Peer Revenue Equity in science

Papers

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

All of arXiv

Friday 6 November

Complex Variables

Optimal domains of Volterra-type operators on Bergman spaces

We consider bounded Volterra-type operators $T_g$ from the weighted Bergman space $A^q_{\beta}$ to $A^p_{\alpha}$. The holomorphic optimal domain $[T_{g},A^p_{\alpha}]$ is introduced and shown to be a quasi-Banach space. We show that the intersection of the optimal domains is equal to $A^q_{\beta}$ if $p\le q$, while if $p>q$, it is strictly larger than $A^q_{\beta}$. Moreover, $A^q_{\beta} \subsetneq[T_{g},A^p_{\alpha}]$ is established for $p\neq q$. If $p=q$, by contrast, $A^p_{\beta}$ may coincide with $[T_{g},A^p_{\alpha}]$. The multiplier algebras and strict inclusions with respect to varying parameters are established. For example, we determine exactly when $[T_{g},A^s_{\gamma}] \subset [T_{g},A^p_{\alpha}]$ for some $g\in \operatorname{Hol}(\mathbb{D})$. We further study the set $V_g^p$ and characterize the coincidence of the holomorphic optimal domains.
Read the full abstract

Robotics

HomeLoop: A Simulation Framework and Benchmark for Robot-IoT Coordination in Everyday Domestic Assistance and Safety

Recent progress in embodied robotics is bringing robots closer to deployment in homes, as vision-language models (VLMs) are demonstrating significant advances in reasoning, making VLM-driven robots promising home assistant agents. These agents will often be introduced into homes that already have cameras, Internet of Things (IoT) devices, and residents with wearables. To fulfill their potential, they need to integrate with existing smart infrastructure and reason through ambiguous scenarios: a smoke alarm may come from a pan about to ignite, a pan already on fire, a resident cooking or a faulty sensor, and each calls for a different response. To that end, we introduce HomeLoop, a simulation framework and benchmark built on Isaac Sim that combines robot bodies, networked devices and wearables, simulated residents and visitors, faulty sensors, and hazards that evolve over time. Our benchmark is composed of 20 tasks across five families grounded in real-world household hazards and needs, and designed so that the same alert can have different causes. A runtime monitor scores each episode with temporal-logic checks on the final outcome and the actions along the way, and every action carries a predefined cost. We evaluate six open-weight and proprietary VLMs. The best achieves a family-balanced score of 61.9%, and across models we see a pattern of over-reaction: failed episodes involving prohibited interventions outnumber those involving missed required interventions by more than two to one, with models often shutting off the stove before investigating the alarm's cause. HomeLoop is an extensible foundation intended for use in research on robot-IoT coordination, long-term memory, proactive assistance, and privacy-safety trade-offs. HomeLoop will be released with 2,100 home layouts that can be automatically furnished in diverse configurations, five robot bodies and 20 networked device types.
Read the full abstract

All authors Ramy ElMallah, Marc Bishara, Hyegang Kim, Yuxi Zhang, Chongyu Zhu, Chi-Guhn Lee

Quantum Physics

Time- and phase-resolved emulation of sign-selective counterdiabatic dynamics in a superconducting qutrit

In cyclic three-level systems, phase-sensitive interference can convert a coupling-sign reversal into a population contrast, providing a controllable analogue of handedness. We emulate this mechanism in a superconducting transmon qutrit using two resonant single-photon drives and a phase-controlled effective two-photon drive. Effective loop phases \(\Phi_L=+\pi/2\) and \(\Phi_R=-\pi/2\), separated by \(\pi\), reverse the auxiliary contribution to the dark–bright coupling, canceling the intrinsic nonadiabatic term for \(L\) and reinforcing it for \(R\). Time-resolved trajectories reveal dark-pathway transfer for \(L\) and bright-manifold participation for \(R\). Measured late-time averages give a target-state population contrast of 0.929. Loop-phase measurements establish phase selectivity, while pulse-area and common-mode amplitude scans characterize the operating regime and control sensitivity. These results provide direct time- and phase-domain access to sign-selective counterdiabatic interference in a programmable cyclic qutrit.
Read the full abstract

All authors Yiwen Li, Luojia Xu, Xinsheng Tan, Yang Yu

Machine Learning

SSF-Net: A Spatial-Spectral Fusion Network for Accurate PDE Operator Learning

Machine learning methods for solving partial differential equations (PDEs) significantly improve computational efficiency over traditional numerical solvers such as finite element method (FEM). However, PDE dynamics inherently span both local and global dynamics: disturbances propagate and decay gradually through the domain, while global coupling effects can influence the whole system simultaneously. Capturing both types of dynamics with a low complexity model remains a key challenge. To address this, we propose the Spatial-Spectral Fusion Network (SSF-Net), built around a dual-branch Spatial-Spectral Fusion (SSF) block comprising a spatial branch and a spectral branch. The spatial branch employs a multi-directional Mamba, which mitigates its inherent causal, single-direction scanning bias and capture local interactions with reduced directional bias at linear complexity. The spectral branch leverages a Fourier transform with a block-diagonal MLP to efficiently capture global spectral features with reduced parameter count and complexity. The SSF block is embedded in a multiresolution convolutional encoder-decoder, which extracts multi-scale features. We evaluate SSF-Net on a comprehensive PDE benchmark spanning multiple task types and assess its long-horizon stability. Compared with attention-based and convolution-based baselines, SSF-Net achieves lower prediction error with fewer parameters while retaining near-linear asymptotic complexity in the number of spatial tokens.
Read the full abstract

Robotics

Polarization calibration of fisheye cameras using skylight in the wild

Skylight polarization provides directional cues for robotic navigation, but fisheye lens effects and analyzer imperfections can degrade polarization measurements across the field of view. We present a joint lens-analyzer calibration method that explicitly models incidence-dependent lens transmission together with analyzer gains and angular offsets in a unified eight-parameter formulation. Given known geometric calibration, rotations between camera frames can be estimated using visual features and cross-frame directional polarized light consistency can be used for polarization calibration. Therefore, our calibration approach requires only a short handheld outdoor sequence under natural skylight, without dedicated polarization calibration equipment or a predefined sky polarization model. On real-world outdoor sequences, the proposed method achieves a weighted cross-frame angle-of-linear-polarization (AoLP) mean absolute error below 0.5{\deg} across all incidence angles, with a 73.4% reduction in the 80{\deg}-90{\deg} incidence angle region, and generalizes to an independent sequence with an overall mean absolute error of 0.466{\deg} without refitting. Preliminary solar-azimuth experiments using a camera mounted on a quadruped robot further show improved alignment with the visible solar center, supporting the potential of calibrated AoLP measurements for relative-heading estimation and future polarization–inertial navigation. The calibration software tool will be released upon acceptance of this paper.
Read the full abstract

All authors Shuo Wang, Shiqi Che, Dianze Dong, Wenxin Liu

Numerical Analysis

Task-Contribution Dynamics of Adam in Physics-Informed Neural Networks: Slow-Manifold Analysis and Gradient Modulation

Physics-informed neural networks (PINNs) often suffer from imbalanced optimization of competing loss components, particularly when approximating solutions with high-frequency or multiscale features. Although existing balancing strategies have proven effective in alleviating this difficulty, the mechanisms governing the evolution of individual task contributions to Adam's parameter updates remain insufficiently understood. We introduce task-wise indicators to characterize this evolution and identify three distinct training regimes: smooth variation, intermittent spikes, and persistent oscillations. Focusing on the early smooth regime, we combine discrete-time analysis with a continuous-time singular perturbation framework under the full-batch setting and the parameter regime $(1-\beta_2)/(1-\beta_1)\ll1$. A first-order approximation of the locally attracting slow manifold reveals two factors governing the step-to-step redistribution of task contributions: curvature-related alignment disparities and differences in the coordinatewise distributions of preconditioned task gradients. Both factors involve Adam's shared second-moment estimate. Inspired by AdaTask, we propose Gradient-Modulated Adam (GMAdam), which modulates incoming task gradients using task-specific second-moment estimates. Numerical experiments on high-frequency Poisson and wave equations, the Allen-Cahn equation, and two-dimensional lid-driven cavity flow ($Re=100,400$) demonstrate consistent improvements in predictive accuracy over AdaTask at comparable computational cost.
Read the full abstract

Machine Learning (Statistics)

Geometry-Aware Score-Repellent Monte Carlo

The recently proposed Score-Repellent Monte Carlo (SRMC) method improves MCMC sampling by tilting the target density $\pi$ in $\mathbb{R}^d$ with the factor $\exp(-\alpha\,\theta_n^\top s(x))$, where the score function $s(x) = \nabla_x \log \pi(x)$, and $\theta_n$ is the score average from past samples. However, a single scalar repellence strength $\alpha\ge0$ cannot handle both the steep and the flat directions in anisotropic energy landscapes at once. We introduce Geometry-Aware SRMC (GA-SRMC), which generalizes the Euclidean alignment to $\theta^\top M s(x)$ for any symmetric positive-definite matrix $M$. A stochastic-approximation central limit theorem under a trace constraint on $M$ identifies the trace-normalized inverse Fisher $M_{\rm opt}\propto S^{-1}$ as the unique minimizer of the worst-direction covariance bound, elevating the Fisher information $S=\text{Cov}_\pi(s,s)$ from a proposal-side preconditioner (e.g., FisherMALA, natural gradient) to the bound-optimal repulsion-side matrix, with a single online estimate $\widehat S_n$ serving both roles at no additional computational cost. The full-Fisher online metric $M_n$ converges almost surely to $M_{\rm opt}$ and satisfies a CLT, providing a theoretical guarantee for the online implementation of the GA-SRMC framework. On an $8$-target simulation spanning the condition number $\kappa\in[1.3,100]$ and dimension $d\in[2,785]$, GA-SRMC dominates SRMC in ESS by $\boldsymbol{4.3\times}$ on anisotropic Gaussian ($\kappa=100$) and $\boldsymbol{6.8\times}$ on MNIST under the MALA baseline, by $\boldsymbol{16\text{--}32\%}$ on three well-conditioned logistic-regression posteriors, and by up to $\boldsymbol{+48\%}$ under FisherMALA; gains peak on the most anisotropic targets, and posterior-mean MSE is lower than with SRMC in all $16$ full-Fisher implementations.
Read the full abstract

Artificial Intelligence

Toward LLM Self-Play Under Data Scarcity: Effectiveness, Generalizability, and Trustworthiness

The rapid development of large language models (LLMs) has been driven largely by scaling model size and human-curated training data, yet this paradigm is increasingly constrained by data scarcity and diminishing returns from human supervision. Self-play offers a promising pathway for autonomous data generation and improvement, but current frameworks remain unsatisfactory due to a lack of a data-centric perspective, limited cross-domain generalization, and growing trustworthiness and security concerns. In this paper, we argue that real-world LLM self-play systems should achieve effectiveness, generalizability, and trustworthiness under data scarcity. To address these challenges, we outline three blue-sky research directions for future self-play systems: (1) data-centric self-play for effective experience generation, (2) generalizable self-play for open-world and unverifiable tasks, and (3) trustworthy self-play for reliable self-evolution. We further discuss research opportunities and solutions within each direction. This paradigm shift will establish a sustainable and safe AI ecosystem by transforming finite human knowledge into inexhaustible high-quality experience through self-play architectures.
Read the full abstract

All authors Zhiyin Yu, Yuchen Mou, Bingzhe Wu, Zhonghai Wu, Shiyang Feng, Xiao Luo

Quantum Physics

Shared Lookups and Lazy Residues: Fewer Toffoli Gates for Factoring RSA-2048

Gidney's 2025 construction for factoring RSA-2048 with few logical qubits spends most of its Toffoli gates on discrete-log table lookups and residue uncomputation. We reduce both costs. Shared lookups return entries for several primes of a residue number system at once because they have the same address. Retained intermediate residues remove recomputation during cleanup and enable lazy arithmetic with a single final modular reduction. We prove output preservation and estimate resources with an open benchmark built on the reference simulator. Against the best reference setting within the same logical-qubit budget, using the published model for the reference and benchmark counts for our circuits, we obtain $15\%$ fewer Toffolis at $1{,}461$ qubits ($5.08 \times 10^{9}$ per factoring) and $46\%$ fewer at $1{,}806$ qubits ($2.55 \times 10^{9}$). A matched ablation finds $28.8\%$ fewer Toffolis from three-prime sharing at fixed window and checkpoint settings. Architecture-specific timing determines how much becomes a runtime saving.
Read the full abstract

Quantum Physics

Quantum Fluctuations in Noether's Theorem

When deriving the expression for the Noether's current one obtains a term that is an integral over the Euler-Lagrange equation. This term vanishes along the classical path and so does not contribute to the current. However, quantum particles do not necessarily follow the classical path; this concept is the basis of Feynman's path integral quantization formalism. Thus, for a quantum system, the Euler-Lagrange term does not vanish and, hence, corresponds to a quantum fluctuation of the Noether's current owing to the system deviating from the classical path. Here, we compute this quantum fluctuation from Noether's theorem and the path integral formalism. We find that, for a time invariant system, the expression for the quantum fluctuations is give by minus the functional expression for the Hamiltonian, $\Delta E = -H(y)$, evaluated at the distance from the classical path, $y$. As an example we evaluate quantum energy fluctuations of a single Harmonic oscillator. In the short time limit we recover the energy-time uncertainty relation and in the long time limit we recover the zero point energy of the quantum harmonic oscillator.
Read the full abstract

Computation and Language

From Procedures to Permissions: Learning Precise Subagent Authorization from Agent Skills

As language agents tackle increasingly complex tasks, they increasingly rely on multiple agents to coordinate specialized work, requiring decisions not only about what each subagent should do but also which tools and workspace resources it should access. Existing agent systems provide mechanisms for controlling tools and execution environments, but do not directly specify the resources required by each delegated subtask. Moreover, task outcomes and execution trajectories alone provide limited supervision for determining whether a subagent has received appropriate permissions. We observe that Agent Skills encode procedural knowledge describing the operations, tools, and files required at each step, providing a natural source of supervision for authorization. Building on this observation, we synthesize tasks and step-by-step procedures from Skills, execute them through subagent delegation in interactive environments, and use verified trajectories to construct training data for task execution and authorization. On ClawArena-Team, our method improves tool permission precision by 13.4 percentage points and workspace permission precision by 24.1 percentage points, while increasing overall task success by 11.6 percentage points for Qwen3.5-35B-A3B. On Terminal-Bench 2.1 and SkillsBench, our method remains competitive with existing agent data synthesis approaches. These results show that the procedural knowledge embedded in Agent Skills provides a source of supervision for precise subagent orchestration that task outcomes alone cannot provide.
Read the full abstract

All authors Wenkai Yu, Ning Ding, Wentao Zhang, Yehui Tang

Computer Vision

MotionCore: A Latent Neural Simulator for Video Generation with Rigid-Body Dynamics

Video generation models can produce highly realistic videos while generating physically inconsistent object motion. This limitation suggests that visual modeling alone may not provide a transferable representation of physical dynamics. We introduce MotionCore, a framework for learning physical dynamics via a learnable partial differential equation solver and focuses on rigid-body motion. We introduce a latent neural simulator learning to predict the evolution of object-level physical states from observed motion. Its predictions are then converted into spatially aligned latent controls for a controllable video diffusion model. The generator serves as a visual interface: it provides appearance and scene detail, while the learned dynamics model specifies the physical trajectory. To preserve this trajectory through the interface, MotionCore combines a diffusion-matching objective with a low-capacity decodability constraint on object identity and position. We evaluate MotionCore on public physical-consistency benchmarks and on paired data with known trajectories. Results indicate that MotionCore improves physical consistency and trajectory obedience in the evaluated rigid-body settings, while maintaining comparable visual quality. A controlled ablation protocol compares learned dynamics with an analytic prior and studies the choice of latent evolution processor.
Read the full abstract

All authors Zheng Wang, Zhengwei Gong, Yuanzhi Liang, Haibin Huang, Chi Zhang

Statistics Applications

Snow Changes Drought Characterization but Not Drought Predictability: A Leakage-Free Multiscale Study in Ohio

We examine whether accounting for snow accumulation and melt, which precipitation-based drought indices ignore, changes drought characterization and one-month-ahead drought forecasting at three stations in Ohio, USA (1953–2026). The Standardized Precipitation Index (SPI) is compared with the Standardized Snow Melt and Rain Index (SMRI), and both are decomposed with the Haar maximal overlap discrete wavelet transform. Nonlinear autoregressive networks without (NAR) and with (NARX) wavelet inputs are compared over 100 paired repetitions in a leakage-free design (past-only wavelet inputs, untouched test period), together with persistence, climatology, and linear autoregressive benchmarks. Snow changes the 3-month index mainly from December to May (mean absolute SMRI–SPI differences of up to 0.17–0.28). These differences are attributable to the snow-adjusted input rather than to the fitted distribution, are robust to the snow-model parameters, and change the drought classification of 21–32% of the December–May drought months. Wavelet inputs reduce NARX error relative to NAR by 6–10% for the 3-month indices in 92–100% of the repetitions, but not for the 6-month indices. A linear autoregression with 7–12 lags is at least as accurate as two-lag NARX for every series, and networks given six lags gain at most 4.5% from the wavelet inputs and only match the autoregression. The forecast gain therefore reflects the memory of the accumulated index rather than multiscale information, and it is similar for the SPI and the SMRI. Snow matters for drought monitoring in winter and spring, but not for one-month predictability.
Read the full abstract

Distributed Computing

Inside the Trap: A Renewal Law for Prefix-Cache Collapse in LLM Agent Serving

Prefix caching lets an LLM agent's follow-up turns reuse cached KV, but a burst can tip an engine into a trap in which every turn misses and the backlog does not drain, at a load it served before. In registered paired tests on two engines, two GPUs and two models (passed at 9 of 10 rates), a 200-session burst left the engine at zero hits to the end in all 30 runs, while every matched no-burst run stayed good and a session cap prevented the trap in all 30 capped runs. We analyse an idealization of the trapped state exactly: for a fluid FIFO server on which every turn misses, we prove that along FIFO orbits its cycles obey a linear renewal recursion. As is classical, the trap drains if and only if arrivals are below the all-miss capacity; above it, a Perron root fixes the dynamics: every trapped state converges to a known self-similar regime in which the gap between turns grows linearly and the turns served are young and short. For our workload the backlog grows at 0.6 to 0.8 of the excess load and sessions complete faster than the all-miss capacity, a youth premium with turn-count and cost-growth factors, so in the model a trap's measured throughput overstates the recovery threshold. Retrospectively, with nothing fitted to trap data, the theory matches 57 stuck runs: the mean served prompt within 3% in 50 runs, completions above the all-miss capacity in every run, and the pooled gap growth. Blind in outcome, 4 runs trapped by a tool slowdown met 4 of 5 pre-set scores, cut traps persisted, and with a 20-step budget, trapped runs completed below the all-miss capacity, as predicted. Of two exits, refusing arrivals sheds sessions; a session cap restores reuse: in registered tests it rescued 9 of 9 established traps at the same load without shedding, and the engine stayed good after release. We do not prove that the engine converges to the model.
Read the full abstract

Logic

A Stone-Weierstrass Theorem for Polish Spaces, Given by Logic

The Stone-Weierstrass theorem obtains every continuous function on a compact space as a uniform limit of functions built from a separating family by finitely many operations. We extend the theorem to every separable metric space, and in particular to every Polish space, by admitting countable suprema and infima according to weights computed from the construction. For a countable, uniformly bounded separating family whose supremum metric makes the space separable, the functions so built are exactly the bounded Lipschitz functions, and their uniform limits are exactly the bounded uniformly continuous functions. The theorem was discovered, and first proved, in infinitary real-valued logic, where the weights are the a priori bounds of formulas and the uniform approximation is equivalent to an approximate form of Beth's definability theorem, which carries Reggio's correspondence from compact spaces to separable metric spaces. The same bounds draw an exact line between the two forms of definability when the values of the coordinates in the models of the implicit definition form a set separable for the supremum metric: an implicitly defined predicate has an explicit definition precisely when the defining function is Lipschitz, and admits uniform approximation by sentences precisely when that function is uniformly continuous. We also give a purely topological-analytic proof of the theorem.
Read the full abstract

Quantum Algebra

Reflections of Nichols algebras in braided monoidal categories

Let $\C$ be an essentially small abelian rigid braided monoidal category. We prove a reflection theorem for Nichols algebras in $\Ind(\C)$ and show that tuples admitting all reflections define connected semi-Cartan graphs. As applications, we prove that a finite Nichols algebra has a generating object of length at most one in $PSU(2)_k$ for odd $k\ge3$, in $SU(2)_{4m}$ for $m\ge1$, and in the representation categories of the Virasoro minimal models $\mathcal M(2,p)$ for odd $p\ge5$. In all three families, we prove that every finite pre-Nichols algebra is a Nichols algebra.
Read the full abstract

High Energy Physics (Phenomenology)

Spiral momentum spectra and spin asymmetry in vacuum pair production driven by phase-modulated elliptically polarized electric fields

We study vacuum electron–positron pair production in phase-modulated elliptically polarized electric fields using the quantum kinetic equations. We analyze how the modulation amplitude, frequency, ellipticity, and modulation form affect the spin-resolved momentum spectra and the spin asymmetry. Cosine-type modulation can deform ring-shaped momentum distributions into spiral patterns. This deformation depends nonmonotonically on the modulation amplitude and frequency and is most pronounced for circular polarization. In the transverse momentum plane, the spin-resolved yields favor the spin channel along the direction of field handedness. Reversing the field handedness would reverse the favored channel while leaving the total yield unchanged. Sine-type modulation preserves ring-shaped and mirror-symmetric spectra approximately, however, it leads to a stronger dependence of the spin asymmetry on the modulation amplitude than cosine-type modulation does. These results show that phase modulation and ellipticity can control the momentum spectra patterns in vacuum pair production. Moreover, phase modulation can also affect the degree of spin asymmetry.
Read the full abstract

All authors Li-Yi Yang, Hong-Hao Fan, Jing-He Huang, Shuo Yang, Bai-Song Xie

Robotics

FATHOM: World-Model-Guided Diffusion Action-Trajectory Selection for Sparse-Marker AUV Tracking

Reliable relative tracking is fundamental to cooperative autonomous underwater vehicles (AUVs), yet remains challenging under nonlinear vehicle dynamics and degraded underwater visual conditions. Active light markers improve robustness to appearance variation, but provide deliberately sparse observations that can become incomplete or corrupted by occlusion, reflections, and detection noise during dynamic motion. To address the resulting partial-observation control problem, this paper proposes FATHOM, a world-model-guided diffusion action-trajectory selection framework for sparse-marker AUV tracking. FATHOM first integrates structured marker observations, inertial measurements, action history, and temporal dynamics into a predictive tracking belief using a multimodal task-supervised recurrent state-space model. A diffusion action prior then generates multiple expert-like multi-thruster trajectories, whose future tracking consequences are predicted through the learned latent dynamics. Rather than directly executing a generated trajectory, FATHOM selects the most favorable candidate under an explicit receding-horizon objective and continually updates the decision from onboard observations. Extensive offline analyses verify the learned predictive belief, multi-step world-model prediction, diffusion trajectory generation, and predictive selection. More importantly, comprehensive real-AUV experiments, including strong pose-based comparisons and component ablations, demonstrate robust autonomous tracking across nominal and challenging motions with naturally occurring partial marker visibility, while maintaining real-time onboard operation without motion-capture feedback.
Read the full abstract

All authors Xiaorui Wang, Zeyu Sha, Mingyang Yang, David Chen Wang, Feitian Zhang

Graphics

Segmenting Geometric Textures of 3D Meshes

Recognizing geometric texture patterns on 3D models is critical for applications such as heritage protection, model editing, and 3D printing. Existing 3D model segmentation methods predominantly focus on identifying high-level semantic parts, thereby failing to capture middle-level textures. While recent Vision-Language Models can perform image segmentation with appropriate text prompts, they still struggle to recognize 3D geometric textures. In this work, we present a method that segments 3D mesh models based on middle-level geometric textures. We further introduce a new geometric texture segmentation benchmark and fine-tune SAM3 accordingly. The segmented texture regions are then clustered based on DINO feature similarity and lifted from 2D to 3D. Additionally, we develop an interactive tool to enable both 3D model editing based on our segmentation results and model retrieval from the dataset based on geometric and texture similarity. Experiments demonstrate that our method significantly outperforms existing methods in segmenting geometric textures and can facilitate various applications.
Read the full abstract

All authors Weidan Xiong, Kaijie Qu, Shihao Feng, Shunqi Fan, Rongyu Xiong, Hui Huang

Functional Analysis

Empirical approximation of $L_p$ norms for $p\in (2, \infty)$

For $p>2$ and an $N$-dimensional subspace $X\subset L_\infty(\mu)$ satisfying \[ \|f\|_\infty\le (KN)^{1/p}\|f\|_{L_p(\mu)} \quad\forall f\in X, \] we show that $m\ge C(p)KN \bigl(\varepsilon^{-p} +\varepsilon^{-2}\log_2(2N)\bigr)$ independent samples $\xi^1,\ldots,\xi^m$ with distribution $\mu$ satisfy \[ (1-\varepsilon)\|f\|_{L_p(\mu)}^p \le\frac1m\sum_{j=1}^m|f(\xi^j)|^p \le(1+\varepsilon)\|f\|_{L_p(\mu)}^p \quad\forall f\in X \] with probability at least $1-\exp(-c(p)\frac{m\varepsilon^2}{KN})$. This gives the optimal sample-size order $N\log_2 N$ for independent random sampling with fixed $K$, $p$, and $\varepsilon$. This result follows from a more general empirical moment estimate over unit balls of $p$-convex norms.
Read the full abstract

Robotics

Localizing and Editing the Goal Representation of a Vision-Language-Action Model

Operators can state safety constraints for robots as prohibitions such as "do not put the bowl on the stove". We study whether OpenVLA follows such prohibitions and where it represents the goal it executes. The 10 tasks of LIBEROGoal share one scene, so we can label the goal that each rollout executes. We combine closed-loop rollouts with activation patching, linear probes, activation steering and runtime monitors. On LIBERO-Goal, negation had no detectable effect on which task OpenVLA executed. The model executed the forbidden task in 79% of negated rollouts and completed 78% of plain tasks. A goal probe gives similar read-outs under "not X" and "X". Activation patching shows that the first action depends mostly on the instruction tokens before layer 12 and on the last prompt token from layer 16 onward. A fixed steering vector at the instruction tokens at layer 8 redirected 12 of 18 rollouts. Fixed vectors and prefill patches at layers 12 to 20 redirected none of 158. From layer 16, only patches at every decoded action token redirected the robot (15 of 18 rollouts at layer 24). When the instruction changes during a rollout, the layer-32 goal read-out detects non-compliance better than failure detectors, and an external probe on image and instruction features performs comparably. These results support enforcing prohibitions outside the instruction and monitoring goal read-outs against the specification. Goal edits in OpenVLA should act before layer 12 or patch every action token.
Read the full abstract

Materials Science

A Real-Space Framework for the Bare and Interacting Longitudinal Susceptibility of Solids

We present an implementation of the bare and interacting longitudinal susceptibility in a full-potential linear muffin-tin orbital code. The bare susceptibility is built in real space from Matsubara sums of Kohn-Sham Green's functions and is resolved by site, spin and orbital degrees of freedom; the spin and charge susceptibilities follow from a Dyson equation, with an on-site exchange-correlation kernel. After testing the solver on the Stoner-like instability of bcc Fe, we apply it to four density-wave materials. The bare susceptibility peaks at H in bcc Cr, where the nonmagnetic state is close to a Stoner instability, and at the stripe vector M in FeSe. In LaTe3 the response of the Te nets peaks close to the observed charge-density-wave vector, at its image in the neighbouring Brillouin zone; which of the two equivalent vectors carries the response is set by the phases of the sites in the cell. In 2H-NbSe2 the response is weakly modulated, and the observed vector lies on the flank of a broad maximum, consistent with the known importance of electron-phonon coupling there. Antiferromagnetic order reduces the bare longitudinal response at the ordering vector by ~8% in Cr and about half in FeSe. The orbital decomposition of the bare susceptibility shows which orbitals contribute the most for the nearest-neighbor bonds.
Read the full abstract

All authors Rohit Pathak, Abhay Mishra, Mikhail I. Katsnelson, Anna Delin, Olle Eriksson, Vladislav Borisov

Complex Variables

Slice Regular Capacity and Rational Approximation

Approximation theory in the complex plane is a well established area of mathematical research, with extensive contributions documented in the literature, notably through the classical results of Weierstrass, Runge, Mergelyan, and Vitushkin. Approximation theory in quaternionic settings, particularly rational approximation on compact sets, which studies when slice regular functions can be uniformly approximated by rational functions, is one of the recent areas of interest in quaternionic rational approximation theory (See [6,13]). In this paper, we discuss Vitushkin approximation theorem in the quaternionic setting by using the quaternionic analogue of analytic capacity and also we establish Davies modified Vitushkin scheme in quaternionic setting. We also discuss quaternionic analogue of $\gamma_{+}$- capacity and Cauchy capacity and the remarkable results of Tolsa [35] and Younsi [45] related to these capacities in complex plane setting.
Read the full abstract

Computation and Language

Richness Is Not Evenness: Ordered Renyi Crossings under Self-Training

Recursive self-training of language models, a setting studied in model collapse, can reduce vocabulary while making frequent words, mostly function words, less dominant. We fine-tune English GPT-2 on its own output for nine generations, vary the training-data sampling temperature, and measure every model with one fixed sampler. As training temperature rises, Renyi entropy changes turn from loss to gain at distinct temperatures: min-entropy first, then collision entropy, Shannon entropy and observed richness. Between the outer crossings, word types fall while the most-common word becomes less dominant. Untruncated training at T = 1 lies in this interval: at 13,000 words, generation 9 has 24% fewer types than generation 0 yet collision entropy is 0.91 bits higher. The ordering persists with a finer grid, different measurement samplers, a fixed word budget and five fresh training seeds. Min-entropy again crosses before richness in four further settings (in Hindi only at interpolated roots) and in 12 of 13 distinct measurements under eight text representations. Next-token identities and a categorical simulation isolate decoding and sampling effects; the crossing order remains empirical. Fine-tuning on human text can produce the same split while likelihood improves, and no single order's sign reliably identifies worse likelihood across the tested settings. We recommend reporting entropy profiles at stated budgets with an independent quality measure.
Read the full abstract

Astrophysics of Galaxies

Predicting Continuum Emission in Massive Star-Forming Regions with Embedded HII Regions - Dust, Free-Free, and Anomalous Microwave Emission with POLARIS

Context: HII regions are a key stage of massive star formation, where ionizing radiation reshapes the interstellar medium. Their continuum spectra trace thermal dust, free-free emission, and anomalous microwave emission (AME). Aims: We investigate how these components evolve as an HII region expands and correlate with gas and electron column densities across spatial scales. Methods: We post-process RAMSES radiation-magnetohydrodynamical simulations of a $10^4 M_\odot$ cloud with POLARIS. The model includes grain charging, rotational excitation and damping, radiative torques, rotational disruption, and emission from dust, ionized gas, and nano-composite grains. We compute maps and SEDs for six snapshots over $1-10^4$ GHz and analyze multi-scale correlations. Results: Ionized emission evolves from a compact, optically thick HII region to an extended, optically thin structure. Around 15 GHz, free-free emission changes from a $\nu^2$ to a $\nu^{-0.1}$ spectral regime as optical depth decreases, and increasingly traces electron column density. The thermal dust peak shifts from mid- to far-infrared as colder material contributes, while dust emission remains strongly correlated with total gas column density. AME peaks at 26-37 GHz and shows scale-dependent correlations with gas and electron distributions. Rotational disruption is the main dust-destruction mechanism affecting mid-infrared emission and the high-frequency AME tail. Conclusions: These continuum components provide complementary, scale-dependent diagnostics of stellar feedback. Self-consistent multi-frequency modeling is essential for interpreting surveys of star-forming clouds.
Read the full abstract

All authors Stefan Reissl, Alessio Traficante, Paolo Suin, Alice Nucara, Ralf S. Klessen, Leonardo Testi

Algebraic Geometry

A counterexample to Manin's Conjecture for almost Fano varieties

We give an example of a smooth rational threefold over a number field with a big anticanonical divisor that does not satisfy the thin set version of Manin's Conjecture. In particular, Manin's Conjecture does not hold for rationally connected almost Fano varieties.

Commutative Algebra

Generalized Edge Ideals and Regularity Offset of Graphs

We introduce generalized edge ideals by assigning pairwise coprime monomials, instead of variables, to the vertices of a graph. We use the lcm-lattice to show that the multigraded Betti numbers and Castelnuovo-Mumford regularity of the quotient by a generalized edge ideal agree with the Betti numbers and resolution regularity of the quotient by an ordinary edge ideal in a non-standard Z-grading. We define the regularity offset as a measurement of the difference between the regularities of the quotients by a generalized edge ideal and its ordinary counterpart, and prove bounds for the regularity offset. We compute the regularity offset for co-chordal graphs and paths, and derive a formula for the regularity offset of forests. We also derive a graph-theoretic criterion for when a forest's regularity offset is always zero, or equivalently, its top-level Betti number is nonzero.
Read the full abstract

Analysis of PDEs

On nonlocal geometric rigidity estimates in bounded uniform domains

We prove fractional geometric rigidity estimates and fractional Korn's first inequalities with general singular weights in bounded uniform domains in $\mathbb R^n.$ The classical choice of the energy well $SO(n)$ for the geometric rigidity estimate fails due to the lack of a notion of orientation for Lebesgue fields. The remedy is to choose $O(n)$ as the energy well. Due to the recently developed theory of local-to-nonlocal passage of rigidity estimates developed by the first author, both estimates reduce to the analogous local estimates on the cube, which we prove in several steps by compactness and linearization arguments.
Read the full abstract

Computer Vision

Decoding spatial and semantic information from large-scale visual cortical activity for natural-movie reconstruction

Reconstructing natural visual experience from cortical population activity is a central challenge in neural decoding, particularly for calcium imaging, where signals are noisy, temporally filtered, and paired neural-video data are limited. Here we present Ca2REAL (Calcium Activity to Reconstruct Events, Appearance, and Layout), a neural-to-video framework that conditions a pretrained video diffusion model through separate layout and semantic controls derived from calcium activity. Using measured responses from 7,863 V1 neurons in one mouse, Ca2REAL reconstructed held-out natural-movie clips with 81.4% all-pairs stimulus-identification accuracy, with its clearest gains over deterministic baselines in perceptual and semantic measures. We further evaluated the framework in a stimulus-aligned virtual population of 116,705 neurons assembled from encoding-model-predicted responses across 11 independently recorded datasets. Perturbation and ablation analyses showed that reconstruction depends on distributed, stimulus-specific population structure, with layout and semantic pathways providing complementary spatial and identity information. Temporal analyses further showed improved preservation of semantic dynamics across reconstructed clips. These results show that large-scale cortical population activity constrains the spatial, semantic, and temporal structure of natural visual experience, providing a generative framework for high-fidelity neural reconstruction of dynamic visual scenes.
Read the full abstract

All authors Yukai Huang, Haoyu Albert Wang, Jiawei Zhang, Jialun Ma, Jiayao Ding, Yiran Lin, Xingyu Chen, Ziqi Gao, Yuguo Yu

Computer Vision

Computer Vision by YOLOs for Detection of Major Diseases in Vietnamese Custard Apple Using Fruit Imagery

Amid Vietnam rapid digital transformation in agriculture, the integration of artificial intelligence technologies for identifying, monitoring, and predicting crop pests and diseases has emerged as an inevitable trend.