Papers

The five major moons of Uranus record a complicated geological history characterized by both tectonic resurfacing and heavy impact cratering, connecting their evolution to dynamical processes in the Solar System. Miranda, the smallest and closest major moon to Uranus, has an unusually large orbital inclination ($4.3^\circ$) and an remarkable surface geology dominated by coronae and tectonic structures. In this paper we explore whether Miranda's inclination and geological history can be explained through past orbital resonances within the Uranian satellite system. Using direct $N$-body integrations with the code SIMPL, we explore the tidal and dynamical evolution of the five major moons while assuming a fast tidal dissipation of Uranus and varying the internal responses of the satellites. Our simulations confirm that Miranda's inclination can be excited through resonant interactions involving Ariel and Umbriel, even when Miranda mean motion is not involved in that resonance. The Ariel-Umbriel 5:3 mean-motion resonance appears to be the most recent major dynamical event capable of driving Miranda's inclination to its current value. We compute the tidal heating associated with these resonance encounters and show that, for a strongly dissipative Uranus ($Q_U\sim600$), the resonance could cause heat fluxes in Miranda approaching those required to form its coronae. These results could provide new constraints on Uranus's tidal dissipation.

In 2018, V. Maz'ya proposed 75 open problems in analysis and PDEs. One of them is about the best constant for the Heisenberg Uncertainty Principle for solenoidal vector fields motivated by questions in hydrodynamics. This was answered by Cazacu, Flynn and Lam in dimension two and subsequently by Hamamoto in all dimensions $N\ge3$ by establishing sharp inequalities with explicit constants. Nevertheless, the much harder stability problem remains open in all dimensions. The first main result in this paper is to establish a sharp stability hierarchy by proving that its deficit controls the distance to the set of extremals, with the sharp stability constant $\frac12(N-\sqrt{N^2-4N+12})$ for $N\ge4$ and $1$ for $N=3$, together with chains of remainder terms measuring the distance to explicit larger families of poloidal and toroidal fields. The second main result is the second order $L^2$-Caffarelli–Kohn–Nirenberg inequality with the weights $|x|^{-2a}$ and $|x|^{-2b}$ on the line $1+a+b=0$, for which we obtain the stability constant $\frac12\min\{4(1+a),\sqrt{(N+2a)^2+4N-4}-(N+2a)\}$. This CKN inequality is sharp whenever the second number is the smaller one. We develop a novel method: the fourth order one-dimensional problems attached to the spherical modes are transformed, by a Fourier–Hankel transform of real order, into first order inequalities whose sharp stability is that of weighted Gaussian Poincaré inequalities. The inverse transform produces the extremals in terms of Kummer's function. As a byproduct of our new approach, we also obtain substantially simpler proofs, with sharp remainder terms and equality cases, of Hamamoto's one-dimensional inequality and of his sharp uncertainty principle for solenoidal fields.

We study simple predictable processes whose coefficients are represented by neural networks. On finite measure spaces, we establish density results for neural networks in Orlicz spaces. For filtrations generated by a stochastic process, measurable random variables, including at stopping times, can be approximated by neural networks depending on finitely many observations. Every stochastic integral with respect to a semimartingale can then be approximated, in the semimartingale topology, by integrals of such simple predictable processes. We show that restricting trading strategies to this class leaves the minimal mean-variance hedging error under partial information unchanged and obtain a no-free-lunch characterization in terms of equivalent martingale measures. Finally, the Bichteler-Dellacherie characterization of semimartingales remains valid even upon restricting the predictable integrands to those whose coefficients are represented by neural networks.

General purpose humanoids require locomotion controllers that are multi-skill, perceptive, dynamic, and robust enough to go anywhere humans can. In this work, we present a two layer locomotion architecture: (1) a perceptive flow matching motion generator plans whole body trajectories from raw depth images while a (2) perceptive tracking policy trained with control-guided RL follows these motions. Both policies are trained on a library of terrain consistent motion clips created with dynamically optimized human data which yields both accurate velocity tracking and terrain consistent references. Our central contribution is a simple yet effective off-policy RL fine tuning loop that improves the motion generator. A structured search method is used with the generator to gather data for advantage weighted regression. This off-policy loop is much more sample efficient than on-policy residual fine tuning and improves terrain consistency on unseen geometries and skill compositions. We find that successful terrain traversals increased by up to 25 percentage points and skill selection improved by up to 80 percentage points. By using raw depth images to perceive the environment no odometry or height maps are needed, and outdoor deployment is easy. With two cameras, the policy can see terrain coming from further away and adjust its velocity regardless of the commanded speed so it can traverse the terrain. A single policy pair enables a Unitree G1 humanoid to walk, run, stand, jump on and off of boxes, and traverse stairs in outdoor environments. Project page: https://zolkin1.github.io/generate-track-improve/

We combine the numerical conformal bootstrap with supersymmetric localization to constrain the spectrum of holographic superconformal field theories in the planar large $N$ limit for finite 't Hooft coupling $\lambda$. We consider the stress tensor multiplet four-point function in $\mathcal{N}=4$ $SU(N)$ super-Yang-Mills (SYM), which is holographically dual to closed string scattering, as well as the flavor multiplet correlator in a certain $ 4d $ $\mathcal{N}=2$ $U\!Sp(2N)$ gauge theory with $SO(8)$ flavor symmetry, which is dual to open string scattering. In both cases, we combine dispersive sum rules with integrated constraints from supersymmetric localization to compute an upper bound on the scaling dimension of the lowest scalar single-trace operator for all $\lambda$. For $\mathcal{N}=4$ SYM, we find that our bounds are close to the known integrability result for all $\lambda$. For the $\mathcal{N}=2$ theory, where integrability is not yet available, we find that our bounds are close to weak and strong coupling predictions in the relevant regimes.

Leaked system prompts are often treated as windows into the hidden values of commercial language models, yet their composition is rarely studied at scale. We analyze a merged corpus of 407 leaked, reconstructed, or officially published system prompts from 62 vendors across four community collections, identifying 29 near-duplicate clusters covering 66 files. Operational content rather than ethical statements dominates the corpus; a deliberately simple block-level classifier assigns roughly 58% of classified words to tool/protocol and roughly 5% to safety policy, while the strictest rule-lines guard tool use and file safety over harmful content by an 11:1 margin. Literal text transfer concentrates in a small set of cross-vendor pairs. Prompts also carry measurable maintenance debt, with version chains turning over thousands of words per release. The evidence supports treating leaked prompts as operational specifications, closer to configuration files than value statements, and treats reuse and prompt rot as engineering and supply-chain concerns. Because most documents are adversarial in origin and the detectors are deliberately simple, all magnitudes are directional; we audit the main classifier's error modes.

A polychromatic $2$-coloring of a triangulation is a $2$-coloring of the vertices such that no face is monochromatic. The discrepancy of a coloring is the maximum difference between the sizes of the color classes. Asayama and Matsumoto (Graphs and Combinatorics, 2022) proved that every triangulation admits a polychromatic $2$-coloring with discrepancy at most $\tfrac{5n-16}{9}$, and that there exists a class of triangulations for which every polychromatic $2$-coloring has discrepancy at least $\tfrac{n}{3} - 2$, where $n$ is the number of vertices. We improve the upper bound, showing that every triangulation admits a polychromatic $2$-coloring with discrepancy at most $\tfrac{3n-16}{7}$ and such a $2$-coloring can be computed in quadratic time. We also show a discrepancy of at most $n-\tfrac{4M}{3}$ for triangulations with a matching of size $M$. This implies, for example, that Delaunay triangulations admit a discrepancy of at most $\tfrac{n}{3}$. We provide a linear-time algorithm to compute a $2$-coloring whose discrepancy is at most $\tfrac{5n-24}{7}$. One of our results shows that any proper four coloring with the largest color class of size $\frac{n}{2}$ would imply a $2$-coloring with discrepancy at most $\frac{n}{3}$. The existence of such a proper coloring has been recently confirmed by Kawarabayashi, Yoneda, and Yoneda (arXiv 2026). Therefore the two results together confirm the discrepancy of at most $\frac{n}{3}$ for triangulations.

Biomedical image segmentation is central to medical image analysis, but practical deployment often faces limited annotations, memory constraints, and cross-site distribution shifts. Implicit Neural Representations (INRs) have recently emerged as a lightweight alternative for semantic segmentation, achieving competitive performance with substantially fewer parameters than conventional architectures. However, the mechanisms, scaling behavior, and domain generalization abilities of INR-based segmentation remain insufficiently understood. In this work, we study these questions in the context of cross-domain brain MRI segmentation. We analyze INR-based segmentation across low-parameter regimes, comparing it with conventional pipelines in both in-domain and out-of-domain settings. Surprisingly, we find that INR-based models do not simply improve with increasing parameter budget. Their advantage is most pronounced under low-parameter and limited-augmentation settings, while U-Net-based models benefit more from larger capacity and standard augmentation. We also investigate how INRs encode semantic information in their hidden features and show that complementary segmentation-relevant structure is distributed across multiple INR layers. Building on this insight, we introduce HierINRSeg, a hierarchical INR-based architecture that aggregates multi-layer representations for improved robustness and generalization. Extensive experiments show that HierINRSeg consistently outperforms MetaSeg, a strong recent INR-based segmentation baseline, with an average improvement of 5.6 percentage points in Dice for the in-domain test set and 8.2 percentage points out-of-domain. Overall, our analysis identifies the conditions under which INR-based segmentation is most effective, providing concrete guidance for model selection and future research.

Uneven rotation and dropped frames make equal-angle assumptions unreliable for turntable reconstruction. We present OC-GS, an object-centric Gaussian splatting that refines each image's angle while maintaining a shared camera, rotation axis, and pivot. This orbit-consistent refinement jointly optimizes image-derived geometry and angles to reconstruct objects from sparse, irregular captures. On rendered objects with 12, 8, and 6 irregularly spaced views, OC-GS achieves mean foreground PSNR scores of 21.26, 19.36, and 15.83dB, respectively, exceeding all four evaluated pose-free Gaussian splatting baselines in each condition. Under a shared trainer, refining image-estimated angles improves mean foreground PSNR by 7.88dB over keeping those estimates fixed. An ablation study shows that both image-derived angle initialization and the shared motion model contribute to the improvement. On real captures, OC-GS's refinement increases mean foreground PSNR by 0.70dB. Results show that refining uncertain angles within a shared motion model improves reconstruction from sparse, irregular turntable captures.

Many LLM training and inference methods, including RL and test-time scaling, depend on repeated sampling, but benefit only when the responses meaningfully differ. Self-training faces the same challenge: training data is typically constructed by sampling IID responses and filtering primarily for correctness, thereby overrepresenting strategies a model already favours. We investigate strategic diversity, or substantive variation among approaches to a problem, as an alternative principle for constructing self-training data. We generate strategically diverse data with two sampling methods: GROOT, a new method which constructs a hierarchical tree of approaches and samples distinct paths, and Verbalized Sampling (VS), adapted to produce an unstructured set of approaches. Across competitive programming and Next-Chapter Prediction domains, models trained on strategically sampled data outperform IID-trained counterparts on difficult tasks and provide strong initializations for RL and test-time scaling. Most strikingly, self-training on strategically diverse but incorrect traces from Qwen3-4B outperforms IID distillation from a 235B teacher. These results challenge prevailing assumptions about what makes useful self-training data and show that diversity of approaches can matter more than correctness or teacher scale.

Deep learning has substantially accelerated the calibration of complex stochastic-volatility models, but neural point calibration alone does not capture the uncertainty remaining after an implied-volatility (IV) surface has been observed. We develop a simulation-based inference framework for rough Heston (rHeston) calibration that learns the posterior distribution of the model parameters conditional on an IV surface. Using neural ratio estimation, we obtain calibrated posterior samples that can be propagated through heteroscedastic neural surrogate pricers for path-dependent exotic options. The resulting posterior-predictive distributions combine residual parameter uncertainty with conditional surrogate uncertainty and yield uncertainty-aware price intervals. We further introduce Hellinger-SHAP, an information-theoretic explainability method for posterior inference. Rather than attributing a single parameter point estimate, it applies local-background Kernel SHAP to a posterior-information functional measuring contraction from the prior to the posterior. This identifies maturity–moneyness regions associated with posterior information gain for individual rHeston parameters. In a simulation study, posterior-predictive intervals provide calibrated or conservative coverage across forward-start, barrier, and realized-variance claims, while point plug-in prices can be materially unreliable for selected contract regimes. Together, the UQ and XAI analyses provide a transparent framework for uncertainty-aware neural calibration and downstream exotic pricing under the specified prior-predictive model.

Conversational AI tools are entering children's everyday experiences, and schools are interested in adopting them. However, successful classroom integration depends not only on the technology but also on the work teachers do to make it usable and appropriate for their students and classroom context. There is little known about how elementary teachers work as they implement conversational AI tools in real classrooms. In this study, we examine three teachers' experiences implementing an AI literacy and English Language Arts (ELA) curriculum built around ToyTalk, a conversational AI toy development platform, over 13 instructional days, a three-week summer camp. Drawing on daily individual reflections, group reflections, and post-camp interviews, we find that teachers' adaptive practices of repair, differentiation, translation, and balancing sit at the intersection of three tensions (technology, learner, and instruction). Teachers' understanding of AI and their role evolved over the camp experiences. From these findings, we contribute design implications and considerations for deploying conversational AI within elementary classrooms.

AI tutoring could markedly improve learning outcomes for students in developing regions such as Vietnam, yet the two obvious paths both fall short. Cloud assistants such as ChatGPT route sensitive student data to foreign servers—violating data-sovereignty laws such as Vietnam's Decree 53—and, pre-trained on Western-centric corpora, are not organized around the national textbook curriculum, so their knowledge of local content is unsystematic and frequently hallucinated. Self-hosting an open model keeps data on-premise but hits a two-fold wall: post-training quantization (AWQ, GPTQ) tames the static weight footprint, yet the dynamic KV cache and prefill latency of long tutoring contexts still cause out-of-memory failures and slow responses on consumer GPUs, while the model keeps hallucinating on region-specific material. We present DeepEdu-v1, an AI-tutoring system for Vietnamese education built on SCALE (Self-improving Context-Aware Learning Engine), a framework with two innovations. First, a long-context inference engine amortizes token selection from per-sub-chunk to per-cluster granularity; on long-context retrieval it issues x7.7 fewer retrieval calls than a state-of-the-art selective-attention baseline, cutting prefill latency (TTFT) by roughly 35% while matching or improving task accuracy. Second, a self-improving agentic layer continuously curates a verified playbook from past interactions instead of fine-tuning, a design intended to progressively reduce reliance on dominant-language priors as trustworthy local knowledge accumulates. In its deployed configuration, DeepEdu achieves a nearly x2 TTFT speedup over standard vLLM serving and lifts agentic accuracy from 70.0% to 79.5% on complex tasks, with the strongest per-track gains across financial-reasoning and interactive-agent benchmarks.

Let $n \geq 4$ and $\rho: \pi_1S \rightarrow PSL(n,\mathbb{R})$ be a Hitchin representation. We study the topology of a cocompact domain of discontinuity $\Omega_{\rho}$ in the flag manifold $\mathcal{F}_{1,n-1}$ of line-hyperplane pairs in $\mathbb{R}^n$ defined by Guichard-Wienhard. In particular, we lift $\Omega$ to its $(\mathbb{Z}_2\times \mathbb{Z}_2)$-cover $\hat{\Omega}$ in the Stiefel manifold $V_2(\mathbb{R}^n)$. The domain $\hat{\Omega}$ is highly connected and smoothly fibers over hyperbolic space $\mathbb{H}^2$ with unknown fiber $\hat{\mathfrak{F}}$, a closed $(2n-5)$-manifold. We determine the homeomorphism type of $\hat{\mathfrak{F}}$ as well as its diffeomorphism type up to connected sum with a homotopy sphere. The classification of $\hat{\mathfrak{F}}$ involves computing both its homology and a certain differential topology invariant needed to invoke Wall's classification of highly connected odd-dimensional manifolds.

There have been recent proposals for human settlements on Mars in 2030s. Any human activity on Mars must be preceded by extensive robotic exploration. However, Mars exploration is bottlenecked by the low bandwidth, intermittent Mars-Earth link. For example, HiRISE, a high-resolution camera onboard the Martian orbiter MRO imaged less than 3% of Mars over eleven years, even though MRO's low resolution Context Camera had mapped more than 99% of Mars in that time. We present a systems case for shared compute for Mars exploration. Such Mars-local compute, paired with advances in computer vision and AI, can enable large volumes of data to be collected and processed on Mars while sending periodic updates, insights, and selective datasets to Earth. To overcome the lack of surface infrastructure on Mars, we propose a two-tier in-orbit deployment of computational satellites that provides consistent coverage and bandwidth. Our analysis shows that the proposed deployment can start small: one areostationary node makes compute reachable from all active Mars missions, two additional areostationary nodes can extend this coverage to roughly 90% of the planet, while low-Mars-orbit nodes add high-rate surface links and compute capacity where demand grows.

As the end point of stellar evolution for most stars, white dwarfs offer powerful insight into stellar magnetism. Multiple formation channels are required to explain the observed properties of magnetic white dwarfs, and the mechanisms that produce unusual objects have potential to unlock new realms of physics. The large samples made possible by spectroscopic surveys enable study of these processes on an unprecedented level. Here we perform a detailed model atmosphere analysis of the magnetic white dwarfs in DESI Data Release 1. We identify 833 magnetic systems, 762 of which we fit using an offset dipole model. Of these, 127 targets require more in-depth analysis as they possess shallower or sharper absorption features than predicted by the offset dipole model. For these we use a patchy offset dipole atmosphere model in which the magnetic field only contributes to the line opacities at certain regions of the visible stellar disk. We compare the observed distribution of magnetic white dwarfs to the mechanisms that generate magnetic fields. We find that a core-convective dynamo that forms on the main-sequence matches the parameters of many of the magnetic systems, while a dynamo that fills the radiative interior on the giant branch reproduces many of the lower mass objects that the core-convective dynamo cannot. This sample is the largest single data release of magnetic white dwarfs to date. The diversity of formation channels and the frequency of objects with unusual field geometries highlight the power of multiplexed surveys to advance our knowledge of stellar magnetism.

We show that neural network weights can be explicilty fintuned to admit a smaller grammar. Weight Pair Encoding (WeightPE) does so by placing a lossy Re-Pair compressor inside a straight-through estimator. The int8 weights of the network are flattened into one string, and near-matching Re-Pair patterns are made exactly equal within a global L2 budget. The network computes with the rewritten weights and trains through them with a straight-through estimator. Unlike a flat codebook of fixed-size entries, a grammar offers variable-length patterns and reuses them hierarchically inside larger ones. On the MLP weights of ViT-B/16 and ViT-L/16 finetuned on CIFAR-10, WeightPE produces a Re-Pair grammar 0.43x and 0.38x the size of the one produced by an equivalent int8 QAT run, at a cost of 1.9 and 1.1 accuracy points. The trend extends to different grammar compressors (LZ78, SEQUITUR), over which the networks has not be finetuned against. To our knowledge, this is the first time grammar size has been used as an explicit training objective for network weights.

Multi-agent LLM systems are often expected to improve as team size increases, yet the scaling behavior may depend on task structure. Our central contribution is to introduce Steiner's taxonomy of group tasks as a framework for analyzing multi-agent LLM scaling and focusing the analysis on disjunctive and compensatory tasks. We model independently sampled agents as conditionally independent given the item, which yields their large-team limits: plurality voting converges to the model's modal answer, and averaging converges to the model's item-level bias. Across selected representative benchmarks, 13 open-weight models, and teams of up to 30 agents, we find qualitatively different scaling behavior. On disjunctive tasks, the probability that at least one agent is correct grows by 5-20 points with team size, but plurality voting over agents that answer directly realises almost none of this potential, as the model predicts to within 0.5 points on average. Multi-round revision raises accuracy considerably, yet the gain is nearly the same with one peer as with 29. In contrast, scaling provides little benefit on Fermi estimation, despite its natural suitability for aggregation: item-level biases shared across the samples of a model account for about 87% of the squared error, so averaging reduces error by only about 6%. Combining model families helps on Fermi estimation but does not surpass the strongest member on disjunctive tasks. These results show that task structure, together with the mechanism combining member outputs, is a fundamental determinant of team scaling.

Recent advances in agentic Artificial Intelligence (AI) systems have marked a shift in AI for Science: moving away from the use of individual AI systems for narrow task execution, toward multi-agent systems capable of orchestrating complex, end-to-end research workflows and performing (semi-)autonomous scientific discovery. The development of multi-agent AI-for-science systems has primarily focused on improving the cognitive capabilities of AI systems, specifically by making advanced reasoning and hypothesis generation more reliable. However, focusing only on cognitive capability improvement could ignore appropriate management of resources, a key bottleneck in scientific discovery. Testing and validating scientific hypotheses and experiments is, physically and economically, resource-intensive and resources are limited. This means that to make significant advancements in autonomous scientific discovery, such as improving human-AI co-scientist complementarity or reaching a truly closed-loop automated process, we must pair ongoing improvements in reasoning capabilities with robust resource management. In this paper, we outline an infrastructure for AI resource management by developing the necessary foundations of scientific agent economies, markets, and institutions. The aim of this infrastructure is to empower AI agents and human scientists to effectively (i) collaborate and establish research priorities, (ii) assign credit, (iii) track accountability and liability, and (iv) safeguard against malicious use and information security risks. Finally, we engage with the macro-level societal implications. of (semi-)autonomous scientific discovery to inform the development of governance policies ensuring an equitable distribution of AI-driven discoveries and derivative future technologies.

Many models, such as fixed-effect models for panel or network data, are hard to estimate because they feature nuisance parameters that are both numerous and estimated imprecisely. This, in general, causes an incidental-parameter problem in the estimator of the parameters of interest. The problem can be alleviated by working with an estimating equation whose expectation is insensitive to the value of the nuisance parameters. We discuss and contrast three notions of insensitivity, also called orthogonality, in the context of likelihood models: Neyman orthogonality, Neyman orthogonality to order q, and full orthogonality. Orthogonal moments are obtained by projecting the estimating equation on nested subspaces, which are spanned by, respectively, the scores of the nuisance parameters, the first q derivatives of the likelihood ratio with respect to the nuisance parameters, and all likelihood ratios of the model. We give explicit constructions in binary-choice, count-data, and nonlinear regression models.

Large neural networks can be compressed by rounding or "quantizing" their weights to numbers that admit representations with fewer bits. One algorithm for quantization, OPTQ, progressively quantizes the weights of a neural network so that the squared quantization error on a specified calibration dataset is as small as possible. We study the performance of OPTQ and a variant algorithm, stochastic OPTQ, in a generalization setting and derive bounds for the expected squared error accrued by the algorithm when a test point is drawn from a fixed distribution. We prove two results. One result relates the generalization error to the error on a calibration dataset comprising independent samples from the same distribution as the test distribution. The other result bounds the generalization error of stochastic OPTQ for all sufficiently nice distributions, regardless of the calibration dataset. In both of these results, the regularization term $\lambda$ plays an important role. We use insights from these results to make a new recommendation for the choice of $\lambda$ and see that this choice of $\lambda$ preforms favorably in experiments when compared to prior recommendations in the literature.

Online learning in non-stationary environments requires models to adapt rapidly from streaming data under strict computational constraints. A principled approach casts online learning as Bayesian state tracking, where model parameters are updated sequentially via Bayesian filtering. However, applying Bayesian filters directly to modern deep models is computationally prohibitive due to the high dimensionality of parameter space, forcing existing methods to rely on restrictive approximations or manually designed low-dimensional subspaces. In this work, we identify the absence of a suitable low-dimensional dynamical representation as the core bottleneck in Bayesian filtering-based online learning. Accordingly, we propose Adaptive Update through Representation Adaptation (AURA), a meta-learning framework that learns offline a low-dimensional latent state-space model governing the evolution of optimal model parameters under distribution shift. Online adaptation is then performed via extended Kalman filtering in this learned latent space followed by reconstruction of the full model parameters through a learned lifting map, enabling efficient single-step online adaptation while preserving model expressiveness. Evaluated on online adaptation of neural wireless receivers under time-varying channels and on non-stationary image classification, AURA shows substantial improvements in adaptation speed, accuracy, and computational efficiency over existing online learning and Bayesian filtering baselines, demonstrating that an adaptation-aware latent geometry is beneficial for effective Bayesian online learning in high-dimensional models.

Contrastive Language–Image Pretraining (CLIP) has emerged as a dominant vision backbone due to its strong transferability and zero-shot capabilities. However, recent studies reveal a critical vulnerability: embedding-space backdoor attacks. By poisoning only a tiny fraction of image–text pairs, adversaries can implant stealthy triggers that induce targeted shifts in CLIP's joint embedding space. Unlike conventional backdoors that manipulate classifier logits, these attacks corrupt representations directly, making them highly effective under extremely low poisoning ratios and difficult to detect. Existing defenses require access to model parameters, gradients, logits, or clean validation data—assumptions that rarely hold in realistic black-box deployments. Moreover, current black-box methods struggle to accurately localize small or out-of-distribution triggers. We propose CLIPGuard, a lightweight and fully black-box defense specifically designed to mitigate embedding-space backdoors in CLIP encoders. CLIPGuard identifies malicious regions by measuring segment-wise embedding perturbations and selectively purifies only suspicious segments via semantic inpainting, preserving benign visual content and alignment quality. Extensive experiments on STL-10, ImageNet, and diverse trigger families—including BadCLIP, BadNets, blended, patch-based, and typographic attacks—demonstrate that CLIPGuard reduces attack success rates to as low as 1.05% while maintaining clean accuracy up to 86.34%, consistently outperforming existing black-box defenses, including CleanCLIP and CleanerCLIP. Our code is available https://github.com/wsu-cyber-security-lab-ai/CLIPGuard.git

Chambolle-Dossal acceleration guarantees $F(x_n)-F^*=o(n^{-2})$ for every fixed smooth convex loss with a minimizer. We show that this qualitative improvement admits no universal quantitative rate. For every damping parameter $\alpha>3$ and positive nondecreasing gain $G(n)\to\infty$, we construct a fixed one-dimensional smooth convex loss whose exact CD orbit satisfies $\sup_{n\ge1} n^2G(n)\bigl(F(x_n)-F^*\bigr)=\infty$. Thus no divergent gain improves the $n^{-2}$ scale for all fixed losses, even with instance-dependent constants. The construction prescribes queried gradients and realizes infinitely many slow blocks within one smooth convex objective. Under local $p$-power growth with $p>2$ and sufficiently strong damping, we also construct a fixed loss whose exact CD orbit satisfies $F(x_n)-F^*\sim Dn^{-2p/(p-2)}$, $D>0$, establishing the sharpness of the known convergence rate. Both main results are formally verified in Lean 4.

We show that real rank zero is not preserved under matrix amplifications of inclusions. This yields real rank zero inclusions that do not approximately factor through real rank zero C*-algebras, thereby resolving two questions of Gabe and Neagu.

We develop a boundary element method for the Laplace equation with Dirichlet boundary conditions in a bounded three-dimensional domain. The method uses flat panels of maximum diameter \(h\) on an interior surrogate boundary and corrects the imposed boundary values by a first-order Taylor expansion along the panel normals. The correction uses the normal flux already present in the Calderón formulation and introduces no additional unknown field. Under uniform assumptions on the surrogate geometry and boundary operators, the penalised form is coercive when \(\gamma\delta_h\le c_0<4\), where \(\gamma\) is the penalty parameter and \(\delta_h\) is the maximum normal offset. The error estimate separates trace approximation from a geometry term of size \(\gamma^{1/2}\delta_h^2\|u\|_{H^3(\Omega)}\). With \(\gamma\simeq h^{-1}\), the sufficient condition \(\delta_h\lesssim h^{(k+1)/2}\) yields an error of order \(h^{k+1/2}\) for \(P_k/P_{k-1}\) Dirichlet/flux spaces under the stated regularity and uniform trace bounds. Thus an offset of order \(h^2\) supports the estimate through cubic Dirichlet approximation. We analyse two penalty variants and present numerical experiments on a sphere and an ellipsoid to examine convergence and sensitivity to geometry, penalty scaling and quadrature.

We investigate the two-dimensional Dunkl–Pauli equation for a spin–$1/2$ particle in the presence of an external magnetic field within an algebraic framework. While previous studies have provided exact solutions of the Dunkl–Pauli system, its underlying dynamical symmetry and algebraic structure remain largely unexplored. By employing Schrödinger factorization, we construct the generators of the $\mathfrak{su}(1,1)$ algebra and establish the associated symmetry of the radial sector. The energy spectrum is derived using irreducible unitary representations, and the corresponding Sturmian radial basis is obtained analytically. Furthermore, $\mathrm{SU}(1,1)$ coherent states are constructed and their time evolution is analyzed, revealing a characteristic radial breathing behavior. The results show that the Dunkl deformation introduces parity-dependent modifications in the spatial structure of the system while preserving its underlying algebraic dynamics.

Ensuring online safety through content monitoring had raised Hate Speech Detection as a crucial task to be addressed. By essence the task demands the capture of contextual cues, which are essential for a precise understanding of the content's intent. Although automated detection approaches for the task have advanced significantly, the scarcity of non-English resources persists, limiting the ability of models to adapt to the subtle, context-dependent, and culturally related nature of multimodal content. In this paper, we introduce MexHat, a video dataset designed to capture the linguistic and cultural cues for the hate-speech detection task in a Mexican Spanish context. Our dataset comprises around 1k video clips annotated across two tasks: a three-way class evaluation (no negative content, offensive content and hate-speech content), and a fine-grained class evaluation including three hate-speech sub-categories. The dataset statistics and the baseline results highlight the inherent challenges associated with the task. Disclaimer: This paper contains sensitive content that may be disturbing to some readers.

As large language model (LLM) inference becomes increasingly expensive, resource-consumption attacks pose a growing threat to model providers. Existing attacks typically amplify cost by inducing abnormally long or repetitive outputs on attacker-controlled or triggered requests, making them easier to detect and limiting their deployment-wide impact when benign traffic dominates. In this work, we uncover a previously overlooked token-level attack surface arising from the many-to-one mapping from token sequences to decoded text. Although standard LLMs predominantly generate the canonical token sequences induced by their tokenizers, the same text can also be represented by substantially longer non-canonical sequences. This representational flexibility exposes a new avenue for resource-consumption attacks: an attacker can train the model to favor such sequences, systematically increasing the number of autoregressive decoding steps without a proportional increase in visible response length. However, we empirically find that directly maximizing token fragmentation substantially degrades model utility, producing conspicuous answer-quality failures that undermine attack stealthiness. To address this challenge, we propose FragToken, a training-time framework that combines source-model self-distillation, capacity-aware filtering and budgeting, and BPE-Aligned Merging to induce fragmented generation under ordinary prompts while largely preserving model utility. We evaluate FragToken on four LLMs across three benchmarks. Across the four models, FragToken achieves a three-benchmark average token inflation ratio (TIR) ranging from 1.99 to 2.46, while causing only minor degradation in model utility. Our work reveals a covert LLM supply-chain threat that increases inference cost without requiring large volumes of attack requests while largely preserving utility.

Disaggregating the two stages, Prefill and Decode, onto separate GPU pools is now a standard optimization for (text-only) LLM serving. However, multimodal LLMs (MLLMs), which add a third phase, Encode, pose new challenges for resource allocation. Encode turns images, video, or audio into embeddings that the language model can consume, yielding a three-stage Encode-Prefill-Decode (EPD) pipeline. Existing frameworks offer only partial answers: text-only PD systems lack Encode, while EPD frameworks expose it as a separate service without regulating downstream request flow. The pipeline also carries a structural resource imbalance: every request enters through Encode before downstream work can begin, yet per-request execution leaves the encode GPU severely underutilized even at high loads, starving the downstream Prefill and Decode workers. Addressing this, we reposition Encode as the control point of the EPD pipeline, exposing three tightly coupled dimensions: when work enters downstream, where prefill executes, and how the GPU is shared. We instantiate this in EAServe across two co-designed layers. Its runtime manages load-adaptive micro-batching, rate-controlled partial offload to a co-resident prefill worker, and dynamic SM partitioning for predictable co-location. The configuration layer, Hybrid Auto Selection (HAS), navigates the joint space of GPU allocation, encode batch size, and offload ratio by pruning unbalanced allocations with per-stage capacity profiling and refining the remainder through TPE-based Bayesian optimization. Evaluated on three MLLM architectures spanning image, video, and audio, EAServe delivers up to 4.3x and 1.7x higher goodput than NVIDIA Dynamo and vLLM, respectively, under identical SLO constraints, sustains more balanced and higher GPU utilization across the EPD pipeline, and reaches near-optimal configurations faster than baseline search methods.