Papers

Modern video generators can synthesize convincing scenes while rendering a few characters incorrectly, omitting requested text, or changing it across frames. Measuring these failures is harder than applying OCR independently to each frame: when a scene contains multiple text targets and multiple noisy OCR regions, the evaluator must first determine which observation belongs to which intended text. We introduce TextVBench, a benchmark and evaluation framework that formulates visual-text evaluation as a multi-target, multi-instance spatiotemporal correspondence problem. TextVBench contains two complementary tracks: TextVBench-T2V measures the creation and maintenance of prompt-specified text, while TextVBench-Ref2V measures preservation of text from reference images. Both tracks use target-conditioned candidate selection rather than unconstrained VLM transcription, support one target mapping to multiple OCR regions, and separate frame-level spatial fidelity from cross-frame temporal fidelity. A length-weighted harmonic score summarizes these factors, while extra-text diagnostics expose failure modes hidden by a single OCR score. The benchmark is organized around controlled text, layout, carrier, and motion attributes so that aggregate scores can be traced back to specific capability limits. Across the evaluated model runs, MiniMax-H3 reaches a T2V TextVScore of 68.32, while the strongest Ref2V score is 56.51; attribute slices reveal a pronounced drop as the number of simultaneous reference text regions increases.

Traditional system identification typically involves two steps: initially, training data is processed to estimate a model of the system of interest; then, the model is used to predict outputs for new input queries. In contrast, recent in-context learning approaches learn a meta model that maps directly from training data and query inputs to outputs in a zero-shot fashion, bypassing explicit model estimation. This approach leverages a meta model trained offline, on large sets of synthetically generated data, to emulate the behavior of a system identification algorithm. In this work, we advance this meta-learning framework for system identification in three directions. First, we develop a unified probabilistic meta-identification architecture that combines a Transformer-based encoder-decoder with tailored mechanisms for handling long training sequences, enabling calibrated predictions over long horizons. Second, we show how a pre-trained meta model can act as an uncertainty-aware synthetic data generator for classical system identification, providing task-specific data that regularize parametric models when real data are scarce. Third, we systematically evaluate the framework on standard nonlinear system identification benchmarks and on a challenging out-of-distribution scenario, demonstrating competitive accuracy with state-of-the-art methods and strong generalization, including cases where input-only adaptation is sufficient to handle significant distribution shifts.

Specific heat of graphite at high temperatures demonstrates significant deviation from the “3R” Dulong-Petite law. It can be explained by accumulation of defects with formation energy as low as 1.45 eV. Based on the density functional theory calculations, we propose a possible candidate of topological defects, whose energy is comparable to the growth activation energy.

End-to-end planning in complex scenes must model diverse driving intentions and select the most appropriate trajectory using scene context and future risks. However, multi-query regression may collapse to one mode, while anchor-based and generative planners share a scaling bottleneck: small candidate sets limit maneuver coverage or discovery of high-quality trajectories, whereas large sets increase generation and scoring costs. We propose ManeuverFlow to generate trajectories for distinct intentions in one step and evaluate and refine them using current and future evidence. The model selects relevant maneuver modes, draws corresponding noise from the Gaussian mixture prior, and maps it to intention-specific trajectories through a maneuver-routed mixture-of-experts MeanFlow field. Current features and future features predicted by the world model are aggregated along each candidate path into trajectory-specific evidence for PDM-factor prediction. Using the same evidence, the refiner predicts local Frenet residuals for selected candidates; refined and retained candidates are then jointly re-scored for selection. With a lightweight TransFuser-style ResNet-34 backbone, ManeuverFlow achieves 90.6 PDMS and 90.1 EPDMS on NAVSIM v1/v2 navtest, respectively. On WorldEngine's recent 3DGS closed-loop benchmark, it attains the highest listed PDMS$^\ast$ scores of 75.19 and 74.58 under logged and IDM-controlled traffic, respectively. Code: https://github.com/wyh-autonomy/maneuverflow.

GPU working sets have outgrown GPU memory, so applications must oversubscribe it, keeping only hot data on the GPU. NVIDIA's Unified Virtual Memory (UVM) supports oversubscription, but its performance degrades once data exceeds the GPU's memory. This stems from its fault-driven data placement, wasting interconnect bandwidth, and underutilizing the GPU. This raises the question: can UVM oversubscription become practical on stock GPUs, exploiting GPU memory speed and minimizing wasted interconnect bandwidth, without placement hints, profiling runs, or application changes? To this end, we present Ember, a UVM management framework for oversubscribed workloads. Ember introduces a hybrid address space; allocations start as mapped into host memory, omitting data migration. Ember's counter-driven region promotion uses GPU's hardware counters to fetch hot regions to the GPU, its decay-based eager reclamation proactively evicts cold regions in batches, while via online region re-placement, Ember tracks the hot set at runtime. We implement Ember within NVIDIA's open-source UVM driver for unmodified applications and as a userspace variant on stock CUDA. At 1.5x oversubscription, it is 6.8-10.6x faster than UVM on six applications across three PCIe GPUs, and generates LLM tokens 4.7-16x faster at sizes where manual placement cannot run at all. Without a profiling run, it beats a profile-guided alternative by 2.1x on kernel time.

Exhaustive optimistic scoring can make action selection costly in linear bandits with a large fixed set of $K$ actions, requiring $\mathcal{O}(K)$ score evaluations per round. We introduce Persistent Candidate LinUCB (PC-LinUCB), which maintains $k$ active actions and samples $q$ challengers uniformly from the remaining $K-k$ actions at each round. By scoring only these $k+q$ candidates, PC-LinUCB reduces the number of score evaluations per round to $\mathcal{O}(k+q)$. We establish problem-independent and problem-dependent high-probability regret bounds consisting of a standard linear-bandit learning term and an additive discovery cost of order $\mathcal{O}\left((K-k)\log(K-k)/q\right)$ at fixed confidence. Notably, this additional cost does not grow with the horizon. For fixed gaps and sufficiently long horizons, we establish an $\Omega\left((K-k)/q\right)$ regret lower bound under uniform challenger sampling, showing that partial candidate access itself imposes an unavoidable regret cost. Experiments show that PC-LinUCB substantially accelerates LinUCB while maintaining comparable regret, and that the persistent candidate mechanism can extend beyond LinUCB.

Forecasting desertification has received far less attention than mapping it. We use a convolutional LSTM (ConvLSTM) trained on ten annual Earth Observation (EO) variables (2000-2023) over the semi-arid state of Ceará, Brazil, on a 256x206 grid at roughly 2.2 km resolution, to measure how much of the year-to-year evolution of a MapBiomas-derived degradation index (IDV) can be predicted from satellite data. The network predicts a bounded change on top of the last observed state, is trained in two stages (single-step, then recursive rollout) and is evaluated on a contiguous spatial and temporal holdout. State-wide mean IDV rose from 0.238 to 0.350 between 2000 and 2023, about 0.005 per year, but cell-level change is dominated by year-to-year jumps, and a gradient-boosting model given all channels explains only 4.7% of its variance. Against this ceiling, the ConvLSTM lowers RMSE relative to persistence by 1.6% one year ahead and 4.4% four years ahead in a held-out southern band (p < 0.001), beats persistence with a constant drift and reproduces 53-69% of the regional drift. Drift-aware baselines only help from the third year onward, by up to 4.4% in RMSE at six years. An earlier formulation that redrew the whole field each year predicted changes unrelated to the observed ones (correlation -0.06) and produced a spurious decline when run recursively. Channel knockouts suggest that elevation and EVI degrade the forecast, while slope and aspect have no measurable effect. Monte Carlo dropout and stochastic scenario maps convey the uncertainty that the expected-value forecast hides. Under climatological forcing, mean IDV rises from 0.350 in 2023 to 0.370 in 2030. EO-based desertification forecasting is therefore reliable for the regional trend, not for individual cells.

We study the recovery of a three-dimensional Riemannian manifold from the areas of minimal surfaces. We show that areas of minimal surfaces determine the metric on manifolds foliated by minimal surfaces, up to boundary-fixing isometry. This resolves the conjecture on boundary rigidity for least-area data in an exterior formulation, even without topological restrictions on the minimal surfaces. We also prove global rigidity for Einstein metrics on compact three-manifolds with boundary, assuming $\(\pi_1(M,\partial M)=0\)$ and a non-trapping condition. The Einstein result includes the

Linear models can predict well in complex domains even when underlying mechanisms are nonlinear. We argue that this can occur because the complexity of the world exceeds the complexity that can be learned reliably from finite, partially observed data. Flexible models can exploit curvature, thresholds, and interactions that fit one hospital, period, population, policy setting, or measurement process but fail elsewhere. When this instability lies mainly in the nonlinear component, omitting it can give linear models lower prediction error in a new environment. We connect this claim to work on simple linear rules, bias and variance, regularization, best linear prediction, measurement error, distribution shift, invariance, Rashomon sets, and underspecification. Proposition 1 gives a simple squared-error case in which a nonlinear rule learned in the training environment raises target error because it transfers too weakly; finite-sample estimation error can strengthen this result. Partial observation alone is not enough: the link between observed variables and unobserved factors must vary across environments. Stable nonlinear structure favors flexible models, and linear models remain vulnerable to linear spurious relationships. Model complexity should therefore be chosen by cross-environment stability as well as training-data accuracy.

Reliable real-time failure detection is critical for deploying learned robotic policies, yet defining failures explicitly is infeasible in open-world settings. This motivates failure detection from nominal demonstrations alone. Existing methods, however, struggle to balance robustness to benign variations with the ability to capture semantic and interaction-level errors. We introduce Robot-WATCHDOG, a self-supervised framework that models manipulation scenes as object-centric spatio-temporal graphs. Building on this representation, we investigate two complementary failure scores: Trajectory Projection and Correlation (TPC), a non-parametric measure of geometric deviation and relational consistency, and GnnT, a Graph Attention Network with a Transformer that predicts future object states. We evaluate both on a public benchmark and on a new dataset of semantically complex manipulation tasks, across six tasks and ten random seeds. Both scores outperform image-based failure detectors, particularly on semantic and interaction-level failures. TPC provides the strongest performance in AUPR and MCC on both dataset averages, with low variance across seeds. Qualitative results are made available at https://cea-list.github.io/Robot-WATCHDOG/.

We propose a coupling-aware tri-hybrid precoding architecture for simultaneous wireless information and power transfer (SWIPT) using reconfigurable parasitic antennas. By integrating multiport network theory and information theory, we develop a third precoding layer that explicitly accounts for mutual coupling and reactive energy storage within the antenna domain. To efficiently optimize this architecture, we propose a low-complexity tri-hybrid solution comprising coordinate descent (CD) for parasitic reactive load optimization, a singular value decomposition (SVD)-based analog precoder, and a protective partial zero-forcing (PPZF) digital precoder. Numerical results demonstrate that the transmit array introduces impedance mismatch that reduces the real power accepted by the antenna array. Moreover, the proposed tri-hybrid architecture significantly outperforms hybrid precoding by constructively exploiting the coupling effects, yielding substantial gains in spectral efficiency (SE) and sum non-linear harvested energy (NLHE) at a fraction of the hardware cost.

The $S=1/2$ Kitaev model provides an exactly solvable setting in which spin fractionalization produces itinerant Majorana fermions coupled to an emergent $\mathbb{Z}_2$ gauge field. In realistic Kitaev magnets, however, non-Kitaev interactions break integrability and make their finite-temperature dynamics difficult to calculate. We develop a finite-temperature parton mean-field theory based on the Abrikosov-fermion representation and incorporate the effects of exchange interactions through the random phase approximation to calculate the dynamical spin structure factor. For the pure Kitaev model, the method reproduces a strong, nearly dispersionless peak at a finite low energy together with a broad continuum at higher energies. It also reproduces the momentum dependence of the dynamical spin structure factor and its temperature evolution, including the downward shift of the low-energy spectral weight, in qualitative agreement with continuous-time quantum Monte Carlo calculations. Within the present theory, the contraction of the continuum is traced to the temperature-induced reduction of the bond amplitudes, which narrows the parton bands and compresses the two-parton phase space. For ferromagnetic Kitaev coupling, the energy of the low-energy peak is minimized at the $\Gamma$ point for a ferromagnetic Heisenberg interaction and maximized there for an antiferromagnetic interaction. In the ferromagnetic Kitaev-$\Gamma$ model, the static response develops a finite-momentum softening, indicating an instability toward the period-three magnetic order reported in numerical studies. For uniform-ferromagnetic states selected by a $[111]$ magnetic field, we find that a broad parton continuum coexists with magnonlike dispersive structures, producing spectral features consistent with a spinon-to-magnon crossover.

How can human researchers use artificial intelligence (AI) to make progress on difficult mechanics problems while retaining control of physical significance, mathematical rigor, and validation? We explore this opportunity through a quantitative stability problem rooted in the Eshelby conjecture. For a uniform dilatational eigenstrain in a homogeneous isotropic solid, we establish a nonlinear two-sided relation between whole-interior strain heterogeneity and a geometry-only distance from the ellipsoid family, with half-order Sobolev scaling that captures morphological length scale. We then develop a complementary finite-shape theory: ellipsoidal subtraction yields an exact residual-field budget from which systematically convergent upper and lower bounds are obtained for a prescribed geometry. Applied to protruded, indented, and necked inclusions, the bounds agree with independent displacement-based finite-element calculations. The study also documents a human-led, AI-assisted research process in which researchers select consequential problems, shape targets, audit arguments, require independent checks, and interpret mechanics, while AI broadens analytical exploration and supports computation. This experience also led us to develop Open Problems in Mechanics, a versioned community-facing map that makes selected mechanics questions, their known/open boundaries, and criteria for progress explicit.

Long-term personalized tutoring requires language models to translate a learner's evolving history into appropriate instructional support. Existing LLM-based tutoring systems use memory and retrieval to personalize responses, but the pedagogical rationale linking historical evidence to teaching strategies is not always explicit. We introduce RegulaTutor, a theory-informed framework for long-term personalized tutoring that connects historical evidence, learner interpretation, and learner-facing instruction. Guided by self-regulated learning (SRL) research, its tutoring harness operationalizes a cognitive, affective, metacognitive, and motivational (CAMM) framework for deriving evidence-supported learner interpretations and teaching policies. The resulting policies retain links to their supporting evidence and are updated as new observations arrive. This staged workflow also supports systematic refinement through PairRoute, an offline procedure that uses intermediate execution records to distinguish harness, model, and mixed failures. PairRoute proposes and evaluates targeted harness revisions, then constructs reviewed stage-level supervision for confirmed model failures remaining under the selected harness. We focus this refinement on historical evidence acquisition, which underpins personalized instruction and allows outputs to be checked against learning records and reference answers. RegulaTutor improves historical-evidence accuracy, learner diagnosis, and judged teaching quality on average across evaluated model configurations. PairRoute results show an endpoint improvement in standalone-model historical-evidence accuracy over the refinement trajectory. Together, these results highlight RegulaTutor's potential to support lifelong personalized education through instructional policies that evolve with learner histories and targeted improvements to the tutoring system.

Tibetan, spoken by some seven million people on and around the Qinghai-Xizang Plateau, remains underrepresented in natural language processing (NLP). We present, to our knowledge, the first peer-reviewed systematic bibliometric survey of Tibetan NLP: 140 publications from 2004 to 2026 in the ACL Anthology and DBLP, annotated for venue type, research area, first-affiliation country, and resource release. Output has grown rapidly, with 84% of the publications appearing since 2018, and has expanded bottom-up along the NLP pipeline, from corpora and word segmentation through script, speech, and applications to large language models (LLMs). Researchers based in China first-authored 87% of the publications, which span all 25 represented research areas, whereas work from other countries concentrates on Classical Tibetan. We argue that this full-pipeline paradigm, realized under the compounded constraints of a complex script, dialectal diversity, and geographic remoteness, offers a reference model for other low-resource languages of the Global South. We also identify gaps in dialect coverage, responsible NLP, native and comparable evaluation, cross-community visibility, and resource release (only 29% of the publications release data, code, or models), and recommend open science, shared standards, and community participation for the LLM era.

We develop a model-free, noise-aware Bayesian clustering framework that constructs a martingale posterior over clusterings from a predictive Bayesian perspective. This avoids explicit likelihood-prior model specification, expensive MCMC sampling, and mode-seeking variational approximations, while accommodating irregular cluster shapes and enabling parallel generation of posterior draws. To interpret the posterior over high-dimensional, noise-aware clusterings, we introduce a Wasserstein-type summarization method. Using a novel noise-aware loss, this approach compresses the large collection of posterior draws into a small set of weighted representatives, capturing distinct cluster structures and noise assignments. Further, we marginalise the uncertainty in the density level, avoiding reliance on a specific choice. We establish theoretical properties of the proposed loss and guarantees for the level-marginal posterior approximation. Results demonstrate the framework's effectiveness in capturing clustering and noise uncertainty.

Large-scale data-intensive workloads powered by iterative graph or machine learning processing have an insatiable demand for memory due to the continuous generation of new intermediate data. Although existing caching strategies strive to keep performance-critical data in memory, they often fail in practice because they are limited to node-local memory management and assume high node availability. In this paper, we introduce ShoreKeeper, a framework that maximizes execution efficiency by treating cluster-wide memory as a unified, elastic pool. ShoreKeeper employs a global optimization model that maps distributed dataflows into a single graph abstraction to minimize the recovery cost of data evicted from cache across the cluster. To handle memory working sets and cluster capacities that fluctuate over time, ShoreKeeper employs a graph-pruning technique that permits the fast refinement of the caching plan. These optimizations allow ShoreKeeper to work seamlessly whether deployed on highly available VMs or cost-effective preemptible instances in the cloud. Our experiments show that ShoreKeeper consistently outperforms state-of-the-art systems.

Coding agents are increasingly used to build programs from scratch as software engineers do rather than to resolve issues within a repository or to generate a repository from a given requirements document. However, scalable verifiable training environments that force agents to discover what is required and to decide how the system should be designed remain scarce. Moreover, existing agent workflows also cannot sustain long-horizon development in which the agent explores the reference program before implementation. To address these challenges, we introduce ProgramGym, a framework consisting of Program-Env and Program-Agent for scalable training environments of long-horizon program reproduction agents. Specifically, Program-Env constructs verifiable training environments from repositories through a feedback-driven loop that repairs runtime sandbox builds and the quality of black-box tests. Using this novel pipeline, we obtain 1,648 training environments spanning C, C++, Go, and Rust. Program-Agent combines a role-isolated workflow of analysis, implementation, and review with revision management for compilation checks and failure recovery. Experiments on ProgramBench show that both Program-Env and Program-Agent improve performance on program reproduction. Fine-tuning on trajectories with these constructed environments raises the performance of Qwen3.8-27B from 34.87% to 42.84%. Program-Agent further raises the base and SFT models to 55.71% and 56.25%, both exceeding some flagship models. We will release the ProgramGym to support further research on coding agents that build programs from scratch.

Traffic efficiency and air pollutants are major issues in urban transport systems, as they are critical points for traffic flows and people living in urban areas, respectively. In response to this challenge, connected and autonomous vehicles (CAVs) have garnered increasing attention, with the potential to address the above challenges through cooperative maneuvering and real-time communication. In this paper, we consider an intersection served by a multi-access edge computing (MEC) where CAVs communicate with a controller to pass the intersection without the need of signals and without stopping unless in presence of heavy congestion. For this purpose, we propose a first-in-first-scheduled (FIFS) algorithm, and we specify the communication protocol for its practical implementation. The solution is validated through simulations with varying traffic densities, showing that it provides significant gains in travel time and emissions compared to both uncontrolled intersections, with or without traffic lights, and a benchmark solution derived from the literature.

Redshift space distortions (RSD) provide a powerful measurement of structure formation in modern cosmology. In this work, we explore the potential of 21cm intensity mapping with the MeerKAT Large Area Synoptic Survey (MeerKLASS) to detect RSD and probe gravity within the redshift range of MeerKAT's UHF band ($0.4<z<1.45$). Using simulated sky observations, we examine the impact from observational effects caused by blind foreground removal and MeerKAT's primary beam, plus $k$-cuts to the measurements. With a multipole expansion formalism, we demonstrate how the quadrupole of the HI power spectrum can be modelled and also show how a foreground transfer function can reconstruct the distorted signal. With 1,115 deg$^2$ of simulated MeerKLASS observations at $0.4<z<1.0$, our proposed analysis pipeline reveals a $6.7\sigma$ detection of an RSD-induced quadrupole relative to one caused by anisotropic observational effects alone. We also explore parameter inference with the full MeerKLASS 10,000 deg$^2$ survey. Within forecasts, we demonstrate how a tomographic analysis across the full UHF redshift range can be leveraged to address certain parameter degeneracies with e.g. the HI bias ($b_{\rm HI}$) and cosmic density of HI ($\Omega_{\rm HI}$). By adopting a redshift-dependent model for the bias and HI density combined with simulation-informed priors, we demonstrate how degeneracies can be alleviated, and indicate the potential for MeerKLASS to reach competitive precision on simple probes of modified gravity.

We study the Yang–Mills flow on $\mathbb{S}^4$ restricted to $\mathrm{SO}(4)$-equivariant connections on the trivial $\mathrm{SU}(2)$ bundle. It was expected that antipodal instanton–anti-instanton configurations would lead to infinite-time singularities along the flow. We show that this does not occur. Moreover, for any smooth equivariant initial condition, the flow converges to the trivial flat connection.

A particle beam with constant intensity that propagates into a plasma triggers a bump-on-tail instability. It has been demonstrated that beam modulation can be used to significantly reduce this instability. However, finding the optimal beam profiles that accomplish this is a complex optimization problem that suffers from an extremely rugged optimization landscape and a relatively high-dimensional parameter space. In this paper, we evaluate a number of optimization algorithms and find that population based methods (in particular, CMA-ES and Differential Evolution) perform best. In particular, we investigate beam profiles where the beam is either completely turned ON or OFF. This configuration is easier to accomplish experimentally, but leads to a higher-dimensional and thus more challenging optimization problem. We obtain ON/OFF beam profiles that reduce the instability by more than 80% and thus perform significantly better than what has been reported in the literature.

Geospatial machine learning often depends on task-specific feature engineering over large collections of sociodemographic, business, and environmental variables. Geospatial foundation models offer reusable representations of place, but leading approaches typically encode dynamic signals, such as mobility and search activity, or visual signals from satellite imagery. Few models represent the static structural composition of places (who lives and works there, which businesses operate there, and which risks they face) as transferable embeddings at fine spatial resolution. We introduce PlacePulse, a structural geospatial foundation model that compresses 572 attributes, expanded to 2,288 inputs through H3 neighborhood aggregates, into 256-dimensional unit-norm embeddings for 1,055,537 H3 resolution-7 cells in the United States. We also introduce a reusable two-axis benchmark that separates intrinsic representation quality from downstream predictive utility across multiple targets and spatial supports. Intrinsically, the embeddings reconstruct standardized inputs with mean $R^2 = 0.853$, use 251 of 256 approximately decorrelated dimensions, and show strong spatial organization, with Moran's $I = 0.93$ on the leading principal component and autocorrelation persisting beyond the input radius. Extrinsically, across 74 held-out targets from 12 external sources and four spatial supports, PlacePulse outperforms a dimensionality-matched feature-selection baseline on targets unrelated to the training inputs and is competitive with Google's Population Dynamics Foundation Model, leading on health, economic, election, economic-mobility and social-vulnerability outcomes, with relative strengths that vary by domain. These findings position structural composition as a distinct and complementary modality for geospatial foundation models, not simply a fallback where behavioral or imagery data are unavailable.

While naively simpler than its higher dimensional cousins, 2d physics still has its own share of puzzles. An example of this is instances of dualities of $(0,2)$ theories, passing all consistency checks save for a mismatch in global symmetry. This is common when said duality can be motivated by dimensional reduction of a duality of 4d $\mathcal{N}=1$ theories where said symmetry emerges in 2d, often from an anomalous 4d symmetry. It is commonly believed that these symmetries are broken in the 2d theory, though often the mechanism for such breaking is not understood. We examine various appearances of this phenomenon and show that in these cases the elliptic genus shows signs of the presence of marginal operators that can break these symmetries. Interestingly, these operators do not appear perturbatively but emerge from the calculation, raising the question of what operators the $(0,2)$ elliptic genus actually counts. This provides a potential mechanism to overcome the discrepancy in symmetry, though we point out some potential subtleties.

Quantum reservoir computing and its special case, quantum extreme learning, are powerful tools for predicting classical discrete time-series with quantum devices. Forecasting continuous signals, however, requires careful discretization as oversampling artificially boosts short-term accuracy but fails over realistic intervals, while undersampling captures too few features and yields poor predictions. Here, we experimentally generate four continuous time-series signals from a chaotic electronic circuit for forecasting with quantum extreme learning. Since simplified circuit models do not fully capture real behaviour, it is unclear whether algorithms that predict simulated time-series also work for experimental signals. We address discretization via two approaches: (i) sampling local peaks/troughs; and (ii) uniform sampling at the autocorrelation timescale. The former captures resonant behaviour, while the latter fully captures chaotic features as well as allowing the reconstruction of chaotic attractors. Our algorithm achieves high forecasting accuracy for both simulated and experimental signals, demonstrating its power for continuous signal processing on near-term quantum simulators, and remains robust against moderate decoherence.

Thyroid nodule malignancy assessment relies on ultrasound imaging, yet current risk stratification systems suffer from substantial interobserver variability, contributing to a high rate of unnecessary biopsies. We propose a radiomics-based framework combining quantitative feature extraction via PyRadiomics with a Generalized Additive Model (GAM) classifier, selected among six candidates for its favorable trade-off between discriminative performance and interpretability. To complement point predictions with uncertainty estimates, Cross-Conformal Prediction with label-conditional guarantees is applied, flagging ambiguous cases for clinical review rather than forcing a binary decision. Experiments are conducted on TN5000, a publicly available ultrasound dataset of thyroid nodules. After Spearman correlation filtering, 36 features are retained and the GAM achieves an AUC of $0.89$, a sensitivity of $0.83$, and a specificity of $0.82$ on the validation set. At a significance level $\epsilon=0.09$, selected to ensure empirical coverage guarantees on both classes, 63% of nodules on the test set receive a certain (i.e., single-label) prediction, with conditional precision of $0.94$ for certain malignant and conditional Negative Predictive Value (NPV) of $0.73$ for certain benign cases, while the remaining 37% are referred to the clinician.

Stance detection remains challenging for large language models (LLMs) due to implicit opinions, sarcasm, and context-dependent expressions of stance. Although automatic prompt optimization offers an alternative to fine-tuning without updating model parameters, existing structured approaches are not specifically designed around stance-label semantics and class-specific ambiguities. We introduce MoPrO-SD, a stance-specific prompt optimization framework that decomposes prompts into functional semantic modules and refines them sequentially using LLM-based feedback. We study three refinement strategies: instruction-only optimization (I), instruction with reasoning (I+R), and instruction with per-example reasoning (I+R+E), using class-conditioned errors for stance-label refinement. We also introduce VaccineStance, a new human-annotated benchmark for vaccine stance detection. Experiments across four datasets and five open-weight LLMs show that MoPrO-SD improves upon the human-written Base prompt in many settings. On SemEval-2016, class-conditioned feedback achieves higher mean Macro-F1 than pooled feedback for four of five backbones. Both reasoning-based strategies, I+R and I+R+E, achieve strong performance, with their relative effectiveness varying across models and datasets, while remaining competitive with recent prompt optimization methods. The MoPrO-SD code and VaccineStance dataset are available on GitHub.

We establish a spectral theorem for compact self-adjoint linear relations on a Hilbert space. The proof relies on the Arens decomposition of a closed linear relation into an operator part and a purely multivalued part, reducing the problem directly to the classical spectral theorem for compact self-adjoint operators. No finite-dimensionality condition is imposed on the multivalued part. We also prove a finiterank approximation theorem for compact bounded linear relations and a Fredholm alternative for compact self-adjoint linear relations. As an application, we employ the theory of boundary triples to show that the inverse of the Neumann Laplacian on a bounded domain is a compact self-adjoint linear relation. The abstract spectral theorem then recovers the classical eigenfunction expansion of solutions to the Neumann problem: an L2-convergent series in terms of the non-constant Neumann eigenfunctions together with an arbitrary additive constant. The framework extends naturally to general elliptic boundary value problems with finite-dimensional kernel.

We present GRHD simulations of Bondi-Hoyle-Lyttleton accretion onto a rotating black hole in Weyl conformal gravity, described by the Kerr-Weyl metric. The equatorial flow is evolved for spins a=0.9M and a=0.5M and several Weyl coefficients, which enter through an effective curvature parameter k, and is compared with Kerr runs sharing the same setup. For the k considered, the horizon, ISCO and test-particle frequencies differ from Kerr by at most ~2%. However, the curvature term grows as kr^2 and acts at the accretion radius R_acc. The flow is controlled by one quantity, eta=|k|R_acc^3/M, the ratio of curvature-induced to Newtonian acceleration at R_acc. For eta<~0.03 a stable shock cone forms, indistinguishable from Kerr. For eta~0.2-0.3 the anti-de Sitter-like term traps shocked gas behind the black hole and drives persistent flip-flop oscillations. For eta>~1.6 the cone is destroyed and an oscillating torus forms. Spin modifies only the inner few gravitational radii and does not change the regime. In strongly deviating models the mean accretion rate is 1.7-3 times the Kerr value and varies by 20-60% rms, versus 0.1-1% in Kerr-like flows. The power spectra show global QPOs with fundamentals at 2.9-4.8 Hz for a 10 Msun black hole, with harmonic ladders whose 3:2 and 2:1 ratios come from one non-sinusoidal oscillation, not resonances. These lie in the low-frequency QPO band and are unrelated to test-particle frequencies (~70-200 Hz). They scale as f_0=xi f_K(R_acc), with xi~0.7-1.1 for stable cones and xi~0.23-0.38 for flip-flop and torus regimes. Strong oscillations would be detectable within seconds for a bright Galactic source. The required |k| exceeds galactic-rotation-curve values by 18-36 orders of magnitude, so a detected shock-driven QPO would measure a local deviation from Kerr, and its absence would constrain |k|.

Nanomaterial properties are governed by atomic structure, and driving structural changes on demand could enable materials with programmable properties. Pulsed light is the most controllable route: wavelength, fluence, polarization and pulse timing can each be tuned to inject energy with nanometre and picosecond precision and mould atomic positions. Yet the spatiotemporal dynamics of these transformations have remained hidden, inferred only from before-and-after snapshots. Here we track gold nanorods in a transmission electron microscope under pulsed optical excitation, quantifying the structural transition frame by frame. The rods enter an unexpectedly dynamic regime: material redistributes within and between particles, twin boundaries nucleate and propagate, rods hundreds of nanometres apart reshape in concert, and individual rods almost vanish before recovering their original shape, indicating a structural memory. Driving and following these transformations with pulse-synchronized in situ electron microscopy could turn light-induced reshaping into a tool for setting structure, and consequently properties, on demand.