Papers

Evidence for excesses observed at LHC tend to cluster mostly around 95 GeV and 650 GeV. We claim that a consequent interpretation of the 650 GeV and some additional indications calls for a Randall Sundrum sequence of Kaluza Klein graviton resonances. A common prejudice claims that such resonances should be very heavy, out of reach for LHC. We argue that this is no so. Although present LHC searches seem to exclude these particles below 2 TeV, we explain why this 650 GeV resonance could be one of them. We also explain why ongoing analyses ignore this scenario and apply selections which are only valid for scalar resonances. This could also apply to ongoing searches for Higgs pairs where the search for resonances decaying into SM Higgs pairs ignores the KK graviton scenario. We expect that with a correct treatment, one could very significantly increase the detection efficiency for ZZ and hh channels and achieve, with present data, a convincing proof of our expectations. Channels like h(95)h(125) and h(95)h(95) should be added to present selections. Such improvements should allow to reach a discovery. For the ZZ and WW channels, one can use semi-leptonic and purely hadronic final states, to confirm the tower of KK resonances indicated up to 3 TeV.

How to efficiently transmit large volumes of Extended Reality (XR) content through current networks has been a major bottleneck in realizing the Metaverse. The recently emerging Generative Artificial Intelligence (GAI) has already revolutionized various technological fields and provides promising solutions to this challenge. In this article, we first demonstrate current networks' bottlenecks for supporting XR content transmission in the Metaverse. Then, we explore the potential approaches and challenges of utilizing GAI to overcome these bottlenecks. To address these challenges, we propose a GAI-based XR content transmission framework which leverages a cloud-edge collaboration architecture. The cloud servers are responsible for storing and rendering the original XR content, while edge servers utilize GAI models to generate essential parts of XR content (e.g., subsequent frames, selected objects, etc.) when network resources are insufficient to transmit them. A Deep Reinforcement Learning (DRL)-based decision module is proposed to solve the decision-making problems. Our case study demonstrates that the proposed GAI-based transmission framework achieves a 2.8-fold increase in normal frame ratio (percentage of frames that meet the quality and latency requirements for XR content transmission) over baseline approaches, underscoring the potential of GAI models to facilitate XR content transmission in the Metaverse.

The concept of counterfactual explanations (CEs) has recently gained increasing attention for deriving explanations of mathematical optimization problems. For a given problem, a counterfactual is an ideally minimal change in the mutable problem parameters such that an optimal solution of the changed problem exists fulfilling some desired criterion. It was shown that so-called weak CEs can be calculated by solving a computationally demanding bilevel optimization problem, which is $\Sigma_2^p$-complete if decision variables and mutable parameters are restricted to integers. Hence, the known exact solution methods are expected to fail on practically relevant problem sizes. In this work we derive a fast heuristic algorithm by approximating the feasible region of the CE problem with a decision tree for which CEs can be calculated more quickly. To this end we analyze the mathematical structure of the feasible region and develop a hand-crafted decision tree which applies splits imitating this structure. Additionally, we develop several techniques to increase the performance of our method. The practicability of our methods is supported by computational experiments on the knapsack problem and the facility location problem.

We investigate gravitational-wave echoes in a horizonless compact object model, a non-perturbative solution of the semi-classical Einstein equation that reproduces the entropy-area law from the bulk structure. By employing a model of anisotropic fluids in classical general relativity (GR), we show that odd-parity gravitational waves are not reflected at the object's surface but are trapped inside due to the exponentially strong redshift, and that the echo-delay time extremely exceeds the age of the Universe. Consequently, the echo signals are effectively unobservable, which is consistent with current null results from echo searches. We finally discuss how this quantum black hole and the conventional classical black hole could be distinguished observationally.

Many works approach continual learning through the lens of infinite context windows. As an agent puts more observation into context (concretely the KV cache), compacting said context is akin to direct memory manipulation, without affecting the base model's weights. Many works pose KV compaction as an optimization problem: learn a smaller set of KV vectors that matches the behavior of the full KV cache. While this preserves base model behavior, optimizing through a frozen base model results in a highly nontrivial optimization problem with a brittle and flat loss landscape. In this paper, we characterize what makes these optimization problems difficult and demonstrate that a heavily simplified Perceiver-based architecture not only matches performance of a full Perceiver transformer in continuous context compaction, but outperforms baselines on compaction utility. Results are presented on MCQ tasks across Finance, Legal, Gutenberg, and Code.

It is an unfortunate fact that even over commutative noetherian rings, localization of K-injective complexes need not be K-injective. However, not all is lost. Given a complex of injective modules $J$ over a commutative noetherian ring $A$, we show that K-injectivity of $J$ can be detected locally: if $(A,\mathfrak{m})$ is a noetherian local ring, then $J$ is K-injective over $A$ if and only if $\operatorname{Hom}_A(\widehat{A},J)$ is K-injective over $\widehat{A}$, the $\mathfrak{m}$-adic completion of $A$; and for an arbitrary commutative noetherian ring $A$, $J$ is K-injective over $A$ if and only if $\operatorname{Hom}_A(A_{\mathfrak{m}},J)$ is K-injective over $A_{\mathfrak{m}}$ for every maximal ideal $\mathfrak{m}$ in the singular locus of $A$. The local result is deduced from a more general theorem on faithfully flat descent of K-injectivity along maps to (possibly noncommutative, possibly non-noetherian) rings. The global result is established via the local-to-global principle in tensor-triangulated geometry. As an application, we generalize a theorem of Grothendieck on faithfully flat descent of regularity, extending it to faithfully flat extensions by noncommutative left coherent left regular rings.

The minimum solid angle of a simplex is shown to be a shape measure in any dimension. While this has been known in dimensions two and three, the higher-dimensional generalization is new. We prove inequalities between the minimum solid angle of a non-degenerate simplex and its fullness. Up to dimension-dependent constants, the fullness is bounded from above by the minimum solid angle, and it is bounded from below by the square of the minimum solid angle. The exponents are optimal for three and higher dimensions. This work therefore generalizes the minimum angle condition. We also review several simplex shape measures and their relationships.

Learning from human demonstrations is a reliable way to teach robots new tasks, but the gains from each additional demonstration shrink as the policy improves. Continued improvement can instead come from supervised deployment, where an operator places the objects and intervenes when the policy fails. We ask how to maximize improvement from a fixed budget of supervised episodes on high-precision manipulation tasks with wide ranges of object placements. We observe that failures can concentrate in a small subset of initial states, so uniform collection spends much of the operator's time on states the policy already handles. Mulligan makes the initial-state distribution a decision, starting each round's episodes at observed failures and untried states. To further improve data efficiency, we augment interactive imitation learning with a value function trained on all data, including failures that imitation discards. Across three real-world tasks evaluated on 2,550 held-out, blinded episodes and two simulated tasks, Mulligan outperforms uniform initial-state sampling at matched collection budgets, and combined with value-based action selection, HiL-IDQL+Mulligan, improves final real-task success by 10-34 percentage points. With operator interventions, the human-robot team completes 98% of collection episodes, remaining productive while the policy learns. Videos, code, and data are available at https://mulligan.page/.

Graphene nanoscrolls form a unique radial superlattice whose geometry enables transport behaviors distinct from their flat counterparts. We theoretically demonstrate that rolling a graphene nanoribbon into a few-turn nanoscroll (1.0-1.2 turns) triggers a giant enhancement in longitudinal conductance by an order of magnitude (nearly twentyfold). We show that this effect stems from interlayer hopping between open boundaries, which not only transforms localized edge states into one-dimensional-like conducting channels and also induces a significant Fermi energy shift. Our results reveal an inverse scaling of conductance with turn number, establishing GNSs as a platform for geometry-driven quantum transport and offering a tunable approach to engineering nanoarchitectural electronic states.

Resting-state fMRI is important in the classification of brain disorders, but highly multimodal and exhibits strong multisite heterogeneity. Existing methods fuse images, BOLD-based functional connectivity, and phenotypic data modalities. Unlike other medical imaging datasets, rs-fMRI datasets rarely include a text modality, so they are generated from phenotypic data or BOLD activations. These text generation methods rely on fixed assumptions for subjects, sites, devices, and protocols, leading to poor generalization across datasets. We propose fMRI-TAMCL, a text-anchored multimodal contrastive learning framework that integrates fMRI images, sparse FC, and generated subject-specific text. Its Subject-Adaptive Threshold Derivation module generates BOLD activation text, while Feature-Value Serialization module generates phenotypic text. All three modalities are encoded as clustered graphs, projected onto a shared unit hypersphere space, aligned using pairwise, text-anchored supervised contrastive learning, and fused with attention. fMRI-TAMCL proves its generalization capability across five datasets outperforming 29 baselines with 78.6%-86.4% accuracy in downstream classification.

We present the Prior-Fitted Language Model (PFLM), a 300M-parameter byte-level transformer pretrained only on samples from a synthetic non-linguistic prior. Given a prefix of real text, it learns to predict the language in context with frozen weights, having never seen a word of any real language. Every training sequence is generated by a recurrent structural causal model drawn fresh from a distribution over such models. The model never sees the same language twice during training, so the only way to predict the continuation is to infer the language from the prefix. Samples from this prior share the statistical signatures of natural text: Zipfian frequencies, slow entropy-rate convergence, and long-range dependence. On Wikipedia in six languages, bits per byte fall from the uniform eight to between 0.9 and 2.4 at one million bytes of context. Given numerals instead of text, PFLM learns to count, to compare magnitudes, and to add approximately. It predicts deterministic sequences like Rudin-Shapiro or the prime indicator, and it compresses six non-text domains, from source code to speech, below gzip and PPMd. The model has not learned a language. It has learned to learn one.

Benchmarks expose vision-language-action (VLA) policies to few canonical instructions, while exhaustive deployment testing is impossible. We introduce Object-Grounded Attention Monitoring (OGAM), connecting systematic testing to runtime assurance: testing reveals attention divergence between successful and failed executions, and OGAM uses this signal to stop failures beyond the finite suite. We generate scene-grounded instructions through pairwise combinations of action templates and objects, and separately test meaning-preserving paraphrases. All 87 out-of-benchmark cases reveal problematic behavior across OpenVLA, OpenVLA-OFT, UniVLA, and $\pi_{0.5}$: none completes any of the 24 feasible instructions, while infeasible or hazardous requests also trigger behavior substitution. At each action query, we project gradient-weighted visual attention through object masks and group it by instruction role for comparison across tasks and policies. Dynamic time warping aligns this course with a successful reference despite speed differences; conformal calibration on successful episodes sets the early-stopping threshold for sustained deviations, with a nominal false-stop target of $\alpha=0.05$. Across four policies, OGAM stops 87-100% of failed episodes at median times of 5-12s within a 20s budget, with observed false-stop rates of 3-5%, without failure-labeled training. Finite testing thus identifies attention patterns that support online intervention before failure fully unfolds.

Exceptional points (EPs) are conventionally identified through eigenvector coalescence, inaccessible in classical platforms lacking a Hamiltonian spectrum. We introduce classical observable coalescence as a diagnostic identifying EPs directly from physical configurations, and demonstrate it in a known non-Hermitian Ising chain with coupling $J$ and staggered imaginary field $\gamma$, chosen for verifiability against established quantum results. A mean-field reduction of the Heisenberg dynamics yields equilibrium branches organising into two symmetry-related families whose fusion at the boundary of the EP supersurface $|\gamma|=|J|$ exhibits the characteristic $\sqrt{J^2-\gamma^2}$ branch-point structure and two-sheeted Riemann topology. Crucially, the degeneracy at $\gamma=0,\eta =0 $ is shown to be a diabolic point, producing no observable coalescence, establishing the diagnostic as discriminating rather than merely suggestive. An exact Jordan-Wigner and Bogoliubov-de Gennes treatment confirms eigenvector coalescence at the same threshold, and a coupled gain-loss circuit realisation is proposed, establishing classical observables as experimentally accessible signatures of non-Hermitian criticality.

Deploying large antenna arrays transitions wireless communication into the radiative near-field (NF), where spherical wavefronts enable beamfocusing in the joint angle-distance domain. However, wideband beamforming with large antenna arrays suffers from beam split across both spatial dimensions. Although fully digital arrays can eliminate this impairment, their prohibitive hardware cost and power consumption motivate the adoption of hybrid array architectures equipped with a limited number of radio-frequency (RF) chains. This paper investigates the minimum number of RF chains required to mitigate the beam split. We derive closed-form expressions for the required number of RF chains in fully-connected (FC) architectures and propose an algorithm for sub-connected (SC) architectures, considering both uniform linear arrays (ULAs) and uniform circular arrays (UCAs). Our analysis reveals that beyond a focal distance of 10L, where L denotes the aperture length, the NF RF chain count converges to its far-field counterpart and becomes a function of the angle alone. Moreover, the FC hybrid architecture mitigates the beam split by using fewer RF chains than the SC architecture, albeit at the cost of increased complexity in the phase shifter network.

Monolithic high-contrast gratings (MHCGs) can exhibit scatter-free perfect reflection despite open higher-order substrate diffraction channels, but the underlying mechanism has not been fully clarified. Using a Fourier-harmonic description, we show that all open higher-order channels in the high-index substrate can be simultaneously extinguished through destructive interference between the radiation fields generated by the direct and indirect source contributions. By contrast, the zeroth-order transmission vanishes when the modulation-induced zeroth-order field interferes destructively with the background transmitted field of the incident wave. Together, these cancellations yield unit zeroth-order reflectance in lossless MHCGs. The near invariance of the internal field harmonics accounts for the remarkable stability of these conditions against changes in the substrate refractive index. We further show that the number of symmetry-compatible open channels determines the generic codimension of the perfect-reflection solution set. Depending on the number of independent design parameters relative to this codimension, the solutions form continuous curves, appear as isolated points, or are generically absent. Introducing an additional geometrical parameter restores isolated perfect-reflection solutions when an additional diffraction channel opens. The same cancellation mechanisms and channel counting apply to both TE and TM polarizations.

Scalable Graph Transformers are commonly trained and evaluated on static large graphs in a transductive setup. Many scalable Graph Transformer components can be formulated as a constant-size shared memory, similar to virtual nodes, providing compressed information about the whole graph. The counterpart of these models in language models and other domains is justified as the input changes, and this mechanism learns to compress some useful information about the input. In transductive learning on a single fixed graph, however, any shared memory can be seen as a constant at test time. This raises the question of what exactly this shared memory does in this static setup. We give preliminary evidence that optimizing a shared memory directly performs similarly to global communication methods, and so normal local message-passing models can embed similar information in their weights. Thus, these settings may be a poor fit for evaluating global communication in graph neural networks.

On-chain transfer volume is often used to describe activity in tokenized real-world assets (RWAs), although recorded transfers may reflect very different economic functions, including token creation, destruction, issuer-related operations, and movements between other addresses. Ownership measures face a similar problem because a large address balance may belong to a pool or intermediary that represents multiple investors. We address these two measurement problems by decomposing transfer activity and making ownership coverage explicit. Transfer turnover is calculated before and after observable event categories are separated, using the same asset, observation window, and supply denominator. For ownership, we compare concentration measured against total supply with concentration among the addresses that remain after identified intermediaries are excluded. Using Ethereum-based tokenized funds from December 2025 to August 2026, we find that transfer classification can materially change measured turnover. We also show that removing an intermediary can lower concentration when measured against total supply while concentration among the remaining addresses increases. Transfer activity and address concentration need to be interpreted with their underlying transaction functions and ownership coverage in mind. These measures provide evidence about liquidity-related activity, but they do not directly measure market liquidity. The code and data used in the analysis are available at https://github.com/atytmr07/rwa-liquidity.

A long-standing goal of AI is a model that can continually learn and improve itself. On post-trained models, supervised finetuning (SFT) on new data often causes poor generalization and catastrophic forgetting. As such, the conventional wisdom is that on-policy training is a prerequisite for continual learning. In practice, however, data containing new knowledge or capabilities are often off-policy. While methods such as on-policy self-distillation (OPSD) try to bridge this gap by converting off-policy data into on-policy signal, they have been shown to cause reasoning collapse. In this paper, we show that off-policy merging beats OPSD for continual learning. We first show that SFT learns a useful signal from new data, but naively applying its update interferes with existing capabilities. We reduce this interference with a simple recipe we term grafting, which changes where the update is learned and how it is applied: (1) learning the update on an earlier donor checkpoint, ideally even before the end of pretraining, and applying the weight update to the post-trained model; (2) scaling the weight update, equivalent to a form of model merging; and (3) optionally, masking the most sensitive update directions when the new data distribution is far from the post-trained model. Across continual learning settings including (1) distilling from expert traces, (2) self-improvement with STaR and Pedagogical RL, and (3) injecting knowledge after pretraining cutoff, grafting Pareto-dominates both SFT and OPSD in new-task and old-task performance, while avoiding expensive on-policy sampling. Therefore, our work challenges on-policy training as a necessity for continual learning on RL-trained models.

The rotation curves of the 175 disk galaxies in the SPARC database are confronted, one by one and under identical conditions, with the two mass models that have dominated the missing-mass debate for four decades: a cold dark-matter halo of Navarro-Frenk-White (NFW) form, and Milgromian dynamics (MOND) with the acceleration constant held at $a_0=1.2\times10^{-10}$ m s$^{-2}$. Stellar mass-to-light ratios, distances and inclinations are treated as nuisance parameters with priors set by the 3.6 $\mu$m photometry and by the SPARC error budget, and every galaxy is sampled with a Markov-chain Monte Carlo. Three findings emerge. (i) Judged by the Bayesian information criterion, the free NFW halo is preferred in 96 galaxies and MOND in 79; once the halo is required to obey the $\Lambda$CDM concentration-mass relation the score becomes 86 to 89, i.e. a draw. The two theories fail in different places: MOND loses in massive, high-surface-brightness, bulge-dominated systems, NFW loses in slowly rising dwarf rotation curves. (ii) The NFW concentrations demanded by the data sit a median 0.24 dex below the $\Lambda$CDM expectation, and 53% of the sample lies more than $2\sigma$ below it, independent of halo mass; this is the core-cusp problem expressed as a scaling-relation offset rather than as a handful of anecdotes. (iii) The radial acceleration relation built from the full sample has $a_0=(1.13\pm0.02)\times10^{-10}$ m s$^{-2}$ and an observed scatter of 0.14 dex, yet the value of $a_0$ preferred by individual galaxies spans more than a decade and correlates with surface brightness. The two statements are not in contradiction: the population-level relation is tight because it averages over exactly the degeneracies with mass-to-light ratio and distance that make $a_0$ ill-determined galaxy by galaxy. The full table of fits is provided.

Generative models can synthesize high-quality inauthentic multimedia content that is already being misused at scale. We evaluate twenty deepfake detectors against ten generators released in the last four years and find accuracy decreasing over time, from near-perfect 99.5% to 76%. Adversarial perturbations further reduce every baseline detector to below 2% accuracy, effectively inverting the detector's assigned label. We argue that this unreliability reflects a fundamental ambiguity: generators can reproduce authentic content exactly (e.g., through memorization), so content alone cannot reveal the true provenance label. For this reason, content produced by a generator must admit a faithful reconstruction by that same generator, and finding such a reconstruction makes synthetic provenance plausible and authenticity plausibly deniable. We therefore propose and evaluate a detection paradigm that outputs a calibrated prediction of whether authenticity is plausibly deniable: a faithful reconstruction by any known generator establishes plausible deniability, while calibration bounds how often content from known generators fails to be reproduced. Our evaluation shows that (i) our detector can be calibrated so that at most 1% of generated content is wrongly certified, an operating point at which most baseline detectors reach near-zero recall, including the strongest with 93% accuracy; (ii) calibrating a stricter security threshold on attacked samples preserves this bound against adaptive adversaries within the evaluated bounded-perturbation attack space, whose perturbations break every baseline, but does not cover arbitrary adversarial transformations; and (iii) post-hoc verifiability is eroding, as 1,116 of 3,000 Reddit images resist reproduction by a 2022 generator, but only 55 to 79 resist reproduction by 2024 generators.

This paper studies online quantile regression for large-scale and streaming data using Stochastic SubGradient Descent (SSGD) with constant learning rates. Classical offline inference for quantile regression is computationally and memory intensive. Existing works of online inference for quantile regression provide only asymptotic guarantees and typically require sub-exponential tail conditions for distribution theory. To bridge these gaps, we introduce new techniques to prove a quenched central limit theorem (CLT) and finite-sample Gaussian approximation for SSGD under a finite-moment assumption. We further show that Ruppert-Polyak averaging with a constant learning rate has a non-vanishing bias and fails to satisfy CLT centering at the population target. Hence we propose suffix averaging to address this issue and establish its finite-sample Gaussian approximation. Based on these results, we provide an efficient online inference method for quantile regression that avoids covariance estimation. Numerical experiments show that our method achieves desirable empirical coverage rates and competitive performance compared to other inference methods. We also apply our approach to U.S. wage data to demonstrate its practical effectiveness.

Xu asked whether the Hoare power space of a co-sober space, and in particular of a $T_2$-space or a metric space, must be co-sober. We give a negative answer. He and Zhao recently constructed a countable $T_{1}$ space $X$ whose lattice of open sets has a non-sober Scott topology. We observe that their topology has a countable subbase. The classical Ponomarev construction then yields an open continuous surjection from a countable zero-dimensional metrizable space $M$ onto $X$. Direct and inverse images then make the Scott space $\Sigma\mathcal{O}(X)$ a retract of $\Sigma\mathcal{O}(M)$, so $\Sigma\mathcal{O}(M)$ is not sober. Finally, we show that, for every $T_{0}$-space $Y$, the Hoare power space $\PH(Y)$ is co-sober if and only if the Scott space $\Sigma\mathcal{O}(Y)$ is sober. Consequently, $\PH(M)$ is not co-sober. Therefore, the same example answers negatively a question of He and Zhao about Scott sobriety of the lattices of open sets of countable Hausdorff spaces.

We study exact support recovery under $(\epsilon,\delta)$-differential privacy in sparse high-dimensional linear regression. We introduce Saturated Propose-test-release, a general mechanism that privately releases the output of a discrete selector with probability one once its stability certificate reaches a finite threshold. Exploiting the coordinatewise geometry of the LASSO, we construct a computable support-stability score. The resulting computationally efficient Saturated LASSO satisfies worst-case $(\epsilon,\delta)$-differential privacy and achieves exact support recovery with high probability under explicit regularity and beta-min conditions. Maximizing sparsity-indexed certificates yields an adaptive procedure requiring no sparsity knowledge and having exactly the same finite-sample exact-recovery risk as the oracle fixed-sparsity procedure under common public tuning parameters. We also establish a minimax lower bound explicitly tracking $\delta$: under its recovery conditions, Saturated LASSO is minimax optimal up to logarithmic factors in $n$ and $1/\delta$ uniformly over $0 < \delta \leq \epsilon/16$; in the broad moderate-$\delta$ regime, it further matches the lower-bound $\delta$-dependence. A complementary information-theoretic construction with known sparsity attains the lower-bound rates up to constant factors under independent Gaussian design, at exponential computational cost. Simulations and a semi-synthetic study using public American Community Survey covariates illustrate the numerical performance of the proposed procedures.

We study rigidity for the weighted $p$-Laplace equation on smooth bounded domains in dimensions three and higher. We show that a smooth positive weight sufficiently close to $1$ is identically $1$ if the derivative of its Dirichlet-to-Neumann map at one affine boundary value agrees with that for the unit weight. We also prove local rigidity at every smooth positive reference weight which is constant in one direction. The reference weight need not be close to $1$, and the unknown weight is only required to be sufficiently close to it, without a directional restriction. No convexity, analyticity, or agreement near the boundary is assumed. The proof combines linearization at a solution without critical points with recent rigidity results for the anisotropic Calderón problem. An identity for the linearized conductivity tensors removes the remaining diffeomorphism. This last step does not require smallness of either weight.

Feature pyramid methods, from FPN to BiFPN, have achieved strong performance in face detection by fusing multi-scale features. However, detecting faces under unconstrained conditions, such as small scale, occlusion, and extreme pose, remains difficult, as it requires global cross-scale dependencies that local fusion cannot model. State space models such as Mamba provide global context with linear complexity by scanning features as a sequence, and therefore offer a promising direction for this problem. Nevertheless, such a scan needs the two pyramid scales combined into a single feature map, and the way they are combined determines whether cross-scale structure is preserved. Summation collapses the two scales before the scan, so the scan has no cross-scale structure to exploit, while concatenation keeps both scales but at far higher cost. To address this, we propose Weave Mamba Fusion (WMF), which interleaves two adjacent pyramid scales column by column so that each step of a horizontal bidirectional SS2D scan moves from one scale to the other. With partial-channel processing and parameter-free de-weaving, WMF enables efficient cross-scale interaction while preserving feature structure. Integrating WMF into every fusion node yields WeaveBiFPN, the neck of our WeaveFace detector. On WIDER FACE, WeaveFace achieves 91.41% mean AP with only 0.34M parameters and 1.16 GFLOPs, outperforming prior detectors under 0.5M parameters. Its largest gains are on the Hard subset, where it reaches 87.14% AP. The code is publicly available at \url{https://github.com/dohun-mat/WeaveMambaFusion}.

We construct a feasible two-step estimator for the elasticity parameter in small-noise CEV-type diffusions. Under joint small-noise and high-frequency asymptotics, we establish its consistency and an $\varepsilon^{-1}$ central limit theorem with an explicit asymptotic variance. We adapt a Lamperti-based transformation strategy to elasticity inference for a CEV process with an attainable, absorbing $0$ boundary. Under a local-to-CEV scaling, an auxiliary CKLS process and a Girsanov change of measure yield a strictly positive CIR-type benchmark in which the elasticity is recast as a linear-drift parameter. A preliminary elasticity estimator based on local realized variance supplies the unknown parameter in the state mapping, yielding a feasible LSE-type estimator. The asymptotic theory combines control of the CKLS-to-CEV path-replacement error, total-variation closeness between the original and changed measures, and asymptotic negligibility of the preliminary plug-in error, thereby transferring the auxiliary CIR limit theory to the feasible estimator under the original measure.

High-quality child care in early life is a critical determinant of children's growth and development. Research on child care quality has been constrained by fragmented, non-research-friendly, and privacy-bound datasets. We present CCQ (Child Care Quality), a large-scale, de-identified dataset for applied data science research at the intersection of AI and early childhood health. CCQ integrates 59,372 child care provider records across 12 U.S. states, covering diverse provider types as well as data schemas. To ensure research utility while protecting privacy, we implement an automated, LLM-based curation pipeline that anonymizes, cleans, and standardizes raw state records into two complementary releases: a cleaned textual release and a fully preprocessed tabular release. We also benchmark traditional machine learning models, tabular foundation models, and language models on quality rating prediction and important features analytics. Within a state, tabular classifiers on the preprocessed tables perform best. Across states, zero-shot transfer is near chance, but modest target-state supervision recovers most of the within-state performance, and pretraining on other states benefits finetuned language models. We release both datasets with all code to accelerate AI-driven research on child care quality and ultimately improve children's health and development.

We extend the theory of Baer, Rickart, and their local, graded, and $*$-analogues for Steinberg algebras. First, we prove that positive definiteness passes from the coefficient field to the Steinberg algebra, and that this, together with graded von Neumann regularity, characterized by Steinberg and van Wyk, yields graded locally Rickart $*$-Steinberg algebras. Then we show that local Baer behavior forces every compact open subset of the unit space to be extremally disconnected. Consequently, when compact open subspaces of the unit space are metrizable, local Baer behavior forces the groupoid to be discrete. In the discrete case, we use the orbit-by-orbit decomposition into finitary matrix algebras over isotropy group algebras to convert annihilator conditions into explicit matrix conditions. For groupoids whose isotropy is trivial or infinite cyclic, these ingredients yield complete local and unital characterizations. In particular, local Baer and graded local Baer conditions are equivalent to discreteness, while the ungraded local Baer $*$-condition additionally requires that every orbit with nontrivial isotropy be a singleton. We then apply the theory to boundary path groupoids. Using the Steinberg algebra model, we obtain the Rickart, Baer, and Baer $*$ characterizations of Leavitt path algebras, including the local and graded variants.

Graphical User Interface (GUI) agents have emerged as a promising paradigm for automating complex digital workflows across diverse applications. However, training highly capable and generalizable agents fundamentally relies on massive, high-fidelity visual-action trajectories, which are notoriously difficult to acquire. While human demonstrations are unscalable, existing GUI world models rely on text descriptions or HTML rendering, discarding crucial pixel-level visual details like icons and layout styles. To address this issue, we introduce Infinite-Dreamer, a simulation-free data synthesis method powered by a pixel-level Image Editing World Model. By conceptualizing GUI transitions as image editing tasks, we leverage Vision-Language Models (VLMs) to describe action-induced UI changes as structured delta-text. We then fine-tune an image editing backbone to controllably synthesize realistic screenshot transitions. We utilize this model to generate both single-frame visual robustness data and multi-step imaginary trajectories. To validate the effectiveness of our approach, we fine-tune the Qwen3-VL baseline solely on the synthesized data to obtain Infinite-Actor, and evaluate it on AndroidWorld, MobileWorld, and AndroidControl-Curated benchmarks. Infinite-Actor consistently outperforms the Qwen3-VL baselines across scales: Infinite-Actor-8B improves AndroidWorld Pass@1 by +4.45 and nearly doubles the MobileWorld Pass@3 success rate, while Infinite-Actor-2B improves Pass@1 by +9.05. Code is available at https://github.com/swaydy-n/Infinite-Dreamer.

Cognitive Diagnostic Models (CDMs) identify which specific skills a student has and has not mastered, the signal a personalized learning path needs and a single aggregate score cannot give. Yet they are rarely deployed. The obstacle is their precondition: the Q-matrix, a mapping from every assessment item to the skills it requires, historically authored by hand. We separate the task into two stages: first construct the curriculum's own knowledge graph, the full space of concepts and skills it contains, independent of any item; then map items against that graph on demand. This paper addresses the first stage only. The item-mapping stage is designed but not implemented here, so the claim that this shifts judgment cost from once per item to once per curriculum is a design rationale rather than a finding. We present Curriculum Brain, a two-repository system pairing a version-controlled knowledge base with an agentic pipeline of eleven single-responsibility agents under a thin deterministic orchestrator. It generates candidate concept-skill mappings from official curriculum documents, checks them against accumulated rules, and compares them with a concept-skill map extracted independently from the textbook, repairing its own failures and escalating to a human only when it cannot resolve a case itself. Across 241 chapter runs (168 distinct chapters), 41.5% produced a Generator output passing both checks without a patch, and 67.6% resolved without escalation. Both are measured against criteria the system itself produced, so both describe internal consistency rather than agreement with an external standard, and both pool two pipeline configurations separated by a single change at run 77; after it the figures are 57.0% and 91.5%. Observed spend was $1.19 per chapter, API spend only, excluding human review. We release both the framework and the resulting curriculum dataset.