Peer Revenue Returns subject to peer review

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Wednesday 2 September

Machine Learning (Statistics)

HyperMC: Multi-Fidelity Hyperparameter Tuning for Stochastic Gradient MCMC

Stochastic gradient Markov chain Monte Carlo (SGMCMC) methods enable scalable Bayesian inference, but their performance depends strongly on hyperparameters such as the step size, mini-batch size, and number of leapfrog steps. Since most SGMCMC algorithms lack a Metropolis-Hastings acceptance rate, standard acceptance-based tuning methods are not directly applicable. We propose HyperMC, a multi-fidelity tuning framework that combines Hyperband-style resource allocation with kernel Stein discrepancy (KSD) evaluation. By running multiple successive-halving brackets, HyperMC balances broad exploration of a continuous hyperparameter space with increasingly accurate evaluation of promising configurations under a fixed computational budget. We further introduce Robust HyperMC, which uses global grid initialization followed by elite-guided local refinement to reduce sensitivity to random candidate generation and noisy finite-budget evaluations. Under suitable approximation and concentration conditions for the estimated KSD, we establish that the successive-halving component selects a near-optimal configuration among the sampled candidates with high probability and derive a sufficient computational budget for successful selection. Experiments on logistic regression, probabilistic matrix factorization, and Bayesian neural networks show that HyperMC improves posterior approximation or predictive calibration relative to MAMBA, grid search, and heuristic baselines, while Robust HyperMC yields more stable and reproducible tuning results.
Read the full abstract

Quantum Algebra

An $\mathfrak{sl}_2$-action on the uri Lie algebra

We construct an $\mathfrak{sl}_2$-action by derivations on the Lie algebra $\mathfrak B$ of swap invariant alternil bimoulds with the uri bracket of Kühn and Schneps. This Lie algebra plays the role for formal multiple Eisenstein series which Racinet's double shuffle Lie algebra $\mathfrak{dm}_0$ plays for multiple zeta values. The kernel $\mathfrak m$ of the lowering operator is a Lie subalgebra, $\mathfrak B$ is the direct sum of its iterates under the raising operator, and this gives Rankin-Cohen type operators on $\mathfrak m$. We define a Lie subalgebra $\mathfrak d$ of $\mathfrak B$ and show that it is isomorphic to $\mathfrak{dm}_0$ extended by an additional element in weight one.
Read the full abstract

Probability

Gamma approximation and Poisson–Gaussian invariance principle on Poisson chaos

We study centered Gamma approximation and a same-kernel Poisson–Gaussian invariance principle on fixed Poisson chaoses. For Gamma approximation, a martingale-core argument extends the carré-du-champ and $d_2$ estimates of Döbler and Peccati (Ann. Probab., 2018) from regular kernels to every fourth-integrable chaos element. In the diffuse regime, characterized by vanishing fourth add-one energy, this yields an exact four-moment criterion under uniform integrability of fourth powers. In the rare-jump regime, ordinary moments do not determine the approximation mechanism: convergence of the full moment sequence may coexist with a nonvanishing fourth add-one energy, and we construct such centered Gamma limits in every fixed chaos order. The invariance principle is independent of the Gamma target. For Poisson and Gaussian multiple integrals with the same kernel, we bound both smooth-test discrepancies and the Wasserstein distance in terms of the variance and the fourth add-one energy. Thus, vanishing fourth add-one energy is an intrinsic Lindeberg condition under which the two chaoses are asymptotically indistinguishable in distribution. Combined with a moment-transfer estimate and the Gaussian fourth-moment theorem, this gives an alternative proof of the qualitative normal fourth-moment theorem on a fixed Poisson chaos. A rainbow example shows that the Lindeberg condition is essential: the Gaussian analogue may be asymptotically normal while the Poisson integral converges to a centered compound-Poisson law. The same comparison also explains the different behavior of even and odd chaos orders for diffuse centered Gamma limits.
Read the full abstract

Information Theory

Embedding Linear Codes into Optimal Self-Orthogonal Codes Based on Constraint-Preserving Genetic Algorithms

We study the problem of finding shortest self-orthogonal embeddings of a given code which have the largest possible minimum distance by employing a constraint-preserving genetic algorithm. We show that the orthogonal group is generated by transvections corresponding to weight-four vectors. Using this result together with the fact that all shortest self-orthogonal embeddings can be obtained through the action of the orthogonal group, we construct a genetic algorithm in which each chromosome is represented by a sequence of weight-four vectors. Applying our algorithm to codes with dimensions $9\le k\le 15$ and lengths $22\le n\le 256$, we obtain 280 inequivalent optimal self-orthogonal codes. In particular, we obtain 34 codes with new parameters and 100 codes which are inequivalent to previously known codes with the same parameters.
Read the full abstract

Robotics

Unified Motion Retargeting for Humanoids with Learned Point Cloud Correspondence

Humanoid learning increasingly relies on transforming vast and diverse human motion data into high-quality robot reference trajectories. However, retargeting human motion to humanoid robots is challenging due to substantial differences in morphology, degrees of freedom, joint ranges, and kinematic constraints between humans and robots. Existing retargeting methods typically address these differences by defining human-robot correspondence through hand-crafted sparse keypoints or body-part pairs. As a result, retargeting quality depends heavily on manual semantic design, limiting scalability across motion sources and robot morphologies and providing only sparse guidance for reproducing detailed poses and interactions. In this paper, we present Unified Motion Retargeting (UMR), a framework that learns dense point cloud correspondence without requiring manually designed human-robot mappings. By treating exterior point clouds as a unified interface between human motion and humanoid robots, UMR decouples retargeting from source-specific skeletal semantics and robot-specific topology. The learned dense correspondence provides fine-grained geometric anchors for constrained point cloud matching optimization, enabling surface-level pose alignment and direct transfer of interaction contacts. Experiments demonstrate that UMR unifies retargeting across heterogeneous motion sources, robot embodiments, and downstream scenarios ranging from locomotion to interaction, while achieving higher motion fidelity and plausibility than state-of-the-art methods. UMR therefore provides a scalable foundation for transforming large-scale human motion references into robot-ready training data.
Read the full abstract

All authors Hanyang Cao, Yuetong Fang, Taesoo Kwon, Runyi Yu, Ji Ma, Jing Tan, Yangchen Zhou, Baoze Du, Yi Gu, Yukang Gao, Ruoli Dai, Lei Han, Renjing Xu

Artificial Intelligence

EmoStance: Response-Side Affective-Orientation Control for Empathetic Response Generation via Emoji Weak Supervision

Empathetic response generation requires models to decide not only what to say, but also how to respond to the previous speaker's affective situation. We formulate this as response-side affective-orientation control and use multi-annotator emoji distributions as weak affective–attitudinal evidence, rather than as output symbols or gold labels, to induce a latent control space that operationally approximates listener stance. We construct EmojiDialogue, an utterance-level extension of EmpatheticDialogues with emoji votes and confidence scores, and propose EmoStance, which models source-side affective expression, predicts a soft response-side orientation from dialogue context and speaker roles, and steers a frozen instruction-tuned LLM through continuous prefix embeddings. In blind pairwise evaluation with 20 annotators and 800 judgments, EmoStance achieves a 62.2% decisive win rate, with the clearest gains in contextual specificity and perceived responsiveness, while remaining complementary to external-knowledge methods. Code, annotation metadata, and reconstruction scripts are available in our GitHub repository: https://github.com/18277390221/EmoStance.
Read the full abstract

Systems and Control

Existential Opacity for Discrete-Event Systems with State Observations

Opacity is a fundamental system property for confidentiality in discrete-event systems (DES). Classical opacity is typically defined under event-based observations, requiring that any secret system behavior remains indistinguishable from some non-secret behavior to an external intruder. However, in many applications such as path planning or opacity-preserving tasks, the intruder observes system states rather than events. Moreover, it often suffices that the system exhibits secret behaviors that can be exploited for opacity-preserving task execution, but such a system property cannot be fully captured by existing notions of state-observation-based opacity. Motivated by this limitation, we propose a relaxed notion of existing state-observation-based opacity, called existential opacity (EO), which only requires the existence of secret behaviors (instead of all secret behaviors) that are indistinguishable from a non-secret behavior under the state observations of the intruder. We show that the notion of EO is more expressive than existing state-observation-based opacity notions. In addition, a class of EO properties together with their corresponding verification approaches are developed, enabling the analysis of existential opacity in discrete-event systems and providing a new criterion for determining the feasibility of opacity-preserving problems.
Read the full abstract

All authors Zhiyuan Huang, Zhao Tong, Jiakai Li, Bingzhuo Zhong

Computation and Language

C$^{3}$T: Counterfactual Causal Reasoning for Sentiment Shifts in Social-Media Conversation Trees

Sentiment in social-media threads does not only vary across posts; it shifts as users react to claims, corrections, evidence, and hostility within a branching reply tree. We study why sentiment changes in rumor-centric conversation trees by treating discourse moves (e.g., denial/correction, evidence/link, toxicity/attack) as candidate interventions and asking (i) what sentiment a reply expresses, (ii) whether the sentiment shifts relative to its parent, and (iii) which prior message most plausibly drove the reply's sentiment. To support this setting, we introduce CaSiRe, a causal sentiment reasoning layer over public rumor conversation datasets that adds post-level sentiment labels, induced parent-child shift labels, calibrated multi-label intervention tags, and explicitly annotated causal-source labels. We then propose C$^{3}$T (Counterfactual Causal Conversation Transformer), a thread-structured temporal model that jointly predicts node sentiment and shifts, learns sparse ancestor attribution, and supports counterfactual queries by forcing conversational intervention embeddings on or off to estimate potential outcomes. Under an event-level split, C$^{3}$T improves out-of-event robustness and attribution over text-only, graph-based, and temporal baselines, and yields interpretable model-based effects: denials/corrections and evidence reduce downstream negativity, while toxicity increases it. We also benchmark open-weight LLM prompting baselines and find that added conversational context helps, but attribution remains less reliable, motivating structure-aware counterfactual modeling for social-media analysis.
Read the full abstract

High Energy Physics (Phenomenology)

Extension of the non-pole technique to the twist-3 gluon distribution contribution in $pp$ collisions

It is known that the Sivers effect is described as a twist-3 effect within the collinear factorization framework and has the characteristic feature that it arises from the pole contribution in a hard parton scattering. The calculation formalism that focuses only on the pole part was established in the 2000s and SSAs were calculated for many processes in $ep$ and $pp$ collisions. On the other hand, there also exists a contribution from twist-3 fragmentation functions regarded as the Collins effect. The Collins effect arises from the nonpole part of the hard scattering and its calculation is formulated in a somewhat different manner from that for the Sivers effect. In recent years, some attempts have been made to revisit the Sivers effect by applying the “nonpole” formalism developed for the Collins effect. For $ep$ collisions, this approach has successfully reproduced the known results. For $pp$ collisions, however, the presence of both initial-state-interaction and final-state-interaction makes the calculation more complicated and the known results have not yet been reproduced. In this paper, we provide a solution to this problem and develop the nonpole method so that it can be applied to the Sivers effect in any process.
Read the full abstract

Artificial Intelligence

Beyond Context Windows: Persistent Discovery Context for Data-Centric Agents

Data-centric agents repeatedly perform a discovery step before planning or execution: identifying the data objects relevant to a task. Yet successful discovery outcomes are typically discarded rather than reused. We introduce persistent discovery context, a lightweight memory layer that stores prior intent-to-object mappings and reuses them to augment future retrieval. Across three structured data environments, persistent discovery context consistently improves retrieval quality over metadata-only search, remains effective with automatically generated memories, and exposes a reproducible interference failure mode. In lexically sparse domains, memory-only retrieval can even outperform metadata-based retrieval. These findings suggest that discovery outcomes constitute a useful form of reusable context for data-centric agents.
Read the full abstract

Fluid Dynamics

Bridging the local and the global: a physically constrained buoyancy–drag model for unified prediction of Rayleigh–Taylor and Richtmyer–Meshkov mixing widths across density ratios

Accurate prediction of the macroscopic width of Rayleigh–Taylor (RT) and Richtmyer–Meshkov (RM) turbulent mixing layers is central to inertial confinement fusion and supernova dynamics. However, bubble–spike asymmetry, density-ratio dependence and unsteady forcing pose a persistent closure challenge: existing low-order buoyancy–drag models struggle to describe different mixing problems accurately with one model and coefficient set. We combine local front dynamics with global mass conservation in separate buoyancy–drag equations for the bubble and spike fronts. Rather than imposing shared or fixed empirical coefficients, the model retains separate inertia, buoyancy and drag coefficients on the two sides and allows them to vary independently with density ratio. Given the bubble-side state scalings, RT/RM similarity relations, a mean-composition profile and endpoint asymptotics jointly constrain all six effective coefficients without case-by-case fitting. A profile-shape parameter $c$ labels distinct internal composition states and is selected a priori from RT spike scaling measured in linear-electric-motor experiments. The model then cross-predicts the RM spike exponent without recalibration to RM spike data and, by construction, recovers low-Atwood-number bubble–spike symmetry and the high-density-ratio free-fall RT-spike and ballistic RM-spike limits. Tests against constant- and variable-acceleration RT mixing, post-impulse RM evolution and Nova laser deceleration show that one closure describes mixing-width evolution across density ratios and acceleration histories without case-specific retuning, while reducing excessive spike growth at high density ratio. This physically interpretable, asymptotically consistent framework enables cross-problem prediction of wide-density-ratio RT and post-impulse RM mixing.
Read the full abstract

All authors You-Sheng Zhang, Ya-Feng Li, Meng-Juan Xiao, Yu-Hui Wang

Cryptography and Security

Stored Is Not Supported: Typed Provenance and Assertion Guardrails for Persistent AI Agents

Persistent AI agents construct autobiographical state through reflection, retrieval, and consolidation. Persistence changes availability, not epistemic standing: stored or retrieved material is not thereby supported. Untrusted inputs, prompt injections, and model inferences can therefore enter persistent state and later be presented as agent history or user commitments. We specify typed provenance and assertion guardrails for autobiographical assertion boundedness, a system-relative release property requiring governed statements about the agent, user, or named relationships to satisfy accepted-evidence, temporal-validity, and disclosure policies. A typed provenance graph separates origin, dependency lineage, epistemic role, validity, and disclosure scope. A resolver evaluates authorized state projections and returns one evidential status, orthogonal conflict, staleness, and withholding flags, and a protected decision witness. A generate-verify-revise mediator then checks candidate semantic units before release and renders policy-authorized status responses. Under explicit assumptions about extraction, predicate correctness, resolution soundness, view declassification, and channel mediation, we prove a conditional assertion-boundedness contract. In an executable suite of 24 hand-authored conformance cases, typed mediation passed none of 19 unsafe opportunities unqualified while preserving all five supported controls. The flat/prior and source-tag comparison rules released 19/19 and 18/19 unsafe candidates, respectively. These results validate the encoded resolver and mediator obligations; they do not constitute an end-to-end evaluation of language models or retrieval systems.
Read the full abstract

Machine Learning

Scalable Bayesian Optimization of Composite Functions for Image-Based Inverse Problems in Materials Characterization

Estimating physical parameters from scientific images is a common inverse problem in materials characterization that often relies on expensive physics-based simulations. In electron microscopy, specimen thickness and crystal mistilt are critical parameters that govern how electrons scatter through the sample, and therefore the accuracy of any atomic-scale structure recovered from it. They are commonly inferred by matching experimental position-averaged convergent-beam electron diffraction (PACBED) patterns to simulated ones, but grid searches scale poorly and neural-network methods require extensive pretraining that may not transfer to new conditions. Here, we propose scalable Bayesian optimization of composite functions (SBOCF), a simulation-efficient method that exploits the known composite structure of the image-matching objective and the intermediate information contained in simulated images. By representing PACBED images with patch-level summaries and two correction terms, SBOCF preserves the original pixel-wise objective while reducing the number of modeled outputs from 24,649 to 11. Under a budget of 50 simulator evaluations, SBOCF outperformed standard Bayesian optimization with expected improvement on synthetic SrTiO3 benchmarks with thick and thin specimens, reducing the median final SSE by up to 290 times in the thick-sample case. On experimental data, SBOCF produced parameter estimates consistent with previously reported values without task-specific pretraining. For a simulated mistilted specimen, using the SBOCF estimates in a downstream ptychographic reconstruction recovered sharp atoms that were otherwise blurred. These results establish SBOCF as a promising approach for inverse problems involving expensive simulators and high-dimensional structured outputs.
Read the full abstract

All authors Dasol Yoon, Poompol Buathong, Chia-Hao Lee, Yujia Zhang, David A. Muller, Peter I. Frazier

Signal Processing

A Survey of Decentralized Physical Infrastructure Network,Research Directions, and Open Challenges

The Decentralized Physical Infrastructure Network (DePIN) represents a transformative paradigm that redefines the construction, operation, and governance of Information and Communication Technology (ICT) infrastructure in the Web 3.0 era. DePIN integrates physical resources, such as networking equipment, storage, and computing power, with decentralized digital governance, forming a self-incentivized ecosystem that is collaboratively built, shared, and governed by the community. It provides a foundational framework for future communication networks, facilitating decentralized edge intelligence, efficient resource sharing, and trustworthy coordination among heterogeneous devices. Focusing on the feasibility of this emerging paradigm, this paper examines the technology landscape in the pre-DePIN era and gaps between existing methodologies and the forthcoming decentralized infrastructure for Web 3.0. It provides a systematic and comprehensive survey of the background, core characteristics, technical architecture, and applications of DePIN across various vertical domains. The paper analyzes the DePIN technology stack from six layers: physical infrastructure, blockchain, interaction, trust, incentive, and application, with special attention to their cross-layer feedback loops, implementation readiness, and deployment limitations. To further bridge conceptual analysis and practical deployment, we propose a DePIN feasibility assessment framework covering technical, governance, and economic dimensions. Moreover, we highlight promising research directions, providing insights and guidance for further exploration and deployment of DePIN.
Read the full abstract

All authors Ming Jiang, Erwu Liu, Xinyu Qu, Wei Ni, Zhongxiang Wei, Ekram Hossain

Data Structures and Algorithms

The Exact Online Threshold for the Asymmetric Binary Perceptron

Let $G\in\mathbb{R}^{M\times N}$ have independent standard Gaussian entries. For a fixed margin $\kappa\in\mathbb{R}$, the asymmetric binary perceptron asks for $\sigma\in\{\pm1\}^N$ such that $G\sigma/\sqrt{N}\ge\kappa\mathbf{1}_M$. We study the online version of this problem, in which the columns of $G$ arrive sequentially and each sign must be chosen irrevocably before future columns are revealed. We determine the exact threshold $\alpha_{\mathrm{on}}(\kappa)$ for every fixed $\kappa$: for $M/N\to\alpha$ with $\alpha<\alpha_{\mathrm{on}}(\kappa)$, there is a deterministic online algorithm, using $O(MN)$ arithmetic operations and polynomial bit complexity, that succeeds with high probability, while for $\alpha>\alpha_{\mathrm{on}}(\kappa)$, no online algorithm succeeds with high probability. The threshold is characterized by a one-dimensional stochastic control problem for Brownian motion. The main difficulty is to upgrade a single-coordinate Brownian limit to simultaneous feasibility of all $M=\Theta(N)$ constraints, which we do with half-line monotonicity and a short final correction block. At zero margin, we give a computer-assisted proof that $0.32747<\alpha_{\mathrm{on}}(0)<0.36664$. In particular, every density below $0.32747$ is achievable online by such an algorithm, more than tripling the best density previously proved attainable by any polynomial-time algorithm, online or offline (the previous bound was $\alpha\le0.1$, due to Li, Schramm, and Zhou). As $\kappa\to+\infty$, the online threshold agrees to first order with the offline storage capacity. As $\kappa\to-\infty$, it has the same asymptotic scale as the best known offline polynomial-time guarantee, while the storage capacity is larger by a factor of order $\kappa^2$.
Read the full abstract

Algebraic Geometry

Groups of Transverse Stratifications and the Hierarchy of Configuration Spaces

We introduce a broad family of groups $G(\Gamma)$ associated with transverse stratifications of moduli spaces. For a wide class of stratifications satisfying a natural local condition (transverse niceness), we construct a homomorphism from the fundamental group of the moduli space to the corresponding group $G(\Gamma)$. This general theorem, which is proved in the companion paper \cite{Manturov2026}, provides a far-reaching generalization of the $G_n^k$-theory. The main goal of the present paper is to discuss numerous examples and situations where the transverse niceness property actually occurs. We describe various moduli spaces and stratifications that are transversely nice. In particular, we study the hierarchy of configuration spaces of points in the plane defined by algebraic curves and establish the transversality property for conic configurations. The general hierarchy problem is formulated in the final section.
Read the full abstract

Computation and Language

AI agents reshape consensus formation in human groups

As large language model (LLM) agents shift from tools to participants in human groups, a fundamental question for collective behavior is how their growing presence reshapes consensus formation. Here we study mixed human-AI groups in a collaborative description game, in which shared conventions emerge through repeated rounds of random pairwise communication. Varying the proportions of LLM agents, we identify three distinct regimes of consensus formation: low agent proportions facilitate human-led consensus, intermediate proportions disrupt convergence, and high proportions restore strong consensus while shifting it toward agent-led conventions. Crucially, these regimes differ not only in the strength of convergence, but also in the semantic grounding and communicative form of the resulting consensus: human-led consensus is more concrete, holistic, and grounded in shared real-world analogies, whereas agent-led consensus is more abstract, less information-dense, and more geometrically segmented. Mechanistically, agent influence arises from a shared linguistic prior that places agents near one another in the expression space, combined with relatively stable expression choices across rounds; humans initially resist adopting expressions from partners perceived as AI but gradually yield to conformity pressure. These findings provide evidence that AI composition can shape the emergence, content, and perceived legitimacy of group norms, making agent proportion and transparency important design variables for human-AI systems.
Read the full abstract

All authors Lin Chen, Ziyi Liu, Xia Hu, Yong Li

Fluid Dynamics

Reynolds-number evolution of wall-pressure statistics beneath canonical turbulent boundary layers

This study investigates the Reynolds-number (Re_\tau) evolution of wall-pressure statistics beneath zero-pressure-gradient TBLs, and links their logarithmic variation to the increasingly energetic large-scale motions in the logarithmic region. The wall-pressure skewness is found to become more negative with increasing Re_\tau, owing to increasing contributions from large-scale wall-pressure fluctuations (that are negatively skewed) and their nonlinear interaction with the statistically invariant inner-scale fluctuations (that are positively skewed). The analysis draws on new, well-resolved simultaneous measurements of wall pressure and streamwise velocity spanning 5000 < Re_\tau < 11300 in the Melbourne tunnel, atmospheric surface-layer measurements at Re_\tau = O(10^6) and a published simulation dataset at Re_\tau = O(10^3). Particular attention is paid to the principal experimental limitations affecting wall-pressure statistics: spatial resolution, Helmholtz resonance, facility noise and statistical convergence. Helmholtz resonance is shown to contaminate inner-scale wall-pressure contributions even after conventional corrections, and reliable estimation of skewness is found to require acquisition durations of O(10^5) eddy-turnover times or longer. The inner-scaled wall pressure spectrum is Reynolds-number invariant over the small-scale regime, in contrast to turbulent channel and pipe flows, whereas at intermediate and large scales it grows substantially with Re_\tau, consistent with these internal flows. Linear and quadratic velocity–wall-pressure coherence link these intermediate- and large-scale contributions to two dynamically distinct coherent structures: the self-similar attached-eddy hierarchy and turbulent superstructures, respectively. These analyses establish the connection between the inertial region and the log variation of wall-pressure variance and skewness with Re_\tau
Read the full abstract

All authors Rahul Deshpande, Balaraman Panneerselvam, Vijaya R. R. Gudla, Joe Klewicki, Ivan Marusic

Information Theory

Instance Optimal Sparse Recovery from Nonlinear Observations: A Unified Framework

This paper develops a unified framework for instance optimal sparse recovery from nonlinear observations. The main ingredient is a signal-dependent restricted approximate invertibility condition (RAIC) of some gradient, which leads to the instance optimality of iterative hard thresholding. Under Gaussian designs, we apply the proposed framework to phaseless, one-bit, and ReLU measurements, which correspond to the problems of sparse phase retrieval, one-bit compressed sensing, and sparse ReLU regression, respectively. For sparse phase retrieval, we propose a variant of thresholded amplitude flow and show its instance optimality under $O(s^3)$ measurements (up to logarithmic factors), where $s$ is the sparsity level. To our best knowledge, this is the first instance optimal efficient algorithm for sparse phase retrieval and complements Gao, Wang and Xu (2016) that achieved this via a computationally intractable program. In one-bit compressed sensing, we establish the instance optimality of normalized binary iterative hard thresholding and strengthen the recent result of Matsumoto and Mazumdar (2024). In sparse ReLU regression, it is shown that a slight variant of the algorithm in Soltanolkotabi (2017) is instance optimal. Moreover, $(\ell_2,\ell_2)$ non-uniform instance optimal guarantees are obtained for these problems. The analysis is built upon a number of high-dimensional concentration bounds, including bounds on restricted eigenvalues and a novel instance-dependent hyperplane tessellation result.
Read the full abstract

Analysis of PDEs

Time-asymptotic stability of generic Riemann solutions for the system of heat-conductive ideal gas without viscosity

This paper is concerned with the time-asymptotic stability of the generic Riemann solution for the one-dimensional system of heat-conductive ideal gas without viscosity, where the generic Riemann solution consists of a shock, a contact discontinuity, and a rarefaction wave. We prove that, as time tends to infinity, the solution of the non-viscous and heat-conductive ideal gas system converges uniformly to a composite wave composed of rarefaction wave, viscous contact wave, and viscous shock wave with a time-dependent shift. Motivated by the recent work of Kang-Vasseur-Wang [Arch. Ration. Mech. Anal. 249: 42 (2025)], we overcome the difficulties arising from the concurrence of shock and rarefaction waves for the partially dissipative hyperbolic-parabolic system with dissipation acting only on a single variable. More notably, the absence of velocity dissipation gives rise to new and intrinsic difficulties when handling the terms associated with the density and velocity. To resolve this, we exploit the precise structure of the governing equations and the additional properties of shock waves. Furthermore, we utilize the wave structure of the system without viscosity and perform separate space-time estimates for the density and velocity.
Read the full abstract

All authors Yilin Guo, Lin-An Li, Jiahong Wu, Xiaojing Xu

Computer Vision

Synergistic Information Disentanglement for Omni-modal Slide Representation Learning in Computational Pathology

In computational pathology (CPath), developing omni-modal self-supervised learning (SSL) models that integrate histology, genomics, and clinical reports enables transferable representation learning for whole slide images (WSIs). Existing approaches implicitly force heterogeneous modalities into a uniform latent space by contrastive alignment, causing modality collapse where unique, synergistic diagnostic signals (termed as $\mathrm{\Phi}$) are discarded in favor of trivial redundancy. We hypothesize that the strongest task-agnostic SSL training signal stems from distilling the synergistic interactions over merely aligning shared redundancy. To this end, we introduce \textsc{$\mathrm{\Phi}$-Omni}, a synergistic information disentanglement framework grounded in Partial Information Decomposition (PID) theory for slide representation learning. Unlike standard contrastive approaches, \textsc{$\mathrm{\Phi}$-Omni} employs a Synergistic Information Bottleneck (SIB) regulated by the proposed $\mathrm{\Phi}\text{ID}$ objective, which explicitly suppresses marginal redundancy while maximizing irreducible synergy, thereby distilling high-order cross-modal interactions. Following pretraining on breast ($n$=1031) and lung ($n$=919) cohorts, \textsc{$\mathrm{\Phi}$-Omni} demonstrates superior few-shot performance across five independent external datasets spanning eight tasks compared to supervised and SSL baselines. Source code is available here.
Read the full abstract

All authors Mingxin Liu, Chengfei Cai, Anwen Lu, Pengbo Xu, Jun Li, Jinze Li, Depin Chen, Jun Xu

Instrumentation and Methods for Astrophysics

Optimization of adaptive optics correction during observations: Algorithms and system parameters identification in closed-loop

The adaptive optics (AO) on the European Extremely Large Telescope, as well as earlier pathfinders like the Adaptive Optics Facility, at the Very Large Telescope in 2014, will no longer be stationary systems. AO is no longer isolated on a bench; some elements are directly in the optical train of the telescope, suffering environment and constrains changes during the observations. To guarantee good performance at any observing time, we investigate a self-calibration strategy. We focus here on one of the most challenging aspects: the identification of system parameters during closed-loop observations without introducing any additional disturbance. Such problem is known in the identification theory to be difficult to solve. We have recently presented (Béchet et al., AO4ELT2 Conference, 2011) an identification method for this, with promising results obtained in simulations. To consolidate these advances, we come back in the present paper to the equations and provide a theoretical analysis to justify the choice of the algorithm. We highlight the benefit of using incremental data and commands to decorrelate the disturbance. We also present 2 implementations of the method, currently studied at the European Southern Observatory.
Read the full abstract

Artificial Intelligence

Semantic Signal-Assisted Inspection and Recovery Allocation in Reverse Logistics

Reverse-logistics operators often decide how to inspect and route returned assets before their condition is fully observed, while full inspection consumes scarce labor. Semantic Signal-Assisted Decision Support converts return notes into a condition factor and a signal-quality score that guide inspection depth and recovery allocation under shared labor capacity. We evaluate the framework in three synthetic benchmark scenarios spanning information technology decommissioning, aircraft maintenance, and consumer-electronics returns. Across 30 paired simulation seeds, the keyword implementation improves net recovery value relative to a structured-feature comparator with noisy full inspection while reducing inspection cost in all three scenarios. A risk-blind comparator that skips inspection altogether still records higher value under the benchmark's purely economic objective. At matched inspection cost, score-guided targeting adds 53.9 thousand United States dollars per batch in the aircraft scenario but has little economic effect in the other two configurations; phrase and large language model extractors provide further gains in the aircraft scenario. These results show how narrative evidence can support inspection allocation before recovery decisions are made.
Read the full abstract

All authors Jiani He, Dingyan Shang, Yihua Xu, Shiqi Huang, Yan Lyu, Jize Li, Shangjing Tang

Computation and Language

text2ql: Multi-Target Natural Language Querying via a Language-Agnostic Intermediate Representation

Natural language interfaces to databases have traditionally suffered from three structural limitations: exclusive targeting of relational SQL, unconditional dependence on large language model (LLM) inference at query time, and absence of any runtime signal when generated queries are semantically incorrect. This paper presents text2ql, an open-source Python framework that addresses all three limitations through a language-agnostic Intermediate Representation (QueryIR) and a pluggable renderer architecture. A single seven-stage detection pipeline serves both SQL and GraphQL targets; a zero-LLM deterministic mode delivers 100% execution accuracy at a median latency of 3.2 ms with no API cost; and every generated query carries a runtime confidence score in [0.15, 0.97] computed from an additive signal model. Evaluated on 50-query random samples from the Spider and BIRD benchmarks (indicative results; full-set evaluation is planned), the LLM-backed mode achieves 62-70% exact match and 84-91% execution accuracy; the deterministic mode achieves 100% execution accuracy with zero parse errors across all 100 test cases. An ablation study isolates schema-aware prompting as the dominant accuracy lever, contributing +18.4 percentage points of exact-match gain over the schema-free baseline on both benchmarks. text2ql is publicly available at https://pypi.org/project/text2ql/ under the Apache 2.0 license.
Read the full abstract

Materials Science

EuIn2Sb2: epitaxially stabilized axion insulator candidate with strong spin-orbit coupling

Eu triangular lattice layer compounds described by the general formula EuA2X2 have attracted growing attention due to a wide range of layered crystal structures and corresponding magnetic topological phases, arising from the coexistence of large Eu magnetic moments and energetically inverted A and X bands. Among the EuA2X2 family, EuIn2Sb2 has been predicted to host robust axion insulator and higher-order topological insulator states by stronger spin-orbit coupling but has not been experimentally realized, including the structural identification. Here we report the epitaxial stabilization of EuIn2Sb2 by adopting the molecular beam epitaxy technique. Structural characterization reveals that EuIn2Sb2 films are based on a unique In-on-In stacking, forming a new trilayer structure distinct from previously predicted structures in the EuA2X2 family. In addition to the in-plane antiferromagnetic ordering at TN=10.5K, the stronger spin-orbit coupling and enhanced In-In hybridization in EuIn2Sb2 make it an ideal candidate of axion insulator and higher-order topological insulator phases. These results establish EuIn2Sb2 as a new member of the EuA2X2 family and offer an advanced platform for exploring these magnetic topological phases.
Read the full abstract

All authors Hsiang Lee, Shinichi Nishihaya, Markus Kriener, Ayano Nakamura, Tadashi Yoneda, Yuki Deguchi, Makuro Goto, Masaki Uchida

Quantum Physics

Logarithmic-scale variational quantum eigensolver for off-lattice protein structure prediction in continuous torsional angle space

Classical and current quantum approaches to protein structure prediction (QPSP) face limitations, notably massive qubit requirements restricting near-term models to simplistic on-lattice simulations. We propose a logarithmic-scale variational quantum eigensolver (VQE) that reduces qubit requirements for N torsional degrees of freedom to O(log2N), enabling off-lattice, all-atom simulations. Our architecture extracts molecular torsions from relative phases in statevector simulations. On quantum hardware, a decoder maps the empirical cumulative distribution function (CDF) from basis-state probabilities to bounded torsional variables. These feed a classical algorithm to build heavy-atom coordinates. We use an EfficientSU2 ansatz and multi-stage relaxation to mitigate barren plateaus. Structures are evaluated via a custom hybrid quantum-classical Hamiltonian, alongside Rosetta and OpenMM benchmarks. Evaluation on chignolin and Trp-cage yielded native-like conformations. Chignolin reached a 0.623 Å C{\alpha} RMSD in retained snapshots and 1.199 Å in final models; Trp-cage achieved a 2.501 Å RMSD among snapshots (3.512 Å in final models). Execution on IBM processors (ibm_cleveland, ibm_miami) successfully recovered native-like structures with a best RMSD of 1.758 Å. The custom energy function performed best overall, though energy-ranking imbalances persisted across sampled landscapes for all functions. This introduces the first all-atom, continuous-space quantum algorithm for QPSP. By converting physical qubit constraints into circuit depth constraints, it proves high-resolution prediction is feasible with exponentially fewer qubits. Despite current limits like computational overhead and energy function sensitivity, it establishes a scalable foundation for hybrid quantum biophysics.
Read the full abstract

All authors Fabio Cumbo, Bryan Raubenolt, Varun Puram, Natalie Katzenmeyer, Jayadev Joshi, Daniel Blankenberg

Number Theory

Summing the reciprocal of the polynomial appearing in Fermat's Last Theorem

Consider the polynomial $f=x^N+y^N-z^N$ where $x,\,y$ and $z$ are positive integers and $N \ge 3$ is an integer. By Fermat's Last Theorem, $f$ is never zero so that its reciprocal, $1/f$, has no singularities. We therefore study the finite sum of the reciprocal: $S(m,N)=\sum_{x=1}^m\sum_{y=1}^m\sum_{z=1}^{m}\frac{1}{f}$. The terms $1/f$ can be positive, negative and their magnitude is less than unity. A key observation is that $S(m,N)$ can be split into two convenient parts: a dominant contribution $D(m,N)$ that has a simple analytical expression and a remainder $R(m,N)$ which is more complicated but negligible compared to $D(m,N)$. Therefore, $S(m,N)$ is almost identical to $D(m,N)$. The analytical expression for $D(m,N)$ is $(2\,m-1)\,H_m^{(N)}$ where $H_m^{(N)}=\sum_{x=1}^m\frac{1}{x^N}$ approaches quickly the Riemann zeta function $\zeta(N)$ as $m$ increases. Therefore, the original sum $S(m,N)$ has a simple expression: it is basically linear in $m$ with slope equal to $2\,\zeta(N)$. Its linear behavior is not an asymptotic result; plots of $S(m,N)$ vs. $m$ for different $N$ show a straight line starting at $m=1$. $S(m,N)$ deviates slightly from a straight line over a small interval $8\le m\le 12$ for the case $N=3$. This slight deviation is due to Fermat near misses where $x^3+y^3-z^3=\pm 1$ (for $z\ne x$ and $z\ne y$); these create a jump in the remainder $R(m,3)$ at $m=9$. We make a numerical and analytical study of the remainder $R(m,N)$. From the numerical analysis, $R(m,N)$ converges for $N\ge 4$ but it was harder to tell whether $N=3$ converged. An analytical study based on a comparison of $R(m,N)$ to its Cauchy principal value integral, shows that $R(m,3)$ likely diverges logarithmically. It also shows that $R(m,N)$ converges for $N\ge 4$ in agreement with the numerical analysis. We discuss in the conclusion some interesting questions for future investigation.
Read the full abstract

Computer Vision

Disease Burden over Skin Tone: Decomposing the Dermatology-AI Generalization Gap

Dermatology artificial intelligence (AI) models are predominantly trained on light-skinned, cancer-focused image collections, yet they are increasingly proposed for deployment in resource-constrained settings where patients differ from training populations along two confounded axes: skin tone and disease distribution. We investigate whether poor generalization is primarily caused by skin-tone underrepresentation or disease-distribution shift. We evaluate a cancer-trained baseline (ResNet-50 fine-tuned on HAM10000 and ISIC 2019), two dermatology foundation models (DermLIP and MONET), and a general-purpose vision model (DINOv3) as frozen feature extractors. Models are evaluated on a tone-stratified disease-matched dataset (Diverse Dermatology Images, DDI) and a disease-shifted tone-diverse dataset (Skin Condition Image Network, SCIN). Our results show that disease-distribution shift contributes more than skin tone in the evaluated settings. The cancer baseline decreases from 0.62 to 0.21 balanced accuracy when transferred to unfamiliar clinical conditions, while the within-disease skin-tone gap is smaller (0.10-0.18) and inconsistent. Label-free representation analysis shows that this failure reflects a representational limitation rather than only missing output labels: cancer-specialized features poorly cluster unfamiliar conditions (kNN purity lift +0.06 over chance), whereas dermatology-pretrained features retain stronger transferable structure (+0.23). Finally, we show that representation quality predicts recoverable performance under lightweight adaptation. Starting from dermatology foundation models, approximately ten labeled examples per clinical category recover most attainable performance. We release the evaluation protocol and code to support reproducible auditing of dermatology AI generalization.
Read the full abstract

All authors Nirajan Kunwor, Sanjaya Poudel, Quoc-Huy Trinh, Jahidul Arafat, Sunil Kumar Gaire

Machine Learning

A Computational Comparison of Fourier Spectral Differentiation and Spatial Automatic Differentiation in Periodic Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) commonly evaluate the spatial derivatives appearing in partial differential equation residuals using automatic differentiation (AD), whose computational and memory costs can become substantial when multiple or high-order derivatives are required. We perform a controlled comparison of spatial AD and Fourier spectral differentiation in periodic physical-space PINNs. Within each paired experiment, the neural representation, temporal differentiation, optimizer, sampling procedure, and training schedule are held fixed, so that the two cases differ only in the spatial differentiation procedure. For the Fourier variant, network outputs are evaluated on a uniform periodic grid and transformed to Fourier space, where spatial derivatives are obtained through spectral multiplication and the same Fourier coefficients are reused across derivative orders. We compare the two procedures in standard PINNs for the Allen–Cahn and Korteweg–de Vries equations and in Causal PINNs for the Allen–Cahn, Korteweg–de Vries, and Kuramoto–Sivashinsky equations. Across these five equation–framework settings, Fourier differentiation yields mean paired end-to-end training speedups ranging from $2.90\times$ to $18.52\times$ and reduces peak allocated graphics processing unit (GPU) memory by $68.7\%$–$94.1\%$. The final relative $L_2$ errors remain of the same order, with neither differentiation procedure showing a consistent accuracy advantage. For the one-dimensional periodic benchmarks considered here, Fourier spectral differentiation therefore provides substantially lower training time and memory usage than spatial AD while retaining comparable solution error, at the cost of requiring a uniform structured spatial grid.
Read the full abstract

Distributed Computing

MeanField Surrogate Modeling for Scalable Runtime Scheduling of Concurrent Heterogeneous AI Inference on Shared GPUs

Deploying heterogeneous AI models concurrently on a shared GPU introduces resource contention that complicates runtime scheduling. While surrogate models avoid costly online benchmarking, their profiling requirements typically grow combinatorially with the number of co-running models, limiting scalability. We propose a MeanField surrogate that predicts per-model performance from local configuration and aggregate GPU state rather than explicitly modeling all joint interactions. Experiments on concurrent LLM and vision workloads across $N \in \{2,3,4,5,6\}$ show high predictive accuracy ($R^2 \approx 0.96$) with an empirical sample budget that grows approximately linearly in $N$, in contrast to the combinatorial cost of fully joint profiling. Integrated into a genetic algorithm scheduler, the surrogate scales to an $N=5$ problem with 78,732 feasible joint configurations, remaining within 0.10% of the exhaustive search with zero SLA violations across eight dynamic workload scenarios, while complete online GA decisions take 26 ms median, about $5\times$ faster than exhaustive surrogate search.
Read the full abstract