Papers

We construct regular integral models for a class of Shimura varieties of symplectic and orthogonal type with maximal parahoric level at an odd prime $p$. These models are defined over the ring of integers of the reflex field and have special fiber with normal crossings and irreducible components of multiplicity one or two. Our construction uses explicit equivariant modifications of the corresponding canonical local models for symplectic and split even orthogonal similitude groups, with the minuscule cocharacters corresponding to maximal isotropic Grassmannians. These modifications are obtained by successively blowing up Schubert varieties in the special fiber. This proves the equivariant modification conjecture of Pappas at maximal parahoric level in the above cases.

How do algebraic relations among unit directions influence the chromatic number of the plane? We introduce monomial Cayley graphs as a framework for investigating this question, beginning with graphs whose allowed unit steps are signed powers of a single complex parameter. Transcendental parameters give bipartite graphs, while for roots of unity we determine both the chromatic and circular chromatic numbers. As a geometric consequence, every graph admitting a unit-distance embedding in the plane in which the angle between any two edge directions is a rational multiple of $\pi$ is three-colorable. For algebraic noninteger parameters, four colors suffice, and three suffice when the primitive integer minimal polynomial has leading coefficient at least three. We conjecture that three colors suffice for every algebraic noninteger direction parameter and that four suffice for every direction parameter.

Time-series anomaly detection (TSAD) identifies deviations from patterns learned from historical data. In non-stationary settings, distribution drift and true anomalies can cause similar local changes, making it difficult to tell whether a deviation reflects abnormality or evolving context. Existing methods typically adapt to detected shifts or learn drift-insensitive representations, but do not resolve this ambiguity. We define this problem as temporal change disambiguation: determining whether a local deviation is explained by broader temporal evolution. We introduce MORA, a drift-robust TSAD framework that reconstructs the same local target from paired short- and long-term views. The reconstruction gap measures contextual support for a local deviation, and a data-dependent correction mechanism conservatively adjusts the primary local anomaly score. Context can only reduce the score when it improves reconstruction of the same target. MORA needs neither drift annotations nor online adaptation. Experiments on four TSAD benchmarks show strong robustness to non-stationarity while preserving sensitivity to genuine anomalies.

Image-size characterization connects coding constraints to the probability carried by source cells and their output images. At second order, this connection must retain the joint fluctuations of all active constraints and distinguish encoder-observable source variation from unobserved channel noise. We extend the single-output posterior characterization to simultaneous posterior and conditional-composition image bounds. The bounds preserve the actual coupling of multiple outputs, quantify conditional likelihood losses for irregular cells, and admit actual-distortion marks. Two further extensions replace disjoint counting by weighted reconstruction-policy images and replace output cardinality by posterior-mass constraints for secrecy. Direct applications yield matching second-order regions for compatible multi-source helper intersections, one lossy reconstruction with several informed lossless reconstructions, Gaussian common/private descriptions, and privacy amplification from a prescribed conditional type. The Gaussian description result also gives joint rate–distortion offsets. For general helper sections, genuinely lossy informed reconstruction, Gaussian Wyner–Ziv coding, and reconciliation followed by extraction, the same image methods give explicit rate bounds with their remaining losses identified. The resulting rate corrections reflect both the image constraints and the observations available to each terminal.

Learning systems adapt quickly inside a familiar family of models. The harder step comes earlier: deciding, from observations that could be noise, an exception, a change within the family or structure outside it, whether opening a richer family is worth its cost. We treat this as a costly sequential decision: prediction failure must be turned into structural evidence, evidence into a value of expansion, and value into action. The Structural Revision Environment produces matched failures from each source, varies the price of expansion and the remaining horizon independently of the evidence, and admits exact Bayesian calculations and an exact normative solution of the one-shot decision. Its solution shows that revision is a value boundary and not an evidence threshold: one history has different optimal actions under different prices, horizons and announced queries, the boundary between local repair and expansion is set by the inputs a rule predicts and a repair cannot cover, and belief in the richer family crosses long before the decision does. Transformers trained in the environment reproduce this boundary from utility alone. Language models of three post-training lineages carry a failure-sensitive signal in their predictions that is not reflected in their revision decisions, and given the gain of expanding they read it without weighing it against price and horizon. Three models allowed to reason weigh the stated gain in the reference's proportions and still do not turn the history into an estimate of what expansion would buy. Controlled post-training of the meta-trained learners moves the prior and the sharpness of predictions, and neither moves the criterion.

Memory can sustain AI companionship, yet even accurate recollection can be inappropriate to use. Two rounds of formative interviews with 14 users (n = 6 exploratory, n = 8 memory-focused) motivate asking what a companion should consider before using past information. Eight themes inform Reconsider, a single-call procedure with five checks and four handling modes, evaluated on 80 scenarios across five models over 400 blinded within-model pairs. Two LLM judges favored Reconsider by net margins of +15 and +23 percentage points, with bootstrap intervals excluding zero for three of five models but not for GPT or Claude. Evaluator analysis linked judge scoring differences to model family, and a preliminary matched-guidance control isolating memory-specific content gave positive margins. We contribute an interview-grounded design framework for memory use and an evaluation that scrutinizes its own evaluators.

Computer-use agents (CUAs) perform tasks across applications (such as desktops, mobile apps, and web browsers) by observing graphical interfaces and issuing commands such as clicks and keystrokes. These interfaces combine trusted controls and content with untrusted content needed for legitimate tasks. An adversary controlling this untrusted content can embed instructions or misleading visual cues to change the agent's intended action or redirect its commands to the wrong interface target. We formalize security requirements for both the agent's decisions and their execution through GUI commands. In an ideal execution model, we show that enforcing both requirements at each step protects execution traces. We instantiate this model in Secure-CUA, our system for secure CUA execution. Its key idea is to commit to an explicit per-action program, called an $\textit{action transaction}$, before accessing untrusted content. Each transaction fixes its queries to untrusted content and the permitted uses of their responses. The system masks untrusted regions and evaluates each transaction to produce the next action, using an isolated query model to answer its queries. It then locates the intended interface target using the masked interface. Under the model's assumptions, Secure-CUA is secure by design, while generating a new transaction at each step helps maintain high task utility by adapting to changing interfaces. We evaluate Secure-CUA under benign conditions on 400 WebArena tasks using three frontier models across $5$ seeds, yielding $6,000$ execution traces. Secure-CUA achieves an average task success rate of $53.55\%$, compared with $55.12\%$ for Vanilla-CUA and $13.17\%$ for CaMeL-CUA.

Drive-by structural health monitoring infers the condition of a bridge from the response of an instrumented vehicle, whose measurements are dominated by the road roughness and the vehicle dynamics. The bridge responses, contact forces and bridge parameters depend on one another through the vehicle-bridge interaction (VBI), so estimating them jointly is desirable. This paper proposes a novel filter in the Universal Filter family, the Unscented Universal Filter (UUF), for joint input-state-parameter estimation of nonlinear systems with unknown inputs. The unknown input is estimated by weighted least squares from the innovation, without any evolution model or statistics, and the system inversion imposes no rank condition on the feedforward matrix. The nonlinear propagation is linearised by the scaled unscented transform, and the linearisation residuals are propagated into the error covariances. For the acceleration-only configuration, auxiliary integrators form pseudo-measurements from the acceleration measurements themselves. The UUF is applied to the coupled VBI system, with the contact forces as the unknown input and the local stiffness of the bridge as the tracked parameter, from the responses of the vehicle body and at most a few bridge responses. In numerical case studies on a simply supported beam and the Old Ada Bridge, a steel truss bridge in Japan, the UUF is compared with classical and state-of-the-art joint estimation filters, all tuned by the same practical strategy. Across the damage severities, locations, road classes and noise levels examined, the UUF yields the lowest total error in most scenarios, and its error varies little with damage severity, location and road class. On the Old Ada Bridge, it identifies the stiffness of a damaged member from acceleration-only measurements, without the suspension parameters of the vehicle, and with a rank-deficient feedforward matrix.

Contextual biasing supplies an ASR system with a list of expected words at inference time, but existing methods rely on word boundaries that Japanese and Chinese do not provide. We present a boundary-free biasing decoder for frozen public CTC models, built on a character-level Aho-Corasick automaton, with no training and no second pass. Two evidence-based mechanisms replace the boundary: a depth-adaptive gate that sets how hard to push from match depth, and reading-space matching for when the audio is right but the characters are wrong. On Aishell-1 NE's hard R1 subset we reach 66.5% recall, above the trained CLAS baseline (64%), transferring to WenetSpeech and to a second architecture without retuning. We release the first open Japanese contextual-biasing benchmark, where biasing lifts rare-word recall by 25 points at precision above 97%, and still by 19 and 22 points against 1,000-word lists.

Unmanned aerial vehicle (UAV) dispatch is beginning to move beyond isolated path planning and optimization-driven resource allocation toward system-level coordination supported by semantic reasoning and LLM-based interfaces. This survey provides a unified characterization of LLM-enabled UAV dispatch systems that bridges semantic intent, symbolic decision-making, and physical UAV execution. Rather than treating LLMs as standalone add-ons, we conceptualize them as a cross-layer semantic orchestration layer connecting human instructions, external solvers, and distributed control modules. We organize the literature into four representative dispatch paradigms: pipeline dispatch, global assignment dispatch, decentralized agentic dispatch, and divide-and-conquer dispatch. For each paradigm, we analyze its decision logic, system structure, control flow, representative methods, and potential LLM roles. We further examine how LLMs support semantic parsing, retrieval-grounded planning, solver orchestration, local agent reasoning, multi-agent coordination, safety assessment, and human-facing explanation. We discuss the implications of these paradigms for scalability, robustness, coordination burden, and verification requirements, and identify open challenges including latency-aware reasoning, grounding reliability, physical feasibility guarantees, edge deployment, privacy protection, and distributed consistency. This survey provides a system-level taxonomy and design perspective for integrating LLMs into safety-critical UAV dispatch systems.

Quantitative electron microscopy requires accurate knowledge of specimen thickness because dynamical scattering strongly affects image contrast and diffraction intensities. In multislice electron ptychography (MEP), specimen thickness is a required input parameter to its forward model, yet the local thickness of a transmission electron microscope (TEM) lamella is often precisely what is unknown. Moreover, thickness is coupled to probe defocus and the number of slices, making its determination from the reconstruction itself nontrivial. Here, we assess whether joint Bayesian optimization of these parameters can provide a physically meaningful estimate of local thickness from four-dimensional scanning transmission electron microscopy (4D-STEM) data. Applied to a wedge-shaped silicon lamella, the optimization reproduced similar local thickness variation measured by electron energy loss spectroscopy (EELS), with a systematic offset of 10-17% relative to the EELS estimates. Using the optimized reconstruction parameters, depth-resolved MEP further separated the crystalline silicon interior from focused ion beam (FIB)-induced amorphous surface layers within the same reconstructed volume. Regions milled at final voltages of 2, 5, 8, and 30 kV yielded amorphous-layer thicknesses of 3.2, 5.0, 7.0, and 26.4 nm, respectively, increasing monotonically with milling voltage and agreeing reasonably with previous cross-sectional measurements. These results show that applying MEP coupled with Bayesian optimization on a single 4D-STEM dataset can provide both local thickness estimates and depth-resolved characterization of FIB-induced damage, offering a route toward more self-consistent, thickness-aware quantitative 4D-STEM analysis.

Political discourse often uses rhetorical and persuasive language to frame narratives, influence public opinion, and mobilize audiences. While Bangla natural language processing has made progress in sentiment analysis and opinion mining, systematic benchmarking of transformer models for fine-grained rhetorical and persuasion technique detection in Bangla political speech remains largely underexplored. This paper presents a benchmark study of transformer-based models for detecting rhetorical form and persuasive intent in Bangla political discourse. Using BanglaRhet, a manually annotated corpus of 30,289 Bangla political speech segments collected from publicly available political news sources, we formulate two supervised single-label classification tasks: rhetorical technique detection (contrast, repetition, exaggeration, metaphor, rhetorical questions) and persuasion technique detection (blame assignment, call to action, unity call, moral, emotional, and logical appeals). We evaluate four transformer-based models, BanglaBERT, BanglaBERT-Base, SahajBERT, and XLM-RoBERTa-Base, against classical TF-IDF baselines. BanglaBERT achieves the highest performance, with 65.40% macro-F1 for rhetorical technique detection and 66.46% for persuasion technique detection, outperforming the best tuned classical baseline by 19.2 and 13.8 macro-F1 points, respectively. Class-level analysis indicates that errors are mainly associated with semantic overlap among labels, figurative language, and class imbalance. The results provide initial benchmark baselines for Bangla rhetorical and persuasion-aware political discourse analysis and highlight the need for context-aware and multi-label modeling.

Population (Pop) III stars, the first generation of stars in the Universe, have eluded direct detection for decades. The advent of the James Webb Space Telescope, with its unprecedented sensitivity to high-redshift sources, has renewed interest in the prospects for identifying these primordial stars. While previous studies have explored the direct detection of Pop III stars through gravitational lensing, here we investigate an alternative approach: identifying Pop III stars through gamma-ray bursts (GRBs) produced by the deaths of massive Pop III progenitors. We generate source-frame mock observables and find that, although Pop III and Pop I GRBs share many common properties, several observables can potentially distinguish between the two populations. In particular, light-curve timescales, isotropic-equivalent energies, peak energies, and the locations of GRBs in empirical prompt-emission correlations depend sensitively on the mass and explosion energy of the progenitor. We find that Pop III GRBs tend to exhibit longer characteristic timescales and higher isotropic-equivalent peak energies than their Pop I counterparts. Finally, we show that Pop III GRBs may provide a promising origin for the population of soft X-ray prompt emission recently detected by the Einstein Probe, offering a complementary avenue for probing the first generations of stars.

Backdoored vision-language-action (VLA) policies can preserve benign task performance while producing malicious actions when a trigger appears. Detecting such activation is difficult because malicious behavior can comprise individually plausible actions, while unfamiliar tasks introduce legitimate changes in observations and behavior. We introduce TMT, a runtime backdoor detector based on Token Manifold and latent Transition modeling. Trained on benign rollouts, its two branches assess input-token structure and prediction errors in adjacent-layer latent dynamics. A suspicious rollout identified by the token manifold branch, once confirmed through latent deviations, guides transition selection for subsequent monitoring. We further explore policy purification through self-distillation: a frozen copy of the backdoored policy provides benign-input actions to supervise a student on paired benign and triggered observations, without requiring a separate clean reference policy. For evaluation, we adapt traditional backdoor detectors and repurpose anomaly and failure detection methods as VLA backdoor detectors. In a post-hoc comparison with ten baselines, TMT achieves state-of-the-art backdoor detection performance on unseen tasks across three VLA backdoor attacks. Our project page is available at https://zzr42.github.io/tmt/.

Computational design of polycrystalline microstructures for enhanced mechanical performance requires repeated high-fidelity simulations over stochastic morphologies, rendering conventional crystal plasticity finite element (CPFE) approaches prohibitively expensive for optimization. Here, we develop a deep learning-driven framework for optimizing polycrystalline microstructures under competing quasi-static and dynamic performance requirements. A 3D U-Net surrogate maps polycrystalline copper microstructures directly to full-field velocity histories from plate-impact simulations, preserving the spatial and temporal resolution needed to evaluate dynamic performance and interrogate underlying wave interactions and failure mechanisms. The surrogate is coupled with stochastic microstructure generation and derivative-free optimization to efficiently navigate the design space. To monitor surrogate reliability during optimization, we introduce a distribution-consistent reliability region (DCRR) based on maximum mean discrepancy (MMD) to quantify distributional shift from the training data. The framework is demonstrated for multiple design problems, including constrained optimization of spall and yield strengths and inverse design. The optimal designs are validated using CPFE simulations with less than 3% prediction error across all optimization cases, while the surrogate-driven optimization reduces cumulative evaluation time by approximately four orders of magnitude compared with direct CPFE-based optimization. Although demonstrated here for the design of polycrystalline microstructures targeting spall and yield strengths, the framework is broadly applicable to optimization problems where high-fidelity full-field simulations are prohibitively expensive but spatially and temporally resolved response information is essential for evaluating, interpreting, and optimizing candidate designs.

As Large Language Models (LLMs) are increasingly adopted for automated grading and feedback in higher education, their structured outputs, including multi-dimensional rubric scores, detailed feedback comments, and improvement suggestions, create an appearance of thorough analytic evaluation. This study examines whether these outputs deliver what they appear to offer. Using the JorGPT dataset of 3,041 student responses to 50 open-ended computer science questions, scored by both human instructors and three commercial LLMs, we identify three systematic discrepancies between the apparent and actual quality of LLM-generated grading and feedback. The sub-dimension scores are highly correlated (r = 0.82-0.99, VIF up to 45), providing redundant rather than independent diagnostic information. The textual feedback rarely detects student misconceptions (5-7% vs. 15-31% for teachers), functioning as a coverage checklist rather than a diagnostic instrument. The feedback tone remains uniformly positive regardless of response quality, lacking the severity modulation observed in human feedback. Additionally, grading accuracy varies significantly by knowledge domain, with procedural topics most reliable. These findings provide empirically grounded guidance on which aspects of LLM-generated grading and feedback can be relied upon and which require continued human oversight.

This paper addresses the prediction of adjacency between pairs of 2D fragments based on their contours. We improved the two-stage architecture proposed in Beaulac's thesis, in which a rotation-equivariant Siamese convolutional neural network scores pairs of local image windows along the two contours of two fragments. The scores are gathered in an adjacency matrix in which a ResNet detects the partial anti-diagonal band that reveals the adjacency of two fragments. In the current work, we keep the pipeline and replace the local score by a comparison of tangent-angle profiles of contour windows, making it, by construction, invariant to fragment rotation and agnostic to the selected contour-starting point. These adaptations may be either a training-free likelihood ratio or a small one-dimensional convolutional model trained on corresponding points. We also replaced the final classifier by a band U-Net that segments the band and classifies the pair, so that the shared arc is obtained along with the decision. In the synthetic data set of the original thesis, the tangent descriptor performs as well as or better than the image-window approach in all tested configurations. The proposed pipeline reaches an accuracy of 98%, vs 93% to 95% for the original approach once its evaluation is corrected. We tested our pipeline, with models trained only on synthetic data, on the PairingNet benchmark, and obtained an AUC of 0.93. Furthermore, under the PairingNet pair-searching protocol conditions, our learned descriptor obtains a Recall@10 of 0.82 on the real set against 0.56 from the best model of the original paper.

When an LLM answers a state-specific policy question wrongly, it may be hallucinating, or it may be returning a real value that holds in another state. We test this with a minimal-set design: the question wording is fixed and only the jurisdiction varies, across the 50 U.S. states and the District of Columbia (51 jurisdictions) and three exactly defined Medicaid income-eligibility quantities. Gold values come from an official data book and agree with an independent source in 101 of 102 checked cells. Under a pre-registered protocol, Claude Sonnet 5.5 and GPT-5.6 Sol reproducibly give another state's current value, identical across two independent repeats, for 10 and 25 of 153 items. The count depends on the reference set: crediting any wrong answer that equals another state's value yields 3-5x more reproducible substitutions than checking every number in the asked state's own records, because many apparent cross-state answers are the asked state's own values under another convention or from an earlier year. Claims about cross-jurisdiction error need a complete same-state reference set. We will release the protocol, gold table, and all model outputs.

Methods that recover individual latent variables of the world, from nonlinear ICA to dictionary learning and causal representation learning, anchor the latents to observations through reconstruction, auxiliary supervision, or distributional asymmetries such as non-Gaussianity. Methods without these anchors, including joint-embedding predictive architectures (JEPAs), identify the latent state only up to a linear transformation, so individual latents remain mixed. We close this gap: individual world latents can be provably recovered with no reconstruction, no decoder, and no labels. The key condition is Structural Diversity: different latents leave distinct dependency footprints on observations, just as no two snowflakes are alike. Building on the linear identifiability that LeJEPA provides, we prove that under Structural Diversity, DSReg (Dependency-Sparsity Regularization) recovers individual world latents up to signed permutation, without reconstruction or a decoder. It applies post hoc to any linearly identified representation, reusing trained checkpoints at no loss over joint training, and establishes the first fully identifiable JEPA that recovers every world latent. Moreover, as a condition on dependency footprints, Structural Diversity is strictly weaker than all structural conditions of prior identifiable latent variable models. Across synthetic regimes, world model probes, learned visual encoders, and external renderers, DSReg preserves dense prediction while improving individual-latent recovery and downstream use with scales.

Reconstructing fine-resolution column-averaged dry-air CO2 (XCO2) fields from sparse satellite observations requires models to infer spatial structure that is only weakly constrained by direct measurements. Existing learning-based methods typically treat environmental covariates as ordinary numerical inputs and must therefore learn heterogeneous source-sink relationships largely from sparse supervision. We introduce GeoPrior-Mamba, a multi-directional Mamba framework augmented with offline language-model-induced structured process priors. Rather than using a language model to predict XCO2, we use it before training to organize relative process knowledge for biospheric uptake, ecosystem respiration, and anthropogenic emissions into deterministic prior tables. These priors are spatially instantiated using geographic, ecological, emission-related, and seasonal information and are adaptively injected into the reconstruction backbone through a lightweight knowledge adapter. Using OCO-2 observations from 2018-2020, GeoPrior-Mamba achieves an RMSE of 0.81 ppm and an R2 of 0.93 on held-out observations, reducing RMSE by 48.2% relative to CAMS background interpolation and by 3.1% relative to Trans-XCO2 under the same evaluation protocol. Ablation experiments show a measurable contribution from the knowledge-prior branch and substantially faster convergence than the knowledge-free Mamba backbone. Independent TCCON evaluation further supports the consistency of the reconstructed fields with ground-based column CO2 measurements. These results suggest that language models can provide a practical mechanism for constructing structured process priors when globally consistent process-response representations are difficult to obtain directly, while remaining outside the numerical prediction loop.

Recently, approaches that leverage human video datasets for robot policy training have become increasingly prevalent. However, most existing hand trackers regress pose from cropped frames with limited priors on hand motion and object interaction, resulting in inaccurate and physically inconsistent estimates. Moreover, the lack of physical cues, e.g., contact and force, limits the use of human videos for robot policy training. To this end, we propose RLHND, a video foundation model-based hand tracking model that jointly estimates hand pose and realistic tactile information from monocular egocentric videos. RLHND turns the pre-trained Cosmos 3 video diffusion backbone into a deterministic clip-level feature extractor via clean-latent conditioning, carrying its learned priors on hand motion and hand-object interaction into tracking. For pose estimation, RLHND (i) predicts hand poses with anatomically plausible joint angles and (ii) enables optional conditioning on the shape parameter to maintain consistent hand shape within the same video and even across videos recorded by the same actor. For tactile estimation, a separate tactile expert stream, trained with the pose stream frozen, predicts dense contact and force over the hand surface. We further adopt LBS-based feature spreading to enable vertex-wise feature extraction without costly per-vertex attention. RLHND achieves state-of-the-art performance across various benchmark datasets for pose estimation, while also achieving state-of-the-art performance in contact and force estimation. Moreover, we demonstrate the utility of RLHND for robot learning through retargeting results and real-world robot experiments. The code will be publicly available at https://seungjun-moon.github.io/rlhnd/.

Robotic manipulation videos are increasingly used as visual plans for embodied agents, but optimizing purely for visual plausibility often fails to capture the fragile physical manifold of real-world interactions. Even minor physics-violating errors at the interaction boundary, such as interpenetration or premature object motion, can completely invalidate the inferred timing and pose needed for downstream execution. Because standard supervised fine-tuning lacks the direct pressure to penalize these localized failures, we introduce AgiBot-PhysPref. This rigorously curated 10,000-sample preference dataset isolates condition-matched physics violations, turning the generator's own failure distribution into a foundational signal for physical consistency. Building upon this, we propose RobotAPO, an adversarial physics preference optimization framework operating in the continuous flow-matching denoising space. To prevent the policy from merely memorizing static curated failures, RobotAPO employs a lightweight adversarial counterfactual proposer that learns a condition-dependent, physical-failure-biased direction in denoising space. This encourages the model to explore and better respect the physical interaction boundary, all while maintaining a pure prompt-and-reference inference interface without requiring external structural conditioning. Comprehensive evaluations demonstrate that explicitly correcting these localized physics violations improves downstream robot execution from generated videos. On held-out AgiBot conditions, RobotAPO outperforms the strongest controlled internal baseline in physical consistency by 6.8% hard score and 10.0% soft score. Crucially, in real-robot replay, it translates these physical-consistency gains into a 37.4% relative improvement in task success over the strongest controlled internal baseline.

Mott gap closing induced by an external potential gradient is studied by considering the Hubbard model on a square lattice. Our model consists of hole- and electron-doped domains forming the p-n junction where the on-site potential varies smoothly across the interface. The effect of the potential drop direction, either along square edge or diagonal, is studied by describing the correlation effects of inhomogeneous 2x2 plaquette impurity clusters within a dynamical mean-field theory framework. It is found that the metallic phase is favored over the Mott insulators when the potential drop becomes comparable to the energy eigenvalue difference of an open 2x2 plaquette between the ground state and the first excited state. Given the same associated electric-field strength, the excitation energy is almost independent of the drop direction, which amounts ~0.35t at U=5.8t with the hopping parameter t. Upon insulator-to-metal transition, the correlated metal exhibits a coherent peak at zero energy whose intensity is greater for the drop direction along the square diagonal than along the edge. The main origin is attributed to the preserved degeneracy of two molecular orbitals for the former case, while it undergoes splitting in the energy levels for the latter. Our result is discussed regarding the experimentally measured Mott gap closing thresholds.

Conditional density estimation describes how the entire distribution of a response changes with covariates, and in imaging studies also with location. Logistic Gaussian processes (LGP) give a flexible prior for such densities. However, the normalizing constant of an LGP has no closed form, so existing methods approximate or replace the likelihood. We propose the exact likelihood finite-rank LGP (ExFR-LGP), which writes the log density as the sum of two bivariate functions, one of the response and a location-varying linear index of the covariates, and one of the response and the location. Each function is assigned a finite-rank Gaussian process prior that is piecewise linear on a regular grid, enabling the normalizing constant, the conditional mean and the quantiles to enjoy closed form representations. Posterior samples are drawn from the exact posterior under this prior via a Gibbs sampler that updates the Gaussian components by elliptical slice sampling. When the true log density is the sum of two such bivariate functions, we show that the posterior contracts at the minimax rate of an $\alpha$-smooth bivariate density up to a logarithmic factor. Numerical experiments using synthetic data demonstrate the effectiveness of ExFR-LGP in conditional density estimation. Furthermore, application to fractional anisotropy responses along the corpus callosum in the Alzheimer's Disease Neuroimaging Initiative data yields covariate-adjusted percentile bands with uncertainty, elucidating how diagnosis changes the distribution along the tract.

Vision-Language-Action (VLA) models are effective at understanding what task to perform, but provide limited control over how it should be executed, such as moving quickly or slowly. We introduce TempoBridge, a lightweight framework that uses frozen VLA representations to modulate actions according to tempo cues in the instruction at each task phase, without additional tempo-conditioned robot demonstrations or tempo-specific base-policy fine-tuning. TempoBridge extracts tempo cues from contextual VLM representations, aligns them with task progress through a causal phase router, and modulates nominal motion commands during execution. Across LIBERO tasks, TempoBridge improves Tempo Success Rate from 52.6% to 89.7% under canonical tempo instructions while retaining high task success. It also preserves near-baseline performance when no tempo cue is present and generalizes to unseen tempo expressions without additional training. Experiments on a physical robot further demonstrate language-conditioned tempo modulation in real-world manipulation.

Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained detail matters. We test this claim along both routes to a pixel-space backbone. We pretrain Iris-3B, a 3B-parameter pixel-space text-to-image transformer, from scratch through a $256\to512\to1024$ curriculum, after first ablating the prediction target and representation alignment at $256^2$ to decide what to scale. We also convert a pretrained latent model, FLUX.2 Klein base 4B, to pixel space. We fine-tune both families for monocular depth estimation and for image restoration/super-resolution. We find no significant improvement from using a pixel-space generative prior. Fine-tuned for depth with one matched direct-regression recipe, Iris-3B is level with the latent FLUX.2 Klein and the converted pixel FLUX.2 Klein falls behind it, and on $4\times$ DIV2K restoration neither pixel model beats a latent FLUX.2 Klein fine-tune, the converted one trailing it slightly. We document the recipes, the failure modes and the remaining confounds behind this negative result. Nevertheless, Iris-3B shows that pixel-space pretraining with the pixel-transformer (PiT) head of PixelDiT scales to 3B parameters and to text-to-image quality competitive with latent models, matching Qwen-Image on OneIG under the official evaluators at $1024^2$. We release its weights and training code in the hope that they help pave the way for further work on pixel-space generation.

Zero-shot Chinese character recognition (ZS-CCR) aims to recognize characters whose categories are never observed during training, and typically relies on the compositional structure shared between seen and unseen characters. Recent CLIP-style methods represent this structure with the Ideographic Description Sequence (IDS) and align it with glyph images in a shared embedding space. However, they rely on a single global image–IDS similarity that discards the spatial layout of radicals and, being learned only implicitly from seen classes, generalizes poorly to unseen ones; moreover, global matching often retrieves the correct character within the top candidates yet fails to rank it first when characters differ only in subtle local radicals. To address these issues, we propose a global-to-local two-stage framework. In the first stage, STG-CLIP augments the IDS with explicit tree-position and radical-level geometric priors, yielding a spatial-aware prototype that provides a consistent spatial description across seen and unseen categories for high-recall global retrieval. In the second stage, the Radical Verification Module (RVM) uses the radical instances of each retrieved candidate as queries to verify whether the corresponding radicals can be matched to spatially compatible regions in the input glyph. A margin-based gating rule activates the RVM only when the leading global candidates receive similar similarity scores. Experiments on the ICDAR2013 benchmark demonstrate that our method achieves state-of-the-art performance under the character-level zero-shot setting, obtaining 83.06% top-1 accuracy with 2,755 seen classes. Ablation studies further show that the explicit geometric priors and radical-level verification provide complementary improvements.

Unlike autoregressive models, discrete diffusion-based models for zero-shot text-to-speech generate speech tokens in parallel and can revisit earlier predictions. Mask-and-replace training extends mask-only training by randomly replacing some tokens, and its gains are commonly attributed to self-correction, the ability to revise previously generated tokens. However, exposure to randomly perturbed context during training may itself improve generation, raising the question of whether these gains require inference-time token revision. To investigate this question, we use DeMaR, which combines mask-and-replace training with confidence-ranked mask-only sampling while preserving the total training corruption probability. Trained from scratch on LibriTTS, DeMaR achieves lower word error rates (WER) than autoregressive and mask-only diffusion baselines using the same speech tokenizer. This advantage persists when each token remains unchanged after first being unmasked. Matched training conditions on two heterogeneous speech tokenizers show that both noisy-context augmentation and replacement supervision improve WER under this restriction. These findings demonstrate training-side benefits of replacement beyond enabling inference-time token revision.

The selection of benchmark workloads is of paramount importance in computer architecture, as it establishes the yardstick against which architectural innovations are measured and guided. Yet for decades, the SPEC benchmark suites, comprising merely tens of workloads, have been the de facto standard in academic architectural research, where they are frequently treated as a principal evaluation and optimization target. When a suite this small is relied upon so heavily, it risks architectural overfitting; as our research and prior studies demonstrate, an overly narrow focus can mislead design decisions by overvaluing certain innovations, producing cores that excel on SPEC benchmarks yet underperform on broader, realistic workloads. To mitigate this overfitting, adopting a large, comprehensive benchmark suite is the natural solution. However, the immense engineering effort required to strip software into the clean, interference-free binary executables demanded by simulators often makes this highly impractical. In this work, we demonstrate that AI agents provide an elegant solution to this challenge. Rather than manually curating yet another static benchmark suite, we introduce an agent-driven workflow capable of autonomously transforming arbitrary open-source repositories into simulator-ready executables. This automated approach makes workload collection highly scalable, allowing us to rapidly harvest hundreds of diverse applications from public repositories into our benchmark suite. Through a comparative analysis of our agent-generated suite against SPEC, we show that it not only achieves higher-fidelity performance assessments but also uncovers novel architectural insights that traditional, static suites fail to expose.

Large language models (LLMs) should abstain from scientific multiple-choice questions when no option is valid, but frequent abstention alone does not demonstrate sensitivity to answer availability. We introduce ArcticQA, a dataset of 194 questions derived from primary Arctic research, with automated checks of answer support and distractor contradiction against source evidence. We further develop ArcticAbstain, a paired benchmark comparing answer-present and answer-absent conditions, with the correct answer replaced by a distractor in the latter and an explicit abstention option in both. We evaluate eight models from the Gemini, Claude, and ChatGPT families at high reasoning effort, with three trials per condition, yielding 9,312 recorded responses. Answer-present abstention rates range from 0.0% to 63.0%, whereas replacing the correct answer increases abstention by 5.05 percentage points on average. These findings highlight substantial baseline differences and the need to evaluate abstention frequency and responsiveness jointly. The dataset and benchmark are available at https://github.com/BenWilcox8/arctic-qa.