Papers

This paper presents a leakage-resistant framework and benchmark (LRFB) for measuring evaluators of AI-generated Oracle PL/SQL specifications. The benchmark contains 494 labelled source-specification pairs across 19 quality metrics grouped into seven dimensions; each metric contains 26 records divided into development and held-out subsets of 13 records each, with three clean and ten defective specifications per subset. Ground truth is mechanically sealed from inference and checked with canary and label-permutation tests. Three LLM judges were evaluated with three runs per record and compared against two trivial baselines. On 247 held-out records, GPT-4o, DeepSeek V4 Pro, and Claude Sonnet 4.5 reached balanced accuracies of 0.87, 0.89, and 0.81, respectively, while a specification-length baseline reached 0.88. Exact paired tests did not separate any judge from that baseline. The judges nevertheless provided information unavailable to length: the labelled defect category appeared among findings in 84%-92% of violated records, and quoted evidence overlapped the labelled span in 92%-93%. The study therefore reframes specification evaluation from binary issue detection toward evidence-linked, typed findings and benchmark-validity controls.

Let $A_n=\mathbb{k}[x_1^{\pm1},\ldots,x_n^{\pm1}]$ be a Laurent polynomial algebra over a field $\mathbb{k}$ of characteristic zero. We determine explicitly the automorphisms commuting with each locally finite derivation \[ D_\mu=\sum_{i=1}^n\mu_i x_i\frac{\partial}{\partial x_i}, \qquad \mu=(\mu_1,\ldots,\mu_n)^T\in\Bbbk^n. \] These are precisely the automorphisms $\rho(x_i)=\lambda_i x^{m_i}$, where $\lambda_i\in\mathbb{k}^\ast$, the exponent vectors $m_i$ are the columns of a matrix $M_\rho\in\operatorname{GL}_n(\mathbb Z)$, and $\mu^TM_\rho=\mu^T$. We also give an explicit parametrization in dimension two and relate these automorphisms to symmetries of diagonal flows on algebraic tori.

Automatic differentiation (AD) is a cornerstone of modern machine learning. Since all AD frameworks implicitly fix a basis of Euclidean space to define gradients and Jacobians, they natively process mathematical objects that are arrays of a fixed set of coordinates. This makes it awkward to express objects such as rotations, volume forms, and curvatures that arise, for example, in applications of AI to science. Bespoke workarounds lack reusability. We show that (i) a single, basis-independent derivative operator can represent many geometric differential operations, and (ii) this operator can be expressed entirely in terms of existing AD primitives. This reusable transformation can be composed with JAX batching, compilation, and hardware acceleration. Geometry enters as algebraic data. We experimentally demonstrate its reuse across different manifold operators, its composition through a spacetime inverse-problem pipeline, and its use in optimization over different rotor subgroups.

The two main traditional player statistics used in soccer are goals scored and assists, with an assist credited when a player's pass immediately leads to a teammate scoring. Attributing credit solely to the player who makes the final pass before a goal fails to account for the contributions of multiple players in the buildup to the goal. Consequently, traditional statistics provide only a limited view of player involvement. In this paper, we use multivariate Hawkes processes (MHPs) to model timestamped sequences of passes and shots, which form continuous-time networks. MHPs provide a branching mechanism where an event (a pass or shot) can be generated as a child of a previous event, which allows us to quantify each pass's contribution to a subsequent pass or shot. This approach distributes credit across multiple passes within an attacking sequence, making it applicable across a wide range of situations. Building on this framework, we propose expected goal involvements (xGI), a metric that measures a player's involvement in creating goal scoring opportunities based on their passes. This metric identifies players who play key roles in creating goal scoring opportunities even if they do not make the final pass, offering a more comprehensive measure of attacking involvement and player performance.

Autonomous small-body proximity operations require continuous-thrust guidance under uncertain gravity, solar-radiation pressure (SRP), and stochastic disturbances. We combine iterative covariance steering (iCS) with $L_1$ distributionally robust adaptive control ($L_1$-DRAC) to provide risk-aware collision avoidance despite mismatch between the nominal planning model and the true dynamics. Iterative covariance steering plans a nominal feedback policy satisfying distributionally robust conditional value-at-risk keep-out constraints, while $L_1$-DRAC compensates for dynamics uncertainty and bounds the residual distributional mismatch through a Wasserstein radius used by the planner. In $\mathbf{1000}$ paired Monte Carlo trials with gravity, SRP, and diffusion mismatch, $L_1$-DRAC adaptive control reduces the keep-out violation rate from $\mathbf{32.7\%}$ to $\mathbf{0.2\%}$ and the physical collision rate from $\mathbf{16.4\%}$ to zero, without actuator saturation.

Time-modulated intelligent reflecting surfaces (TM-IRS) achieve directional modulation by switching their elements on and off at designed instants and durations, so that an undistorted signal reaches the intended user while appearing scrambled elsewhere. We propose a Flow Matching (FM)-based generative framework that designs these parameters for strong scrambling and reliable reception simultaneously. For orthogonal frequency-division multiplexing (OFDM) signals, we introduce a physics-informed reward that maximizes the time-modulation harmonic energy radiated in the null space of the legitimate-user response, subject to a minimum per-subcarrier signal-to-interference-plus-noise ratio (SINR), without requiring knowledge of adversary directions. Unlike prior generative flow network (GFlowNet) approaches, which assign discretized parameters sequentially, the FM model generates complete continuous configurations jointly, decoupling the generation depth from the number of elements and accelerating training. It also learns a distribution over configurations that can be resampled to vary the scrambling pattern. Numerical results show that, compared with a GFlowNet baseline, the proposed approach reduces the time to training loss near-saturation by $99.4\%$ on average for $3{\times}3$ to $12{\times}12$ surfaces, attains higher final rewards, scales to $80{\times}80$, and confines the low-symbol-error region more tightly around the legitimate user.

Humans effortlessly react to the world; they adeptly return a tennis shot, catch a falling cup, step aside from a passing cart, or pass a box tossed during truck unloading. These everyday actions require anticipating a moving target and reflexively responding with the whole body within about a second. Such visual reactive whole-body control remains challenging for humanoids, which must (1) coordinate whole-body control, (2) react to object dynamics in real time, and (3) ground perception in egocentric vision. We study catching thrown boxes as a representative task that demands all three capabilities; existing systems cover only part of it, relying on end-effectors, motion capture, or marked objects. We present Reflex, a three-stage framework that learns these capabilities stage by stage while connecting them through a compact representation of object dynamics. First, a state-based policy learns whole-body catching with reinforcement learning. Second, the policy learns to infer the dynamics of a moving box from delayed and incomplete observations. Third, an egocentric RGB-D encoder learns to recover this dynamics representation directly from visual observations while keeping the whole-body controller fixed. This decomposition provides each capability with a direct training signal and makes visual learning substantially more scalable. In simulation, Reflex achieves 85.7% catch success from egocentric RGB-D, compared with 87.9% when given privileged box observations. Temporal visual history is critical: removing it reduces success by 35.6 and 39.8 points on centered and laterally offset throws. On a real Unitree G1, Reflex catches human-thrown boxes with 65% and 40% success at 2.5 and 5 m, reaching 76% and 59% of human catchers' success at the same distances.

The integration of two-dimensional Ti3C2Tx MXenes into Halide Perovskite (HaP) optoelectronics has shown immense potential for enhancing device performance, yet the fundamental chemical interactions within the precursor ink (a critical stage determining the final material state) remain largely unexplored. This study presents a comprehensive investigation into the chemical stability and surface evolution of Ti3C2Tx MXenes when exposed to the individual and combined components of a model methylammonium lead iodide (MAPbI3) solution. By combining detailed X-ray Photoelectron Spectroscopy (XPS) analysis with Density Functional Theory (DFT) calculations, we decouple the complex interactions between MXenes and the ink constituents. We demonstrate that elemental iodine (I2) and polyiodides (In-) act as aggressive oxidizers, triggering rapid degradation and dissolution of the MXene flakes. Conversely, lead iodide (PbI2) induces a slower oxidation process driven by the strong adsorption of Pb2+ ions onto oxygen terminated sites. In stark contrast, methylammonium iodide (MAI) exhibits a stabilizing effect: it adsorbs onto the MXene surface via hydrogen bonding, effectively passivating the flakes and retarding degradation. In addition, our results reveal that the sequence of precursor addition governs the MXene surface state, requiring extra caution, but also enabling further tunability. These findings provide a critical view on how MXene integration with materials, in general, and HaP in particular, should be perceived, providing a fundamental for their rational engineering.

This paper studies continuous covariant kernel families over hulls of Delone sets of finite local complexity using graded Wiener-algebra methods. The norm given by the least common weighted summable off-diagonal profile is identified isometrically with a graded $\ell^1$-norm. The resulting Banach $*$-algebras are symmetric, spectrally invariant in the reduced groupoid C*-algebra, and stable under holomorphic functional calculus. Exact physical weights give preservation of polynomial and stretched-exponential classes, including a fixed stretched-exponential rate. A configuration with dense orbit has a faithful regular representation. Invertibility on that configuration therefore gives an inverse family with one profile valid throughout the hull; for repetitive sets every configuration has this property. For polynomial weights of order $s>0$, return-vector counting yields a differential estimate with parameter $\kappa/(\kappa+s)$, where $\kappa$ is a growth exponent of the return vectors and may be taken equal to $d$ for Meyer sets. Combined with established norm-controlled inversion criteria, this gives explicit bounds in the common-profile norm. Explicit bounds are also obtained for stretched-exponential physical weights and for gapped Fermi projections.

Unlearning in large reasoning models (LRMs) requires removing forget-set information from both the reasoning trace and the final answer. Existing approaches achieve this objective at the cost of reasoning collapse, where the model fails to produce a coherent chain-of-thought (CoT) and does not properly terminate its reasoning process. To address this challenge, we propose REDaCT (REgenerate via Detection and Corrective Tuning), an inference-time unlearning framework for LRMs. REDaCT consists of two key components: (1) a binary leak detector that identifies sentences in the reasoning trace that leak forget-set information, and (2) an embedding optimization module that learns a corrective embedding offline, such that when injected at the first detected leak during inference, it guides subsequent reasoning towards sentences that do not leak forget-set information. Experimental results show that REDaCT completely avoids reasoning collapse, substantially improving reasoning preservation compared to state-of-the-art unlearning methods while maintaining perfect answer forgetting and near-perfect CoT forgetting. Our code is available at: https://github.com/jmaccoby/redact.

Semantic query engines evaluate natural-language filters with a large language model, and every evaluation is a costly oracle call. Optimizers that reduce these calls face a dilemma: cascades that trust a cheap proxy miss their precision and recall targets on many queries, methods with finite-sample guarantees save little, and all of them discard the oracle's labels when the query ends. We present SemViews, a certified semantic filter that orders tuples by signals that cost no oracle calls (the labels that earlier queries left behind as semantic views, a local embedding, and a helper model's score), refines that order with labels it must pay for anyway, and certifies its answer with fresh stratified oracle samples. The answer meets user precision and recall targets with probability at least $1-\delta$ for every workload and every catalog, so a stale or misleading view can cost oracle calls but never accuracy. On four workloads replayed against 725,000 recorded oracle labels, SemViews meets both targets on 99.63% of queries and saves 28%-45% of the oracle calls, whereas the only prior method in our study with the same guarantee needs 1.32x-1.71x as many calls. When a question returns in different words, views cut its calls by a further 27%-36%, with both targets met on at least 99.8% of such queries; similarity caches and entailment-based reuse meet them on only 17%-47%.

We propose constructions of spherical $5$- and $7$-designs in odd dimensions by lifting balanced halves of even-dimensional designs to several layers of suitable latitudes. For every odd dimension, we obtain an explicit spherical $5$-design with at most $60(d^2-1)$ points, attaining the optimal quadratic order in the dimension. Assuming Zauner's conjecture, this bound improves to $7.5(d-1)^2$. For every odd $d \geq 9$, we also construct a spherical $7$-design with $110M$ points from a complex projective $3$-design with $M$ points in $\mathbb{CP}^{(d-3)/2}$.

Machine-learning models for perovskite solar cells are typically evaluated using random data splits, which we show systematically leak information through massive internal duplication. Because nearly two-thirds of curated devices share their exact recipe with another, random splits primarily reward the mere recall of historical data rather than testing true physical generalisation. To solve this, we mapped the institutional provenance of the Perovskite Database and established a rigorous hierarchy of grouped splits that progressively isolate duplicate devices, publications, and institutions. Benchmarking thirteen models establishes new state-of-the-art performance across all splits and reveals a critical regime shift: while tree ensembles dominate random splits by recognising familiar recipes, their ranking collapses on genuinely unseen institutions and future discoveries, where composition-aware neural architectures excel. Furthermore, feature importance demonstrates that the device architecture and fabrication methodology dictate final efficiency just as strongly as the absorber chemistry. This standardised protocol empowers the community to separate memorisation from true physical discovery.

We study testing a random geometric graph with a fixed kernel against an Erdős-Rényi graph with the same edge density. Under the null hypothesis, the graph has $n$ vertices and independent edges with probability $p$. Under the alternative hypothesis, latent points $X_1,\ldots,X_n$ are independent and uniform on $\mathbb{S}^{d-1}$, and edges are conditionally independent with probabilities $K(\langle X_i,X_j\rangle)$, where $p=\mathbb{E}[K(\langle X_1,X_2\rangle)]$. The goal is to determine a critical dimension $d_\star = d_\star(n,K)$: the total variation distance between the two graph laws tends to one for $d\ll d_\star$ and to zero for $d\gg d_\star$. For dense graphs, [BDER16] established $d_\star\asymp n^3$ for hard threshold kernels, whereas [MWX26] proved $d_\star\asymp n^{3/4}$ for smooth kernels with $K'(0)\ne0$. We connect these results by characterizing the detection thresholds for kernels that need not be differentiable at zero. We also determine the thresholds for smooth kernels whose first nonzero derivative at zero has finite order greater than one. Throughout these classes, signed triangle or quadrilateral counts attain the optimal detection threshold. For certain nondifferentiable kernels, quadrilateral counts are optimal while triangle counts are suboptimal.

Multiple approaches have been implemented for identities that associate meaningful identifiers, such as domain names, with self-authenticating identifiers, such as onion addresses. While these approaches offer various advantages both from identifier meaningfulness and from identifier self-authentication, identifier association itself is subject to blocking, hijack, fingerprinting, and other forms of surveillance and manipulation. So too are the resources associated with an identity comprising a meaningful identifier and a self-authenticating identifier. This paper describes ongoing work on two such approaches to self-authenticating meaningful identity (SAMI): associating a registered domain name with an onion address (onion association) and associating an atproto handle with a decentralized identifier (DID association). We present both the application of already implemented and published means of secure onion association to DID association and the combination of both in a single identity. Finally, we describe grounding identity association and integrity in contextual trust, wherein trusting attestations of identity relies on contextual relevance of an attestor to the identity rather than relying only on structural authority, for example as embodied in the traditional centralized infrastructure.

We show that for a generic compact initial surface in $\mathbb{R}^3$, and any generic compact initial mean convex hypersurface in higher dimensions, the mean curvature flow encounters only finitely many singularities. This resolves the full generic Morse regularity conjectures of Ilmanen and Colding-Minicozzi-Pedersen in the above two cases. More generally, we show that if a mean curvature flow of a compact hypersurface has only multiplicity one spherical and cylindrical singularities, then after a small perturbation of the initial condition only non-degenerate cylindrical and spherical singularities appear.

Hardware datasheets contain engineering specifications essential for component selection, system integration, and hardware compatibility verification. However, manual inspection is tedious and error-prone because datasheets are lengthy and heterogeneous, with dense information fragmented across tables, figures, and text. Although large language models (LLMs) can automate this process, naïve one-shot processing omits relevant specifications even when the document fits within the model's context window. We present an automated LLM-driven pipeline that transforms hardware datasheets into general-purpose hierarchical hardware knowledge graphs through chunk-based property extraction, hierarchical routing, and iterative refinement. The framework uses LLMs for information extraction and semantic reasoning while preserving source provenance and applying graph updates through validated instructions and deterministic local execution. This staged design reduces per-query context and maintains traceable intermediate representations. The resulting knowledge graphs provide semantically organized, source-grounded hardware properties for downstream retrieval and reasoning. We evaluate the framework on 20 hardware datasheets spanning diverse component types and document lengths. We assess property faithfulness and introduce inter-graph completeness to measure relative semantic coverage without exhaustive ground truth. Our framework achieves 98.8% property faithfulness compared with 76.1% for the one-shot baseline, while fully covering 83.6% of faithful baseline properties; in the reverse direction, the baseline fully covers only 10.5% of ours. These improvements provide more reliable and comprehensive engineering information for downstream hardware design and verification, reducing the risk of overlooked constraints and incorrect decisions.

Foundation models are increasingly used for high-level multi-robot task planning, yet independently generated symbolic plans can become jointly infeasible after geometric grounding because of shared spatial and temporal constraints. We study execution-aware coordination of large-language-model (LLM) planners for multi-robot navigation. We compare a centralized architecture, in which one LLM jointly reasons over the robot team, with a decentralized multi-LLM architecture in which each robot independently proposes a task plan. A geometric verifier grounds the candidate plans into robot trajectories, detects predicted inter-robot conflicts, and converts conflict information into structured feedback for targeted replanning. This verifier-mediated architecture separates semantic task reasoning from execution-level feasibility while enabling iterative coordination among LLM planners. Experiments in CARLA and on Crazyflie aerial robots evaluate coordination performance as task complexity increases and characterize the trade-offs between centralized reasoning, decentralized verifier-mediated coordination, and physical executability.

Lymphangions, the valved muscular segments of lymphatic vessels, pump lymph against an adverse pressure gradient. Current models resolve this pumping in one vessel, a short chain or a regional network, and whole-body human models are passive. We extend the passive whole-body network of a companion static study to the active regime on the same anatomical graph, with intrinsic contraction, dynamic valves and a respiration-gated venous outlet on every vessel. Without fitting the waveform or the contraction amplitude, the network settles into a bounded pumping cycle at the lymphovenous junction, at a clinical pressure of 7.5 to 12.3 mmHg and a drainage of 35 to 50 mm^3/s. Respiration adds a swing of 1 to 2.5 mmHg at the junction and contraction about 1 mmHg inside the duct. Pressure generation is limited more by load than by drive, and no single contraction amplitude fits every vessel. Raising the subclavian venous pressure from 3 to 15 mmHg lifts the deep cervical node from 8.6 to 21.1 mmHg at the weakest drive, a transmission of 1.04 that falls to 0.78 at the strongest, so contraction buffers cervical drainage against venous pressure. The same solver gives the human cervical outlet pressure that cerebrospinal-fluid drainage models prescribe.

Let $\mathcal D$ be a product of Drinfeld upper half spaces and let $J$ be the corresponding product of projective linear groups acting on $\mathcal D$. We prove that if $Y$ is a reduced, quasi-compact separated strict $\mathbb C_p$-analytic space and $f: Y\to\mathcal D$ is analytic with $\dim Y<\dim\mathcal D$, then $J f(Y(\mathbb C_p))$ is closed and nowhere dense in $\mathcal D(\mathbb C_p)$. The quasi-compactness assumption is necessary even when the source is a strict rigid analytic curve, while a quasi-compact perfectoid source can still have an orbit closure with nonempty interior. For every finite extension $F/\mathbb Q_p$ and $d\geq3$, we construct a closed smooth irreducible non-quasi-compact curve in $\Omega_F^{d-1}$ with dense orbit under a fixed diagonal one-parameter subgroup. We also construct a $\mathbb Q_p$-defined analytic map from the open unit disc to $\mathbf P^2$ with dense $\operatorname{PGL}_3(\mathbb Q_p)$-orbit, and an analytic map from a one-dimensional affinoid perfectoid torus to $\mathbf P^2$ whose $\operatorname{PGL}_3(\mathbb Q_p)$-orbit closure has nonempty interior.

Language models have been adopted faster than evaluation practices that reliably assess their risks, and a growing share of use is affective rather than informational. Within that traffic users disclose suicide and self-harm (SSH) crises. Existing safety benchmarks score models primarily on refusal, but refusal is not the desired behavior here; the conversation requires a substantive supportive response. We introduce DistressBench, 718 English conversations with clinician-authored user turns spanning 23 SSH subcategories, each paired with a task-specific rubric of weighted atomic criteria across seven dimensions of crisis support and a clinician reference response. A model's score is the weighted average over the criteria its reply satisfies, reading directly as the fraction of required care delivered. The best of 25 frontier models satisfies 88.4% of weighted criteria and the median model 71.4%, against 98.7% for clinician references scored on the same criteria. Pooled across all 25 models, in 35.3% of conversations where a model satisfied every criterion for recognizing the crisis, it failed to route the user toward human help. The shortfall concentrates in the dimensions requiring the model to act rather than withhold. Contrasting each dimension's median model against its best separates capability limits from choices: de-escalation reaches 74.6% in the strongest model against a median of 46.0%, and safety disclaimers 84.6% against 8.3%. Multi-turn conversations score below single-turn ones (68.5% against 74.2%). We release the corpus, rubrics, judge calibration labels, and a companion annex of 20 long-horizon red-team transcripts. These results argue for evaluating crisis safety on the care a model delivers rather than on what it declines to say, and for treating de-escalation and resource routing as deployment requirements.

Task-oriented conversational agents remain fragile under realistic conditions, especially when users exhibit non-cooperative behavior. Existing function-calling benchmarks underrepresent adversarial conversation trajectories, thereby limiting the training resources available for developing robust agents. We present MADS (Multi-Agent Dialog Simulation), a synthetic data generation pipeline that produces structured, diverse dialogue trajectories. We inject adversarial conversation, producing validated conversation trajectories that capture realistic, unhappy paths. Using MADS, we construct MADS-Bench: a benchmark spanning six domains, 55 personas, and 1095 conversations, with approximately 66% of conversations containing failure-prone conversation scenarios. We evaluate the robustness of the trained models using a two-stage training approach that combines supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR).

We present the first systematic study of the resilience of large language model (LLM) agents to random hardware-induced faults. Prior work on hardware fault resilience in LLMs focuses on model-level outcomes, such as incorrect outputs or generation failures. Yet LLM agents repeatedly interact with tools and environments, creating loops that can change how hardware faults manifest. We conduct an extensive evaluation across three agent benchmarks, comprising 6,814 agent executions: 2,400+ clean runs and 4,300+ fault-injected runs with independently sampled random two-bit weight faults. We trace how these faults propagate through agent execution. Unlike prior studies, we find that agentic execution transforms model-level corruption into distinct system-level outcomes: (1) agents can recover from faulty model outputs and still complete tasks correctly; (2) this recovery increases computational and execution costs; (3) a few faults cause subtle but semantically meaningful errors while the agent otherwise behaves normally; and (4) for some models and scenarios, faults can alter security and alignment behavior.

Aggregated academic search interfaces combine resources from heterogeneous sources, such as scholarly libraries and reference encyclopedias, into a single workspace. Such tasks frequently span multiple sessions, requiring learners to return to and resume work after hours or days. End-of-session consolidation, the activity of organizing and synthesizing gathered information so it can be resumed later, is important in this setting but underexplored in search-as-learning. Generative AI (GenAI) has the potential to make consolidation more efficient, but it may also shift cognitive work away from the learner. This study compares two consolidation scaffolds embedded in the workspace framework of an existing academic search interface that aggregates and clusters saved search results from two different sources (our university library and Wikipedia), which was developed in earlier work. STAIS (Steerable AI Summaries) provides a workspace enhancement that produces GenAI summaries that users can correct, reorganize, and fact-check, automating parts of the consolidation activity. NT (Note-Taking) provides a workspace enhancement that includes rich-text note editors that learners can use to author their own consolidations of what they have found. In a controlled between-subjects study (n=36), participants completed two sessions one week apart while searching an unfamiliar topic using either STAIS or NT. The two conditions produced contrasting post-resumption outcome profiles: STAIS supported faster task resumption and was associated with broader conceptual coverage and higher self-reported domain knowledge, whereas NT was associated with richer inter-concept connections and higher self-reported certainty.

We propose NV-Tesseract-AD, a curriculum learning-enhanced conditional diffusion framework for multivariate time-series anomaly detection. Our central hypothesis is that reconstruction difficulty can be explicitly structured along two complementary axes, masking severity and diffusion noise level, and that training from easier to harder tasks along both axes simultaneously stabilizes optimization and improves residual discriminability. We integrate a Transformer-based diffusion backbone with a three-stage dual curriculum over masking ratios and diffusion timesteps, retaining anomalous samples throughout training and relying on structured corruption to drive self-supervised representation learning. We evaluate NV-Tesseract-AD under two regimes: in-distribution training on the TSB-UAD-M corpus and strict zero-shot transfer to ten EasyTSAD benchmark families without per-dataset tuning. The dual-axis curriculum produces masking-invariant reconstruction, confirmed by a near-zero loss gap between complementary masking strategies within the first training stage, and yields an approximately 3.7-fold improvement in F1 and a 56 percent reduction in MSE relative to a non-curriculum baseline. Zero-shot performance reaches F1 = 0.984 on NAB and F1 = 0.948 on NormA. Weaker results on heterogeneous families are traced to unsupervised threshold calibration rather than reconstruction quality, identifying calibration as the dominant remaining detection bottleneck. Accelerated inference via DPM-Solver++ reduces sampling cost by a factor of 3.5 to 5.7 at the dataset level but couples to thresholding behavior, establishing that the sampler and threshold must be validated jointly. Together, these findings demonstrate that aligning diffusion denoising difficulty with curriculum learning principles provides a principled and practical approach for robust anomaly detection in complex multivariate time series.

Multiterminal Josephson junctions (MTJJs) extend conventional two-terminal Josephson physics to devices defined by multiple superconducting phases, where rich phenomena have been predicted. However, in conventional MTJJs, superconducting transport is often dominated by local coupling between neighboring leads, obscuring physics that goes beyond simple connectivity. Here, we realize a multiterminal Josephson network in a graphene Hall bar geometry with micron-scale dimensions, enabling substantial nonlocal coupling between distant terminals. Our key observation is a pronounced negative differential longitudinal resistance (NDLR) in the four-probe response near switching currents, while no such feature appears in two-terminal measurements. We reproduce the NDLR response with a network containing the interplay between two distinct forms of nonlocality: nonlocal Josephson coupling, which determines the switching sequence of the superconducting network, and nonlocal normal-state conductance, which redistributes current after different regions switch. Our results highlight the potential of the emergent NDLR for characterizing superconductor–semiconductor hybrid systems and for nonlinear circuits. Further, this Josephson Hall bar geometry provides a platform for exploring the interplay between mesoscopic superconductivity and many-body ground states in graphene-based heterostructures.

In this report, 15 variants of nine parametric models for non-directional ocean wave spectra are reviewed. They are compared against observation-based spectra created by taking spectra measured by moored buoys, binning by wind speed, and averaging. Surface Stokes drift (SSD) and Stokes drift profiles are computed using the model and observation-based spectra, and these are compared. This 'parametric wave model' strategy for estimating Stokes drift is contrasted with other methods for estimating Stokes drift, such as using spectra from a dynamical wave model. Advantages and shortcomings are reviewed. One practical application for Stokes drift profiles is in ocean circulation modeling. Here, the information is used to compute wave-enhanced mixing ('Langmuir turbulence'). We present examples of such cases, in the context of a global ocean model, tested in two modes (coupled and uncoupled) and with two types of forcing (Stokes drift approximated using wind speed and Stokes drift computed using a full ocean wave model). A more comprehensive review of the uncoupled tests with the full ocean wave model can be found in a separate report, Fan et al. (2026). We evaluate simple Stokes drift models that use wind speed and/or wind stress but do not require a dynamical wave model. We find some are sufficiently accurate for routine prediction of SSD. A subset of these models can also predict Stokes drift profiles (via parametric spectra), but none are accurate in this role. The SSD values computed from a very large drifting buoy dataset are compared with the total drift calculated using positional change of the same buoys. The SSD is roughly 25% of the total drift, consistent with expectations based on 'rules of thumb'. However, the scatter is large, indicating that such rules have predictive skill only in the mean sense.

Consider the Mathieu operator $L=-d^2/dx^2+2a\cos(2x)$ for fixed $a\in\mathbb{C}\setminus\{0\}$. For large enough $n$, the periodic ($n$ even) or anti-periodic ($n$ odd) problem on $[0,\pi]$ has exactly two algebraically simple eigenvalues $\lambda_n^-,\lambda_n^+$ in the disc $|\lambda-n^2|<1$. We prove that, as $n\to\infty$, $\lambda_n^+-\lambda_n^-=\pm\frac{8(a/4)^n}{[(n-1)!]^2}\exp\left[-\frac{a^2(n+2)}{4(n^2-1)^2}-\frac{15a^4}{128n^7}+O(n^{-8})\right]$, where the sign depends on the labelling. For real $a$ with $\lambda_n^-<\lambda_n^+$, this difference is the length of the $n$th spectral gap. Expanding the exponential determines the relative corrections of orders $n^{-4}$ to $n^{-7}$, beyond the correction $-a^2/(4n^3)$ obtained by Anahtarci and Djakov (2012) with a relative $O(n^{-4})$ remainder. By isospectrality, the same formula holds for the potential $ce^{-2ix}+de^{2ix}$ with $a^2=cd\ne0$.

The lymphatic system returns 2 to 3 L of interstitial fluid per day to the venous circulation and drains the central nervous system. Most models of lymph transport describe one lymphangion or a short chain, so they cannot show how body-wide topology, calibre hierarchy and the venous outlet set pressure and flow. We build a passive steady hydraulic model on an anatomically derived graph of the human lymphatic system, with 1010 vertices and 1126 edges expanded to 89,060 sub-edges. Radii follow a Murray cube law, derived both from minimum work and from wall-shear-stress homeostasis. Pressure and flow follow from nodal hydraulic balance over Hagen-Poiseuille conductances, with an exponential wall law, hysteretic valves, Starling leak, node retention and a Windkessel outlet. Active contraction is off, and only the inflow scale is calibrated. Two quantities that are not fitted land in clinical range at the same operating point. Radii fall inside the vessel-class range, from 1.50 mm in the thoracic duct to 0.16 mm in the cervical collectors, and pressures stay below the 50 mmHg cuff scale on all sub-edges. About 49% of the gross load is removed before the outlet, nearly independent of the outflow scale and of the outlet parameters. On the route that drains the central nervous system, 30% of the superficial cervical load reaches the vein, at 10 to 18 mmHg along a segment that cervical-scale models replace with a prescribed outlet pressure. The model is a passive whole-body baseline for contraction-driven and clinical extensions.

The growing diversity of two-dimensional (2D) materials has expanded opportunities in electronics, energy technologies, catalysis, and optical applications. Within this landscape, understanding how corrugated architectures accommodate metallic bonding and influence the stability and properties of monoelemental niobium sheets remains an open research direction. Here, we propose a corrugated niobene nanosheet, denoted c-Nb, and investigate its structural, energetic, vibrational, mechanical, electronic, and optical properties using density functional theory simulations. The optimized structure adopts a centrosymmetric P-3m1 symmetry, with six atoms per primitive cell and a geometrical thickness of 4.279 Angstrom. Its cohesive energy of -6.03 eV/atom indicates strong binding relative to isolated atoms, although bulk BCC Nb remains energetically favored by 2.49 eV/atom. The absence of imaginary phonon frequencies along the investigated high-symmetry path and the positive-definite in-plane stiffness matrix support dynamical and mechanical structural stability. Despite its pronounced corrugation, the nanosheet exhibits isotropic in-plane elasticity, with Young's and shear moduli of 43.66 N/m and 13.66 N/m, respectively, and a Poisson's ratio of 0.598. Electronic states near the Fermi level are dominated by Nb 4d orbitals, establishing multiband metallicity. These findings identify a corrugated structural candidate that combines strong cohesion, elastic compliance, and metallic electronic behavior, extending the structural diversity of monoelemental 2D metals and providing a basis for further exploration of niobium nanosheets in electronic and ultraviolet-responsive systems.