combined with the faithful continuous global symmetry of the ultraviolet fixed point, which we determine from the superconformal index, the same data decides whether the theory possesses a 2-group symmetry or a mixed 't$~$Hooft anomaly. We illustrate the method in gauge theories, including $\mathrm{SU}(4)$ and $\mathrm{USp}(4)$ with antisymmetric hypermultiplets, and $\mathrm{Spin}(7)$ and $\mathrm{Spin}(8)$ with vector hypermultiplets, as well as in several families of non-Lagrangian theories, for which we obtain new results: notably, the effective prepotentials of the $B_N$ and $B_N^{(1,2,3)}$ families, the cubic 1-form anomalies of the rank-two theories $\mathbb{P}^2\cup\mathbb{F}_3$ and $\mathbb{P}^2\cup\mathbb{F}_6$, and several mixed anomalies involving the 1-form and flavour symmetries.

Berends--Giele currents have recently attracted renewed attention across the amplitude community, appearing in new contexts ranging from gravitational amplitudes to cosmological correlators and string theory. They can be derived from first principles using the perturbiner method, a thirty-year-old framework that organizes tree-level scattering data in the form of classical multi-particle solutions to field theory equations of motion. After its rediscovery about a decade ago, it has been systematically studied and applied in many different contexts. This review aims to provide a pedagogical introduction to the method, covering its basic formulation across a broad (but certainly not exhaustive) range of models and progressing to some of the most recent findings in the literature. This includes several unpublished results, many of which represent tacit knowledge of the community that has not previously appeared in print.

We derive positivity bounds on the complete dimension-eight effective field theory of photons and a massless dark photon. The mixed gauge sector contains twelve CP-even Wilson coefficients and an enlarged helicity-amplitude structure. Using a modified forward-limit dispersion relation, we analytically obtain non-trivial linear and non-linear elastic positivity constraints that define a spectrahedral geometry. We analyze implications of these bounds on non-forward amplitudes and discuss where kinetic-mixing and dark-axion-portal UV completions populate this geometry.

I construct a microscopic ADHM/chiral-Wishart representative of the reduced charge-sector coefficient $W_ν[b]$ that enters the Type-IIB axion--dilaton wormhole partition function $Z_{\rm wh}(θ;b)$. I fix the axion-charge sector, equivalently the form-field-flux sector, which plays the exact same structural role as a fixed-topology sector in QCD: it supplies the domain where I compute the microscopic coefficient before forming the final theta sum. At $E=0$, the axion--dilaton radial family reaches its BPS instanton endpoint. After I impose the Hamiltonian constraint, gauge quotient, charge-sector boundary condition, and collective zero-mode quotient, the physical quadratic fluctuation operator at that endpoint becomes a positive adjoint square. Its non-zero spectrum is therefore a squared singular-value spectrum, and its microscopic endpoint is a Laguerre/chiral hard edge. The D(-1)/D3 super-ADHM collective-coordinate integral supplies the Type-IIB microscopic representative of this same coefficient. In the large$N$ result of Dorey et al., this super-ADHM measure becomes the $k$-D-instanton measure on $AdS_5\times S^5$, multiplied by a centered zero-dimensional supersymmetric matrix-model factor. The chiral/Wishart ensemble forms the hard-edge limit of the rectangular block inside this super-ADHM integral. Fermionic ADHM variables and supergravity fermions remain part of the coefficient: in protected sectors they cancel paired non-zero modes, impose zero-mode saturation, and determine which reduction data \(b\) give a non-vanishing $W_ν[b]$.

Motivated by recent and forthcoming experimental progress, we provide updated predictions for the total decay rates of $b$-baryons, their lifetime ratios and their lifetimes normalised to that of the $B^0_d$ meson within the framework of the heavy quark expansion (HQE). We include, for the first time, next-to-next-to-leading-order QCD corrections to the free $b$-quark decay, which significantly reduce the theoretical uncertainties in the total decay rates. In addition, we also include, for the first time, the complete next-to-leading-order QCD corrections to the dimension-five contributions. While these corrections have only a minor effect on the total decay rates, they induce a sizeable shift in the lifetime ratios, improving the agreement between HQE predictions and experimental data. Overall, we find excellent agreement between the HQE predictions and current experimental measurements for both total decay rates and lifetime ratios within the quoted uncertainties.

We present a study of Lorentz invariance violation (LIV) in neutrino oscillations, with primary emphasis on the interplay between CPT-violating and CPT-conserving Standard Model Extension (SME) coefficients. We find that $a_{ee}$ and $a_{ττ}$ dominate among the diagonal LIV coefficients, whereas $a_{eμ}$ and $a_{eτ}$ provide the most significant off-diagonal contributions. In contrast, the corresponding $c_{αβ}$ coefficients have comparatively sub-leading effects. Taking DUNE as a representative long-baseline experiment, we show that the sensitivity to CP violation is modified by LIV in a parameter-specific manner. We find correlations between off-diagonal LIV parameters, and non-trivial dependence on the new phases. In absence of LIV, DUNE is expected to establish CP violation at $5σ$ for a limited fraction of $δ_{CP}$ values. We find that the presence of $a_{ee}$, $a_{ττ}$, $a_{eμ}$ and $a_{eτ}$ weakens the CP discovery potential, reducing the achievable significance to below $3σ$ over a substantial fraction of $δ_{\rm CP}$. The independent impact of $c_{αβ}$ terms also results in suppression, but in combination with $a_{αβ}$, they introduce degeneracies, complicating the extraction of $δ_{CP}$. This generally results in a deterioration of CP violation sensitivities below $5σ$. These findings emphasize the importance of incorporating LIV effects in precision oscillation studies at upcoming long-baseline experiments.

We have investigated a fermionic system from the perspective of an effective quantum field theory defined on a nontrivial topology in the presence of an external magnetic field. Using the proper-time representation, we obtained one-loop expressions for the corresponding effective Lagrangian, taking into account all Landau levels in the propagator. We also computed the significant boundary-induced contributions to the system's magnetization. To verify the reliability of our results, we examined the limit of zero temperature and infinite spatial extent, which correctly reproduces the celebrated Schwinger result.

We study in detail a recent proposal stating that M-theory observables arising from the quantisation of M2-branes are naturally computed in a fixed $A_3$ ensemble. In a holographic setting this implies that in a semiclassical approximation 11d bulk observables in an asymptotically AdS$_4$ background are computed in a grand canonical ensemble of fixed chemical potential $μ$ conjugate to the integer $N$ determining the rank of the gauge group in the dual CFT. We provide detailed precision tests of holography in the fixed $μ$ ensemble by focusing on the two leading terms in the derivative expansion of the 11d theory. In addition, we study one-loop logarithmic corrections which we compute in detail. For asymptotically AdS$_4\times S^7/\mathbf{Z}_k$ 11d backgrounds our results are in agreement with supersymmetric localisation calculations in the dual ABJM SCFT. Moreover, we employ the logarithmic corrections in the $μ$ ensemble to determine the measure of the integral Laplace transform and use this to compute dual SCFT supersymmetric partition functions, like the superconformal index and the squashed $S^3$ partition function, to all orders in the perturbative $1/N$ expansion in terms of Airy functions in harmony with results and conjectures in the dual SCFT.

Within the framework of Becchi-Rouet-Stora-Tyutin (BRST) formalism, we invoke the beauty of the basic canonical (anti)commutators to prove the nilpotency property of the Noether (anti-)BRST charges for the D-dimensional BRST-quantized version of the free Abelian 2-form gauge theory which is endowed with a non-trivial Curci-Ferrari (CF) type restriction. In this proof, we use only the theoretical strength of the Gauss divergence theorem. We demonstrate that, under the off-shell nilpotent (anti-)BRST symmetry transformations, the Noether conserved (anti-)BRST charges are not invariant and they are also found to be not off-shell nilpotent (if we exploit the standard relationship between the continuous symmetry transformations and their generators as the Noether conserved charges). However, these charges become (anti-)BRST invariant and nilpotent if we use (i) the appropriate equations of motion at suitable places, and (ii) the Gauss divergence theorem. We derive the consistently modified versions of the Noether (anti-)BRST charges which are invariant under the off-shell nilpotent (anti-)BRST transformations. We prove the (anti-)BRST invariance of these modified versions of charges by using the basic canonical (anti)commutators, too. We discuss the physicality criteria w.r.t. (i) the conserved Noether (anti-)BRST charges, and (ii) the modified (anti-)BRST invariant versions of the Noether (anti-)BRST charges. We prove the superiority of the latter over the former (in view of the consistency with the Dirac quantization conditions for the gauge theories).

{We study low-energy photon propagation in a constant magnetic field within the one-loop Heisenberg--Euler theory, retaining the refractive-index normalization $γ_s$ without expansion. Here ``finite-field'' denotes exact dependence on $B/B_{\rm cr}$ within the one-loop, constant-field approximation. The resulting birefringence is propagated into magnetar polarization transport. In a centered-dipole model, the polarization-limiting radius is unchanged to better than $10^{-12}$ because mode decoupling occurs at $\sim10^2R_{\rm NS}$, where $B\ll B_{\rm cr}$. Near the surface, however, the weak-field Cotton--Mouton expression overestimates the accumulated birefringent phase by up to a factor $2.9$ at $10^{15}$~G. At the plasma--vacuum resonance, finite-field corrections reduce the resonance density by $32\%$ and raise the adiabatic conversion energy by $14\%$ for 1E~1547.0$-$5408; the corresponding changes are factors $2.6$ and $1.37$ for 1RXS~J1708$-$4009, and factors $9.7$ and $2.13$ for SGR~1806$-$20, the latter controlled by the strong-field asymptote. The resummed one-loop parallel-mode magnetic response remains positive and develops a broad maximum near $17B_{\rm cr}$. The strictly truncated $\mathcal O(α)$ response is monotonic; therefore the maximum is a structural prediction of the resummed one-loop constitutive model, while its detailed profile and precise location require higher-loop validation. These results identify vacuum-resonance observables as the most sensitive channel for testing finite-field QED in magnetars.

We construct a correspondence between a broad class of unitary matrix models and vacuum correlation functions in quantum spin-chain Hilbert spaces. The key step is to lift symmetric functions to operators acting on the $N$-magnon sector in a way that preserves the relevant ring structure. For any unitary matrix model whose integrand admits a factorized expansion in symmetric functions, the Schur orthogonality of the unitary group integral is then translated into the inner product of quantized Schur states. We illustrate the construction for the Gross--Witten--Wadia model, superconformal indices of $\mathcal{N}=4$ super Yang--Mills theory with classical gauge groups, and Toda tau functions. The resulting operator formulation provides a unified algebraic bridge between unitary matrix integrals, quantum integrable systems and symmetric function theory.

Mathematicians have embraced interactive theorem provers with growing enthusiasm -- building large shared libraries and machine-checking a string of landmark results. Theoretical physics is different: most of its results are not theorems but justified by arguments the community trusts without a rigorous proof. For many -- the one we treat here among them -- no rigorous proof is within reach. For 4d Yang--Mills theory, deriving exact rigorous results from first principles would first require constructing the interacting theory nonperturbatively, which is a sizable piece of one of the Clay Millennium prize problems. We argue here that an interactive theorem prover can be used to verify some non-rigorous physics arguments. The method is to postulate a short list of explicit, named physical postulates, which imply the physical results by virtue of a machine-checkable proof. The trust that remains then rests on that short, inspectable list, and the prover can report, for any downstream result, exactly which assumptions it used. We carry this out for the Seiberg--Witten solution of ${N}=2$ $SU(2)$ super-Yang--Mills -- the genus-one case -- formalized in Lean 4; the higher-genus $SU(N)$ generalization is developed in the same repository as an axiomatized skeleton and left to future work. We describe what is proved, what is assumed, how the assumptions are checked -- external review and an independent numerical oracle -- and why this discipline is a sound standard for validating AI-generated results in theoretical physics. What we offer is a discipline, reviewable on its own terms: a reader may take the Seiberg--Witten mathematics on trust and still assess the formalization method.

An intermediate scale R-axion faces an immediate obstruction from the Dine-Festuccia-Komargodski (DFK) bound on the superpotential, $2|\langle W\rangle|\leq f_R F$, since for a nearly Minkowski vacuum it typically follows that $f_R\gtrsim M_{\rm Pl}$. We show that this lower bound on $f_R$ can be relaxed in an effective construction with the scalar potential tuned near zero via a mixed $F$- and $D$-term uplift, leading to a metastable vacuum in which the usual Planckian-$f_R$ inference from the DFK argument is avoided locally. Validity of the effective field theory and metastability of the small $f_R$ vacuum generically both imply a relaxed bound: $ f_R \gtrsim \sqrt{m_{3/2}M_{\rm Pl}} $. We also highlight that if the R-symmetry has a QCD anomaly, this potentially permits the R-axion to play the role of the QCD axion. TeV-scale supersymmetry permits $f_R\sim10^{11}$ GeV, this not only evades certain astrophysical and cosmological axion constraints, but notably lies in the window for which the observed dark matter abundance can be reproduced by the R-axion via the misalignment mechanism.

We discuss a recent proposal for constructing ``syzygy solutions'', which play a crucial role in integration-by-parts reductions for multi-loop scattering amplitudes. We highlight a relationship between syzygies and the leading Landau singularities of Feynman diagrams and discuss how this implies that integral relations are controlled by infrared singularities. We use this insight to systematically construct highly compact, determinantal syzygy solutions, providing an explicit, simple example at one loop. We demonstrate the power of this approach by applying it to a two-loop pentabox family relevant for the NNLO corrections to $pp \rightarrow t\overline{t}H$ at the LHC.

The CMS experiment at the Large Hadron Collider employs Gas Electron Multiplier (GEM) detectors, a technology based on gaseous ionization, as one of the muon detectors. The muon spectrometer is being upgraded to handle the increased muon flux in the forward region. This study analyzes muon detection efficiency in the GE1/1 triple-GEM detector, using 2023 proton-proton collision data at $\sqrt{s}=13.6$ TeV. A dataset enriched with muons from Z boson decay, with a total recorded luminosity of 17.8 fb$^{-1}$ has been used for this study. The detection efficiency of 137 GEM detectors are measured using muon trajectories established using other detectors in the tracking and muon systems, without use of the GEM detectors. The average efficiency of 137 GEM detectors is $\sim$93.3$\%$. A subset of 108 detectors that had no shorts were operated at the nominal HV working point with average efficiency of $\sim$96$\%$. Efficiency is found to be unaffected by the number of p-p interactions per bunch crossing (pile-up).

Using an up-to-date compilation of $σ_{\mathrm{tot}}(e^+e^- \to hadrons)$ data we estimated the LO hadronic contribution to the muon anomalous magnetic moment, $a_μ(had,LO)$. Incompatibilities between $σ_{\mathrm{tot}}(e^+e^- \to hadrons)$ measurements by independent experiments are mitigated by extra systematic uncertainties estimated using as a guideline a requirement of uniformity of $χ^2$ distribution over degrees of freedom in joint fits of $σ_{\mathrm{tot}}$. Tensions in the $e^+e^-$ input data translate into an expanded uncertainty of the $a_μ(had,LO) = (697.7 \pm {9.8}_{e^+e^-} \pm 3.6_{sys}) \times 10^{-10}$. Given this, we obtain the SM prediction for the muon anomaly $a_μ^{\mathrm{SM}} = 11659185.5(10.6) \times 10^{-10}$, below the experimental world average $a_μ^{\mathrm{exp}}$ at $2σ$ level.

We investigate the $K X(3872)$ interaction and the corresponding correlation function, assuming the $X(3872)$ to be a molecular state of $D \bar D^*$ and $D^* \bar D$ with isospin $I=0$ and positive $C$-parity. The interaction is treated within the fixed-center approximation (FCA) to the Faddeev equations, in which the $X(3872)$ is taken as a cluster of its constituents and the kaon interacts with the $D^*$ and $D$ components. The three-body scattering amplitude is evaluated using the Fixed Center Approximation (FCA) to the Faddeev equations, improved by taking the FCA amplitude as an optical potential which is later unitarized by means of the Lippmann-Schwinger equation. We find a narrow resonant structure about 50~MeV below the $K^+ X(3872)$ threshold with a width of approximately 1~MeV, and determine the $K X(3872)$ scattering length $a = (0.39 - i\,0.00)$~fm and effective range $r_0 = (1.16 - i\,1.66)$~fm. The corresponding correlation function is evaluated and shows a clear deviation from unity at low momenta, characteristic of a strongly attractive interaction leading to a bound state. These predictions are tied to the molecular nature of the $X(3872)$ and can be measured experimentally via measurements of the $K X(3872)$ correlation function and three-body invariant mass distributions.

We study the $J/ψ\to \barΛ Λπ^0$ reaction, an isospin violating reaction, by looking at the trios of baryon-antibaryon-pseudoscalar meson that conform a singlet of $\text{SU}(3)$ in the $u, d, s$ quarks. These terms conserve isospin; however, once the final-state interactions of meson-baryon and meson-antibaryon are taken into account, isospin is violated due to the different masses of particles within the same isospin multiplets. Since the reaction is tied to the interaction of particles, only resonances that are dynamically generated by these interactions show up in the reaction. In this sense, our approach produces the $Σ(1430)(1/2^-)$ state, but not the $Σ(1385)(3/2^+)$. Comparing with the BESIII data we observe that, within the limited statistics of the experiment, the data show a structure {around $M_{πΛ} = 1430$ MeV} that is reproduced by the theory, and no signal is seen for the $Σ(1385)(3/2^+)$ as the theory predicts. We call for a future update of the experiment once better statistics become available.

In this work, we analyze the flux-integrated inclusive and pionless differential cross sections for neutrino scattering off argon. The cross sections are calculated using the relativistic distorted-wave impulse approximation (RDWIA), taking into account the contribution of the two-particle-two-hole (2p-2h) meson-exchange currents (MEC). We find that the measured single- and double-differential cross sections can be well described within the experimental uncertainties using this approach. We also compare the differential cross sections for pionless neutrino scattering on carbon and argon, measured with the Booster Neutrino Beam flux in the MiniBooNE and MicroBooNE experiments, to study nuclear effects in these nuclei.

This paper reports simultaneous measurements of the efficiencies with which $b$-jets are identified, and $c$-jets are misidentified as $b$-jets, by the single-transformer-based GN2 tagger that is currently used in the ATLAS experiment at the LHC. The measurements use $t\bar{t}$ events where one of the $W$ bosons decays into an electron or muon and a neutrino, and the other decays into a quark-antiquark pair, together with the known $t\to bW$ and $W\to cs$ branching fractions. The data were collected from $\sqrt{s} = 13.6$ TeV proton-proton collisions by the ATLAS detector and correspond to an integrated luminosity of 56 fb$^{-1}$. Events are reconstructed using a kinematic likelihood technique which maps the jets to the $t\bar{t}$ decay products. The $b$-jet tagging efficiency is measured from jets arising from leptonically decaying top quarks and the $c$-jet mistagging efficiency is obtained from jets arising from hadronic $W$ decays. The efficiencies are measured as a function of jet transverse momentum in intervals of the tagging-variable distribution. The $b$-jet ($c$-jet) tagging efficiency measurements reach a precision of better than 2% (5%). Each $b$-jet efficiency is combined with an independent measurement using $t\bar{t}$ events where both $W$ bosons decay leptonically.

Near-threshold structures have been observed in the cross sections for $e^+e^- \to p \bar p$, $e^+e^- \to n \bar n$ and several non-baryonic final states in the vicinity of the $N \bar N$ thresholds. We investigate whether these structures can be understood as manifestations of a common $N \bar N$ final-state interaction. The strong $N \bar N$ interaction is taken from the chiral EFT description of the coupled ${}^3S_1$-${}^3D_1$ system constrained by low-energy $N \bar N$ scattering data. With this interaction fixed, the $p \bar p$ and $n \bar n$ cross sections are described by fitting only short-distance electromagnetic production sources, which are assumed to vary slowly over the near-threshold region. The resulting $N \bar N$ production amplitudes are then used as input for five inelastic hadronic channels. A simultaneous description of the near-threshold cross sections is obtained, indicating that the observed structures can be consistently interpreted as consequences of the same underlying $N \bar N$ dynamics, without introducing separate narrow resonances in individual channels.

The distribution of particles is highly anisotropic in the initial stage of a heavy-ion collision. In this paper we demonstrate that this anisotropy induces a sizable effect on the spin alignment of vector mesons. We study two different production mechanisms for $φ$ and $K^{*0}$ mesons, on one hand the coalescence of quarks and on the other that of pseudoscalar mesons. In the quark-coalescence picture where $φ$ and $K^{*0}$ are produced via a bare vector coupling to quarks, a negative $δρ_{00}^y$ of order $10^{-3}$ is observed. In contrast, when $φ$ and $K^{*0}$ are produced via quark coalescence with a vertex with spin-orbit coupling, or when they are produced via pseudoscalar-meson coalescence, a positive $δρ_{00}^y$ emerges. In all cases, the magnitude of the spin alignment is directly proportional to the degree of anisotropy. The sign difference between the cases provides a possibility to clarify the production mechanism for vector mesons.

Future long-baseline neutrino oscillation experiments aim to establish leptonic CP violation and determine the neutrino mass ordering with unprecedented precision. However, these measurements can be significantly affected by possible deviations from the unitarity of the PMNS mixing matrix, which introduce additional CP-violating phases capable of generating fake CP-violating signals. We investigate the impact of non-unitary leptonic mixing on CP-violation and mass-ordering measurements at DUNE and Hyper-Kamiokande using GLoBES simulations with a custom non-unitary probability engine. We analyze the energy dependence of the neutrino--antineutrino CP asymmetry, quantify the fake-to-genuine CP asymmetry ratio, evaluate the CP violation discovery sensitivity, and study the hierarchy--CP--non-unitarity degeneracies in the $(δ_{CP},φ_{21})$ parameter space. We demonstrate that non-unitary mixing can generate sizeable CP asymmetries even for CP-conserving values of the standard Dirac phase, thereby mimicking genuine leptonic CP violation. While the fake contribution remains below $5\%$ of the genuine signal near each experiment's oscillation maximum, it exceeds the genuine signal in specific intermediate energy windows ($\sim130\%$ at $1.5$--$1.6$~GeV), demonstrating that fake CP violation can dominate over the genuine contribution in these regions. The combined DUNE and Hyper-Kamiokande analysis reduces the allowed $(δ_{CP},\varphi_{21})$ parameter space by a factor of $7$ at $1σ$ and $\sim16$ at $2σ$--$3σ$, substantially suppressing the degeneracy that neither experiment resolves individually and providing a robust strategy for distinguishing genuine from fake CP violation while improving sensitivity to the neutrino mass ordering.

The two-zero minor structure predicted by flavor symmetries is usually discussed as a tree-level relation among low-energy neutrino parameters. We point out that this relation can be significantly modified by radiative corrections even when the two-zero minor structure is enforced by an underlying symmetry at tree level. As a concrete example, we analyze the minimal $\mathrm{U}(1)_{L_μ-L_τ}$ model and compute the universal one-loop threshold corrections associated with the type-I seesaw sector. These corrections generate flavor-dependent contributions to the Weinberg operator and violate the exact two-zero minor conditions once the $\mathrm{U}(1)_{L_μ-L_τ}$ symmetry is spontaneously broken. This effect relaxes the tree-level lower bound on the total neutrino mass and thereby weakens the tension between the model and cosmological neutrino-mass constraints.

We present the C++ library ggxy, which can be used for the fast and flexible calculation of partonic and hadronic cross sections to gluon-initiated top-mediated processes such as $gg\to HH$, as well as the newly derived processes $gg\to ZH$ and $gg\to ZZ$, at NLO QCD including full top-quark mass dependence. The two-loop virtual amplitudes are implemented using analytical approximations in different kinematic regions, while all other parts of the calculation are exact. This implementation allows to freely modify all input parameters, such as the top-quark mass and its renormalization scheme and the masses of the external $Z$/Higgs bosons. In addition, ggxy has been interfaced to Powheg, which allows the matching to parton showers.

Using $(10.087\pm0.044)\times10^9$ $J/ψ$ events and $(2.712\pm0.014)\times10^9$ $ψ(3686)$ events collected by the BESIII detector operating at the BEPCII collider, we report the first observation of the hadronic decays of $J/ψ\to p \bar p K^0_S K^0_S$ and $ψ(3686) \to p \bar p K^0_S K^0_S$, both with statistical significance greater than $10σ$. Their branching fractions are determined to be $\mathcal{B}(J/ψ\to p \bar p K^0_S K^0_S)=(1.60 \pm 0.02 \pm 0.09)\times10^{-5}$ and $\mathcal{B}(ψ(3686) \to p \bar p K^0_S K^0_S)=(3.93 \pm 0.24 \pm 0.34)\times10^{-6}$. The ratio of their branching fractions is $\mathcal{B}(ψ(3686) \to p \bar p K^0_S K^0_S)/\mathcal{B}(J/ψ\to p \bar p K^0_S K^0_S)=(24.6 \pm 1.5 \pm 2.1)\%$, which deviates from theoretical expectation by 4.6$σ$. Here the first uncertainties are statistical and the second systematic. We have also examined the $p\bar p$ invariant mass distributions in these decays, and no significant enhancement around the $p \bar p$ near threshold is found.

This dissertation develops a quantitative framework that provides a physical picture for non-perturbative QCD based on the topological structure of the QCD vacuum. By integrating out the ultraviolet degrees of freedom, the infrared gluon configurations are modeled as a liquid ensemble of instantons and anti-instantons, which induce effective interactions among quarks. This framework captures the origin of trace and axial anomalies through the infrared distributions of QCD and dynamically breaks chiral symmetry. More specifically, this framework can be formulated in two ways: in one, we construct a statistical ensemble with weights defined by the instanton action and Dirac determinant, while in the other we formulate an effective field theory (EFT) by rewriting the determinant as effective quark interactions. By reformulating the EFT on the light front, we explicitly construct the light-front wave functions and calculate various parton observables in linear factorization. We further embed the framework into transverse momentum dependent factorization and establish a vacuum origin for rapidity evolution by computing the soft functions. We also extend this approach to various form factors in light hadrons, including scalar, pseudoscalar, and energy-momentum tensor (EMT), as well as higher-twist color force and multigluon correlations, with applications to hadron mass and spin decomposition, near-threshold quarkonium production, and strong CP problem, highlighting the importance of the vacuum origin in hadron structures. Overall, this work demonstrates that the QCD vacuum provides a quantitatively crucial description of hadronic structure from low to moderate resolution, bridging nonperturbative vacuum physics with partonic phenomenology.

As we approach HL-LHC, there is a growing need for increased precision in theoretical predictions so that meaningful comparisons with experimental data can be made. It is no longer sufficient to include only QCD higher-order corrections, with EW effects becoming increasingly important. Even at hadron colliders, QED radiation provides large corrections to some observables. In this paper, we present the first automated matching of NLO QCD+EW to an interleaved QCD+QED parton shower using the MC@NLO matching method in the Catani-Seymour dipole formalism. When considering such a matched parton shower, the presence of resonances can lead to spurious higher order terms, originating in the recoil assignment, within the standard dipole construction. We therefore develop a resonance-aware modification to the MC@NLO algorithm that can be applied to QCD- and QED-singlet resonances. We validate our interleaved matching and its resonance-aware modification against fixed-order NLO QCD+EW and pure MC@NLO QCD combined with YFS resummation. Finally, we present resonance-aware MC@NLO QCD+EW predictions for Drell-Yan lepton pair production, a vital precision process at hadron colliders.

Within the framework of nonrelativistic effective field theory, we calculate the radiative decay width of the process $X(3872) \to \bar{D} D γ$ while taking into account the $D\bar{D}$ final-state interactions. In this work, the $X(3872)$, with the spin-parity quantum numbers $J^{PC}=1^{++}$, is treated as a $D^*\bar{D} +\rm c.c.$ bound state comprising equal proportions of neutral and charged components. Our numerical calculations predict the tree-level partial decay width of approximately $11.0$ keV for the decay process $X(3872) \to \bar{D}^0 D^0γ$, while the partial width for $X(3872) \to D^- D^+γ$ is less than $1.0$ keV. It is found that the $D\bar{D}$ rescattering effect enhances the tree-level width of $X(3872) \to \bar{D}^0 D^0γ$ by $6\%$. In contrast, the rescattering effect makes a suppression to the charged channel $X(3872)\to D^- D^+γ$ by roughly $38\%$. We expect that the present predictions based on the molecular picture of the $X(3872)$ can be tested by future experiments.

Cryogenic semiconductor detectors are widely used in dark matter direct detection. Electronic excitations in these materials can be caused either by direct dark matter electron scattering events, or indirectly via dark matter nuclear scattering events. For nuclear scattering of light dark matter, the event rate is enhanced due to an inelastic process known as the Migdal effect. Both the electron recoils and the Migdal effect in semiconductors depend on the dielectric function of the target material via the so-called energy loss function (ELF). In the standard approach found in the literature, the ELF is approximated as isotropic to simplify the calculation of the event rate. We introduce a practical method for computing the event rate for a general anisotropic ELF. We find that the daily modulation of the Migdal rate arises solely due to the quadrupole component of the ELF, whereas in electron scattering all spherical harmonic components affect the rate. We apply the formalism to study the daily modulation in silicon and gallium arsenide detectors.

We discuss energy scales in soft hadron scattering and possibility of their separation: the one relevant to the region where the total cross-sections increase begins and another one related to the asymptotic region. The latter scale is most sensitive to the unitarity effects and former one -- to the gross features of absorption of the initial wave due to multiparticle production. Transition from the shadow to reflective scattering sets a relation between the two energy scales.

We develop an analytical framework to constrain Yukawa-type fifth-force interactions using the Lamb shift in hydrogen and deuterium. The sixfold coordinate-space integral for the energy shift is transformed into a one-dimensional momentum-space convolution of the atomic form factor and the Yukawa potential, eliminating the need for approximations. This formulation directly incorporates empirical nuclear charge distributions through their form factors, making the implementation of experimental results straightforward. Applying this method to the latest hydrogen and deuterium spectroscopy data, we derive constraints on the coupling constant $α$ in the interaction range $λ\in [10^{-4}, 10^{6}]$~fm. A resonance-like cancellation between the $2s$ and $2p$ orbital shifts is identified at a critical range $λ\sim 10^{5}$~fm, where the energy shift vanishes and the constraint on $α$ diverges, marking a transition between attractive and repulsive regimes. The constraints are shown to be independent of the nuclear charge distribution model. Our results indicate that the transition frequency shifts induced by such new forces lie well below current spectroscopic resolution, establishing spectroscopic precision as the primary bottleneck for probing these interactions.

These notes give a cross-dimensional crash course in supersymmetric field theory. The goal is to provide enough common language to recognize the basic setups used in talks and papers, while also working through representative computations that make the terminology concrete. We begin with supersymmetry algebras, multiplets, protected quantities, moduli spaces, deformations, and anomalies, then follow these ideas through examples in two to ten spacetime dimensions. The emphasis is on recurring structures, such as holomorphy, supersymmetric vacua, dualities, extremization principles, indices, BPS data, and anomaly constraints, rather than on a classification of theories in any one dimension.

The liquid scintillator (LS) detectors are widely used in reactor neutrino experiments. To precisely measure the oscillated neutrino spectrum, it is crucial to understand the LS energy response, especially the quenching effect. Numerous bench-top measurements have been conducted on quenching effects for $α$ particles, protons, and electrons. These results have typically been described by Birks' law, but with different Birks' coefficients required for different particle species. In this study, we find that if more secondary electrons are allowed to be generated in Geant4 simulations, the bench-top results for all particles can be well fitted using a single Birks' coefficient of about $0.013~{\rm g\,cm^{-2}\,MeV^{-1}}$. The underlying reason is that a large fraction of the primary energy is deposited through the generation of $δ$-electrons with energies below 4~keV, which should be tracked separately due to their different quenching behavior compared to primary particles. This study provides a unified framework for comparing Birks' coefficients from different bench-top measurements and helps experiments like JUNO and SNO+ better tune their simulations for precision energy measurements.

This article describes the deployment of a germanium detector at Oak Ridge National Lab's High Flux Isotope Reactor (HFIR) for the purpose of understanding the energy and spatial distribution of the gamma field in the experiment hall where the Precision Reactor Oscillation and Spectrum Experiment (PROSPECT) took data and future neutrino experiments could be located. The sources from both the reactor and the neutron beamlines are described in detail, along with their temporal variations due to reactor power and their spatial variations due to the geometry of the beamlines and building materials in the vicinity. Additionally, a shielding study was performed to assess the amount that backgrounds in tens of keV range can be mitigated. This work helps inform backgrounds for future experiments at reactors such as IBD-based neutrino measurements and CEvNS measurements.

We construct a $\frac{\mathbb{Z}}{2}$-graded meromorphic open-string vertex algebra from a finite-dimensional vector space with a nondegenerate symmetric bilinear form, together with its canonically twisted module. This algebra is generated by suitable noncommutative generalizations of bosonic and fermionic fields and is a noncommutative generalization of the free boson-fermion vertex operator superalgebra. Similarly to the purely bosonic and purely fermionic cases in the early works by the second author in [H1], by Fiordalisi and the third author in [FQ], and by the third author in [Q3], the usual super-commutatitive relations between creation and annihilation operators still hold while no relations exist among creation operators. In particular, normal-ordering remains well-defined. As in [H1], [FQ], and [Q3], we prove a generalized Wick's theorem in this case, which gives a formula for a product of two normal ordered products of bosonic and fermionic generating fields. Using this generalized Wick's theorem, we construct the $\frac{\mathbb{Z}}{2}$-graded meromorphic open-string vertex algebra and its canonically twisted module in this case. The construction in this paper is the algebraic part of our construction of suitable Dirac-like operators from spin manifolds.

We propose a Coulomb-like parameterization in terms of bag radius of the short-range chromo-electric interaction between heavy quarks and strange quarks within the framework of the MIT bag model for hadrons including multiquark systems and re-examine mass spectra of doubly and fully heavy baryons self-consistently and variationally. Building upon this, we apply this approach to systematically investigate the mass spectra and $S$-wave strong decay stability of singly heavy tetraquark systems, including $nQ\bar{n}\bar{n}$, $nQ\bar{s}\bar{n}$, $Qs\bar{n}\bar{n}$, $sQ\bar{s}\bar{n}$, and $sQ\bar{s}\bar{s}$ (with $Q=c,b$). Choosing the bag confinement energy as a compactness criterion, the bag radius is shown to be close to the confinement uplimit of the radius, $R_c = 5.615\,\text{GeV}^{-1}$, for the compact singly charmed systems,and for these systems some high-spin states are unlikely to form compact structures, while several others do exhibit potential for compactnessNotably, computation indicates that the state $T_{nc\bar{s}\bar{n}}(0^+,2.925)$ emerges as a compact tetraquark with relatively large strong decay width, a plausible candidate of the observed tetrquark $T_{c\bar{s}0}^{a}(2900)$.

In this work, we investigate the emergence of dipole ghost structures in gauge theories at both the classical and quantum levels. Traditionally, dipole ghosts are introduced through an auxiliary field $χ$ satisfying $\Box^2χ=0$, whose presence is reflected in the appearance of double poles in the propagator. We show that such dipole ghost sectors can instead be understood as a consequence of the multiplicity of solutions of the classical equations of motion. To establish this connection, we develop a constructive method and first apply it to covariant Maxwell theory in $(2+1)$ dimensions, where the essential ingredients can be identified in a transparent manner. We then extend the analysis to the constrained higher-order Chern-Simons theory, demonstrating that the same mechanism gives rise to dipole ghost sectors and associated double poles in the propagator. Our results provide a unified perspective on the origin of dipole ghosts, relating them directly to the degeneracy structure of the underlying classical dynamics. As a physical application, we investigate spontaneous symmetry breaking and compute the one-loop effective potential. We show that symmetry breaking removes the degeneracy underlying the dipole ghost sector, leading to a spectrum of ordinary massive gauge excitations that determine the quantum vacuum structure of the theory.

The proton's spin structure has proven to be far more complicated than was originally believed, and has been the subject of a number of experimental investigations. %Early measurements of the proton's spin structure function $g_1$ showed that the proton does not solely derive its spin from the spins of its quarks, starting the `proton spin crisis'. Of particular interest are the spin structure functions $g_1$ and $g_2$, which can be used to generate moments to directly compare experimental results to Chiral Perturbation Theory and other theories of Quantum Chromodynamics (QCD). The proton's $g_1$ structure function has been the subject of two other recent low momentum transfer experiments, but there are currently no published low momentum transfer measurements which collected data on the proton structure functions using both a longitudinally-polarized and a transversely-polarized target at the same kinematics. In this paper, we present the longitudinally polarized results of the Jefferson Lab E08-027 experiment, along with linked moments which combine this new result with the previously published transversely-polarized data from the same experiment. These results provide a proton $g_1$ extraction measured with very high precision across the resonance region, and provide new information on the value of $g_1$ dependent sum rules and moments.

The integration kernels for polylogarithm functions on a compact Riemann surface of arbitrary genus $h$ are shown to close as the surface undergoes a non-separating degeneration to one of genus $h{-}1$. Explicit formulas are obtained for the non-separating degeneration of the multivariable Enriquez connection for genus $h$ with an arbitrary number of variables to the Enriquez connection for genus $h{-}1$ with two additional punctures whose Lie algebra generators are related to the original ones by the characteristic Bernoulli generating functions known from the degeneration at $h=1$. Analogous degeneration formulas are obtained for the single-valued DHS kernels at the leading order in the real degeneration parameter that is adapted to relating modular tensors at genus $h$ and $h{-}1$.

Natural supersymmetry (SUSY) models remain viable even in the face of LHC Run 2 sparticle search limits. However, the LZ experiment has placed strong limits on light higgsino dark matter even when the higgsinos carry only their thermally-produced abundance, with the bulk of the dark matter composed of axions. One way out is the possibility of WIMP direct detection blind spots where cancellations in direct detection (DD) couplings lead to tiny DD rates. We examine natural SUSY models with mu <0 and μ>0 but find that the surviving blind spots all lie in the unnatural region where the superpotential |mu | parameter is much greater than the weak scale gaugino masses; the few natural candidates are excluded by LHC soft-dilepton searches and by the measured Higgs mass. Within NUHM2/NUHM3-type gravity-mediated models with positive gaugino masses and assuming a thermally produced neutralino fractional abundance, direct-detection blind spots do not rescue stable light higgsino dark matter in the electroweak-natural region. Thus, within this framework, stable light higgsino dark matter is disfavored, although special circumstances like large entropy dilution of all relics is still possible. This points to SUSY models with {\it unstable} light higgsinos as perhaps the preferred alternative.

The analysis of the production of strange $K^{*}(892)^0$ resonances allows us to better understand the temporal evolution of high-energy nucleus--nucleus collisions. In particular, the ratio of $K^{*}(892)^0$ to charged kaon yields is used to determine the time interval between chemical and kinetic freeze-outs. In this paper, the first measurements of $K^{*}(892)^0$ production in central $^{40}$Ar+$^{45}$Sc collisions at the CERN Super Proton Synchrotron are reported. They were performed by NA61/SHINE at collision center-of-mass energies per nucleon pair $\sqrt{s_\mathrm{NN}}$ = 8.8, 11.9, 16.8 GeV. The obtained $\langle K^{*}(892)^0 \rangle/\langle K^{+} \rangle $ and $\langle K^{*}(892)^0 \rangle/\langle K^{-} \rangle$ mean multiplicity ratios are compared with corresponding results in $p$+$p$ collisions, allowing for an estimate of the time interval between chemical and thermal freeze-outs in the $^{40}$Ar+$^{45}$Sc system. These are the first such results reported for $^{40}$Ar+$^{45}$Sc collisions.

We track the chain of dualities relating five-brane webs in Type IIB to M-theory on noncompact Calabi-Yau threefolds, and follow the effect of adding $(p,q)$ seven-branes. T-dualizing along the seven-brane, a junction of D5-branes ending on a D7-brane becomes a single smooth D6-brane wrapping a smooth holomorphic curve, which uplifts directly to an M-theory geometry. Treating the branes as coherent sheaves (tachyon condensation), we obtain a spectral-curve dictionary that maps non-Abelian (T-brane) partition data to explicit complex-structure deformations of the threefold. As applications we (i) give a physical derivation of these deformations, (ii) exhibit a simple geometric prototype of an s-rule violation, and (iii) link T-brane data to geometry.

Assuming the existence of axion-like particles (ALPs), the standard Quantum Electrodynamics (QED) amplitudes for the linear Breit-Wheeler (BW) pair production process are shown to be supplemented by an additional ALP-mediated amplitude. We compute the coherently combined cross section and show that, for pseudoscalar ALPs, interference with QED produces an asymmetric Fano-type resonance. We propose experimental observation of the BW process with real photons as a complementary probe of MeV-scale ALPs, coupling jointly with photons and electron-positron pairs, on the basis that it occupies a kinematic region not readily accessible to other ALP search methods.

Large-$N$ three dimensional QCD with a level $k$ Chern-Simons term and $N_f$ massless flavours admits $N_f+1$ degenerate vacua. We discuss various aspects of the domain walls that interpolate between those vacua. Using a bosonic dual for the case $k \ge N_f/2$ we write down the profile of the composite wall. By fluctuating the profile, we find the 2d field theory that lives on the wall, a level-$N$ gauged WZW model, and compare it with a theory that was previously obtained by using anomaly inflow considerations. We also discuss the case $k < N_f/2$. Finally, we remark on the realisation of the wall as a D-brane in a holographic dual.

Massive quantum fields usually mediate Casimir interactions that are exponentially screened beyond their Compton wavelength. We show that an isolated normalizable zero mode in a classical background opens a low-frequency collective channel that bypasses this screening. The effective interaction is universal, controlled only by the zero-mode profile. Kinks and planar domain walls provide robust realizations, with the sine-Gordon/Pöschl--Teller kink as an exactly solvable example. Localized zero modes thus define a generic infrared mechanism for modifying massive vacuum forces.

We construct equivariantly closed polyforms for the fluxes and the action of (massive) type IIA and type IIB supergravity directly in ten dimensions. We illustrate applications of our formalism to two distinct classes of solutions in massive type IIA. Our analysis straightforwardly reproduces and vastly generalizes all known results for on-shell actions in these classes, with no need for a consistent truncation. Where known, we match the corresponding field theory partition functions at leading order, and our results extend easily to other solutions of type II supergravity.

Metal-oxide-semiconductor field-effect transistor (MOSFET) test structures are investigated to characterize p-stop isolation implants between n-type electrodes in p-type silicon sensors. The device transfer characteristics are measured as a function of the voltage applied to the backside to extract the threshold voltage, which quantifies inter-electrode isolation, and the field-dependent mobility parameters. We present a methodology to reconstruct depth-dependent doping profiles from the threshold voltage characteristics, accounting for the localized space-charge effects and electric-field screening induced by the p-stop implants. The study evaluates the sensitivity of this technique to various p-stop geometries and doping concentrations across different wafer types. The results demonstrate the potential of MOSFET-based structures as a non-destructive diagnostic for monitoring p-stop consistency and inter-electrode isolation properties in silicon detectors.

We analyse semileptonic and non-leptonic decays of the $B_c$ meson into P wave charmonium states. The analytic expressions for the transition form factors are taken from NRQCD, while their normalisations and shapes are constrained directly using experimental data, without introducing any additional model-dependent inputs. Using radiative decay data of $χ_{c0}$, $χ_{c1}$, and $h_c$, we extract the derivatives of their radial wave functions at the origin and update the $B_c \to (χ_{c0}, χ_{c1}, h_c)$ form factors. We present predictions for the semileptonic branching fractions, the lepton flavour universality ratios $R(χ_{c0})=0.185(3)$, $R(χ_{c1})=0.147(26)$, and $R(h_c)=0.068(2)$, as well as selected non-leptonic $B_c$ decays and P-wave charmonium production in $e^+e^-$ annihilation and $Z$-boson decays.

Global monopoles, which can also be understood as the zero-gauge-coupling limit of gauged monopoles, can form in the early Universe and evolve following a scaling network, as do other topological defects. However, only a limited number of numerical studies have investigated their scaling behavior. In this Letter, we show that the monopole number density parameter, $ξ$, does not follow the commonly assumed constant scaling. Instead, it exhibits a logarithmic-like evolution, despite the fact that the energy of an isolated global monopole grows linearly, rather than logarithmically, with the infrared cutoff. This behavior is found using the fat-monopole prescription and is supported by a conventional fixed-core simulation. We characterize the deviation by the fractional response $γ\equiv d\log ξ/d\log(m_r/H)$, and find $γ\sim 0.5$ for blue-tilted initial spectra. These results suggest that analytic studies based on the assumption of constant scaling should be revisited. They are also relevant to cosmological scenarios involving monopole dark matter, axion and dark-photon dark matter produced by monopole networks, monopole-induced primordial black holes, and gravitational-wave production from monopole dynamics.

We extend the spacetime duality programme of Burgess \textit{et.al.} to massive theories in 1+1 dimensions. For the massive scalar, a heat-kernel computation tracking three contributions to the conformal-mode effective action reveals that the naive leading correction $\sim m^2(e^φ -1)$ to the Liouville action cancels exactly, with the genuine leading deformation being $-\frac{m^2}{16π}(e^φ-1)^{2}$. This breaks self-duality and renders the dual theory for the Lagrange multiplier field $Λ$ non-local. For the massive Dirac fermion, two independent derivations establish that the fermion mass dresses under conformal scaling as $m \to m\, e^{φ/2}$, reflecting the Weyl weight $\frac{1}{2}$ of the two-dimensional spinor. Via the Coleman-Mandelstam bosonisation, this transfers to the mass bilinear as $μ\cos(β\vartheta) \to μe^{φ/2}\cos(β\vartheta)$, producing a coupled Liouville-sine-Gordon system as the natural starting point for the fermionic construction. Both results are interpreted in terms of the determinant line bundle over Met($Σ$)/Diff($Σ$).

Big Bang nucleosynthesis (BBN) predicts the primordial deuterium abundance to a precision now limited by the nuclear reactions that burn deuterium. For the simplest of them, proton-deuteron radiative capture, d + p -> γ+ 3He [d(p,γ)3He], the precise LUNA data sit below the ab initio benchmark, and BBN reaction networks split on which to adopt. We develop an effective field theory (EFT) expanding in the finite size of the nuclei, building the amplitude with modern on-shell methods that enumerate every tree-level structure consistent with symmetries without the need for an explicit Lagrangian. A global Bayesian fit to the capture data and nuclear-theory priors returns S(0) = 0.209 +/- 0.008 eV b and traces the offset from the ab initio benchmark to a single natural-sized next-to-leading contact term (t_E1 ~ -0.15, the fractional shift of the electric-dipole amplitude) -- equivalently a ~15% lower effective 3He asymptotic normalization. We estimate the leading EFT truncation errors and identify an elastic d-p observable that would separate them. Our results suggest that amplitude methods enable systematic and complete tree-level construction and matching of EFTs for low-energy nuclear reactions.

Self-interacting dark matter (SIDM) influences halo structure through collisional heat transport and may solve several small-scale puzzles in structure formation. SIDM creates thermalized cores in low-mass haloes, which may account for the observed cored dwarf galaxies. During late-time gravothermal core collapse, SIDM can produce dense low-mass DM haloes and substructures detected through perturbations to cold stellar streams and strong gravitational lenses. In this work, we present a new Monte-Carlo SIDM implementation in the moving-mesh code AREPO-2, designed for efficiency, scalability, and extensibility. The central feature of the implementation is a dedicated DM-only neighbour-search tree that decouples the scattering solver from gravity. This preserves compatibility with the hierarchical time integration used by AREPO-2 while leaving the optimized gravity solver unconstrained. A pairwise communication scheme between MPI tasks allows tracking multiple scattering events in a single timestep while conserving momentum and energy and maintaining parallel consistency by construction. This is complemented by a per-pair timestep criterion that significantly reduces unnecessary timestep restrictions. The implementation natively supports velocity-dependent cross-sections and inelastic interactions, while a compact interface is designed for additional SIDM physics to be implemented without knowledge of the parallelization layer. We validate the implementation for isotropic, elastic scattering using a suite of idealized and cosmological tests. We assess performance and scalability in isolated core-collapse simulations and in cosmological boxes, both DM-only and with baryons. Except during the late stages of gravothermal collapse, SIDM simulations incur only modest overhead relative to the corresponding CDM runs and are substantially faster than the previous SIDM implementation in AREPO-1.

We study the low-impact-parameter bootstrap for gravitational effective field theories (EFTs) using the stringy dispersion relation (SDR). After incorporating eikonal spectral support, we ask where the residual spectrum is supported in partial-wave space. On finite grids, the extremal spectra are far from featureless: a cap-saturated low-impact band tracking an order-one Giddings-Porto rotating black-hole scale appears across a substantial part of the allowed EFT boundary being examined, accompanied by a separate Regge-like high-spin ridge. A broad region between this black-hole-band and the eikonal input layer is available to the optimizer but remains largely empty. The gap is available but expensive; the black-hole-scale band is capped and useful. Thus the pole-subtracted bootstrap acts as a microscope for the small-impact-parameter completion of gravity: it resolves organized black-hole-scale and Regge-like structures rather than a featureless residual continuum.

We propose a systematic framework for constructing six-dimensional supergravity theories with eight supercharges that respect all known consistency constraints, including anomaly cancellation and the non-perturbative ${\it H}$-string constraints recently discovered by Kim, Vafa, and Xu. The basic objects in this framework are ${\it supergravity\,blocks}$, which are minimal collections of tensor multiplets consisting of a single little string theory sharing the ${\it H}$-string charge together with additional tensors whose string charges intersect it positively. A characteristic feature of each supergravity block is that its Gram matrix has exactly one positive eigenvalue, and therefore it necessarily contains gravitational BPS strings that cannot become tensionless anywhere in tensor moduli space. Any consistent 6d $(1,0)$ supergravity theory can then be obtained by gluing compatible blocks and subsequently enhancing the gauge algebras and matter content. As a first step toward establishing this framework concretely, we provide a complete classification of the ${\it non\text{-}Higgsable\,supergravity\,blocks}$, (or ${\it non\text{-}Higgsable\,gravity\,blocks}$ for short) namely those built from tensor multiplets that support only non-Higgsable gauge algebras.

Rare top-quark decays offer a sensitive probe of charged Higgs bosons with masses below the top mass, owing to the large $t\bar t$ production rate at the LHC and the distinctive final states involving leptons and $b$-jets. While existing searches target the $H^\pm\to τν$, $cs$, and $cb$ modes, the decay $H^\pm\to tb$ has been studied only for heavier charged Higgs bosons with an on-shell top quark in the final state. The low-mass off-shell channel $H^\pm\to t^*b$ therefore remains essentially unconstrained, even though it can become the dominant decay mode below the top-quark threshold owing to the large top Yukawa coupling. We study charged-Higgs production from the rare top decay $t\to H^\pm b$, followed by the decay $H^\pm\to t^*b$. Top-antitop production and decay hence give rise to a $t\bar t b\bar b$-like final state, which we constrain by reinterpreting recent ATLAS fiducial measurements in dileptonic events. We obtain model-independent limits of 1.9%-2.9% on the product of branching ratios ${\rm Br}(t\to H^\pm b)\times{\rm Br}(H^\pm\to t^*b)$ for charged-Higgs masses between 110 and 165 GeV, and interpret these bounds in several two-Higgs-doublet scenarios. While dedicated searches for conventional charged-Higgs decays dominate for canonical $Z_2$-symmetric models, the $t^*b$ reinterpretation becomes competitive near the top-quark threshold in the up-type aligned limit and provides relevant direct constraint in a top-philic scenario. In the top-specific Two-Higgs-Doublet Model, it additionally excludes a low-$\tanβ$ region not covered by conventional searches. These results thus establish $t\bar t$ production with additional $b$-jets as a complementary probe of light charged Higgs bosons and nonstandard top-quark decays, and motivate dedicated analyses by the LHC collaborations.

Due to the exponential growth in the number of terms, computing $α'$-corrections to string theory's low-energy effective actions is a challenging matter. In order to fix all the couplings, one has usually to deal with a large number of string scattering amplitudes. This difficulty can be overcome by exploiting T-duality, which severely constrains the allowed structure of the effective action. It is then convenient to work in a formulation where T-duality is a manifest symmetry. Building on a series of previous works by some of the authors, we present a refined version of an elegant and effective procedure that allows to obtain all the higher-derivative corrections of the NS-NS sector of sting theories at order $α'$ and $α'^2$, up to an overall coefficient. We dub this approach $α'$-bootstrap, since it is based only on consistency conditions and avoids the direct computation of scattering amplitudes. The procedure relies on an infinite dimensional algebraic structure that we present in full detail, and it is conjectured to work at all orders. Although, at the moment, it still misses the $ζ$-like corrections starting at order $α'^3$, the ease with which it can be generalized is promising for future developments in this direction.

We provide a new formal extension of the BFSS matrix model by an additional 5-bracket. Maximal supersymmetry leads us to promote the BFSS 2-bracket structure constants to a dynamical field governed by a Chern-Simons-like kinetic term, as well as a novel potential self-duality relation with respect to the 5-bracket. We show that the full action is invariant under maximal supersymmetry and that the associated supersymmetry algebra closes. This result hinges on a conspiracy of properties of the $SO(9)$ gamma matrices and the 2- and 5-brackets. Compellingly, the resulting model seems to realize some features expected of a theory containing M5-branes and opens up the possibility of further including higher brackets for the M6- and M9-branes.

We present a geometric derivation of Liénard--Wiechert fields in flat-space and AdS, emphasizing the origin of antipodal matching. In flat-space, the field of a uniformly moving charge is rewritten in coordinates centered on the source timelike geodesic. In this frame the charge is at rest and the solution is Coulombic, so the matching of the leading data at null infinity arises from describing a static field in a non-centered frame. We extend this construction to global AdS, where uniform motion is replaced by motion along a timelike geodesic. Starting from the static Coulomb solution at the center, we reconstruct the field of a freely moving charge in arbitrary global coordinates using embedding-space invariants. The resulting closed-form field obeys exact antipodal covariance in the bulk, and its boundary null-fringe limit reproduces the usual flat-space antipodal matching relation. We also describe an image-charge interpretation: the flat-space Coulomb field is represented after conformal compactification by an image singularity at spatial infinity, while the AdS Coulomb seed may be viewed as a charge together with an opposite image charge in a reflected copy. Together, these perspectives give a unified picture of Coulombic Liénard--Wiechert fields, antipodal matching, and the AdS-to-flat-space limit.

In this paper we will examine if nonlocal quantum field theory will suffer from the fermion doubling pathology. We find that for a nonlocal Dirac theory, that no additional fermion species are introduced. This is provided that the form factor is nonvanishing at every point. The proof follows from the invertibility of the entire-function operator, which implies that the nonlocal Dirac operator has exactly the same kernel and finite-momentum zero set as the original local Dirac operator. We will distinguish this result from the standard Nielsen-Ninomiya theorem, which applies local lattice Fermions on a compact Brillouin zone. We provide a general criterion for fermion doubling, a test for Genuine and false doubling, and then a test procedure for mathematical and physical fermion doubling. We then will go on to distinguish this result from finite derivative truncations, which can introduce spurious polynomial zeros. We then conclude that fermion doubling is absent in the full continuum nonlocal theory.

We revisit the old problem of analytically continuing QFTs to fractional dimension $d\in\mathbb C$. We observe that common theories like QCD and QED have branch cuts in the complex $d$ plane. In particular, many operators in their low-energy CFT have OPEs and scaling dimensions that jump as a function of $d$.

The IceCube and ANTARES collaborations have recently reported evidence for high-energy neutrinos associated with the Galactic plane and the Galactic Ridge, offering a new pathway to search for dark matter (DM). Deep inelastic scattering of cosmic rays with sub-GeV DM in the Galactic halo produces a distinctive neutrino signature from meson decays. Using detailed Galactic cosmic-ray maps and ANTARES observations, we derive 99% C.L. upper limits on the DM-nucleon cross section that extend down to keV-scale masses. These results establish Galactic neutrino telescopes as a powerful, complementary probe of light DM, with substantial improvements expected from upcoming IceCube-Gen2 and KM3NeT observations.

This paper presents the performance of the identification of jets containing $b$-hadrons ($b$-jets) for the GN2 algorithm, a transformer-based model for jet flavour tagging, using data collected by the ATLAS detector at the LHC. The analysis uses proton-proton collision data recorded in 2022 and 2023 at a centre-of-mass energy of $\sqrt{s} = 13.6$ TeV, corresponding to an integrated luminosity of 56 fb$^{-1}$. The $b$-jet identification efficiency and jet flavour composition are extracted simultaneously from a sample enriched in top-quark pair events ($t\bar{t}$). This efficiency is measured as a function of the jet transverse momentum in the range of 20-400 GeV and across six intervals of cumulative efficiency as derived in simulated $t\bar{t}$ events: [100%, 90%], [90%, 85%], [85%, 77%], [77%, 70%], [70%, 65%], and [65%, 0%]. The GN2 algorithm demonstrates significant performance improvements over its predecessor DL1d, achieving up to a factor of two (three) higher rejection of light-flavour (charm-flavour) jets at the same $b$-jet efficiency. The measured efficiencies in data are compared with simulation to derive correction factors ranging from 0.9 to 1.3. The total uncertainty is around 1% for jets with transverse momentum larger than 60 GeV in the [65%, 0%] interval of cumulative efficiency.

Parametric resonance in a Higgsed Abelian sector provides an efficient mechanism for producing vector dark matter, but its viability depends crucially on the origin of the initial dark-Higgs displacement that seeds the resonance. In this work, we investigate this initial-condition problem in a weakly coupled Abelian-Higgs theory with potential $V=λ_4(φ^2-v^2)^2/4$, using the calibrated nonlinear broad-resonance relic map together with a stochastic inflationary analysis of the dark-Higgs condensate. We show that a minimal light-spectator realization fails under standard inflationary duration: while broad resonance and isocurvature constraints require \( φ_0/H_I \gtrsim 3.3\times10^4, \) the stochastic equilibrium and finite-duration random walk produce only \( φ/H_I=\mathcal O(1). \) This large displacement mismatch is robust against order-of-magnitude variations in the resonance efficiency and broadness threshold, establishing a model-independent obstruction to the stochastic branch. We then identify a distinct classically sourced branch, generated by a negative Hubble-induced mass, in which the condensate tracks a time-dependent minimum, \( φ_0=κH_*/\sqrt{λ_4}, \) and the radial fluctuation remains heavy during inflation. In this case, the fixed-$e/λ_D$ relic scaling shifts from \( m_X\propto λ_4^{5/8}H_I^{-3/2} \) to \( m_X\propto κ^{-3/2}λ_4 H_*^{-3/2}. \) We derive the simultaneous consistency conditions for this sourced branch, including broad resonance, adiabatic tracking, perturbativity, sub-Planckian displacement, thermal non-erasure, spectator backreaction, and control of inflationary vector fluctuations. Our results establish that Higgsed-vector resonance is not merely a dark-matter production mechanism, but a sensitive probe of the inflationary and reheating dynamics that determine its initial conditions.

Despite compelling observational evidence for dark matter (DM), its fundamental physical properties remain poorly understood. In this report, we propose a three-step light-shining-through-walls (LSW) experimental scheme utilizing the high-brilliance, high-energy X-rays from the ID21 Hard X-ray Imaging Beamline at the High Energy Photon Source (HEPS) to search for signatures of dark photons (DPs) and other weakly interacting slim particles (WISPs). The scheme includes three steps of LSW experiments: a short-term (several days) dedicated exposure experiment, a long-term (several years) synchronous accompanying experiment, and a WISP detection with strong magnetic fields. Projection results show that this HEPS-based LSW experiment can effectively constrain DP parameters in the 1 eV--400 keV mass range, covering unexploited parameter space of the existing X-ray LSW experiments. It provides a least model-dependent and most purely-laboratory approach for probing dark sector particles and advancing new physics research beyond the Standard Model gradually.

The Payload for Ultrahigh Energy Observations (PUEO) is a NASA balloon-borne instrument for the detection of ultra-high energy (UHE) neutrinos with energies above $10^{17.5}~\textrm{eV}$ via either the Askaryan effect or geomagnetic emissions from an upward-going air shower. The main instrument trigger system for PUEO is a fully digital supersample rate beamformer based on 24 Xilinx Radio Frequency System-on-a-Chip (RFSoC) digitizers sampling 192 channels operating at $3~\textrm{GSa/s}$ and a system clock frequency of $375~\textrm{MHz}$. The trigger implements frequency band conditioning, dynamic radio-frequency interference (RFI) rejection, and matched filtering, with significant emphasis on optimization to reduce both the power and resource usage while maintaining sensitivity. The system implements 48 total synthetic antenna beams with up to 8 antennas each, covering a $\sim25^\circ$ range in zenith and $\sim60^\circ$ range in azimuth. Preflight testing demonstrated a trigger performance of a minimum signal-to-noise ratio (SNR) of $\sim1.5$ using simulated signals while consuming between $5-7~\textrm{W}$ in the trigger logic.

We study the internal spin polarization of Proca stars formed by gravitational Bose--Einstein condensation of a three-component nonrelativistic vector field. In idealized periodic-box simulations, we decompose the aperture-averaged spin into a coherent net fraction, a local polarization fraction, and their ratio, thereby distinguishing genuine coherent core polarization from local spin density whose direction cancels inside the aperture. For independent vector components, condensation produces Proca stars that are sizably but not maximally polarized. Across an independent-component simulation ensemble, the coherent core-spin fraction has mean $\langleχ_{\rm net}\rangle\simeq0.62$, with substantial realization-to-realization scatter. We interpret this scatter as the outcome of random elliptical polarization of the dominant component-space mode, rather than as evidence for a universal Proca-star spin fraction. This interpretation is supported by the core polarization matrix: its leading eigenvector provides an estimate of the ideal single-mode spin fraction, while the difference between this estimate and the directly integrated coherent spin tracks the departure of the core from a rank-one component-space state. The measured leading-eigenvector spin fractions are broadly compatible with an isotropic random-complex-vector reference and less compatible with an equal-amplitude random-phase reference. Correlated and circular initial data drive the dominant component-space mode toward the circular-polarization bound, giving the ordering independent $\rightarrow$ correlated $\rightarrow$ circular. These results show that internal polarization is a genuine vector degree of freedom of gravitationally condensed nonrelativistic Proca stars, and that the resulting core spin is controlled by the polarization of the dominant condensed mode rather than by a fixed universal value.

We analyze the finite-temperature two-point function of the large-$N$ SYK model at intermediate couplings away from the infrared fixed point. Specifically, we examine its analytic structure in the complex time plane, tracking the complex-time singularities over a range of temperatures. The location of the leading singularity lies on the imaginary axis. It controls the short-time dynamics of operator complexity, defining an `effective temperature' for the correlator. The next-to-leading singularity lies outside the thermal strip set by the above effective temperature. It has been argued that this could be interpreted in terms of bouncing null geodesics in the emergent black hole geometry. Both these singularities persist all the way down to zero temperature. We discuss our observations and motivate the related emergent geometry using a kinematic space perspective.

We present results from a comprehensive multi-wavelength monitoring campaign of the changing-look active galactic nucleus 1ES 1927+654 during the onset and evolution of a radio jet (May 2022-August 2025). Using observations from XMM-Newton, Swift, TNG, ZTF, VLA, and VLBA, we characterize the spectral evolution of the source. Soft X-ray emission lines at ~0.56 keV and ~1 keV have appeared with variable strength and width during the formation of a nascent jet, with the ~1 keV feature persisting since the post-2017 flare phase. We also report the detection of a broad (~800 eV) Fe K emission feature at ~(6-7) keV in ~70 ks of stacked EPIC-pn spectra, marking the first such detection in this historically featureless source. Joint spectral fitting of XMM-Newton EPIC-pn and RGS data reveals the presence of ionized absorbers in 2022, followed by weaker absorption from 2023 to 2025. The emergence of emission features concurrent with the decline of ionized absorption suggests a transition in the inner accretion and outflow processes, indicating the reflection and reprocessing of X-rays from the inner accretion disk during jet formation. The apparent weakening of ionized outflows as the jet develops supports a scenario in which accreting material is preferentially channeled into the jet rather than expelled as winds. Furthermore, both the 0.3-2 keV soft X-ray and 5 GHz radio fluxes, which have increased by factors of ~10 and ~60, respectively, since 2022, have recently plateaued at elevated levels. Combined with steady optical emission, this indicates a stabilized accretion disk, corona, and jet configuration. Finally, the absence of broad optical emission lines suggests that the broad-line region is either not along our line of sight or insufficiently illuminated by the central source.

We describe the variable flavor number scheme to three-loop order, which modifies the massless parton densities by single- and two-mass effects and introduces heavy-quark parton distribution functions for charm and bottom. A renormalization group analysis shows the validity of this picture at large scales $Q^2$, where it resembles the non-power-suppressed heavy-flavor corrections completely. We also provide numerical implementations of a series of charged and neutral current Wilson coefficients.

The development of Group Theory in Mathematics as well as its impact in Physics has been spectacular during the twentieth century, and more particularly these last fifty years. If its contribution to Particle Physics deserves a special consideration, its usefulness cannot be neglected in other domains like Statistical Physics for example, and might also be of interest in Theoretical Biology. Some of these aspects will be examined.

A direct detection of the cosmic neutrino background (CNB) in laboratories on Earth has been called the ``holy grail'' of experimental neutrino physics, but a still more glorious prize awaits. Beyond simply detecting the presence of relic neutrinos and measuring their flux, one may aspire to measure their energy distribution, polarization, anisotropies, temporal variation, and other properties. In this work we focus on the CNB wind, which is the approximately dipolar anisotropy in the CNB flux resulting from the relative velocity of the CNB rest frame and the lab frame. We consider a CNB detection strategy based on measuring the angular distribution of recoiling electrons at the tritium $β$-decay endpoint. In order to quantify the difficulty of detecting the CNB wind, we calculate the required exposure (detector mass times observation duration) for a $3σ$ discovery. We find that detecting the CNB wind would require an exposure that is at least $10^{5}$ times larger than what's required for detecting the CNB flux alone. Additionally if the experimental energy resolution were to exceed the neutrino mass scale, then an exceptionally good control of systematic uncertainties would also be required. For nonrelativistic neutrinos, the Majorana wind signal is suppressed relative to the Dirac case by the cancellation of the leading helicity-odd angular-correlation term, leading parametrically to an exposure penalty of order $(m_ν/T_ν)^2$.

A search for nonresonant triple Higgs boson (HHH) production in the final state with six bottom quarks is performed using proton-proton collisions at $\sqrt{s}$ = 13 TeV corresponding to an integrated luminosity of 138 fb$^{-1}$ recorded by the CMS experiment. No significant excess of events over the background prediction is seen. Observed (expected) 95% confidence level upper limits on the signal cross section are set at 44 (43) fb, corresponding to 588 (572) times the standard model expectation. The observed (expected) constraint on the trilinear coupling modifier $κ_3$ is $-$7.4 $\lt$ $κ_3$ $\lt$ 12.4 ($-$6.4 $\lt$ κ_3 $\lt$ 11.2), assuming the quartic coupling modifier $κ_4$ = 1. The corresponding constraint on $κ_4$ is $-$177 $\lt$ $κ_4$ $\lt$ 185 ($-$180 $\lt$ $κ_4$ $\lt$ 190), assuming $κ_3$ = 1. This analysis provides the most stringent constraint to date on nonresonant HHH production and excludes, for the first time, part of the $(κ_3,κ_4)$ space allowed by the perturbative unitarity bound.

Like the Gell-Mann--Low coupling $α$ of QED, the effective charge $α_{g_1}$ is a physical, observable-defined running coupling that can serve as the QCD coupling $α_s$. It can be represented by an analytic form consistent with renormalization-group evolution, light-front holographic QCD, and the world data on $α_s$. As an observable, $α_{g_1}$ has a physically relevant analytic structure that reflects the underlying partonic dynamics. It displays, in the long-distance regime, two imaginary conjugate singularities at $Q^2=\pm iΛ_s^2$, with $Λ_s$ the QCD scale. To connect this structure to confinement, we use known relations between $α_s$ and the parton dressed propagators and vertices. Since, unlike vertices, propagators characterize a field variation between separate locations, the two singularities are assigned to the propagators. This results in the parton propagators displaying a long-distance behavior $e^{-Λ_s|x|/\sqrt{2}}|x|^{-5/2+d_a}$ where propagation is suppressed beyond distances of order $1/Λ_s$, as expected from confinement. This offers an intuitive interpretation of confinement as the suppression of QCD Green's functions at long distance.

We extend the coadjoint orbit approach to bosonization systematized by Delacretaz-Du-Mehta-Son to double-scaled non-relativistic Fermi liquids with central inverted harmonic oscillator potential. In 1+1d, this reproduces the Das-Jevicki collective field theory describing quasiparticle excitations of the $c=1$ matrix model. In 2+1d, this yields an effective field theory for noncritical M-theory, namely the 2+1d Fermi liquid proposed by Hořava and Keeler as a unified framework for characterizing noncritical string vacua. With a suitable gauge choice this theory reduces to a continuous family of 1+1d chiral bosons resembling Das-Jevicki collective fields coupled by additional interactions. Utilizing the underlying integrability of noncritical M-theory, we compute select fermion density correlation functions order-by-order in perturbation theory, and provide evidence that they agree with their Fermi liquid counterparts in a suitable semiclassical limit. This represents a step toward an effective spacetime gravity description of noncritical M-theory, which we expect to shed light on both the landscape of 2d noncritical strings and 3d quantum gravity in general.

We establish a rigorous geometric correspondence between thermal partition functions of discrete-spectrum quantum systems with bounded ground energy and the Chern character of "virtual physical sheaf" over spacetime. By interpreting Hamiltonian dynamics as a $U(1)$-equivariant flow on the quantum phase space $\mathbb{CP}^n$ and pushforward to spacetime, we show that the finite-temperature partition function emerges as the integral of the Chern character of "virtual physical sheaf" over spacetime. The construction extends naturally to infinite dimensions through trace class guaranteed by Weyl's aspmtotic law. Using the Grothendieck-Riemann-Roch formalism, we prove pushforward invariance of the Chern character under thermal compactification on arbitrary manifolds, providing a topological foundation for thermal traces in quantum field theory. This framework unifies spectral theory with characteristic class theory, offering a geometric interpretation of partition functions based on operator-algebraic approach.

Fully developed isotropic turbulence exhibits a dual nature: a continuous, scale-invariant energy cascade coexists with discrete, intense vortex filaments. We show that this duality arises from a spontaneously broken SO(3) gauge symmetry. By identifying the specific angular momentum $\mathbf{L} = \mathbf{r}\times\mathbf{u}$ as a non-Abelian gauge connection and the radial velocity $u_r$ as a Higgs field, the turbulent vacuum is described by the SO(3) Georgi-Glashow model. When the radial strain condenses, the symmetry breaks SO(3) $\to$ U(1), generating a topological mass gap $M_W = gv$. This gap partitions the energy into a massless U(1) sector (the solenoidal background) that sustains the Kolmogorov cascade, and a massive SO(3)/U(1) sector that is confined to vortex filaments. Using high-resolution DNS data (JHTDB, $Re_λ\approx433$), we empirically verify three key predictions: (i) the energy spectra obey a strict 1:2 equipartition over the inertial range, with a sharp divergence at $M_W \approx 40$; (ii) the radial Higgs field extracted around isolated vortex cores follows the exact BPS monopole profile $H(r)=\coth(r/η)-η/r$ with $η= 0.0093$ domain units and the VEV $v = 0.338$, identifying the ubiquitous "worms" as macroscopic 't Hooft-Polyakov monopoles; (iii) the Wilson loop computed from the velocity field exhibits a clean area law $\langle W_C \rangle \sim e^{-σA}$ with string tension $σ= 0.303 \pm 0.009$, directly confirming the confining nature of the turbulent vacuum.

In generalized, also known as $p$-brane Galilei limits, the speed of light $c$ becomes infinite in the directions transverse to a $(p+1)$-dimensional Lorentzian worldvolume. In this paper, we explain that allowing the worldvolume to be Euclidean, and thus time to be transversal, $p$-brane Galilei limits turn into Carrollian ones, that we refer to as "Euclidean $p$-brane Carroll limits", in which $c$ goes to zero in the $p+1$ worldvolume directions. This leads to a unified approach to taking Galilean and Carrollian limits, whose consequences we explore for $p=0$. We show that the spacetime symmetries that arise from the Euclidean 0-brane Carroll limit can be centrally extended to what we will call the Alice algebra, similar to how the Bargmann algebra centrally extends the Galilei symmetries. This gives rise to the novel notion of an Alice particle, and we obtain the Bargmann and Alice particle actions from a unified limit of the action of a relativistic massive particle or tachyon, suitably coupled to a one-form gauge potential. In the presence of a cosmological constant, we find that the Bargmann and Alice particles undergo stable motion for negative and positive cosmological constant, respectively. Finally, we show that the Bargmann and Alice particle actions can be obtained from null reduction of a massless particle action in a relativistic spacetime with one and two times. Our results indicate that in 10 dimensions, the Alice particle is a decoupling limit of a D$0{}^*$-brane in Hull's type IIA${}^*$ theory, similar to how the Bargmann particle is related to a D$0$-brane in type IIA string theory.

We review in the present talk a series of recent works on the flavor decomposition of the gravitational form factors of the proton and its mechanical properties within the framework of the chiral quark-soliton model. Starting from the energy-momentum tensor operator derived from the QCD instanton vacuum, we carry out the twist projection of the energy-momentum tensor operator into its twist-2 and twist-4 components, which enables us to isolate the $\overline{c}$ form factor originating from the twist-4 operator. We present the flavor-decomposed mass, spin, pressure, and shear-force distributions of the proton, together with the corresponding form factors. While the up quark dominates both the mass and the spin of the proton, the strange quark is found to contribute sizably to the $D$-term form factor. We also discuss the mechanical stability of the proton governed by the pressure and shear-force distributions.

By analyzing $(2712.4 \pm 14.3) \times 10^{6}$ $ψ(3686)$ events collected with the BESIII detector operating at the BEPCII collider, the decays $χ_{cJ} \to pK^{-}\barΛη+ \mathrm{c.c.}$ ($J=0,1,2$) are observed for the first time, with statistical significances exceeding $5σ$ for all three $χ_{cJ}$ states. The measured branching fractions are $\mathcal{B}(χ_{c0} \to pK^{-}\barΛη+ \mathrm{c.c.}) = (5.3 \pm 0.7 \pm 0.5) \times 10^{-5}$, $\mathcal{B}(χ_{c1} \to pK^{-}\barΛη+ \mathrm{c.c.}) = (9.8 \pm 0.6 \pm 0.6) \times 10^{-5}$, and $\mathcal{B}(χ_{c2} \to pK^{-}\barΛη+ \mathrm{c.c.}) = (9.3 \pm 0.6 \pm 0.6) \times 10^{-5}$, where the first uncertainties are statistical and the second are systematic. Structures consistent with the known hyperon resonances $Λ(1520)$ and $\barΛ(1690)$ are seen in the $pK^{-}$ and $\barΛη$ invariant mass spectra, respectively. The reported branching fractions include both resonant and non-resonant contributions. These results provide new experimental information on hadronic decays of $P$-wave charmonium states and contribute to the understanding of baryon production and hadronization dynamics in the nonperturbative QCD regime.

Wavelength-shifting (WLS) techniques enable particle detectors based on noble gases, whose scintillation light is predominantly emitted in the vacuum-ultraviolet. We investigate WLS fibers coated with tetraphenyl butadiene (TPB) for scintillation light detection in gaseous xenon and argon at pressures up to 8.5 bar, motivated by future high-pressure xenon time-projection chambers of the NEXT program. Two detector configurations are studied: an elongated high-pressure vessel with four PTFE panels equipped with WLS fibers read by temperature-stabilized SiPMs, and a compact box-shaped detector operated at 1 bar Xe with WLS fibers read out by PMTs. Both operate with continuous gas purification. The detector response is characterized using cosmic muons and alpha particles from a $^{241}$Am source. With the SiPM setup, we measure a light collection efficiency (LCE) of ${1.18 \pm 0.01~\mathrm{(sta.)}~^{+0.07}_{-0.09}~\mathrm{(sys.)}~\%}$ for xenon and ${1.07 \pm 0.01~\mathrm{(sta.)}~^{+0.06}_{-0.08}~\mathrm{(sys.)}~\%}$ for argon. With PMT readout, we measure a LCE of ${0.45 \pm 0.01~\mathrm{(sta.)} \pm 0.05~\mathrm{(sys.)}~\%}$ in xenon, in agreement with the SiPM result once photon detection efficiency is accounted for. Average scintillation waveforms in xenon and argon are studied to assess the time structure of the emitted light. Cosmic-muon measurements yield a mean energy required to produce a scintillation photon $45\pm7~\mathrm{(sta.)}~^{+4}_{-5}~\mathrm{(sys.)}~\mathrm{eV}$ at 1.5 bar, in agreement with the literature. The results demonstrate that TPB-coated WLS fiber systems can reliably detect scintillation light in high-pressure gaseous noble detectors, with a LCE representing an upper limit for realistic large-scale TPCs, where additional photon losses from materials and fiber attenuation are expected.

Current and future neutrinoless double-beta decay ($0νββ$) searches are usually characterized by half-life limits or projected sensitivities. For operator-level interpretation, however, the relevant question is not only the strength of the limit, but also the stability of the inferred constraints on low-energy lepton-number-violating (LNV) coefficients under changes of isotope and nuclear matrix-element (NME) method. Using $ν$DoBe, we convert half-life limits or projected sensitivities into limits on low-energy LNV coefficients for several choices of isotope, operator class, and NME method. We then compare the resulting limits across isotopes and NME methods. We quantify the stability by the spread of coefficient limits in non-observation scenarios and, for single-operator post-discovery benchmarks, by the spread of isotope-to-isotope half-life ratios. For the operator examples considered in this paper, the light-Majorana exchange mechanism gives the most stable interpretation. The selected dimension-six operators remain reasonably robust, while the short-range dimension-nine examples show a stronger dependence on isotope and NME-method choices. The same hierarchy appears in post-discovery benchmarks based on isotope-to-isotope half-life ratios: the reliability of such tests depends on whether the predicted half-life pattern remains stable across NME methods. A multi-isotope $0νββ$ program therefore provides information beyond half-life reach by testing the stability of the operator interpretation. This spread-based analysis provides a practical way to identify which low-energy operator scenarios are suitable for isotope-ratio tests after a possible discovery, and which require more care before being connected to ultraviolet models of lepton-number violation.

We study anomalous Wess-Zumino-Witten terms associated with the chiral anomaly in dense QCD matter under magnetic fields and rotation. By introducing electromagnetic, baryon number, and isospin background gauge fields, we write down the topological couplings of neutral mesons for the $N_f=2$ and $N_f=3$ cases. The resulting terms contain characteristic contributions proportional to $\vec{B} \cdot \vec{\nabla} φ$ and $\vecΩ \cdot \vec{\nabla} φ$, where $φ$ denotes $π^0$, $η$, or $η'$. These results are relevant to chiral soliton lattices in dense rotating matter.

We present the first functional QCD calculation of the pion distribution amplitude (DA) using the large-momentum effective theory within the functional renormalisation group (fRG) approach. With only the strong coupling and current quark masses as inputs, we compute the quasi-DA from first-principles QCD correlation functions. By pushing the pion momentum up to $P_z = 4.5\ \mathrm{GeV}$, the quasi-DA becomes fully saturated, rendering the extrapolation errors to the light-cone limit negligible. The resulting second-order moment $\langle ξ^2 \rangle_π= 0.267$ is significantly smaller than existing lattice-LaMET determinations and lies in a range consistent with other nonperturbative approaches.

Quantum Chromodynamics (QCD) is explored at the Large Hadron Collider (LHC) over a wide range of energy scales, from soft hadronic interactions to hard parton scattering. This contribution presents a selection of recent measurements from the ATLAS and CMS experiments probing QCD in proton--proton collisions at the LHC. The results span a broad range of momentum scales, from studies of charged-particle event-shape observables in minimum-bias events, which are sensitive to soft QCD dynamics and hadronisation, to measurements of inclusive di-jet production that provide stringent tests of perturbative QCD. Recent studies of double-parton scattering in same-sign $WW$ production are also discussed, offering new insight into multi-parton interactions inside the proton. Finally, the capabilities of the forward proton detectors operated by the ATLAS and CMS collaborations are briefly reviewed, highlighting their contribution to the LHC physics programme through measurements of diffractive and photon-induced processes.

In this article, we have presented one of the very important observations, regarding the behavior of mutual information of two different sets of subsystems in the after Page time scenario. This provides us with some deep insights about the conservation of geometrical entanglement. In our earlier works, we have shown how the saturation of mutual information between two specific subsystems plays a vital role in obtaining the correct Page curve for the eternal black hole. In those works, we have shown that the mutual information between $B_+$ and $B_-$, that is, $I(B_{+}: B_{-})$, vanishes at scrambling time, which leads to the correct Page curve. That means that at scrambling time, there is no correlation between $B_+$ and $B_-$. Remarkably, it is observed that at this particular value of observer's time, the mutual information between $\mathcal{I}$ and $R$, that is, $I(\mathcal{I}:R)$, becomes infinity. This indicates that the regions $\mathcal{I}$ and $R$ become maximally entangled. This provides us with a notion of conservation of geometric correlation between different regions on the Cauchy slice. In this work, we have also provided a way to calculate the tripartite mutual information of regions $\mathcal{I}$,$R_+$ and $R_-$, that is, $I(\mathcal{I}:R_+:R_-)$ on the Cauchy slice using the earlier results involving the bipartite regions. This is a new result which was missing in the earlier literature.

The molecular states in coupled channel system of $B_s^0N-\bar BΛ-\bar B^*Λ-\bar BΣ-\bar B^*Σ$ are investigated within one-boson-exchange model that includes $S$-$D$ wave mixing and a tunable short-range $δ(r)$ term. Bound, resonant, and virtual states are searched by analytically continuing the $S$ matrix in the complex energy plane. In the single-channel analysis, the $\bar BΣ$ and $\bar B^*Σ$ systems with $J^P=1/2^-$ and $3/2^-$ are found to be attractive and form three bound states with reasonable cutoff region, corresponding to $1/2^-(\bar BΣ)$, $1/2^-(\bar B^*Σ)$, and $3/2^-(\bar B^*Σ)$ . When coupled-channel dynamics is included, these states evolve into near-threshold poles that produce enhancements in the $B_s^0N$, $\bar BΛ$, and $\bar B^*Λ$ invariant mass spectra. The pole below the $\bar B^*Σ$ threshold in the $J^P=1/2^-$ system depends strongly on the treatment of the short-range $δ(r)$ term and becomes a virtual state when this contribution is removed, while the other two resonant poles and their line shapes are only moderately affected. For representative parameter region, the predicted masses of these states lie in the $6.44-6.52$ GeV region and have widths of a few to several MeV. The pole coupling and partial-width analyses indicate that the poles are generated mainly by the $\bar BΣ$ and $\bar B^*Σ$ channels interaction, and their observable signals are expected mainly in the lower open channels $B_s^0N$, $\bar BΛ$, and $\bar B^*Λ$. These results support the existence of near-threshold molecular pentaquark $P_{\bar b s}$ and provide useful guidance in future experimental searches.

The equation of state (EoS) of strongly interacting matter at low temperature and high baryon density is a central ingredient in the physics of compact stars, but it is still difficult to determine directly from first-principles QCD calculations because of the sign problem. Two-colour QCD (QC$_2$D) provides a useful and controllable laboratory: for an even number of fundamental flavours the fermion determinant is real and positive, while the theory shares the nonperturbative properties with three-colour QCD at least zero chemical-potential regime. In this short review we summarize recent lattice progress on dense QC$_2$D, with emphasis on the phase structure, the emergence of superfluidity, the Bose--Einstein-condensation (BEC) to Bardeen--Cooper--Schrieffer(BCS) crossover, and thermodynamic quantities. A particularly interesting outcome is that the sound velocity in the cold dense superfluid regime can exceed the conformal value $c_{\rm s}^2/c^2=1/3$. This behaviour, now seen in the simulations for several QCD-like theories, gives a stiff strongly interacting matter. It is expected to offer insight into at least some aspects of the physics realized inside neutron stars.

Recently, Calkins and Pate calculated the multi-particle operator product expansions (OPEs) of single-particle celestial operators with two-particle celestial operators in pure Einstein gravity. We apply their construction to the graviton, gravitinos, graviphotons, graviphotinos and scalars in ${\cal N}=8$ supergravity. By performing various contour integrals on these multi-particle OPEs, we obtain ninety-five (anti)commutators for the single-particle contributions from the celestial soft current algebra (derived from the single-particle OPEs) in a two-dimensional boundary. Moreover, these ninety-five multi-particle OPEs provide nontrivial relations between the corresponding celestial amplitudes. The triple-collinear limits between these bosonic and fermionic particles from the four-dimensional bulk are also described as a double check. Finally, we implicitly propose both the multi-particle OPEs of single-particle celestial operators with $(N-1)$-particle celestial operators and the corresponding (anti)commutators for the modes of celestial operators that have linear terms (or single-particle exchange terms due to the sum of $N$-particle factorization channels from amplitudes) on the right-hand sides.

We revisit the species scale in quantum gravity from the viewpoint of effective field theory (EFT). Two characterizations are commonly used: one defines it as the energy at which the perturbative description of quantum gravity breaks down, as inferred from the one-loop correction to the graviton propagator; the other identifies it with the suppression scale of higher-derivative operators in the gravitational effective action. We clarify the relation between these characterizations by analyzing the cutoff-scheme dependence of the one-loop tower contribution. For a large number of species, the leading tower-enhanced local correction is regulator dependent and should be interpreted as EFT matching data, while the subleading logarithmic contribution is universal within the class of proper-time cutoff profiles considered here. As a concrete application, we compute the full four-derivative one-loop corrections from Kaluza-Klein towers. Our results separate regulator-stable logarithmic data from scheme-dependent local threshold contributions, providing a controlled EFT interpretation of the relation between perturbative-breakdown and higher-derivative definitions of the species scale and explaining why their coefficient-level identification is not regulator independent.

We study exclusive electroproduction of dijets in the collinear factorization framework as a probe of generalized parton distributions (GPDs). For quark dijet production, we extend previous analyses by including contributions from helicity GPDs and by assessing an additional leading-order electromagnetic channel governed by elastic nucleon form factors. Furthermore, we investigate exclusive gluon dijet production. We compare our prediction to HERA data and provide projections for measurements at the future Electron-Ion Collider.

In the present study, we extend our previous analysis of the proton electromagnetic form factors (FFs) extracted from exclusive photon leptoproduction (EP) measurements in kinematic regions where the Bethe-Heitler (BH) process dominates the cross section by including all currently available high-precision EP data from the CLAS and Hall~A Collaborations. Using the same phenomenological framework, we investigate the consistency among the different data sets, determine the proton electromagnetic FFs within several fitting scenarios, and extract the corresponding charge and magnetic radii. A significant tension is observed between the CLAS 2018 measurements and the remaining EP data. We show that excluding this data set, or restricting its kinematic coverage by imposing suitable low-$|t|$ cuts, leads to stable fits with good quality and consistent FFs. For all analyses, the extracted proton charge radius is smaller than the Particle Data Group average and most determinations based on elastic electron-proton scattering. However, the results are consistent, within uncertainties, with the PRad measurement and muonic hydrogen spectroscopy. In contrast, the magnetic radius is found to be compatible with the current world average. These results demonstrate that BH-dominated EP measurements provide an independent and complementary approach to determine the electromagnetic structure of the proton and offer additional support for the small-radius solution of the proton charge-radius puzzle.

Neutrino oscillations provide compelling evidence for physics beyond the Standard Model, while the weakly interacting nature of neutrinos makes them powerful probes of new interactions and hidden sectors. In this work, we investigate a \textit{dark neutrino portal} scenario in which neutrino mass generation is linked to a light dark sector charged under a new $U(1)_D$ gauge symmetry. While Standard Model fields remain neutral under $U(1)_D$, the dark neutrino sector is charged and communicates with the Standard Model exclusively through active--dark neutrino mixing. The associated neutrinophilic mediator induces ultra-long-range interactions, whereby electrons and neutrons in the Earth, Moon, Sun, Milky Way, and the cosmological matter distribution generate sizable matter potentials that modify neutrino oscillations. We explore the sensitivity of the upcoming Deep Underground Neutrino Experiment (DUNE), whose long baseline and pronounced matter effects make it uniquely suited to probe such interactions. We show that DUNE can access previously unexplored regions of parameter space and demonstrate that the same underlying coupling can simultaneously give rise to sizable neutrino self-interactions, including regions relevant for alleviating the Hubble tension, while remaining consistent with current neutrino oscillation constraints.

Arroyo-Ureña, Hernández-Tomé, López-Castro, Roig, and Rosell (AHLRR) take the structure-dependent (SD) amplitude for $τ\to P ν_τγ$ from the resonance chiral theory result of Guo and Roig (GR10), obtaining $δr_{SD} = +0.15\%$ in the pion channel. The convention dictionary in footnote 2 of GR10, repeated in footnote 1 of AHLRR, has the wrong sign in its axial entry: the printed amplitudes imply $F_A^{DF} = -2\sqrt{2}\, m_P F_A^{GR}$, not $+2\sqrt{2}\, m_P F_A^{GR}$. With the corrected dictionary, the GR10 amplitude agrees with the $O(p^4)$ chiral prediction in the axial sector but carries the opposite sign to the chiral anomaly in the vector sector; the relative sign fixed in the Decker-Finkemeier addendum is confirmed in both sectors. The identification in footnote 6 of AHLRR of their value with the pre-addendum result rests on the erroneous dictionary and does not hold. The numerical impact is small: $δr_{SD}$ shifts from $+0.150\%$ to $+0.146\%$ in the pion channel and from $+0.18\%$ to about $+0.16\%$ in the kaon channel.

We present an analysis of one-loop partition functions for scalars in AdS$_{d+1}$ obeying the Neumann boundary condition and explore phases of scalar field theories in several dimensions at zero and finite temperature. The partition function computation involves an analytic continuation from $Δ_+$ corresponding to the Dirichlet boundary condition to $Δ_-$ corresponding to Neumann boundary condition. We show that this can be implemented by deformations of the contour for integral over eigenvalue ($λ$) of the Laplace operator as compared to the integral over ${\mathbb R}$ in [arXiv:2201.09043] for the Dirichlet boundary condition. We further contrast these phases with those appearing for the case of scalars obeying the Dirichlet boundary condition and corroborate the occurrence of these phases by studying the long range behaviour of correlators.

We study the energy-momentum tensor of stationary rotating topological solitons in a (2 + 1)-dimensional $\mathbb{C}P^2$ nonlinear sigma model with a stabilizing potential term. We evaluate the distributions of the corresponding shear forces and pressure and study the stability criteria for these solutions. Our results suggest that these solitons become classically unstable for some range of values of the parameters of the system.

We construct an effective field theory description of relativistic superfluids with nonzero angular momentum density. At first order in the derivative expansion, the effective action contains a term with a quantized coefficient, which encodes the Berry phase for the angular momentum of the superfluid condensate. From this single term we derive a number of physical effects, including the relativistic Mermin-Ho relation, an anomalous Ettingshausen effect (exchangeable for an anomalous Hall effect), and an anomalous Hall viscosity.

We study the processes $e^+e^-\toΥ_J(1D)η$ and $e^+e^-\toΥ_J(1D)π^+π^-$ at center-of-mass energies $\sqrt{s}$=(10.73 -- 11.02) GeV using a $142.5\,\mathrm{fb}^{-1}$ data sample, including 122~fb$^{-1}$ near the $Υ$(10860) peak ($\sqrt{s}$ = 10.866 GeV), collected with the Belle detector at the KEKB asymmetric-energy $e^+e^-$ collider. From the peak sample, the products of Born cross section times branching fraction are obtained for $σ_{\rm Born}(e^+e^-\toΥ_J(1D)η)$ or $σ_{\rm Born}(e^+e^-\toΥ_J(1D)π^+π^-)$ and ${\cal B}(Υ_J(1D)\toχ_{b1}γ)$ or ${\cal B}(Υ_J(1D)\toχ_{b2}γ)$ for each $Υ_J(1D)$ state. The corresponding branching fractions for $Υ(10860)$ decays are also obtained. The significances of the $Υ_1(1D)$, $Υ_2(1D)$, and $Υ_3(1D)$ signals are 4.8$σ$, ${>}10σ$, and 3.0$σ$, respectively, including systematic uncertainties. The mass for $Υ_2(1D)$ is measured to be $(10167.0\pm 1.0\pm 0.2)$ MeV/$c^2$, where the first and second uncertainties are statistical and systematic. The mass splittings $Δm_{12}=m(Υ_2(1D))-m(Υ_1(1D))$ and $Δm_{23}=m(Υ_3(1D))-m(Υ_2(1D))$ are $(11.8\pm1.5\pm0.4)$ MeV/$c^2$ and $(7.6\pm2.4\pm0.6)$ MeV/$c^2$, respectively.~We determine the energy dependence of the cross sections for $e^+e^-\toΥ_J(1D)η$ and $e^+e^-\toΥ_J(1D)π^+π^-$ for the $Υ_1(1D)$, $Υ_2(1D)$, and $Υ_3(1D)$ states, combined.

We calculate the magnetic dipole moments of doubly strange hidden-charm pentaquark states with spin-parity $J^P = 1/2^-$ and $3/2^-$ using QCD light-cone sum rules (LCSR), presenting the first systematic LCSR investigation of the electromagnetic multipole structure in the $S = -2$ sector. To assess the model dependence, we employ independent interpolating currents in diquark-diquark-antiquark form, which probe different assumptions about the internal color-spin correlations. For the spin-$3/2$ states, we also compute the electric quadrupole and magnetic octupole moments. The magnetic dipole moments exhibit a considerable spread across currents, ranging from $-2.15\,μ_N$ to $5.74\,μ_N$ for spin-$1/2$ pentaquarks and from $-4.25\,μ_N$ to $-0.43\,μ_N$ for spin-$3/2$ states, reflecting the sensitivity of magnetic moments to the internal wave function. A quark-level decomposition reveals that the charm quark dominates in most configurations, while strange quarks play a decisive role only in currents favoring axial-vector diquark structures. The electric quadrupole moments lie between $-2.01\times10^{-2}$ fm$^2$ and $5.55\times10^{-2}$ fm$^2$, and the magnetic octupole moments are typically an order of magnitude smaller. The current dependence of the magnetic dipole moments provides a quantitative measure of the theoretical uncertainty arising from the choice of interpolating operator. The pronounced isospin sensitivity of $J^3_μ(x)$ across all three multipole moments arises from its axial-vector diquark structure, which isolates the light quark from spin averaging and allows the charge asymmetry $e_u/e_d = -2$ to propagate directly into the electromagnetic moments; the ratio $μ_u/μ_d = -2.00$ confirms this mechanism exactly. Our predictions offer benchmarks for future experiments and lattice QCD calculations, and may help discriminate among competing structural models.

We present a comprehensive analysis of the $π^+p$ femtoscopic correlation functions measured by the ALICE Collaboration in high-multiplicity $pp$ collisions at $\sqrt{s}=13$ TeV. Using the Koonin-Pratt formula with a Gaussian source and data-driven $πN$ partial-wave amplitudes, we account for the contributions from $π^+p$ scattering and $Δ(1232)^{++}$-decay, thereby successfully reproducing the measured data and their transverse-mass ($m_T$) dependence. The scattering contribution yields a peak near the relative momentum $k\approx140$ MeV/$c$, whereas the decay contribution peaks around $k\approx220$ MeV/$c$. The observed correlation peak results from a weighted sum of the two contributions, with $m_T$-dependent relative weights. We find that the 140 MeV/$c$ peak originates from quantum interference between the incident and scattered waves-a mechanism previously unnoticed in femtoscopic studies. This finding resolves the peak-shift puzzle in $π^+p$ correlations and provides a novel perspective for quantum interference effects in femtoscopy.

This article provides a comprehensive overview of the application of relativistic fluid mechanics to describe the collective evolution of the Quark-Gluon Plasma (QGP) formed in ultra-relativistic heavy-ion collisions. We map out the chronological transformation of spatial eccentricities in the initial interaction volume into measurable anisotropic azimuthal momentum distributions, parameterized by the harmonic flow coefficients $v_n$. Utilizing multi-particle correlation techniques developed within the ATLAS experimental framework, we dissect the event-by-event fluctuations of the participant planes and evaluate non-linear hydrodynamic responses across higher harmonics. Furthermore, we embed local rotation fields into this continuous description by solving the covariant transport equations for subatomic vorticity. We demonstrate that while the Helmholtz-Kelvin theorem guarantees the topological conservation of vortex lines within the ideal medium, the collective multi-dimensional expansion forces a systematic 1/t power-law geometric dilution of the local rotational magnitude. Finally, we contrast different pre-equilibrium generation mechanisms and evaluate their final signatures on differential spin alignment observables.

The Super Tau-Charm Facility, operating at high luminosity, will produce high event rates and high data throughput, imposing stringent requirements on fast track finding and data reduction and compression algorithms in the High-Level Trigger. Local track segment finding in the Main Drift Chamber underpins subsequent segment combination and full track reconstruction, yet high background rates and limited detection efficiency can significantly increase the risk of false triggers and signal loss in pattern matching algorithms. This paper presents a neural-network-assisted framework for constructing binary template libraries used in matrix-based pattern matching for MDC local track segment finding. The framework formulates template construction as a differentiable multi-objective optimization problem, employing a neural network to jointly learn template parameters under multiple constraints. After training, only binary template pairs are exported and deployed into the existing bitwise pattern matching routine, requiring no neural network inference at runtime and thus preserving the deterministic, fast, and parallelizable nature of the online algorithm. Experimental results based on simulation samples demonstrate that, under limited detection efficiency, the resulting template library maintains relatively high signal retention across different transverse momentum ranges and background levels, and can be flexibly tailored to adjust the coverage range according to practical requirements. The proposed approach decouples the physics performance from the computational speed by combining the improved physics performance brought by offline neural-network-based optimization with the determinism and high speed of a conventional online algorithm, suggesting a new research direction for artificial-intelligence-enhanced online data processing in high-luminosity particle collider experiments.

We compare the size of dimension-8 effects on $W^+W^-$ production at the LHC arising from the $q\bar q$ and $γγ$ initial states. In particular, we consider bosonic operators, which contribute to the anomalous triple and quadruple gauge couplings. The relevant dimension-6 and dimension-8 operators are matched to the anomalous triple gauge couplings that contribute to the $q\bar q$ channel, allowing for the resummation of large logarithmic contributions arising in the presence of a jet-veto through the program MCFM-RE. For the $γγ$ channel, which receives contributions from both triple and quadruple gauge couplings, such resummation can be performed through the program MadGraph, by simply setting the factorisation scale to the jet-veto scale. We find that, when neglecting fermionic dimension-8 operators, the $γγ$ channel has a dominant bosonic dimension-8 contribution and this channel can be used to understand the range of validity of the Effective Field Theory. With this in mind, and carefully considering theoretical and experimental uncertainties, we provide constraints on higher dimensional operators using current and future data.

We update the analysis of minimal supersymmetric SO(10) grand unification with a universal, constrained-MSSM-like soft-breaking spectrum. A pair of Yukawa matrices in the ${\bf 10} \oplus \overline{\bf 126}$ Higgs sector fixes the charged fermion masses and quark mixing, the baryon number-violating dimension-five operators that mediate proton decay, and the right-handed neutrino Majorana masses of the type-I seesaw. The simultaneously imposed constraints include: (i) gauge coupling unification and vacuum stability; (ii) dimension-five proton decay (the dimension-six gauge mode also computed and negligible), the 125-GeV Higgs mass, and the LEP chargino limit; and (iii) $μ\to eγ$, the muon $g-2$, and electric dipole moments. We perform a global scan over the soft-breaking parameters, complemented by a constrained-$μ$ cross-check that fixes the electroweakino sector from the GUT-scale boundary conditions through radiative electroweak symmetry breaking. Subject to a single effective colored-Higgs scale $M_{H_C}$ for the proton decay normalization and to a constrained-$μ$ treatment, what survives is a narrow mini-split region: heavy, multi-tens-of-TeV scalars with $m_0 \gtrsim 7.7$~TeV and $\tanβ\lesssim 9$. We also examine the fate of the lightest supersymmetric particle as a dark matter candidate, in both the universal model and its free-$μ$ non-universal-Higgs-mass extension. The surviving region is sharply bounded and testable in the coming decade, most directly by Hyper-Kamiokande proton decay and electroweakino searches, with dark matter direct detection and electric dipole moment experiments.

We investigate the spectroscopic properties of the double strange hybrid baryon with quark content $ssqg$ within the framework of QCD sum rule. Using an interpolating current with explicit gluonic degrees of freedom, the two-point correlation function is analyzed in terms of two independent Lorentz structures, $\slashed{q}$ and $I$. The operator product expansion is carried out by including vacuum condensates up to dimension ten, and the corresponding sum rules are derived for both structures. By taking the average of the results obtained from the two Lorentz structures, we extract the masses and pole residues of the ground and first excited states. For the ground state, we obtain a mass of $\widetilde{M } = (1593.44\pm 130.29)~ \mathrm{MeV}$ and a residue of $\widetilde{λ} = (2.57\pm 0.40) \times10^{-3} ~\mathrm{GeV}^5$. For the first excited state, the corresponding values are $M = (1897.47\pm 124.44)~ \mathrm{MeV}$ and $λ= (2.88\pm 0.62) \times10^{-3} ~\mathrm{GeV}^5$. The obtained results provide theoretical predictions for the double strange hybrid baryon spectrum and may be useful for future experimental searches as well as further nonperturbative studies of hybrid hadrons.

The cross-section for photoproduction of a vector meson on a nucleus is usually determined in nucleus-nucleus collisions, which introduces a two-fold ambiguity as to which nucleus emitted the photon. This can be resolved in asymmetric collisions of protons and nuclei, where the ratio of photoproduction cross-sections on the proton and nucleus directly measures the nuclear suppression factor with minimal model dependence. The feasibilty of using existing and future proton-lead collisions at the LHC is evaluated.

Using the quantum field theory, we derive a Breit-Wigner-type formula for the $e^+ e^-$ annihilation into a vector meson and its antiparticle, and relate the formula parameters to observable quantities. The formula's soundness is checked by fitting the $e^+ e^- \to D^{\ast+}D^{\ast-}$ data published by the BESIII Collaboration in 2022.

We study integrable boundary states of the alternating SU(4) spin chain arising in ABJM theory. Starting from the $KT$-relation, we directly solve the integrability constraints for states with $n$-site translational invariance. For odd and even $n$, these constraints are reduced respectively to state equations and operator equations for the elementary $n$-site block and the matrix $K(u)$. We analyze chiral and achiral cases for $n\leq4$. In the 1-site case we allow general operator-valued integrable pairs with an internal space, while in the remaining cases we focus on $c$-number solutions.

Quasi-PDFs provide a connection between Euclidean spatial correlations in lattice QCD and lightcone parton distributions. Their perturbative expressions contain both the infrared divergence required for matching and the scheme-dependent contributions associated with the renormalization prescriptions. The separation of these two ingredients is not always transparent. In this work we use a transverse-momentum cutoff as a simple setting in which these ingredients can be systematically decomposed into a scheme-dependent sector and a remainder for the nonsinglet quark quasi-PDF at one loop. We choose the minimal transverse-momentum-cutoff scheme, where the scheme-dependent sector is identified by its explicit cutoff dependence, while the remainder contains the full collinear infrared divergence and the finite contribution relevant for matching to the lightcone PDF. After expressing the quasi-PDF in terms of distributions, we show how to deal with the linear divergence and the logarithmic terms in the counterterm, and discuss the dependence of the renormalization-group behavior on the renormalization prescriptions. This organization clarifies how scheme dependence enters the quasi-PDF before the final matching is performed, and provides a benchmark for examining analogous separations in other renormalization schemes.

The axion, a hypothetical particle proposed to solve the strong CP problem and considered a viable candidate for dark matter, has prompted extensive experimental efforts for its detection. This study presents a novel approach combining photon regeneration techniques ("light-shining-through-a-wall") with high-sensitivity Penning trap technologies to enhance the search for axions. Penning traps offer significant advantages, including precise electromagnetic field measurement, strong magnetic fields, and single-particle detection capabilities. By integrating these traps with resonantly enhanced photon regeneration using microwave cavities, our proposed method significantly increases sensitivity to axion-photon couplings. Preliminary calculations demonstrate an unprecedented achievable sensitivity, reaching an axion-photon coupling constant limit of $g_{aγγ} \le 7.10\times 10^{-8}\mathrm{GeV ^{-1}}$ in just one day, specifically targeting axion energies below 1 MHz. This experimental setup presents a robust and controlled platform, circumventing astrophysical uncertainties, and represents a substantial advancement in laboratory searches for axions and our understanding of dark matter.

We give an operator-algebraic and representation-theoretic description of the Hilbert spaces finite-$N$ bilocal holography. This work is a sequel to the finite-$N$ Hilbert space construction of arXiv:2602.20788 [hep-th]. The central result is the establishment of an invariant dual-pair operator algebra: before imposing the singlet constraint the Fock space carries commuting actions of the color group and of a bilocal Lie algebra, while the projection to the singlet sector selects a single irreducible representation of the invariant Lie algebra, which we call a master algebra. The finite-$N$ trace relations, beginning with the quadratic identities studied here, are shown to become representation-theoretic identities of the selected irreducible representation. We summarize the orthogonal, symplectic and unitary cases, identify the corresponding finite-$N$ constraints, compute the singlet Casimirs, and explain how finite traces and partition functions are obtained through characters of the resulting irreducible representations. This provides a novel, previously unknown mathematical description of the singlet space.

PQ solution to the strong CP problem probably encounters the axion domain wall problem. In this article, we propose a simple and testable solution, assuming that the $U(1)_{\rm PQ}$ possesses mixed anomaly to a hidden $SU(N_c)$ color. Then, the axion field receives a new cosine potential from the hidden instantons, which breaks the $Z_{N_{\rm DW}}$ subgroup explicitly. The new potential lifts the vacua degeneracy, but also drives the effective $θ$ angle away from the origin, re-incuring the strong CP problem. However, we find that the dark QCD scale within the 0.1 to 3 MeV window survives, maintaining a delicate balance. Two observational signatures are explored: gravitational waves from domain wall collapse, already probed by current PTAs, and di-photon signals from axion-dark-glueball mixing, which require next-generation MeV telescopes. The scenario favors a cold dark QCD sector consistent with dark glueball relic constraints.

Standard routes to a light composite Higgs either rely on tuning a single channel near criticality or protect a pseudo-Nambu--Goldstone coordinate of a coset. We introduce a third mechanism class in which the protected object is an \emph{eigenvalue} of the renormalized multi-operator scalar kernel of the strong sector. Two basis-invariant diagnostics, a sector participation number $\Psec$ and a sector gap ratio $\Rcoll$, identify collective lightness, but they cannot distinguish an accidental small determinant from a protected zero mode; the missing discriminator is the microscopic sensitivity $Δ_g=\left|\partial\ln|m_H^2|/\partial\ln g\right|$. A protected collective Higgs is defined by $\Rcoll\gg1$, $\Psec>1$, and $Δ_g=\mathcal{O}(1)$. We prove that this class is nonempty. A rank-one TC--DTC locking invariant forbids tree-level aligned curvature, while universal vectorlike DQCD bridge fermions, massless in the microscopic Lagrangian but acquiring a common DQCD constituent mass, obey $\partial_h^2\sum_A\Tr\mathcal{M}_A^2\big|_0=0$. The aligned scalar is therefore lifted only at joint two-spurion order, $m^2_{\mathrm{br}}=-(d_X/2π^2)g_T^2g_D^2(f_B^4/f_H^2)L_X$, giving $Δ_{g_T}=Δ_{g_D}=2$ at leading logarithmic order. The same DTC topology admits a collective top completion and a vector-decoupling route to reducing the positive technicolor contribution to $S$. The mechanism is falsifiable by sector-restricted lattice spectroscopy and coupling-response scans.

Two-sided octonionic $E_6^L\times E_6^R$ unification carries a nominal second colour $SU(3)_{c'}$ which, gauged, would colour the charged lepton and must be suppressed by hand. We argue no such device is needed: the left-right symmetry acts in spacetime, not internally -- in the gravi-weak reading $SU(2)_R$ is the gravitational frame group and the exchange is ordinary parity on the Dirac field; a spacetime operation cannot double the internal colour, so $SU(3)_{c'}=SU(3)_c$, one vector-like colour with a colour-singlet electron. The right sector contributes one colour-blind datum, $\sqrt m$ (values: the exceptional-Jordan trace split, an input; $N_R$ fixes only the colour representation). An anomaly no-go proven here shows every anomaly-free $U(1)$ on visible fermions lies in span$\{Q,B-L\}$, which the $\sqrt m$ pattern does not, in any sign or chirality assignment. The gauged dark-electromagnetic $U(1)_{dem}$ -- the mirror electroweak chain's unbroken remnant -- therefore carries the parity-mirror of electric charge on the dark sector: the visible fermions are dem-neutral (kinetic mixing the sole portal; no visible fifth force), and $\sqrt m$ enters visible physics only as the spectral label of the Jordan mass operator. The same recognition recovers hypercharge as the consistency relation $Y=Q-T^3_L$, $Q=N/3$, with no right-sector generator. As an application, the spontaneous parity forbids the QCD vacuum angle ($θ_{QCD}=0$, conditional on the gravi-weak identification), while the flavour rotors have real determinant, so $\arg\det M=0$ at tree level (exact for the Cabibbo rung, texture-contingent in full), coexisting with a nonzero CKM phase. Loop stability -- where minimal gauged-$SU(2)_R$ parity solutions fail -- is plausibly evaded (no gauged $W_R$) but open, pending the Higgs-bridge matrix elements. We tag throughout what is derived, what is input, and what is open.

We show that with the introduction of new right-handed neutrinos $N_R$, the Peccei--Quinn ($PQ$) transformation in the DFSZ (Dine--Fischler--Srednicki--Zhitnitsky) model depends only on the $PQ$ charge of the scalar singlet $φ$, namely $PQ_φ$, and is independent of the sine and cosine of $\arctan(v_u/v_d)$. The model consists of four new scalars: a charged Higgs boson $H^\pm$, a CP-odd scalar $A$, and two CP-even scalars with masses at the EW scale (of a few hundred GeV), and a superheavy inflaton $Φ$. The scalar fields have been presented in the form of unitary matrices, in which there is no mixing between the axion and the Goldstone boson $G_Z$. As a result, there are two kinds of couplings between the axion and fermions, namely Yukawa-like interactions and anomaly-induced ones associated with derivative axion couplings. {The model belongs to the DFSZ I kind since the axion - photon coupling with $\fr E N = \fr 8 3$.} The trilinear Higgs self-coupling is investigated. We have shown that, in the parameter region where new scalar fields such as the singly charged Higgs boson $H^\pm$ and the CP-odd scalar $A$ have masses around 150 GeV and the vacuum expectation value of one Higgs doublet ($v_u$) is of the order of a few GeV, the coupling modifier $κ_λ$-a probe of new physics-is consistent with the current experimental constraints.

We present a relativistic study of the exclusive double-charmonium processes $e^-e^+ \to γ^*γ^* \to χ_{c1}+η_c$ and $e^-e^+ \to γ^*γ^* \to h_c+h_c$ at $\sqrt{s}=10.6$ GeV in the framework of the $4\times4$ Bethe-Salpeter equation. Since these channels are forbidden in single-photon annihilation, they proceed purely through two-photon quark-rearrangement mechanisms and thus provide a clean probe of relativistic quarkonium dynamics. Using parameters fixed from mass spectroscopy in our earlier work, we obtain leading-order predictions for the total cross sections and their energy dependence. The predicted cross sections are of order $10^{-3}$ fb and exhibit a hierarchy that originates from the radial structure of the meson wave functions, which determines the overlap integrals contributing to the production amplitude. While the absolute magnitudes of the cross sections differ from NRQCD estimates, the qualitative ordering of the channels is consistent with NRQCD expectations. We find a smooth power-law falloff of the cross section for both $χ_{c1}+η_c$ and $h_c+h_c$. These results provide benchmark predictions for future searches at high-luminosity $e^-e^+$ colliders.

The Super Tau-Charm Facility (STCF) is a proposed next-generation high-luminosity electron-positron collider operating at center-of-mass energies of 2-7 GeV for precision studies of tau-charm physics. Its high event rate, detector occupancy, and background level impose stringent requirements on real-time track reconstruction in MDC, particularly for low-transverse-momentum particles with strongly curved or multi-turn trajectories. To address this challenge, we develop a GPU-accelerated matrix-based Hough transform method for online track reconstruction in the STCF MDC. Following an algorithm-architecture co-design paradigm, the data representation and computational workflow of the conformal Hough transform are reformulated for GPU execution. The original irregular parameter-space computations are organized into regular matrix-based operations, and the core computations are adapted to CUDA thread organization and the GPU memory hierarchy to exploit the inherent parallelism of the Hough transform and reduce computational and data-transfer overhead. Tests on five representative simulated physics channels with nominal background overlay show an average signal retention ratio of 93.04%, while reducing the retained hit volume to 34.92% of the original level. The GPU implementation processes 1,000 events in approximately 0.14s, achieving a speedup of 151.57 x compared with the CPU baseline. These results demonstrate that the proposed method substantially improves track reconstruction throughput while preserving track-associated hits, providing a new methodological perspective for real-time track reconstruction in future high-luminosity particle-collider experiments.

The precise determination of fragmentation functions (FFs) of hadrons relies on the accurate description of the differential cross sections obtained from both experimental high-energy hadron colliders and theoretical predictions at higher orders in quantum chromodynamics. Various phenomenological strategies have been employed to extract FFs. In this work, we analyze the use of kinematical cuts for reactions including pions and kaons in proton-antiproton collisions to isolate individual FF contributions. This study examines the feasibility of using a similar approach as in proton-proton colliders to analyze FF flavour separation~\cite{Ochoa-Oregon:2023ktx}. In particular, we study photon-hadron production at colliders, including NLO QCD and LO QED corrections to reconstruct the partonic momentum fractions.

We present a framework connecting the algebraic consistency of the Operator Product Expansion (OPE) in Conformal Field Theories to observable signatures in the stochastic gravitational wave background (SGWB). By interpreting the conformal 6j symbols as crossing kernels, we extract the anomalous dimensions $γ_{n,J}$ of double-trace operators from the singularities of the crossing equation. Through the AdS/CFT correspondence, these algebraic shifts are holographically reinterpreted as effective mass shifts in the Anti-de Sitter bulk. We identify a critical anomalous dimension, $γ_c$, which marks the threshold where the effective mass violates the Breitenlohner-Freedman bound. Upon analytic continuation to de Sitter space ($L_{\text{AdS}} \to iL_{\text{dS}}$), we demonstrate that crossing this threshold ($γ< γ_c$) forces the dual scaling dimensions into the complex principal series, $Δ_{\text{dS}} = d/2 \pm iμ$. This complex dimensionality induces characteristic logarithmic oscillations in the four-point correlators. In the cosmological context, these oscillations manifest as a log-periodic modulation in the SGWB, characterized by a specific spacing $Δ\ln f = 2π/μ$. Detecting this log-periodic structure with next-generation interferometers such as LISA would directly measure the parameter $μ$, closing the loop between abstract OPE associativity and observational cosmology.

Pre-supernova neutrinos, emitted during the last day prior to core collapse of a massive star, could provide a unique probe at MeV energies of scattering interactions between dark matter and neutrinos. Due to attenuation of the flux of electron anti-neutrinos from their scattering on dark matter on the way to Earth, we expect a deficit of inverse beta decay events at large-volume, low-threshold detectors such as KamLAND, JUNO, and Super-K. which we use to derive upper limits on the dark matter-neutrino reduced scattering cross section. Though seemingly weaker than limits from post-bounce supernova neutrino events, these sensitivities provide an important cross-check, could help determine the energy dependence of the cross section, and may even be a distinct probe of certain models. Further, pre-supernova neutrinos may test hints reported of dark matter-neutrino scattering in the early universe suppressing small-scale power as seen in Lyman-alpha data.

We report the performance of low gain avalanche Silicon detectors (LGADs) for instantaneous electron and proton beam monitoring across dose rates ranging from conventional radiotherapy to the FLASH regime, benefiting from the fast response of these detectors of a few nanoseconds. The beam sources provide a dose rate greater than 40~Gy/s through pulses of widths 0.5, 1, 2 and 3~$μ$s for electron beams and 3, 5, 10 $μ$s for proton beams. Two different LGAD devices and silicon diodes are tested, yielding a linear dose response for electron beams up to $\sim$450~Gy/s and for proton beams up to $\sim$12~Gy/s. Beyond the linear regime the response continues to increase with a reduced slope and no true signal plateau is observed, at least up to 1800 Gy/s for electrons and 150 Gy/s for protons. This study contributes towards the instantaneous monitoring of increasingly intense flash beams for radiotherapy using fast detectors such as LGADs since measurements can be performed every fraction of $μ$s.

Quantum states of Neon-20 are generally agreed to lie in rotational-vibrational bands of a cluster of five alpha particles. However, more than one cluster shape has been proposed as dominant at low energy. As relative motion within a cluster is soft in certain directions, we investigate how the low-lying rotational bands of Neon-20 can arise from a circle of clusters connecting favoured shapes: a triangular bipyramid, a square pyramid, and a $D_{2d}$-symmetric distorted tetrahedron -- a twisted bow-tie. Motion around the circle extends the Berry pseudo-rotation that connects differently oriented bipyramids.

Many black-hole wave equations reduce to radial ordinary differential equations with regular singular points at horizons, where local solutions split into distinct branches. We propose a simple branch-selection principle to isolate the physically distinguished branch: removing the leading horizon behavior and imposing \(C^{\infty }\) smoothness on the residual field. Using an exactly solvable scalar field in \(JT/AdS_{2}\), we demonstrate that this single principle successfully organizes three seemingly distinct structures: smooth continuation across the black-hole horizon, the ambiguity of the boundary retarded Green function (pole-skipping), and the lowest-weight structure of an \(SL(2,\mathbb{R})\) representation. Rather than treating these phenomena as isolated boundary or horizon properties, our results unify them into a cohesive bulk-to-boundary dictionary, proving that the boundary pole-skipping ambiguity is the precise holographic manifestation of horizon-residual smoothness. Furthermore, the group-theoretic identification reveals that this physical branch selection is not a mere mathematical artifact, but is fundamentally governed by the underlying spacetime symmetries. Finally, we apply this framework as a diagnostic tool to a static holographic superconductor model, revealing that a genuine pole-skipping point requires both simultaneous residual smoothness and strict linear independence of the boundary branches to prevent deceptive branch-collapse artifacts.

We revisit a minimal fermion-portal scalar dark matter model consisting of a real singlet scalar dark matter candidate and additional vector-like singlet and doublet charged fermions stabilized by a discrete $Z_2$ symmetry. In light of the latest dark matter direct-detection constraints, the conventional Higgs-portal interaction is severely restricted, motivating a detailed investigation of fermion-mediated dark matter annihilation channels. We perform a comprehensive analysis of the model parameter space by incorporating theoretical constraints from vacuum stability and perturbative unitarity, together with experimental bounds from relic density measurements, direct-detection experiments, Higgs invisible decay searches, lepton-flavor-violating processes, and anomalous magnetic moments. We show that the observed dark matter relic abundance can be successfully reproduced over a wide mass range through Yukawa-driven $t$- and $u$-annihilation and co-annihilation processes involving the new fermions, while remaining consistent with current direct-detection limits. Motivated by the viable parameter space, we investigate the discovery prospects of the lightest charged vector-like fermion at future muon colliders operating at center-of-mass energies of 3 TeV and 10 TeV. Focusing on the process $μ^+μ^- \to E_1^+E_1^- \to e^+e^- + \cancel{E}_T$, we perform a detector-level analysis including realistic Standard Model backgrounds. We demonstrate that the clean experimental environment of a muon collider provides excellent sensitivity to charged fermion masses extending into the multi-TeV regime, significantly improving the exploration prospects of this class of fermion-portal dark matter scenarios.

In the functional-integral formulation of Euclidean field theory, exact renormalization group (ERG) transformations are realized by functional-integral kernels. Unlike the ERG flow equations that describe infinitesimal ERG transformations, these ERG kernels explicitly depend on the global topology of the configuration space. This paper explores this topology dependence for multiply connected configuration spaces. We show that the ERG kernel is in general given by a weighted sum of kernels on its universal covering space, where the weight factors are determined by a one-dimensional representation of the fundamental group. These weight factors are shown never to be renormalized under the ERG. We also show that these factors can be interpreted as Aharonov-Bohm phases with respect to a background magnetic flux penetrating the infinite-dimensional configuration space. From this viewpoint, a normalization condition for ERG transformations corresponds to a flux-quantization condition, which is equivalent to the level-quantization condition for Wess-Zumino-Witten terms in nonlinear sigma models. Finally, we present an alternative gauge-equivalent form of the ERG flow equation that incorporates this topological information locally.

This article provides a self-contained bridge between classical vortex dynamics and the relativistic, subatomic domain of the Quark-Gluon Plasma (QGP) produced in ultra-relativistic heavy-ion collisions. While the QGP is widely studied for its near-perfect fluidity, we focus on its role as the most vortical medium in the Universe ($ω\sim 10^{22}\text{ s}^{-1}$). The originality of our approach lies in isolating the explicit physical competition between non-linear vortex stretching and violent relativistic volumetric dilatation. By solving the covariant transport equations and tracking the comoving enstrophy density, we demonstrate that the explosive kinematics of the QGP provide an innate geometric shield that naturally regularizes the continuous flow, suppressing the self-amplification of vortex tubes before any microscopic viscosity is required. Furthermore, we connect this expansion phase back to the highly non-equilibrium initial state. We quantitatively predict that under a peripheral dipole initial topology, the global mid-rapidity hyperon polarization vanishes ($P_Λ\lesssim 10^{-4}$), establishing that the definitive signature of the QGP's rotation must be sought in azimuthal differential measurements within the LHC and RHIC experimental programs.

To best reproduce observed multiply-lensed lensed images, lens models usually incorporate shear attributed to objects unrelated to the lensing galaxy (i.e., external shear): whether it be neighbouring galaxies not explicitly included in the lens model or other cosmic structures along the sightline. When constrained solely by the positions of image counterparts, such lens models, even those utilising simple ellipsoidal mass distributions, can satisfactorily -- if not near perfectly -- reproduce the observed image positions, but often leave significant differences in flux ratios between the predicted and observed images. For the narrow-line regions (NLRs) of quasars, which are too large to be affected by micro-lensing from stars in the lensing galaxy, the flux ratio anomalies thus left are commonly attributed to small-scale structures (sub-structures) in Dark Matter associated with the lensing galaxy. Here, we show that external shear can always resolve, among the quadruply-lensed quasar NLRs studied, position anomalies in lens models constrained solely by the observed image positions, and in addition reduce although not fully resolving flux ratio anomalies when constrained by both the observed image positions and flux ratios -- provided, usually, that the external shear incorporated have strengths that far exceed (as is the common practise) those typically inferred from weak lensing along general sightlines (i.e., cosmic shear). Our work highlights the subversive role of excessive external shear in concealing lensing anomalies, undermining inferences on the characteristics of Dark Matter sub-structures -- and, correspondingly, the nature (mass and temperature) of the Dark Matter particle -- when not sensibly incorporated into lens models.

Detecting pulsars in short-period binary systems, which are unparalleled laboratories for fundamental physics and tests of general relativity, is a prime objective of radio astronomy. Their rapid orbital motion, however, presents a formidable computational challenge. Conventional searches are therefore limited to simplified signal models (e.g., constant acceleration) that remain valid for only short integrations ($\lesssim 4$-$10$% of an orbital period). This fundamental limitation severely degrades search sensitivity, placing much of the faint, relativistic pulsar population beyond the reach of current surveys. We present a novel hierarchical search framework based on extreme pruning that overcomes these limitations by progressively eliminating improbable regions of parameter space across successive coherent integration stages. The algorithm achieves $>90$% detection probability at the sensitivity threshold, with near-unity recovery for stronger signals, while reducing the computational complexity of full circular-orbit searches by up to 10 orders of magnitude relative to an unpruned hierarchical baseline. The resulting efficiency enables, for the first time, fully coherent integration over an entire orbital period and beyond. Compared to conventional acceleration searches, the proposed method delivers a 3- to 5-fold improvement in sensitivity, dramatically increasing the discovery potential for high-value targets such as pulsar-black hole binaries.

The Balmer break (BB) is a key spectral feature for constraining stellar population ages, star formation histories, and redshifts of high-redshift sources. The redshift evolution and distribution of BB strength, together with the properties of BB galaxies, constrain stellar population characteristics and the nature of star formation across cosmic epochs. However, a systematic and unbiased characterization of BB strengths across the full galaxy population remains limited with the James Webb Space Telescope (JWST). We aim to characterize the redshift evolution of BB strength over $z=3.5$-10 and its distribution across different epochs using photometry. We also examine correlations between BB strength and key physical parameters within these redshift intervals. We further assess the implications of BB galaxies for the nature of star formation and stellar populations at $z>3.5$. We used the JWST NIRCam photometric observations taken as part of various programs, including CEERS, JADES, FRESCO, and PRIMER. We estimated the BB strength of the objects with two adjacent broadband filters in various redshift windows between redshifts 3 and 10, which exclude strong line contamination. We employ the SED-fitting code CIGALE for both SED fitting and the generation of mock galaxy simulations. We find that the median Balmer break strength (expressed as a flux ratio) increases from cosmic dawn to cosmic noon, from 1.1 to 1.5, primarily driven by the age of the stellar population. These estimations are in agreement with the latest spectroscopic estimations in the literature. We identify objects with extremely large BB strengths (BB$>3.0$) at $z=3.5$-4 and $z=7$-10, indicating strong extinction combined with an old stellar population and the presence of Little Red Dots (LRDs) in the former, and predominantly LRDs in the latter.[abridged]

Star-forming galaxies are generally believed to be the main drivers of cosmic reionization. However, the relative contributions of bright and faint galaxies to this process remain unclear. As the most luminous transient phenomena in the universe, long gamma-ray bursts (LGRBs) provide a unique opportunity to probe star formation occurring in both detectable and undetectable galaxies. In this Letter, we present new estimates of the cosmic star formation rate density (SFRD) at $4<z<10$ using Swift LGRBs detected over the past two decades, by considering LGRBs as unbiased tracers of total star formation at high redshifts. Crucially, we find that the new LGRB-inferred SFRD can naturally explain current measurements of hydrogen reionization without invoking extreme ionizing photon production efficiencies or escape fractions from galaxies. Using these LGRB-inferred SFRD values, we further investigate the faintest magnitude limits of high-redshift galaxies, finding a redshift evolution of the limiting magnitudes from $M_{\mathrm{lim}}\sim-14$ to $-15$ at $z\sim6$ to $M_{\mathrm{lim}}\sim-10$ to $-11$ at $z\sim10$. This result provides one independent piece of evidence for the presence of a large population of faint galaxies at redshifts $z\gtrsim6$, as an important complement to our understanding of the ionizing photon budget in the early universe.

We present our numerical computations of broadband radiation spectra forming in a layer of high-temperature (kTe ~ 50 keV) semitransparent (with a Thomson optical depth ~1-3) plasma with an electron density Ne ~ 10^{17}-10^{19} cm^{-3}, typical for the accretion disk regions surrounding a black hole in X-ray binaries. The computations take into account the bremsstrahlung processes of photon production and absorption and their Comptonization. We show that the intrinsic radiation of such a plasma is sufficient to explain the X-ray spectra observed in the low (hard) state of Galactic black holes. No commonly assumed additional soft (with E < 1 keV) photons to maintain Comptonization are required; moreover, their presence would lead to severe distortions of the spectrum compared to the observed one. In the hard X-ray range the forming power-law spectrum with a photon index alpha ~ 1.4-1.7 and an exponential cutoff at E > 50 keV, exceeds considerably the bremsstrahlung flux that might be expected from such a plasma layer in the limit of its small depth. This is a result of the multiple Compton scattering of bremsstrahlung photons. It is important that the power-law spectrum of such a plasma should extend in an invariable form downward along the energy axis to the UV and OIR ranges (1-3 eV). At energies E < 1 eV the depth for bremsstrahlung absorption grows and the spectrum becomes the Rayleigh-Jeans one. To explain the steeper alpha ~ 2.1-2.5 X-ray spectra observed from black holes in their high (soft) state, it is indeed necessary that a large number of external photons enter a hot layer. Such photons could be emitted by the surface of an outer cold disk whose inner edge during these states approaches the black hole. The OIR emission from systems in these states is associated with that of the outer disk, whereas during their low state it can be produced directly in its central region.

We searched for transient and variable X-ray sources in the elliptical galaxy NGC 4552 using the {\it Chandra} and {\it XMM-Newton} observations from 2001-2012. We detected 14 transient and variable X-ray sources within the 4$R_{e}$ region of the galaxy, which exhibited peak state and undetected flux behaviour in the {\it Chandra} observations. Among them, two sources (T1 and T14) exceed the X-ray luminosity $10^{39}\rm~erg~s^{-1}$ in the detected observations and have the peak state and undetected luminosity ratio of $\geq 14.3$ and $\geq 8.7$, respectively, which we classified as transient and variable ultra-luminous X-ray sources (ULXs). For the third relatively bright transient source, the flux varies by a factor of 3 across the detected observations, and the hardness ratio analysis suggests a normal X-ray binary system. The remaining sources have luminosities in the $10^{37}-10^{38} \rm~erg~s^{-1}$ range, and the spectral analysis indicates that the sources may belong to neutron star or black hole X-ray binary systems, either in the hard or thermal-dominated state. The identified transient and variable ULXs exhibited high variability and soft spectral characteristics, which are consistent with ULXs accreting at super-Eddington rates. The near-infrared (NIR) counterpart search using {\it JWST} observations identified potential NIR counterparts for three mildly variable sources, and their magnitudes in different NIR bands suggest a red supergiant companion for each source. Deep X-ray monitoring observations can shed further light on the new transient and variable ULXs.

Prestellar cores are the sites of the earliest stages of star formation. Dust continuum observations are often used to identify and characterize their properties yet only a small fraction of them was observed and studied in terms of their composition and dynamical status. We explore the chemical diversity of prestellar cores and protostellar cores residing in the Orion B giant molecular cloud selected on their dust continuum emission to provide an unbiased view of their line emission properties and how they vary as function of the core parameters and environment. We make use of the large scale maps of Orion B in 25 molecular lines from which we extract information for a sample of 1001 cores selected using positions extracted from \textit{Herschel} dust continuum observations. The main properties of the core sample are derived using the Principal Component Analysis and additional maps of physical parameters: column density $N_{\rm{H_2}}$, far-ultraviolet (FUV) radiation field $G_0$ and mean volume gas density $n$. Additional high spectral resolution observations of $\rm C^{18}O(1-0)$ serve to evaluate the dynamical status of cores. The average line width of the cores is larger than what is typically expected for prestellar cores of closer star forming regions, which suggests that cores in Orion B are subjected to stronger turbulence affecting their stability. The first factor of the PCA analysis explaining the variation of the detected line intensities is the core column density of molecular gas. The second factor explains how the core chemical composition is strictly linked to their environment, which can be traced by the ratio of the external FUV radiation field over the core volume density, $G_0/n$. The third factor explaining the core chemical diversity is the mean density along the core line of sight, which is also associated with freeze-out and fractionation signatures.

We present a velocity-resolved three-dimensional map of local atomic hydrogen (HI) within 1.25 kpc of the Sun, tackling the challenge of converting emission from position-position-velocity space into true 3D structure. Our method combines the HI4PI full-sky survey with the Edenhofer et al. (2024) 3D dust map in the framework of Information Field Theory, enabling a joint reconstruction of the local HI density, radial velocity field, and effective line width while also separating emission arising inside the mapped local volume from more distant Galactic HI. The inference is driven by morphological matching between dust and HI structures together with kinematic coherence in 3D space. Synthetic data tests show that the method recovers the local density and velocity structure, even in the presence of substantial contamination from distant emission. The resulting map reveals a smoother, more diffuse local HI distribution than the dust, a declining HI-to-dust ratio toward high dust column densities consistent with the atomic-to-molecular transition, and a velocity field that captures both large-scale Galactic rotation and local non-circular velocities. Independent comparisons with maser and young stellar cluster velocities agree with the recovered kinematics. This HI map provides a new three-dimensional, kinematically resolved view of the nearby atomic interstellar medium and a foundation for localising other velocity-resolved Galactic emission in physical space.

Measurements of the HI content of galaxies serve as an important tracer for probing the impact of environment on galaxy evolution. More specifically, the HI deficiency (defined as the difference between expected unaltered and observed HI content of a galaxy) is closely related with environmental effects, which are most significant in large groups and clusters. In this work, we aim to estimate the HI deficiency of ALFALFA galaxies and investigate its relation with galactic environment. Using a random forest machine learning algorithm, we developed a predictive model capable of estimating the original HI content of a galaxy based solely on its optical properties. The model was trained on a subsample of 6 982 isolated ALFALFA galaxies with optical photometric data from the Sloan Digital Sky Survey (SDSS). Our predictive model outperforms the traditional approach, in which HI mass is linearly related to optical size (both on a logarithmic scale). The model achieves RMSE $\approx$ 0.22 dex and $R^2 \approx$ 0.80, compared with RMSE $\approx$ 0.26 dex and $R^2 \approx$ 0.70 for the traditional method. We applied this model to predict the expected HI content for non-isolated ALFALFA galaxies, enabling the calculation of HI deficiency. Controlling for the effects of internal factors, like stellar mass and presence of AGN, we find an increase in binned median HI deficiency of 0.15 dex attributable to environmental effects. In addition, we evaluate the temporal evolution of the predicted HI mass, and associated HI deficiency, due to the evolving stellar populations, following a gas removal event.

Using 200,388 red clump stars from LAMOST and APOGEE, we investigate the radial metallicity gradients of the Galactic disk as a function of vertical height and stellar age. The thin disk displays a pronounced negative radial metallicity gradient near the Galactic mid-plane that progressively flattens with increasing $|Z|$, following $Δ\mathrm{[Fe/H]}/ΔR$ = $-$0.0784 $+$ 0.0776 (1 $-$ exp ($-$ $|Z|$/1.4244)). The thin disk also exhibits a clear age dependence in radial metallicity gradients, evolving smoothly from a strong gradient regime for young stars to a weak gradient regime for old stars, following $Δ\mathrm{[Fe/H]}/ΔR$ = $-$0.0438 $+$ 0.0233 tanh (($τ$ $-$ 11.2908)/4.2130). The thick disk shows weakly positive radial metallicity gradients that remain statistically invariant with respect to both vertical height and stellar age, following respectively, $Δ\mathrm{[Fe/H]}/ΔR$ = 0.0038 $+$ 0.0009 $|Z|$ and $Δ\mathrm{[Fe/H]}/ΔR$ = 0.0146 $-$ 0.0007 $τ$. These results indicate that the thin disk retains radial metallicity gradients shaped by relatively ordered inside-out growth and long-term secular evolution processes. The thick disk exhibits spatially and temporally homogeneous radial metallicity gradients, supporting a picture in which it formed rapidly during an early phase when the disk was dynamically hot and efficiently mixed, potentially in chaotic mergers of gas-rich systems and/or the turbulent ISM.

Polarization observations at millimeter wavelengths can be used to study magnetized plasma in jets launched by supermassive black holes. We used multiband Atacama Large Millimeter Array (ALMA) data to study the polarization structure of both the nucleus and the kiloparsec-scale jet of the archetypal quasar 3C 273. We modeled the wavelength-dependent polarization of the nucleus using multiband observations (2, 1.3, and 0.85 mm) and applied QU-fitting to constrain Faraday rotation models with one or more polarized components. We also produced total intensity and polarization maps of the kiloparsec-scale jet to study the evolution of the magnetic field structure along and across the jet. The data of the nucleus were best fit by two Faraday components: a high rotation measure (RM) component with RM = +(2.6 +/- 0.1) x 10^5 rad/m^2 and comparable RM dispersion, and a second component with RM = +(1.5 +/- 0.2) x 10^4 rad/m^2 and lower dispersion. The high RM is comparable to previous ALMA observations at 1.3 mm, although we found clear evidence of variability when compared to previous single-band ALMA studies, highlighting the importance of broad wavelength coverage. On kiloparsec scales, the millimeter polarization structure closely resembled that observed at centimeter wavelengths, revealing a complex magnetic field configuration around the jet head and brightest hotspot region. Our results indicate a multicomponent, time-variable Faraday screen in the nuclear region of 3C 273 that is likely associated with a dense, magnetized environment close to the jet base. Future spatially resolved millimeter polarimetry with the Event Horizon Telescope will be crucial to disentangling these components and directly localizing the high-RM emission region.

The DESI collaboration's Data Release~2 (DR2) provides baryon acoustic oscillation (BAO) measurements from over 14 million galaxies and quasars, and a joint analysis of DESI BAO, CMB, and Type~Ia Supernovae reveals a preference for time-evolving dark energy. We quantify this preference relative to SDSS BAO and report three key results. First, DESI+Planck favors a higher $w_0 = -0.41^{+0.21}_{-0.22}$ than SDSS+Planck ($w_0 = -0.71^{+0.19}_{-0.18}$). Second, DESI+Planck prefers a deceleration parameter whose median lies on the decelerating side ($q_0 = 0.10^{+0.21}_{-0.23}$, consistent with $q_0 = 0$ at $1σ$), while SDSS+Planck prefers a negative value ($q_0 =-0.22^{+0.20}_{-0.21}$) indicating accelerated expansion. Third, we argue that this discrepancy arises from the difference in the lowest effective redshift probed by each survey: $z_{\rm eff} \approx 0.295$ for DESI versus $z_{\rm eff} \approx 0.15$ for SDSS. As present-day quantities, $w_0$ and $q_0$ are sensitive to the lowest probed redshift: data near $z = 0$ constrain them directly, whereas higher-redshift data rely on extrapolating the dark energy parametrization (here CPL). Reaching $z_{\rm eff} \approx 0.15$, SDSS constrains $w_0$ and $q_0$ in a data-driven way, finding consistency with $w_0 = -1$ and acceleration. Limited to $z_{\rm eff} \gtrsim 0.295$, DESI relies more on extrapolation, driving $q_0$ positive and $w_0$ well above $-1$. Adding the Pantheon+ supernova sample restores low-redshift information, returning $q_0$ to negative values and reducing tension with $Λ\text{CDM}$. We therefore propose that the apparent DESI preference for a non-accelerating present epoch in the BAO+CMB combination reflects redshift sampling rather than new physics, and suggest future DESI analyses adopt finer tomographic binning of the Bright Galaxy Survey sample to access lower mean redshifts and test this conclusion.

The growing volume of spectroscopic data in modern surveys motivates data-driven approaches that complement traditional emission-line diagnostics for active galactic nuclei (AGN) identification. We present a machine learning framework that exploits the full optical spectrum using unsupervised representation learning within a semi-supervised classification scheme. We use the SPENDER autoencoder to compress DESI galaxy spectra into a low-dimensional latent space and classify sources through a k-d tree nearest-neighbor search. The model is trained on 50,222 DESI Main Survey spectra from the Guadalupe dataset and released as part of Data Release 1 (DR1), restricted to z <= 0.5. We validate the performance using labels derived from FastSpecFit's emission line measurements defining seven galaxy classes: AGN, broad-line (BL), composite, star-forming, passive, retired, and Other. The method achieves high accuracies for AGN (0.952) and broad-line AGN (0.965), reliably identifying these sources even in low signal-to-noise spectra and recovering AGN missed by standard single-diagnostic methods. Our classification metrics are benchmarked against traditional diagnostics, and we show they represent lower limits of the model's true performance. We also find that the learned latent space correlates with key galaxy properties such as stellar mass and star-formation rate, demonstrating that it captures physically meaningful information. These results show that unsupervised spectral representation learning, implemented within a semi-supervised classification framework, provides a scalable and effective approach for constructing more complete AGN catalogues for current and future spectroscopic surveys.

We present and investigate the properties of a sample of radio-selected, Euclid-dark galaxies, identified from LOFAR HBA observations at 144 MHz within the Euclid Deep Field-North (EDF-N). Starting from radio sources lacking optical counterparts in previous surveys, but detected with Spitzer/IRAC, we identified 166 galaxies with no emission at a more than $3σ$ level in Euclid Quick Release 1 (Q1) images, and no matches in the Euclid Q1 catalogue. To minimise contamination from nearby sources, we selected a sub-sample of 88 isolated galaxies. By exploiting multi-band images and catalogues available for the EDF-N, we inferred the physical properties of our sample via SED fitting. The resulting redshift distribution spans $0.4 \leq z_\mathrm{ph} \leq 5.0$. We used recent sub-arcsecond imaging from the International LOFAR Telescope to constrain the nature of the compact radio emission through brightness temperature estimates. By combining this information with the radio excess relative to the infrared/radio correlation (IRRC), we searched for possible active galactic nuclei (AGN) activity. Approximately 40% of our sources show evidence of AGN activity. The Euclid-dark sources detected in the far-infrared are consistent with a population of heavily obscured, massive star-forming galaxies with high star formation rates. Their location above the star-forming main sequence is consistent with similar near-infrared-dark galaxy populations reported in the literature. We also performed a UV-to-radio median stacking analysis, finding that the two subsamples exhibit similar global physical properties and differ primarily in their radio emission. These preliminary results indicate that the wide area covered by Euclid enables the identification of a higher fraction of systems in which intense star formation and AGN activity coexist, likely capturing a key phase of galaxy--black hole co-evolution.

Cosmic voids provide a distinct low-density region for studying the environmental effects of galaxy properties. Using the SDSS DR7 catalog, we identify non-spherical voids via Voronoi tessellation and the watershed algorithm, and classify void galaxies based on their local volume. We compare and find that void galaxies classified by this method are systematically less massive, fainter, bluer, and have higher specific star formation rate (sSFR) than non-void galaxies and all galaxy samples. We then divide void and non-void galaxies into stellar mass bins to focus on the environmental dependence of $g-r$ color and sSFR. By further classifying galaxies into blue/red and star-forming/quiescent populations, we calculate the ratio of blue to red and star-forming to quiescent for void and non-void galaxies separately. Comparing the ratio of the void value to the non-void value for both metrics presents an overall decreasing trend with stellar mass $M_*$ over the $9.4-10.4$ range in $\log[M_*/\mathrm{M}_\odot]$, indicating a stronger environmental effect in lower-mass systems. These results show that our classification of void galaxies in non-spherical voids based on local volume offers a robust approach for quantifying the influence of underdense environments on galaxy evolution.

Binary stars are key tracers of the dynamical evolution of star clusters and provide important constraints on stellar populations and mass functions. The Magellanic Clouds host clusters spanning a wide range of ages and masses, offering an ideal laboratory to investigate these properties in regimes poorly sampled in the Milky Way. We aim to characterize the binary populations, mass functions, blue straggler content, and structural parameters of intermediate-age Magellanic Cloud clusters, and to explore their dependence on global cluster properties. We analyze HST photometry for 16 clusters obtained with ACS/WFC and WFC3/UVIS. Structural parameters are derived from stellar density profiles. Binary fractions are measured using the binary map technique, focusing on systems with mass ratios q > 0.7. We derive MFs accounting for unresolved binaries and identify candidate BS populations from color-magnitude diagrams. The fraction of binaries with q > 0.7 ranges from 5% in NGC 2121 up to 13% in NGC 2173, with a mass-ratio distribution that is consistent with being flat on average. By combining our results with literature data, we confirm a clear anti-correlation between the core binary fraction and cluster mass, while no significant dependence on cluster age is found. The clusters follow the established relation between age and core radius, although with substantial scatter at fixed age. Within the narrow age range explored here, clusters exhibiting steeper MFs are found to have smaller core radii. We find no evidence for a correlation between the fractions of binaries and BS fractions. These findings are consistent with a scenario in which dynamical evolution plays a primary role in the formation of binary populations. The connection between MF slope and structural parameters provides new constraints on cluster evolution and suggests a link between MF slope and structural evolution.

GalPort is a Python package for analysing the orbital dynamics of evolving disc galaxy numerical models in action-angle space. The package implements novel numerical methods for efficiently estimating actions, angles, and frequencies across different, particle-specific timescales: on the scale of radial or vertical oscillation and on the resonant libration/circulation timescale. The algorithm allows calculation of these dynamic quantities simultaneously at all time steps of the simulated galactic evolution. With this tool, one can trace orbital behaviour within time-varying galactic potentials and classify orbits (resonantly trapped, circulating, or passing through a resonance) based on their angle evolution. GalPort also includes specialised options for analysing the phase-space structure of a galactic bar in the disc plane and along its major axis. We demonstrate the package's performance on a typical N-body model of a barred galaxy, obtaining the global distributions of actions and frequencies and performing a detailed orbital decomposition. The code is publicly available under the MIT license at: https://github.com/vdzozulia/galport

We present a detailed study of X-ray spectral properties observed in 4 classical TeV (tera-electron volt) photon-emitting high synchrotron-peaked BL Lacertae objects using the simultaneous data of NICER and NuSTAR satellites. We analyzed 13 spectra in total from four BL Lacertae objects: Mrk 421, Mrk 501, PG 1553+113, and PKS 2155-304. We fitted all the spectra using the absorbed Log-Parabolic (LP) model first. While 7 spectra were fitted well using the absorbed LP model, we observed that 4 spectra of Mrk 421 and 2 spectra of Mrk 501 were not fitted satisfactorily using the absorbed LP model. The investigation of the flux states of the sources revealed that Mrk 421 was in a moderate to low-flux state during the 4 epochs and Mrk 501 was in a low-flux state during the 2 epochs. We concluded that there was a contribution from the disk in these 6 spectra. The moderate to low-flux state can justify the contribution of disk emission in the X-ray spectra. In the case of 4 spectra of Mrk 421, we observed a Gaussian feature between 1.42 and 1.70 keV.

We present the most densely sampled multi-band optical photometric observations of the peculiar BL Lacertae object OJ 287 from 2015 to 2025 with a focus on its optical activity on diverse timescales. We present a total of 2296, 10927, 11484, and 2982 data points in B, V, R, and I bands, respectively. The densely sampled observations allow us to keep track of the source evolution that it has exhibited since the start of the predicted major optical flaring activity at the end of 2015. The study reveals clear and persistent bluer when brighter trends in both the long-term and short-term variations. Different bands were cross-correlated with discrete correlation functions, which peak at zero lag, implying co-spatial emission. Using eight optical spectra in the low flux states of OJ 287 taken from 2017 October 21 to 2017 November 22, from Steward Observatory, we estimate the central black hole mass to be at least 3.89 $\times \ \rm{10}^{9} \ \rm{M}_{\odot}$ from the [O III] line width. The emission mechanism of the binary black hole blazar, and its possible implication in various aspects of multi-messenger astronomy are briefly discussed.

We report here the detection of signatures of a quasiperiodic oscillation (QPO) and a short flare in the optical light curve of the blazar PKS 0735+178, observed in two sectors, 71 and 72, spanning around 49 days with the Transiting Exoplanet Survey Satellite. The modest flare in sector 71 lasted ~4.3 days and appears as a combination of two sub-flares. In sector 72, a transient QPO with a period ~11.2 hours is detected at local and global significance levels of 4.11$σ$ and 3.06$σ$, respectively. We used weighted wavelet z transform, Lomb-Scargle periodogram, and phase dispersion minimization analysis techniques to look for and confirm the QPO feature. We also performed a segment-wise statistical inspection of these light curves and discuss here possible mechanisms that could explain the observed flux behavior.

We study the spatially resolved relation between dust attenuation ($A_V$) and star formation rate surface density ($Σ_{\mathrm{SFR}}$) in galaxies from the MAGPI survey ($0.25 < z < 0.42$). Using Balmer-decrement-based attenuation maps for 178 galaxies, we investigate whether the locally calibrated resolved $A_V$--$Σ_{\mathrm{SFR}}$ relation remains valid at intermediate redshift by comparing MAGPI with the local relation measured from MaNGA. We find a clear positive correlation between $A_V$ and $Σ_{\mathrm{SFR}}$ in MAGPI, with systematically higher attenuation than in MaNGA at fixed $Σ_{\mathrm{SFR}}$. After matching galaxies in stellar mass ($M_{*}$) and offset from the star-forming main sequence ($Δ$SFMS), MAGPI galaxies remain more attenuated than MaNGA galaxies at fixed $Σ_{\mathrm{SFR}}$. The attenuation excess is strongest for galaxies below the SFMS ($ΔA_V \sim 0.40$ mag), weaker for galaxies on the SFMS ($ΔA_V \sim 0.28$ mag), and minimal for galaxies above the SFMS ($ΔA_V \sim 0.07$ mag). The dependence of the offset on $Δ$SFMS suggests that nebular attenuation on kpc scales is regulated not only by local star formation activity, but also by the global evolutionary state of the host galaxy. Together, these results indicate that the resolved $A_V$--$Σ_{\mathrm{SFR}}$ relation is not universal, and that locally calibrated attenuation relations may not fully describe galaxies at intermediate redshift. This highlights the need for attenuation calibrations that account for galaxy population and redshift when interpreting spatially resolved galaxy properties.

The gravitational lensing of multi-messenger signals from binary neutron star mergers (BNSs), including gravitational waves (GWs), short Gamma-Ray bursts (sGRBs), kilonovae, and afterglows, can serve as a unique probe to constrain the mass of the graviton and cosmological parameters. In this paper, we estimate the detection rates of lensed electromagnetic counterparts associated with lensed BNS GW events detected by Cosmic Explorer and Einstein Telescope. For kilonovae and afterglows, we further consider a complementary pointed follow-up strategy targeting pre-identified galaxy-scale lens candidates within the GW localization region. By utilizing both numerical and observational constraints on BNS mergers, we find that: (1) Future $γ$-ray telescopes, even with a sensitivity more than ten times better than that of Fermi-GBM, may only detect lensed sGRB prompt emission at a rate $\sim 0.1$ yr$^{-1}$, corresponding to $\sim 2\times 10^{-3}$ of detectable lensed BNS GW events. (2) For the known-lens pointed strategy, the identifiable lensed-host fraction is approximately $0.15-0.30$ for the fiducial deep lens-catalog case considered, suggesting a possible gain in per-lens sensitivity for faint kilonovae and afterglows. (3) An RST-like near-infrared facility could detect lensed kilonovae at rates of approximately $\sim 0.45^{+0.81}_{-0.34}$, $0.55^{+0.98}_{-0.41}$, and $0.078^{+0.139}_{-0.059}$ yr$^{-1}$ in the F106, F158, and F213 bands, respectively. (4) Lensed afterglows remain difficult to detect in the optical and radio bands, while ATHENA-like X-ray observations may detect $0.5-5$ events over ten years.

The recent discovery by the LHAASO collaboration of a variable ultra-high-energy (UHE; $E_γ\ge$ 100 TeV) $γ$-ray source associated with the microquasar Cygnus X-3, with a spectrum extending to several PeV, provides compelling evidence for a hadronic super-PeVatron operating within the binary system. Inside the binary, the accelerated protons lose only a small fraction of their energy; upon escaping into the interstellar medium, they propagate diffusively to form a vast gamma-ray ``halo" structure extended to hundreds of parsecs. We argue that this halo has already been detected and corresponds to the Cygnus Bubble, an extended UHE $γ$-ray source reported by the LHAASO collaboration -- which possesses an angular extension of $\approx 6^{\circ}$ and an energy spectrum reaching 1 PeV. While the Cygnus Bubble is generally attributed to the star-forming region Cygnus X (specifically the Cygnus OB2 association at 1.4 kpc), we demonstrate that an association with Cygnus X-3 is physically more natural at energies above 400 TeV. This is supported by the cosmic-ray radial distribution, derived from the $γ$-ray and gas distributions, which points to continuous injection from a point-like source. The energetic requirements of the central accelerator are reasonably affordable and feasible. This reassignment identifies the Cygnus Bubble as a member of the recently discovered population of microquasar UHE $γ$-ray halos.

O- and B-type stars are young and hot, thereby serving as vital tracers of the star formation and spiral arm structure of the Milky Way. At the dusk of the \textit{Gaia} DR3 era, a high-confidence and accurate catalogue appears timely. Here we have characterized a population of 105,971 OB-type stars (T$_{\rm eff} >$ 10,000 K; hereafter OB stars) within 2 kpc from the Sun, using an astro-photometric Bayesian inference tool. Our resulting map unveils a complex view of the young stellar populations across the thin disk, with prominent large-scale features such as the Cepheus Spur, the Giant Oval Cavity, and a segment of the Sagittarius-Carina spiral arm all visible. Their inhomogeneous spatial distribution implies that massive star formation has taken place clustered across a few highly concentrated regions. We find a correlation between the overdensities of OB stars and young open clusters ($<$20 Myr), although OB stars can be better detected in high-extinction regions. We identify over 4200 OB stars as core-collapse supernova (ccSN) or direct-collapse black hole (BH) progenitor candidates, and therefore targets of interest for spectroscopic follow-up. Furthermore, we find no OB-type star ccSN progenitor to explode within the next 1 Myr within 100 pc, at which such an event could be harmful to Earth's biosphere. Finally, we identify more BH progenitors to collapse within the next 1 Myr than ccSN to explode, despite the former's much scarcer number - which could be indicative of a recent massive star formation burst in the local Milky Way.

We infer the initial conditions of Milky Way (MW) globular clusters (GCs) from present-day observations, through the coupling of recently updated rapid cluster evolution models with multimass equilibrium models. This novel method is validated by fitting to simulated observations of a large grid of star-by-star Monte Carlo models, demonstrating that we are able to recover cluster properties like the total mass, half-mass radius/density and black hole (BH) mass fraction, both initially and at the present day, across a large region of parameter space. We apply this framework to a sample of 40 MW GCs, fitting to a suite of observed radial profiles of number densities, proper motions, line-of-sight velocities and stellar mass functions. From these fits we infer a distribution of initial half-mass densities with a median and $1σ$ width, across our sample, of $ρ_{h,0} = 10^{6.4\pm0.9}\,{M_\odot pc^{-3}}$, higher than what is found for young massive clusters in the Local Universe and in line with young clusters at high redshift. We also find stellar initial mass functions that are bottom-light in comparison to canonical prescriptions, and relatively small present-day BH mass fractions ($\lesssim 1.5\%$). We discuss the implications of these initial cluster densities for observations of high-redshift proto-GCs, binary BH merger rates and intermediate-mass BHs (IMBHs) in GCs. Finally, we quantify how these densities may depend on assumptions typically made surrounding BH formation and natal kicks.

Magnetohydrodynamical (MHD) simulations of isothermal compressible turbulence report that the density distribution is well described by a lognormal with a variance proportional to the flow's Mach number. The distribution of magnetic field strength also has a lognormal component, but includes long, power-law-like tails. In this work, we use semi-analytical arguments to predict the distributions of density and magnetic field strength in compressible turbulent flows. Specifically, in the Lagrangian description of the continuity and the induction equations, we model the velocity gradients of the turbulent flow as a simple random process, essentially turning these equations into stochastic differential equations. Integrating them leads to a lognormal distribution for the density field and the strength of the magnetic field. The power-law tails in the magnetic field PDF appear when we introduce intermittent shocks due to sampling rare events. Gradually increasing the frequency of these events, essentially going closer to a continuous process, leads to lognormal-like distributions again. The asymmetry is connected to the relative abundance of slow and fast shocks. An overabundance of fast MHD shocks produces a high-value tail, while the contrary produces low-value tails. We propose that the appearance of power-law tails along lognormals in turbulent flows is the signature of the co-existence of continuous, diffusion-like propagation combined with localized, intermittent events.

We present the results of a deep 251 ks XRISM/Resolve observation of the cool core of the galaxy cluster Abell 2199. From the integrated spectrum of the central $3' \times 3'$ Resolve field of view ($104 \times 104 \mathrm{~kpc}^2$), we find that the intracluster medium (ICM) redshift is consistent with that of the brightest cluster galaxy, within the optical-redshift uncertainty. This indicates that they form a kinematically coherent core system, which offset from the mean cluster redshift by $\sim200~\mathrm{km~s^{-1}}$. The observed velocity dispersion of $\sim100~\mathrm{km~s^{-1}}$ corresponds to a three-dimensional Mach number of $M_{\mathrm{3D}}=0.16$ and a non-thermal pressure fraction of $P_{\mathrm{NT}}/P_{\mathrm{tot}}=1.4\pm0.2$%. Abell 2199 is one of the most dynamically quiescent relaxed clusters observed with XRISM, despite the presence of radio jets and a plume-like structure possibly associated with sloshing motions. Order-of-magnitude estimates suggest that turbulent dissipation could offset a non-negligible fraction of the radiative cooling losses, with $Q_{\mathrm{turb}}/Q_{\mathrm{cool}}\approx0.2$ for a large-scale driver such as sloshing and larger values for smaller AGN-feedback scales. Finally, we detect a localized enhancement of the Fe XXV He$α$ $y$ line in the southeast region, which spatially coincides with a Chandra surface brightness discontinuity.

Reconstructing the cosmic distance ladder directly from observations is a crucial issue in cosmology. In this paper, we present a novel method for modeling the cosmic distance ladder and estimating cosmological parameters through the use of Kolmogorov-Arnold networks (KAN), Long Short-Term Memory (LSTM), and Transformer networks (collectively referred to as KLT-Net), based on the apparent magnitude data from the Pantheon SN Ia compilation. As a data-driven, non-parametric method for reconstructing the distance modulus $μ(z)$, KLT-Net is shown to be highly effective in capturing the intricate, nonlinear measurement distributions. After validating against various statistical and machine learning models, we have identified it as the most effective choice among the considered alternatives and ablation experiments. Subsequently, the statistical inference of $H_0$ and $Ω_{\rm m}$ adopts the flat $Λ$CDM framework. Moreover, we introduce the Most Frequent Value (MFV) approach to evaluate the absolute magnitude, $M_B$, from existing literature data. In addition, we employ the Hessian matrix to validate the Bayesian method, demonstrating that the Hubble constant can be precisely constrained from the KLT-Net predictions within the flat $Λ$CDM framework. The integration of KLT-Net, the MFV approach, and Bayesian statistics establishes a robust framework for inferring cosmological parameters. This methodology facilitates future cosmological research, particularly in the analysis of complex datasets and the exploration of high-dimensional parameter spaces.

We present a methodology for constructing a mock galaxy catalog for the All-sky SPECtroscopic survey of nearby galaxies (A-SPEC) using the Machine-assisted Semi-Simulation Model. The model is trained on the cosmological magnetohydrodynamical simulation IllustrisTNG to predict baryonic properties of subhalos from dark-matter-only features and is applied to our own N-body simulation tailored to satisfy the requirements of A-SPEC. We have improved the model's accuracy by introducing additional features such as subhalo anisotropy parameters and modified definitions of the subhalo environment, which result in the coefficient of determination R^2=0.96, 0.90, 0.70, 0.79 for stellar mass, gas mass, star formation rate, and gas metallicity, respectively. The resulting mock galaxies reproduce the luminosity-dependent clustering of the target galaxies when tuned to match the number density. We discuss avenues for further improvement, including the role of environment in the predictions. We release the mock galaxy catalog with the baryonic properties predicted from the model.

Understanding the cycling of interstellar medium (ISM) between the galactic plane and off-plane regions is crucial for tracing the evolution of disk galaxies. We present $^{12}$CO($J=1-0$) multi-pointing observations of the edge-on, Milky Way-like galaxy NGC 4565 obtained with the Nobeyama 45-m telescope, with an angular resolution of 14 arcsec, corresponding to about 0.8 kpc. Along a prominent dust filament, we detect significant CO emission at three off-plane positions above the galactic plane. After evaluating possible beam-pattern contamination, the detections remain robust. The derived per-beam molecular masses are $M_{\rm mol} \simeq (2.1-4.3) \times 10^7 M_\odot$. While the off-plane spectra show broader effective line widths, $σ_{\rm eff}=83-115$ km s$^{-1}$, than the disk spectra, they contain CO components consistent with the local disk rotation. The mean observed off-plane CO intensity fraction is about 0.34. Comparison with geometrically thin-disk models suggests that this large fraction is best explained by gas above the disk. The large $σ_{\rm eff}$ values are partly attributable to a high-velocity component with molecular gas mass $M_{\rm mol}^{\rm HV} \sim 10^7 M_\odot$ that is offset by about 100 km s$^{-1}$ from the local disk velocity. The kinetic energy of this component is estimated to be $E_{\rm kin} \sim 10^{54}$ erg. Such a large energy requirement is difficult to explain by disk-driven feedback in NGC 4565, which has a Milky Way-like star formation rate. External inflow is therefore one possibility.

We compare the apparent directions of rotation in the plane of the sky of active galactic nuclei (AGNs) and their host galaxies. The direction of rotation of the galaxy was inferred from the direction of the spiral arms, while the direction of rotation of the AGN was inferred from spectropolarimetry, where the change in relative polarization position angle (PA) across broad lines is believed to be caused by equatorial scattering. The numbers of co-rotating and counter-rotating AGNs are equal. Studies of the relative position angles of radio jets have implied that there is a "zone of avoidance" where jets avoid being in the plane of disk galaxies. We point out that bi-conical narrow-line-region outflows also avoid the plane of the host galaxy. The equal numbers of co-rotating and counter-rotating AGNs exclude the hypothesis that the "zone of avoidance" is due to a lack of large tilts of the black hole rotating axis relative to the host galaxy rotation axis. Our results imply that the relative orientations of spin axes are random, at least for the black hole mass range we consider. We propose that changes in the broad-line polarization PA with wavelength that do not closely follow the predictions of the simple equatorial scattering model are a consequence of the scattering dust being clumpy. We note a couple of cases of possible changes in PA over several years, which, if real, could be due to motions of the dust clumps or changing anisotropy of the continuum emission.

Ground-based seeing imprints a size-dependent bias on galaxy structural parameters, yet the space-based imaging needed to remove it currently covers only a small fraction of the sky. We close this gap with a generative model that translates DESI imaging of Bright Galaxy Survey (BGS) targets into Euclid VIS images. A Fourier-domain analysis confirms that it recovers structure down to 0.37'' (from the 1.41'' DESI r-band baseline), an approximately 3.8-fold improvement in resolution. Although it stops short of the 0.16'' Euclid VIS resolution, this recovery already de-biases the structural parameters relative to the DESI r-band structure measurements: the Petrosian radius bias falls to +0.075'' (from -0.870''), independent of galaxy size; the Sérsic-radius bias drops to -0.018'' (from -0.322''); and the Sérsic-index bias to +0.093 (from +0.262). We release these translations over the Euclid DR1 footprint as the Euclid-resolution BGS (E-BGS), which can be blindly validated once DR1 is public.

Outflows in active galactic nuclei (AGN) are recognized as a fundamental mechanism driving the co-evolution of supermassive black holes (SMBHs) and their host galaxies. Although powerful outflows are frequently detected in gas-rich, active star-forming galaxies, their existence and potential impact within gas-poor, quiescent galaxies remain poorly understood. We report the first detection of a powerful ultra-fast outflow (UFO) in a nearby quiescent galaxy KUG 1208+386, providing a multiscale analysis of AGN winds from nuclear to galactic scales. We detect a nuclear X-ray UFO with a velocity of $v_{out} \approx -0.07c$ and a kinetic power of $\dot{E}_{\rm UFO}=(0.8\mbox{--}6.5) \times 10^{43}$erg/s, sufficient to drive effective AGN feedback ($\dot{E}_{\rm UFO}/L_\mathrm{Edd}=(1\mbox{--}8)\%$) and far exceeding the galactic [OIII] outflow power $\sim 10^{40}$ erg/s. Host galaxy analysis reveals a massive quiescent system (specific star formation rate $\sim3\times10^{-12}\mathrm{yr}^{-1}$) that has been quenched $\sim$9 Gyr ago. The central AGN is obscured by a line-of-sight (LOS) column density of $\log (N_\mathrm{H}^\mathrm{LOS}/\mathrm{cm}^{-2})\sim23$ and the circumnuclear scattering material is Compton-thick $\log(N_\mathrm{H}^\mathrm{scatter}/\mathrm{cm}^{-2})=24.7^{+0.8}_{-0.5}$. The discovery of a nuclear UFO in a long-quenched massive galaxy challenges the paradigm that UFOs are exclusive to gas-rich, star-forming systems, suggesting instead that they are governed by the local circumnuclear environments, rather than the global gas reservoir. Our results indicate that episodic, powerful winds can maintain the quiescent state of KUG 1208+386 over Gyr timescales, supporting a wind-driven `maintenance' mode of AGN feedback that is distinct from the classical jet mode.

Recent observations of young galaxies in the high-redshift Universe reveal signs of early enrichment of nitrogen. Extremely massive stars ($M \gtrsim 10^{2}-10^3\,M_\odot$) with strong stellar winds have been proposed as a potential driver of this phenomenon. Here, we show that the observed fraction of nitrogen rich stars with masses $\gtrsim 100M_\odot$ cannot be explained solely by mass loss, requiring significantly more efficient mixing beyond their convective cores than accounted in present evolutionary models. We compile a representative sample of 122 stars in the Tarantula Nebula of the Large Magellanic Cloud (LMC) with masses $M \gtrsim 30\,M_\odot$. Nearly all stars with masses $M \gtrsim 100\,M_\odot$ exhibit strong nitrogen enrichment, by factors $\gtrsim 5-10$. We demonstrate that this trend cannot be reproduced by varying assumptions on binary fraction, star formation history, or mass-loss rates within ranges predicted by current empirical and theoretical models. In contrast, enhanced core overshooting of $α_\text{ov}\gtrsim 1$ can account for the observed enrichment, but leads to quasi-chemically homogeneous evolution that is inconsistent with the observed Hertzsprung-Russell diagram. While the origin of this discrepancy remains unclear, our results, in combination with observational and theoretical constraints on mass-loss rates, suggest the presence of efficient early mixing operating during or shortly after the formation of very massive stars. Such mixing models are currently not included in stellar evolution models. These findings have immediate implications for the formation, radial expansion, evolution, and final fates of stars at the upper mass end, and provide a potential pathway to explaining the rapid nitrogen enrichment observed in the high-redshift Universe.

We present multi-epoch, multi-frequency radio observations of the tidal disruption event (TDE) AT 2022wtn, obtained with the Karl G. Jansky Very Large Array (VLA) and Giant Metrewave Radio Telescope (GMRT), spanning 97-866 days after optical detection. The peak radio flux density increases until 300 days post optical discovery, flattens out for several hundred days, then begins to decrease at 534 days. Utilizing an updated equipartition analysis framework, we estimate several physical parameters of the event and the surrounding medium. We model AT 2022wtn with two different geometries: a spherical and a conical emitting region. The spherical outflow model gives an expansion velocity of $v\approx0.21c$ and a kinetic energy of $\sim3.8\times10^{49}$ erg, and the conical outflow model yields a higher energy ($\sim1.8\times10^{50}$) and velocity ($v\approx0.41c$) than the spherical case. After ruling out the possibility of a relativistic jet, we consider several potential origins for sub-relativistic outflow regions in TDEs including unbound debris streams, collisionally-induced outflows, an accretion-driven wind, and an outflow from an accretion disk state transition, and find only an accretion disk state transition outflow to be consistent with the high energy and velocity found in our equipartition results. AT 2022wtn is a uniquely powerful non-relativistic radio-emitting TDE, and joins a growing population that display a diverse range of outflow properties.

Determining the energy, size, and velocity of synchrotron-emitting outflows is essential for testing models of their formation and evolution, but these quantities are often poorly constrained by observations alone. Equipartition analysis, therefore, provides a widely used framework for estimating these properties. Prior works have developed refinements to account for additional physical effects and other sources of energy (e.g., self-absorption, hot protons, and deviations from strict equipartition); however, these corrections are typically applied independently of one another, resulting in internal inconsistencies. In this work, we derive a self-consistent equipartition framework that accounts for the interdependence of various correction factors for Newtonian outflows and on- and off-axis relativistic jets. We implement our framework in an easy-to-use, publicly available code and apply it to study the tidal disruption events ASASSN-19bt and AT2019dsg, fast X-ray transient EP240414a, and active galactic nucleus J0231-0433. The interdependence of the corrections can increase energy estimates by a factor of ~5, suggesting that the energies of other synchrotron sources may be similarly underestimated in the literature. These results indicate that simultaneously incorporating these correction factors is essential for determining accurate outflow properties and constraining launch mechanisms.

Strong gravitational lensing combined with stellar dynamics provides a complementary route for testing gravity on kiloparsec scales and probing the internal structure of massive galaxies. However, such studies remain limited by degeneracies among the mass-density profile, stellar orbital anisotropy and external convergence, and by modelling assumptions, especially when only single-aperture velocity dispersions are available. Here we develop a hierarchical Bayesian framework to disentangle gravity and stellar orbital anisotropy from other effects at the population level. By reconstructing the lens mass distribution with a flexible broken power-law model and propagating its posterior uncertainty into the predicted velocity dispersion, we obtain a likelihood for each lens in the plane of stellar orbital anisotropy and an effective mismatch parameter. This parameter encapsulates projection bias, external convergence, cosmological distance ratios and deviations from general relativity via the post-Newtonian parameter $γ_{\rm PPN}$. Applying this framework to 121 galaxy-scale lenses, we find $γ_{\rm PPN}=1.027^{+0.099}_{-0.095}$, consistent with general relativity, and obtain $2σ$ evidence that the stellar orbits of massive galaxies have become more radially biased over the past $\sim6$ Gyr. Forecasts show that future samples of order $10^5$ lenses could enable sub-percent tests of gravity, precise measurements of orbital-structure evolution and complementary constraints on the cosmological matter-density parameter.

Cosmic rays (CRs) likely provide dynamically important non-thermal pressure support in the circumgalactic medium (CGM), but how their transport physics shapes observable absorption signatures remains uncertain. We investigate whether absorption-line diagnostics can distinguish between different CR transport regimes in CR-pressure-dominated halos. Using high-resolution simulations, we generate synthetic spectra along large ensembles of sightlines and measure column densities, equivalent widths, covering fractions (CFs), velocity widths, abundance ratios, and stacked absorption profiles for ions tracing cool, warm, and hot gas. We find that the effective CR transport speed strongly regulates the multiphase structure of the CGM. Efficient CR transport enhances the formation of cool ($T\sim10^4$ K) and warm ($T\sim10^5$ K) gas, leading to deeper and broader absorption lines of low- and intermediate-ionization species. The two-moment CR transport model produces the strongest MgII and SiII absorption and reaches MgII CFs consistent with the range inferred for star-forming galaxies. In contrast, slow CR transport underproduces cool, low-ionization gas and yields substantially weaker absorption. We also find that the origin of CIV-bearing gas changes with CR transport: slow transport mainly produces extended warm halo gas, whereas efficient transport shifts much of the CIV absorption into mixing layers around cool clouds. The high-ionization tracer OVI responds more weakly, indicating that CR transport primarily regulates the cool condensed phase and its interfaces rather than the volume-filling hot halo. These findings suggest that absorption-line measurements of cool and transition-phase gas can provide valuable constraints on the effective transport of CRs through the CGM.

Gravitational waves from binary black hole mergers remain among primary sources for ground-based and space laser interferometers. Primordial binary black holes (PBHs) can form binaries in the early Universe or at the matter-dominated stage in virialized dark matter halos at redshifts $z\sim 7-15$. We study enhancement of primordial binary PBH merger rate in virialized dark matter halos due to baryon accretion in the post-recombination epoch within two accretion models onto individual PBH -- Bondi-Hoyle-Littleton (BHL) and Park-Ricotti (PR), representing two extreme cases of the accretion efficiency. We calculate modification of the initial PBH mass function and their fractional contribution to dark matter. A significant enhancement in the binary PBH merger rate is found for the late three-body channel, which was previously treated as subdominant. In the case of delayed virialization onset at $z<10$, merger rates of dynamically formed PBH binaries are limited by the existing LVK binary BH detections. An increased merger frequency is revealed for PBH binaries with highly asymmetric mass ratios. We also calculate expected detection rates of binary PBH mergers for the TianQin space-based gravitational-wave interferometer and show them to be optimistic in the BHL accretion case even for very low initial PBH abundances $f\sim 10^{-6}$.

Core-collapse supernova gravitational waves may carry information about the dense matter equation of state (EoS), which describes the relation between pressure, density, temperature, and composition. This work tests a crucial question for physical inference: can a machine learning model trained on a finite simulation catalogue predict EoS parameters for an EoS family that was absent during training? Under standard random cross-validation, a LightGBM regressor appears highly successful, yielding $R^2=(0.70,0.67,0.60)$ for the nuclear incompressibility, symmetry energy, and slope parameter $(K_0,J,L)$. However, under Leave-One-EoS-Out (LOEO) validation, where all waveforms from a single EoS are withheld, the model fails, yielding mean absolute errors of $(44.57,3.19,30.54)$ MeV and negative pooled $R^2$ scores, performing worse than a baseline mean predictor. This generalisation gap persists across linear models, random forests, neural networks, and gradient-boosted trees. Restricting inputs to physical features (bounce amplitude, bounce width, peak frequency) reduces template leakage, the memorisation of related templates shared across random splits, but does not restore reliable EoS extrapolation. In contrast, a progenitor mass case study shows that classification generalises to unseen rotation speeds, while continuous mass regression compresses predictions towards the catalogue interior. These results demonstrate that while machine learning successfully interpolates within current waveform catalogues, this does not imply robust physical inference for unseen EoS models. Future pipelines should adopt leave-family-out validation, wider simulation coverage, and physics-aware inference frameworks.

Recent baryon acoustic oscillation (BAO) measurements from the Dark Energy Spectroscopic Instrument (DESI), in combination with CMB and supernova data, show a mild preference for dynamical dark energy over flat $Λ$CDM. They can also drive the best-fit effective neutrino mass to unphysical negative values, motivating models that mimic this effect through late-time expansion physics. The sign-switching cosmological constant ($Λ_s$CDM) model addresses this by introducing an AdS-like negative vacuum energy density at high redshifts ($z > z^\dagger$) that transitions to a dS-like positive density at low redshifts. While original fits assumed a discontinuous step function transition, we generalise the dynamics using a smooth hyperbolic tangent parametrisation, $Λ(z) \propto \tanh[Δ(z^\dagger - z)]$. We constrain the transition redshift $z^\dagger$ and smoothness parameter $Δ$ using background distance data from DES-SN5YR (with Dovekie recalibration) and DESI 2024 BAO. We further incorporate a compact redshift space distortion (RSD) $fσ_8$ compilation. The joint data do not prefer the smooth sign-switching model over flat $Λ$CDM, yielding nearly identical $χ^2_{\text{min}}$ values, while information criteria penalise the additional parameters. The transition smoothness remains unconstrained and dominated by the prior volume ($Δ= 24.75^{+17.42}_{-17.48}$), indicating that current data cannot distinguish between a sharp transition and a smooth dynamical crossover. This work does not directly constrain the physical sum of neutrino masses $\sum m_ν$; rather, it shows that current late-time background and growth data alone cannot resolve the transition details. We discuss the implications of this null detection and present projections for next-generation surveys.

Modern and next generation supernova cosmology analyses rely on end to end simulations to train photometric classifiers, characterise selection effects, validate light curve models, and calibrate distance bias corrections. Standard closure tests based on global Hubble diagram summaries can miss multivariate structure that remains in post correction residuals. We introduce a supervised machine learning closure audit that tests whether measured observables can predict the bias corrected Hubble residuals $Δμ$. We apply the audit to LSST Type Ia supernova simulations from two independent analyses: the M23 mock data sets of Mitra et al. (2023), with spectroscopic redshift and photometric redshift samples, and the predominantly photometric LSST like M25 simulation of Mitra et al. (2025). Standard one dimensional Redshift binned diagnostics explain <1% of the residual variance ($R^2 < 0.01$), suggesting apparent closure. In contrast, out of fold LightGBM models recover up to 98.2% of the variance in the simulated residuals ($R^2 = 0.982$), revealing structured residual predictability. Applying the same audit directly to the real Dark Energy Survey 5 Year (DES 5YR) spectroscopic sample yields a held out $R^2 = 0.725$. SHAP feature attribution rankings are highly consistent between independently trained M25 and DES models (Spearman $ρ= 0.802$), with apparent magnitudes, signal to noise ratios, and redshift dominating the shared predictive hierarchy. The resulting scorecard provides a diagnostic protocol for comparing mock ensembles and real observations, identifying residual non closure before cosmological parameters are unblinded.

We present new HI observations of Cloud-9 with the Five-hundred-meter Aperture Spherical Telescope (FAST), measuringvtotal flux of 0.35 +- 0.03 Jy km/s. Combined with Very Large Array (VLA) data, the HI shows a two-component structure: a compact, kinematically quiescent core and an extended envelope. The outer component exhibits a coherent velocity gradientvof ~25 km/s across ~7 kpc. The position-velocity structure does not support rotation; instead, the gradient is consistent with ordered large-scale motion, induced by environmental interaction with M94. The coexistence of a compact, kinematically quiescent core and an extended structured envelope is difficult to reconcile with a unbound or purely gaseous system. While the extended component is more naturally interpreted as environmentally shaped gas, the compact core suggests gravitational confinement beyond that provided by the observed baryonic content, without requiring strict long-term dynamical equilibrium. Deep optical limits further rule out a significant stellar component. Taken together, our results indicate that Cloud-9 is most naturally interpreted as a dark matter-dominated system undergoing environmental interaction. The observed morphology and kinematics are consistent with a stripped RELHIC scenario, in which a compact gas-rich core is embedded within an extended envelope shaped by ram-pressure interaction with the circumgalactic medium of M94.

The three-dimensional galaxy power spectrum is a powerful probe of primordial non-Gaussianity (PNG) and additional general relativistic (GR) effects on large scales, which can be constrained by current and upcoming large-scale structure surveys. In this work, we forecast the measurability of local PNG and linear-order relativistic effects in the spherical Fourier-Bessel (SFB) power spectrum for DESI, Euclid, and SPHEREx surveys. A Chebyshev-decomposition scheme is employed to accelerate the multi-tracer SFB power-spectrum calculations. Fisher forecasts establish baseline constraints and test the sensitivity of $f_{\rm NL}$ constraints to SFB mode cuts and to the marginalization over primordial cosmological parameters, while simulated Bayesian inference is used to quantify the impact and measurability of relativistic effects. We find that neglecting GR effects can bias $f_{\rm NL}$ constraints at the $1$-$3σ$ level for Euclid and SPHEREx. The degeneracy between GR and PNG terms is strongly tracer dependent, with the degradation of $σ(f_{\rm NL})$ ranging from a few percent to nearly a factor of two when the GR amplitudes are varied. Lensing can be detected at high significance for several tracers, while multi-tracer analyses substantially improve the measurability of the Doppler term. Assuming GR, we show that relativistic clustering partially breaks the exact $b_φf_{\rm NL}$ product degeneracy present in Newtonian linear power-spectrum analyses, although the resulting $b_φ$ constraints remain weak for the survey configurations considered. The joint PNG-GR inference consistently propagates uncertainty in $b_φ$ into the marginalized $f_{\rm NL}$ constraint. This firmly establishes the path toward extracting cosmological information from ultra-large-scale galaxy clustering.

Primordial non-Gaussianities (PNGs) are a broad class of features in the initial density field that are connected to the particle physics of the early Universe. Measuring the amplitude of these features directly constrains fundamental physics from these earliest epochs and lends insight into energy scales that cannot be probed with terrestrial experiments. Using a new class of simulation methods, we propagate these signatures to their impact on the formation of non-linear structure and quantify the constraining power in non-Gaussian summary statistics of weak lensing, thermal Sunyaev Zeldovich (tSZ), and X-ray surveys. We use semi-analytic baryon models that consistently include astrophysical effects across all these observables, and use foreground modeling approaches that explicitly fold in correlations between the various components. We find that the tSZ and X-ray fields have significant information about PNGs, and additionally can help self-calibrate a broad set of nuisance parameters/models by breaking parameter degeneracies. Using the second and third moments of the lensing, tSZ, and X-ray fields, we find a factor of 2 improvement in PNG constraints relative to using lensing alone. Larger improvements are expected when including more scales and other complementary summary statistics. Our multi-wavelength map maker can be found at https://github.com/DhayaaAnbajagane/Vaanam. The simulations and software pipelines used in this analysis are publicly available.

The forthcoming Nancy Grace Roman Space Telescope will revolutionize astrophysics by generating huge amounts of data of unprecedented quality. To properly address the data deluge and fully realize its potential, analysis tools that are both efficient and optimal are needed. {\sc Effortless} (previously known as Fast {\sc Imcom}) is a new algorithm for linear image reconstruction. Like its predecessor {\sc Imcom}, it offers control over point spread functions in output images; by avoiding laborious calculations, it is tens of times faster and has a smaller memory footprint. In this paper, I apply {\sc Effortless} to simulated Roman images and present promising first results. With ideal point sources, I illustrate that a single image reconstructed by {\sc Effortless}, combined with the post-measurement calibration procedure described in a companion paper, can lead to better measurements than a set of $\sim 6$ images coadded by {\sc Imcom}. While both algorithms were originally designed for weak gravitational lensing cosmology, {\sc Effortless} can benefit studies of static features and dynamic aspects of the Universe alike. Moreover, the efficiency and interpretability of {\sc Effortless} provides new possibilities for further reducing errors in measurements. The implementation of {\sc Effortless} is detailed in the companion paper.

We re-analyse near-infrared and optical spectroscopy from the Subaru, Gemini, and Keck telescopes of the bright gravitational arc seen in the z=0.62 X-ray cluster eMACSJ2229.9-0808. The 22 strongest spectral features we identify uniquely determine the redshift of the galaxy as z=3.20, as previously reported by Ebeling et al. (2025), not z=0.82 as claimed by Wagner & Falco (2026).

Ubiquitous nature of accretion disks and associated jets in modern astrophysics is extreme for black holes. The current state-of-the-art of black hole activities lies with modeling of underlying general relativistic magnetohydrodynamic (GRMHD) flows. These simulations have shown the importance of magnetic fields in the generation of outflows/jets and the overall dynamical evolution of the accretion flow. They also reveal critical insights into mechanisms that influence accretion dynamics, jet formation and stability. This further sheds light on the underlying magnetic field configurations based on magnetic field saturation leading to Standard and Normal Evolution: SANE, and Magnetically Arrested Disk: MAD. By employing SKA's high-resolution imaging and sensitivity, we can directly compare simulation outcomes with observational data, validating our models and enhancing our understanding of these phenomena. Key to this investigation is the examination of magnetic fields and their associated polarization signatures. Comparing the observational data from SKA with GRMHD simulations will facilitate a deeper analysis of the polarization properties, which can reveal the magnetic field geometry and dynamics in these extreme environments. The VLBI capabilities of SKA will prove instrumental in understanding jet morphologies and spectra of these systems due to its high spatial resolution. Collating these observations with GRMHD simulations will lead to a better understanding of the jet generation mechanisms and their interaction with ambient medium. By integrating advanced GRMHD simulations with SKA's capabilities, we aim to bridge theoretical predictions and observations, ultimately contributing to a more comprehensive understanding of the behavior of accreting black holes and their jets.

We present spectral classifications and photometric redshifts for 79.2 million sources up to an r-band magnitude of 22 in Data Release 4 of the Javalambre Photometric Local Universe Survey (J-PLUS). Leveraging the 12-band J-PLUS filter system, we compare a template-fitting approach (LePhare) against LeMoNNADE, a morphology-blind machine learning pipeline that uses spectral mixing augmentation to overcome training set limitations. LeMoNNADE consistently outperforms template fitting in precision, robust scatter, and outlier rates. Including WISE infrared photometry breaks optical degeneracies between stars and quasars, reducing the catastrophic outlier rate for quasars from ~40% to ~23% and constraining systemic redshift bias to <1% up to z = 4. We find LeMoNNADE is also less susceptible to redshift aliasing, particularly when adopting the probability density function median. Because the spectroscopic training samples severely under-represent stars, we apply an Expectation-Maximization Bayesian calibration to recover unbiased class probabilities for the magnitude-limited sample. This reveals that extragalactic counts agree with the literature down to the r ~ 20.5 completeness limit. The inferred redshift distribution for r < 21 extragalactic sources peaks at z ~ 0.3, showing broad agreement with existing literature up to z ~ 0.6. The resulting catalogues represent a significant milestone for local Universe science, offering probabilistically calibrated classifications and distances while explicitly characterising faint-end limits and contamination.

Context. The propagation of ultra-high-energy cosmic rays (UHECRs) and ultra-high-energy gamma-rays remains an open question in astroparticle physics, with the intergalactic magnetic field (IGMF) playing a crucial role in deflecting charged particles and shaping electromagnetic cascade spectra. Characterizing the IGMF across cosmic large-scale structure is therefore essential for interpreting multi-messenger observations and constraining the magnetogenesis scenarios that seeded it. Aims. We aim to provide accurate IGMF models to the astroparticle physics community and test their properties and robustness. Methods. We analyze IGMF models derived from the constrained cosmological simulation SLOW alongside a set of rescaled magnetic field models. We further introduce a novel algorithm to determine an "ideal position" for galaxies lying below the constraining power of the initial conditions, enabling accurate line-of-sight magnetic field extraction toward relevant sources. Results. The models span a wide range of filling factors and sample distinct regions of the electron density-magnetic field strength phase space in filaments, while converging in the cores of galaxy clusters; the simulated field from SLOW best reproduces the IGMF derived from the electromagnetic gamma-ray cascade. Models extracted using the introduced "ideal position" yield improved accuracy and may benefit multi-messenger studies more broadly. The large-scale structure drift of simulated clusters exploited by the algorithm also offers a potential route to refining the simulation's constrained initial conditions.

We explore the dependence upon rotation rate alone of various supernova observables simulated to their saturation for the explosion of a 9-$M_{\odot}$ progenitor. We find that the explosion energy is non-monotonic with, and weakly dependent upon, spin across a broad range of initial spins. The asymmetries of the blast depend weakly on spin, with faster spins leading to only slightly greater asymmetries. There is little significant pole-equator neutrino heating asymmetry during explosion, even for rapid rotation, and only for the fastest rotator does the neutrino heating rate diminish noticeably. Hence, the effects of rotation alone on all salient aspects of supernova dynamics are not large. We find that the recoil kick and spin are clearly aligned only for the most rapidly rotating model. Interestingly, for the fastest rotator, we witness a $T/|W|$ corotation instabilities near a value of 0.05 and spiral arm modes emerge. We find that the nucleosynthetic yields depend little upon the rotation rate and determine that the ratio of initial to final core spin period is near $\sim$4000, implying, given the modest inferred radio pulsar periods at birth, that the initial spin periods of most supernova cores are likely quite long. However, we focus on only one progenitor and do not include magnetic fields. Nevertheless, at least for low-mass progenitors which explode early, we find muted consequences of rotation in most major particulars across a wide range of initial spins.

By inspecting data from the LOFAR Two-meter Sky Survey (LoTSS), we noticed a peculiar bright and filamentary radio source at low-galactic latitude ($b \approx 0.5 °$). This source, detected also in previous radio observations, was originally believed to be a pulsar until Green et al. (2004) suggested that it is located in a heavily obscured galaxy cluster behind the Galactic plane. In this paper, we characterize for the first time the main properties of the host cluster (redshift, mass, temperature, X-ray luminosity, and dynamical status) by using X-ray observations performed with Chandra and SRG/eROSITA. In addition, by combining new uGMRT follow-up data with observations from the e LOFAR LBA Sky Survey (LoLSS), we perform a multifrequency, spatially resolved spectral analysis of the filamentary radio source (VLSS J0318.9+5755, nicknamed here the "Treble Clef" due to its morphology). We conclude that this source is a radio phoenix belonging to a massive, merging galaxy cluster in the Zone of Avoidance. We speculate that its complex morphology is shaped by gas motions generated in the intracluster medium during the ongoing merger, which are also likely responsible for the generation of the candidate radio halo tentatively observed in the cluster center. Owing to its highly filamentary morphology, brightness at $\lesssim$1 GHz, and extremely steep spectrum, reaching values of $α> 4$ between 400 and 650 MHz, this source represents an ideal target for high-resolution, very-low-frequency follow-up observations with LOFAR2.0.

We show that angular cross-correlations between distance indicators and galaxy redshift catalogues, when interpreted within an assumed cosmological model, can constrain potential biases in distance measurements induced by calibration systematics. As a test case, we consider a simple scenario in which a constant calibration offset $ΔM$ shifts all observed distances by a multiplicative factor, and we produce Fisher forecasts for the constraining power on $ΔM$ and $H_0$ from existing and upcoming surveys. In our most optimistic scenario, based on the expected number of SN Ia observed by LSST and the DESI final data release, we find $σ_{ΔM} \approx 0.05$ and $σ_{H_0} \approx 1.81~\mathrm{km~s^{-1}~Mpc^{-1}}$. This has important implications for the Hubble tension, since explaining the discrepancy purely as a calibration systematic in low-$z$ measurements would require $ΔM \approx 0.13$.

Compact symmetric objects (CSOs) are subkiloparsec radio sources with two-sided emission about a core resulting from jets that are not relativistically beamed towards the observer. This relative simplicity makes them attractive targets to study the launching and evolution of relativistic jets. We use radio surveys and spatially resolved VLBA observations to estimate the minimum energies and magnetic field strengths of a subset of edge-brightened CSOs (CSO-2s). These are necessary to test models of CSO-2 formation via stellar capture and evolution via synchrotron cooling. By treating the observed X-ray emission of CSO-2s as inverse Compton emission from synchrotron and external photon fields, we estimate a mean departure from the minimum energy magnetic field strengths of ${\sim}2\times$, suggesting that CSO-2 lobes are close to minimum energy. Typical lobal minimum energy magnetic field strengths of $20$ mG suggest that once the jets shut off, luminous CSO-2s should fade at GHz frequencies within ${\sim}10^{3}$ years. We find that CSO-2 minimum energies are systematically larger than previously estimated. If luminous CSO-2s result from tidal disruption events, a majority would require the capture of massive stars $>1 \ M_{\odot}$ assuming jet launching efficiencies less than $100\%$.

Isotopic ratios are key tracers of Galactic chemical evolution because different isotopes are synthesized through distinct stellar nucleosynthesis processes. While the $^{12}$C/$^{13}$C ratio increases with galactocentric distance across the Galactic disc, measurements in the Central Molecular Zone (CMZ) have historically yielded low values ($\sim$3-30), often affected by high optical depths. We aim to determine the initial $^{12}$C/$^{13}$C ratio of the parent material of the CMZ molecular cloud G+0.693$-$0.027 using optically thin molecular tracers and correcting for isotopic fractionation. We analyzed an ultra-high-sensitivity spectral survey obtained with the IRAM 30m and Yebes 40m telescopes, detecting single and double $^{13}$C isotopologues of HC$_3$N and HC$_5$N. Column densities and isotopic ratios were derived and compared with astrochemical models including $^{13}$C- and $^{15}$N-isotopologues to quantify isotopic fractionation. We derive $^{12}$C/$^{13}$C ratios of $36.7\pm1.0$ for HC$_3$N (using double $^{13}$C isotopologues) and $38.8\pm1.5$ for HC$_5$N, significantly higher than previous CMZ estimates based on simpler molecules. The models indicate low to intermediate isotopic fractionation at early times ($<3\times10^4$ yr), implying an initial $^{12}$C/$^{13}$C ratio of $\sim$48 for the gas from which present-day CMZ molecular clouds formed. The inferred range (37-48) is consistent with values observed at Galactocentric distances of 3-5 kpc, supporting a scenario in which the CMZ is replenished by gas inflows from the Galactic disc driven by the Galactic bar, with a possible contribution from less chemically processed material accreted from external systems such as dwarf galaxies.

Neutral hydrogen (HI) traces the dark matter distribution of the Universe. Upcoming surveys such as the Square Kilometre Array Observatory (SKAO) will trace neutral hydrogen up to z < 6 using several detection techniques including Intensity Mapping, which offers a unique window to explore the post-reionization Universe. Beyond two-point statistics promise to extract additional non-Gaussian information but require an accurate modelling of observational systematics such as foregrounds and the telescope beam. This work develops a theoretical model for the HI one-point probability density function (PDF) in spherical cells based on large-deviation statistics and spherical collapse for dark matter along with a nonlinear tracer bias and stochasticity parameterisation. It incorporates foreground removal and telescope beam effects that are validated against high-resolution simulations. We show that, despite these observational systematics, the HI PDF is able to capture additional non-Gaussian information from HI intensity maps compared to the power spectrum and can thus tighten constraints on cosmological parameters, breaking the degeneracy between the linear bias and the clustering amplitude.

The analytical modeling of a plasma wakefield driven by a relativistic proton beam is an element in optimizing advanced plasma-based acceleration schemes. In this work, we present a 1D nonlinear fluid framework under the quasi-static approximation to describe the wake potential excited by a positively charged proton driver. We examine our model using a two-bunch pump-probe configuration, demonstrating close agreement between the analytical invariants and adaptive numerical integrations. The distinct geometric curvature changes observed at the micro-bunch boundaries are shown to be physical consequences of step-discontinuities in the second derivative of the wake potential across the beam interfaces. Furthermore, by scaling this numerical framework to a train of $N=100$ micro-bunches undergoing seeded self-modulation (SSM), we model the physical parameters of the CERN AWAKE facility ($n_0 = 7.0 \times 10^{14}\text{ cm}^{-3}$). Our model replicates the characteristic linear growth envelope and matches the calibrated field envelope boundaries of approximately $\pm 0.75\text{ GV/m}$ inferred from the experiment. This piece-wise framework provides a computationally efficient foundation for investigating customized, asymmetric micro-bunch profiles designed to optimize the transformer ratio beyond the fundamental symmetric limit of 2.

Integral Field Spectroscopy (IFS) is a key technique to study galaxies, stellar and planetary systems, and transients, by spatially splitting the scene and dispersing this light onto the detector. This technique is now central for many modern telescopes, including the Lazuli Space Observatory which will host a low spectral-resolution image slicer spectrograph. To prepare for future IFS-equipped instruments and to better understand the systematics of current ones, realistic simulations are needed. We present slicersim, a Python package designed to simulate IFS observations. slicersim is a modular and flexible tool that allows the user to build a scene, define a telescope, a spectrograph, and a detector, and to simulate the resulting observation, including various sources of noise and instrumental effects. The package is designed to be easily extensible to any IFS, with a current focus on that from the Lazuli Space Observatory. In this paper, we present the structure and logic of the package, detailing its main components, and providing an example of its usage. Using slicersim we demonstrate the feasibility of a large cosmology campaign on Lazuli -- a sample of 8000 Type~Ia Supernovae uniformly distributed between $z=0$ and $z=1.5$, observed within the first few years of its operations for a typical average signal-to-noise of 25 per resolution element rest-frame optical wavelengths.

The detection of multi-TeV radiation from a handful of black-hole X-ray binaries (BHXRBs) has positioned the source class as promising candidate to explain the cosmic-ray flux in the PeV energy range. To determine what distinguishes the systems detected in the gamma-ray range from the rest of the population, we build on existing catalogues of BHXRBs by including information on the radio and gamma-ray emission from these systems. We then compare the $>100$ TeV gamma-ray luminosity to different properties of the systems and evaluate the strength of the (rank) correlation between them. Additionally, we compare the distribution of BHXRBs with the positions of unassociated gamma-ray sources in the 1LHAASO and 1HGPS catalogues. We find no correlation with the mass of the compact object, the companion mass or the inclination angle. Instead, we find evidence for a strong correlation between the gamma-ray luminosity and the radio and hard-state X-ray luminosities in our sample. The limited number of gamma-ray detections means that this correlation cannot yet be claimed as statistically significant. We find indications that a high (>20\%) duty cycle (defined as fraction of time spent in outburst) might be a requirement for $>100$ TeV gamma-ray loudness. We find significant evidence ($>3σ$) that BHXRBs are located near unassociated gamma-ray sources in the region of the sky not covered by LHAASO. This evidence is only significant when comparing to BHXRBs detected in the radio band. We conclude that radio-bright BHXRBs with a high duty cycle are more likely to be detected as gamma-ray sources than sources which are not detected in the radio band or those with low duty cycles. This result indicates that the particles responsible for the gamma-ray emission are associated with the jets, either accelerated within them or via their interaction with their surroundings.

This chapter proposes a research framework to quantitatively investigate non-thermal components in the Intracluster Medium (ICM) of galaxy clusters, which are critical ingredients for governing energy transport, structure formation, and particle acceleration. Turbulence, primarily driven by cluster mergers, is the leading mechanism for re-accelerating cosmic ray electrons (forming radio halos) and amplifying magnetic fields (via the turbulent dynamo). Observational understanding of both the turbulence and magnetic fields is rapidly evolving: the high-resolution X-ray spectrometer XRISM is directly measuring the velocity properties of the thermal ICM, providing insights into the kinetic energy of turbulence. Concurrently, high-sensitivity low-frequency radio observations, including SKA pathfinders, are mapping non-thermal components and magnetic structures through diffuse synchrotron emission and high-density Faraday Rotation Measure (RM) grids. The synergy between XRISM and SKA offers a decisive paradigm shift. XRISM's velocity maps, with its high energy resolution (<7 eV FWHM), combined with SKA-Mid's capability to deliver high-resolution RM grids ($\sim 100$--$200~\rm deg^{-2}$) and high-dynamic-range imaging, will allow for the first direct, multi-wavelength comparison of the turbulent energy properties (from X-ray) and the magnetic field properties (from radio). This joint analysis will validate Magnetohydrodynamic (MHD) simulation predictions, clarify the process of turbulent energy cascade and decay, and ultimately lead to a comprehensive understanding of the co-evolution of turbulence, magnetic fields, and cosmic rays in the largest laboratories of the Universe.