Peer Revenue Equity in science

Papers

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

All of arXiv

Friday 28 August

Solar and Stellar Astrophysics

Dynamically driven collapse of a thermally stable accretion disk in a nova-like system

While standard disk instability theory predicts thermal-steady, outburst-free disks in nova-like variables due to their high mass transfer rates, mass transfer variations are prevailingly invoked to explain the occasional flaring/fading phenomena observed in these systems. Here we report the observational evidence for a dynamical collapse of a thermally stable accretion disk, captured serendipitously by TESS during a fading episode of the VY Scl-type nova-like system MASTER OT J072703.91-631952.8, and traced by the emergence of an unusual negative superhump that evolves toward its orbital frequency. Simultaneously, the system underwent an anomalous eruptive event featuring a remarkably symmetric 45-day light-curve profile, indicative of a mild energy-release mechanism fundamentally distinct from documented eruptive events in cataclysmic variables. Notably, the concurrence of the eruption and the disk collapse is difficult to reconcile with the paradigm of mass transfer variations: the eruption implies enhanced mass transfer, whereas the disk collapse indicates a decline. The morphological disk evolution over $\sim 700$ days, characterized by both negative and positive superhumps, indicates a transition from a circular disk to an eccentric disk, followed by a tilted state and ultimately a minimal disk configuration. This evolutionary sequence provides evidence for a previously unrecognized dynamics-driven cycle operating in the thermally stable accretion disk of a VY Scl-type nova-like star.
Read the full abstract

All authors Wentao Li, Jie Lin, Tinggui Wang, Yongkang Sun, Chengyuan Wu, Heran Xiong, Krystian Iłkiewicz, Luca Casagrande, Michael Bessell, Ning Jiang, Chengyi Wang

Computer Vision

Relational Knowledge Distillation Brings DNN Representations Close Enough to Humans to Be Aligned Without Supervision

Linking the internal representations of deep neural networks (DNNs) to human mental representations is important for using DNNs as computational models of human vision. Existing DNN representations remain insufficiently similar to human mental representations, which are not directly observable and are therefore commonly measured through large-scale similarity judgments of object images. A natural approach to narrowing this gap is to directly transfer the relational structure of human representations into DNNs, and previous studies have reported improved human-DNN representational similarity. However, whether this improvement holds under stricter evaluation remains untested in two respects: fine-grained alignment at the individual-object level, and generalization to a human embedding derived from a dataset independent of the training data. Here, we employ an unsupervised comparison method, Gromov-Wasserstein optimal transport (GWOT), which estimates human-DNN correspondences from the internal distance structure alone and thereby tests fine-grained alignment. We further assess generalization on a curated test set of concepts non-overlapping with the training data. We show that fine-tuning pre-trained DNNs with Relational Knowledge Distillation (RKD), an established relational transfer method, brings DNNs close enough to humans to be aligned at the individual-object level on this test set. We also show that this improvement is driven by a more human-like global structure, as reflected in the ordering of distances among coarse categories, while the local human-DNN nearest-neighbor overlap rate remains largely unchanged. These findings indicate that relational transfer from humans brings the global structure of pre-trained DNNs close enough to the human structure to enable fine-grained human-DNN alignment without supervision.
Read the full abstract

All authors Yuria Shimizu, Soh Takahashi, Takato Horii, Masafumi Oizumi

Metric Geometry

Equilibrium Laws for Julia's Zero and the Hyperbolic Zero of Binary Forms

For a real binary form with no real roots, Julia's zero and the hyperbolic zero are two SL(2,R)-equivariant points in the upper half-plane. We show that they admit closely related equilibrium characterizations, with radial weights given respectively by the hyperbolic tangent and hyperbolic sine of the distances to the roots. This common framework gives geometric criteria for coincidence of the two zero maps. They always agree for binary quartics; for binary sextics they agree exactly when the three upper-half-plane roots form an equilateral hyperbolic triangle, or are collinear with one root the hyperbolic midpoint of the other two. We also obtain a corresponding result for collinear binary octics. The two equilibrium laws further reveal a sharp difference in the influence of distant roots. A strict majority of roots confined to a compact set keeps Julia's zero in a compact set, and the threshold one-half is optimal. In contrast, an escaping minority can force the hyperbolic zero to escape at linear scale. We also illustrate the computational advantages of the explicit formula for the hyperbolic zero.
Read the full abstract

Artificial Intelligence

HyQuant: Hybrid-Precision Quantization for LLM Attention

Quantization has been widely adopted in LLM training and inference to reduce cost and improve efficiency. However, low-bit quantization of the attention module often introduces large errors at very low bit-widths, causing performance degradation. Existing methods mainly rely on smoothing techniques to handle outliers, while we propose a hybrid quantization design to better balance accuracy and efficiency. Specifically, we propose HyQuant, an efficient hybrid quantization framework for LLM attention. HyQuant quantizes most attention states into low-bit formats while retaining a small set of vertical-line tokens and local-window states in high precision. These accuracy-critical regions are selected using lightweight vertical-line-aware attention-pattern signals, reducing quantization error with limited overhead. In the Prefill stage, HyQuant uses a hybrid-precision quantized attention operator that preserves vertical-line tokens and a local sliding window in full precision while quantizing the remaining context. In the Decode stage, HyQuant applies the same principle to KV-cache compression and fuses KV dequantization with attention computation to improve memory and hardware efficiency. Across diverse tasks, models, and datasets, HyQuant maintains nearly lossless accuracy with an extremely simple design, demonstrating the efficiency and practical feasibility of hybrid quantization for LLM attention. Code is available at: https://github.com/jerrysfls/HyQuant .
Read the full abstract

All authors Jiatong Ding, Bingxin Xing, Yu Zhang, Dian Ding, Xiaodong Yi, Xianbin Ouyang, Feihu Zhou, Kun Zhang, Zhenyu Guo, Hao Pan, Guangtao Xue, Yiming Zhang

Distributed Computing

TerraceMoE: A Cost Model for Hierarchical MoE All-to-All Communication

Hierarchical two-hop dispatch can reduce slow-fabric traffic in expert-parallel Mixture-of-Experts training, but it adds a second collective and an arrival-side operator chain. We present a cost model for screening that trade at the communication-call level, bounded by validation gates that withdraw a capability in code when they fail rather than reporting a caveat. At a reference geometry with 16 groups of 8 ranks, $q=3$, $H=2048$, and 4096 tokens per rank, the corrected effective breakeven hierarchy ratio is 3.98 for the measured PyTorch arrival chain, 1.49 for a hypothetical fused target, and 1.10 at zero implementation overhead. These are ratio-only sensitivity results, not deployment predictions: platform A measures 1.03, platform B has no separated fast/slow measurement, and neither machine measured here reaches the hierarchical regime. Four communication-level corpora pass their gates; a drift probe and the step-level gate fail. The latter failure is enforced in code, so we make no training-throughput prediction. The enabling routing constraint fixes per-token fan-out and per-selected-group quota, while aggregate per-peer counts remain data-dependent. Its measured validation-loss cost is small but nonzero (+0.0034 nats); downstream equivalence is reported with incomplete estimator provenance and is therefore not independently reconstructible from the artifact. Code, calibration constants and the validation gates are at https://github.com/weich97/TerraceMoE-simulator.
Read the full abstract

All authors Weicheng Xue, Bingqiang Wang, Li Yuan, Huihui Zhou, Yonghong Tian

Optimization and Control

Anchored Scenario Coverage for Failure-Aware First-Hit Batch Inverse Design

Early discovery of at least one valid design satisfying a target requirement is a central objective in failure-prone closed-loop inverse design. A natural batch baseline ranks candidates by a product-form marginal valid-hit score, but selecting the highest-ranked candidates independently can produce redundant recommendations under predictive uncertainty and waste the experiment budget. We introduce ARC-SC(Anchored Risk-Constrained Scenario Coverage), a batch acquisition method that preserves strong marginal candidates as anchors and allocates the remaining batch positions by maximizing complementary coverage over predictive target scenarios under a risk-support constraint. In frozen-oracle closed-loop simulations on superconductivity and JARVIS materials-property benchmarks, ARC-SC yields a statistically supported improvement in first-hit discovery and remains competitive with directionally favorable first-hit performance on more challenging design space. These results establish ARC-SC as a POF-anchored, scenario-aware batch strategy for improving early valid-target discovery under structured experimental failure.
Read the full abstract

All authors Chuhan Yang, Chenxi Wang, Linhan Wu, Yuyang Liu

Quantum Physics

Hybrid Heralded Noiseless Amplification with Finite-Cutoff Quantum Scissors

A noiseless linear amplifier (NLA) can probabilistically amplify an optical state without the noise required by deterministic phase-insensitive amplification. We study a hybrid amplifier in which a finite-cutoff quantum-scissor NLA is placed between two single-mode squeezers. Analysis based on an ideal (infinite cutoff) NLA finds a gain enhancement from the squeezing. We explore the physics of this enhancement as the cutoff of the quantum scissors is increased. The low-cutoff sequence shows how this gain enhancement emerges from the truncated Fock space. Cutoff 3 is the lowest order at which the additional even- and odd-photon components can both contribute to gain enhancement, albeit with some skewing of the coefficients which reduces the fidelity. As the cutoff is increased the fidelity improves. However, unlike the ideal transformation, the finite-cutoff device depends on the phase of the coherent amplitude relative to the squeezing axes, with states aligned with the anti-squeezing requiring higher cutoffs to achieve high fidelity. We track behavior to high cutoffs and eventually see the performance predicted in the ideal theory emerge.
Read the full abstract

Computer Vision

Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding

Long-video understanding remains challenging for Multimodal Large Language Models (MLLMs) due to limited context length. Uniform sampling may miss crucial moments, while agent-based frame video understanding methods often evaluate frames independently, overlooking the temporal organization of videos. Ideally, evidence selection should mimic how humans answer questions about long videos: first locating the relevant segment from the global context, then zooming into local objects and details. We propose Temporal Tree of Thought T^3, a training-free framework for adaptive coarse-to-fine long-video understanding. T^3 constructs a question-agnostic hierarchical temporal tree via recursive temporally constrained clustering, where each node represents a contiguous segment with an informative key frame. During inference, T^3 performs an answer-retrieve-explore loop: it reasons over coarse representative frames, generates a search statement when evidence is insufficient, and expands relevant branches for finer-grained evidence. This process adaptively shifts the search target from temporal regions to specific objects and visual details to help video understanding. Experiments on VideoMME, LongVideoBench, and LVBench show that T^3 improves Qwen2.5-VL-7B by 0.5%, 4.6%, and 4.4%, respectively, under the same frame budget, demonstrating the effectiveness of structured temporal reasoning.
Read the full abstract

Combinatorics

On cluster deep loci in double Bruhat cells

We study deep loci for the reduced-word Berenstein-Fomin-Zelevinsky cluster atlas on type-A double Bruhat cells. As defined in CGSS24, for a seed collection on a cluster variety X, the deep locus is the complement of the union of all corresponding cluster tori. We focus on double Bruhat cells G(u,v) for G = SL(n). Our main results are several theorems describing the structure of these deep loci, including their interaction with the Poisson structure, the relationships between the deep loci of different cells, and a tracking formula that allows us to compute the reduced-word deep locus of a cell from a lower-dimensional one. The paper also includes explicit low-rank calculations for double Bruhat cells in SL(3) and Borel double Bruhat cells in SL(4). Finally, we discuss the relationship between these results and some general properties of deep loci.
Read the full abstract

Artificial Intelligence

See, Hypothesize, Validate: Multimodal Agentic Framework for Discovering Governing PDEs

Discovering governing partial differential equations (PDEs) from observational data remains a core challenge across the sciences. Existing sparse-regression, symbolic-regression, and LLM-based approaches can be constrained by predefined libraries, noise sensitivity, hallucination, or limited iterative refinement. We introduce MAGE (Multimodal Agentic Governing Equation Discovery), an agentic framework that organizes PDE discovery as a confidence governed hypothesis validation loop inspired by the scientific cycle of observation, hypothesis, and falsification. Four role-specialized agents collaborate: a Differential Observer computing derivatives and diagnostic visualizations; a VLM-powered Phenomenology Extractor distilling qualitative cues from multimodal diagnostics; an LLM-driven Governing Law Synthesizer proposing candidates without a predefined library; and an Equation Arbiter fitting coefficients and assigning confidence scores. Discovery iterates until the top candidate clears a user-specified threshold, providing a structured process with an explicit accept-reject protocol. On the evaluated canonical PDE suite, MAGE obtains 8/8 exact structural recovery and the lowest coefficient error among the compared methods on 7/8 systems, with improvements of up to 4 orders of magnitude and a geometric-mean improvement of approximately 3 orders of magnitude. The pipeline also recovers the expected operators in two complex geometries and, on one laboratory sensor record, selects a cubic restoring-force model with held-out $R^2=0.98538$. These results support further study of structured agentic reasoning for library-free governing-law discovery, while broader generalization remains to be evaluated.
Read the full abstract

All authors Sarang Manoj Pekhale, Amartya Roy, Rajat Sarkar, Souvik Chakraborty

High Energy Physics (Phenomenology)

Vector-boson-fusion timing references for four-dimensional displaced-vertex searches

Long-lived particles that decay inside the tracker are searched for via the spatial displacement of their decay vertices. However, the same decays also arrive late, providing a handle made available by a precision timing layer at the HL-LHC that displaced-vertex searches do not exploit. We study the exotic decay of a Higgs boson produced through vector-boson fusion (VBF), $pp\to hjj$ with $h\to ss$, in which the long-lived scalar decays as $s\to b\bar b$, and use the VBF tag jets to identify the hard-scatter vertex. Its prompt central tracks, timed by a CMS-MTD-like barrel layer that also times the central tracks of the displaced vertex, define a per-event start time, turning each displaced vertex into a four-dimensional object. Because a fast simulation cannot determine the absolute normalization of the heavy-flavour and instrumental displaced-vertex backgrounds, we take as the primary result a normalization-free figure of merit: the ratio of the timing-enhanced sensitivity to the spatial-only sensitivity. Adding the timing layer improves this ratio by a factor of $\simeq2.8$ at a nominal working point and by up to an order of magnitude for a tighter delayed-vertex requirement, because the heavy-flavour background that survives the spatial selection is prompt in collision time. The improvement increases with decay length, extending sensitivity into the long-lifetime regime in which purely spatial searches lose tracker acceptance. The gain is essentially mass-independent at a decay length of $\ctau\simeq100\mm$, where the trend with scalar mass reverses: lighter scalars benefit more at short lifetimes because of their larger boost, whereas heavier scalars benefit more at long lifetimes as the lighter states leak out of the tracker.
Read the full abstract

Artificial Intelligence

CoRe-MoE: Compact Reusable MoE for Continual Multimodal Instruction Tuning

Continual multimodal instruction tuning requires multimodal large language models to acquire new task abilities sequentially while preserving previously learned knowledge. LoRA-MoE provides a promising solution by introducing expert-based capacity, but repeatedly learning and maintaining full LoRA experts leads to substantial parameter overhead. This raises a natural question: is full expert expansion necessary for every new task? To answer it, we analyze the SVD of task-specific LoRA updates and observe substantial overlap in their input- and output-side LoRA direction subspaces, with task-specific adaptation largely captured by lightweight coordinates over these subspaces. Motivated by this observation, we propose CoRe-MoE, a Compact Reusable MoE framework for parameter-efficient continual multimodal instruction tuning. CoRe-MoE extracts reusable input- and output-side direction bases from an initial expert bank, and for subsequent tasks trains only compact coordinate experts together with task-specific low-rank routers. Experiments on two representative MLLMs show that CoRe-MoE improves final average performance over the strongest competing baseline by up to 5.90 points, while using less than 1% of the trainable parameters required by sequential LoRA for later tasks. The code is publicly available at https://github.com/runzezz/CoRe-MoE.
Read the full abstract

All authors Runze Liu, Naibin Gu, Mingxu Ai, Yuqing Li, Peng Fu, Zheng Lin, Weiping Wang

Computer Vision

Iron: Intent-Aligned and Retrospective Dual Learning Framework for Enhancing Generalist Virtual Agents

Achieving virtual agents capable of automating tasks across diverse digital environments remains a pivotal challenge in Embodied AI. While Multimodal Large Language Models (MLLMs) offer enhanced visual perception and reasoning, their agentic deployment faces three challenges: costly data annotation, imprecise action-intent alignment, and inefficient exploration from discarded failed trajectories. To address these, we introduce Iron, an intent-aligned, self-improved, and annotation-efficient framework for training GUI agents. Iron employs a novel dual learning strategy that utilizes a stepwise cycle-consistent (SCC) reward to achieve fine-grained alignment between low-level actions and high-level intents, thereby improving instruction grounding and intent understanding. Concurrently, Iron introduces a hindsight reproduction mechanism to repurpose failed trajectories for training, improving both learning efficiency and task diversity. Extensive experiments demonstrate that Iron-trained generalist agents consistently improve performance on cross-environment and cross-device tasks, outperforming models trained with three times more data. Iron also achieves a substantial 25.06% relative improvement on unseen web tasks, with further gains observed on inherently complex tasks, demonstrating the feasibility of building more capable virtual agents.
Read the full abstract

All authors Jiahe Ying, Wendong Bu, Kaihang Pan, Bingchen Miao, Siyu Chen, Wen Wang, Xueming Jiang, Juncheng Li, Siliang Tang

Performance

FFSlim: An Efficient and Lightweight Format for Multi-modal Data Storage and Retrieval

With the rapid expansion of large-scale media-text corpora, multi-modal datasets increasingly require efficient storage and retrieval. Existing formats such as Files, TDP, and FFRecord work adequately for uni-modal data but expose fundamental limitations in multi-modal settings, including storage redundancy, massive small-file overheads, cache-unfriendly layouts, and heavy index structures. These issues jointly inflate storage and memory usage and make I/O the dominant bottleneck in real training workloads. We present FFSlim, a lightweight format for storing and retrieving multi-modal data. FFSlim improves storage efficiency and loading throughput through three components: a unified file format that removes media duplication and avoids small-file proliferation; an adaptive retrieval mechanism that enables low-overhead pair-level access and accelerates repeated media loading; and a redundancy detection and aggregation module that converts existing datasets into the FFSlim layout. The experimental results demonstrate that FFSlim achieves 2.07x and 8.26x higher data loading and write throughput on average than the strongest baseline, with minimal storage and index overhead. Consequently, these underlying I/O accelerations enable FFSlim to reduce end-to-end training time by 5.36%-14.18% across seven diverse multi-modal models.
Read the full abstract

All authors Long Yang, Yu Mao, Yuchen Shao, Yumiao Zhao, Yaqi Li, Xuan Liu, Xiaolong Shen, Tao Yu, Gezi Li, Jing Wang, Chengcheng Wan, Liang Shi

Analysis of PDEs

A Rigorous Derivation of the Vlasov-Navier-Stokes Model from Multicomponent Boltzmann Equations

The rigorous justification of the Vlasov-Navier-Stokes system remains an outstanding open problem, both as a mean-field limit of an $N$-particle fluid-interacting system and as a hydrodynamic limit derived from multiphase Boltzmann equations. Inspired by the work of Bernard, Desvillettes, Golse, and Ricci [Commun. Math. Sci., 15(6), 1703-1741, 2017], we present a rigorous derivation of the incompressible Vlasov-Navier-Stokes system from the two-component Boltzmann system for elastic hard-sphere collisions. To this end, we establish the well-posedness of the rescaled multicomponent Boltzmann system for small initial data. Specifically, we obtain estimates for the solution that hold uniformly with respect to both the thermal speed ratio $\varepsilon$ and the mass ratio $\eta$. Furthermore, under the Vlasov-Navier-Stokes scaling assumption $O(1)\varepsilon^3 \leqslant \eta \leqslant o(1)\varepsilon^2$ as $\varepsilon \rightarrow 0$, we establish the weak convergence of the solution to the fluid-kinetic coupled limit system.
Read the full abstract

Numerical Analysis

A Massively Parallel Hybridizable Discontinuous Galerkin Solver for Direct Numerical Simulation of Compressible Flows on GPUs

Direct numerical simulation (DNS) of compressible transitional and turbulent flows requires numerical methods that combine high-order accuracy, robustness, and computational efficiency to resolve a broad range of spatial and temporal scales. This paper presents a massively parallel hybridizable discontinuous Galerkin (HDG) solver for DNS of the compressible Navier-Stokes equations on GPU-accelerated high-performance computing systems. The proposed solver combines high-order HDG discretization with robust shock capturing, diagonally implicit Runge-Kutta (DIRK) time integration, and an efficient Newton-GMRES solution strategy accelerated by additive Schwarz preconditioning and reduced-basis approximation. A distributed implementation of these methods based on GPU-aware MPI, Kokkos, and CUDA/HIP libraries enables scalable execution on heterogeneous computing platforms. The solver is demonstrated on three canonical benchmark problems covering a wide range of Mach-number flow regimes: subsonic transitional flow over the Eppler 387 airfoil, the supersonic Taylor-Green vortex, and hypersonic boundary-layer transition. Numerical results are compared with available experimental measurements and published DNS data, showing good agreement across distinct flow regimes. The results demonstrate the ability of the proposed solver to resolve laminar-turbulent transition, strong compressibility effects, shock-associated flow structures, and fully three-dimensional turbulent dynamics.
Read the full abstract

All authors Andrew Welter, Thea Collin, Ngoc Cuong Nguyen, Jaime Peraire

Mathematical Physics

Holomorphy of the ground state and energy for bosonic quadratic Hamiltonians

We study the holomorphy of the ground state and its energy with respect to the coupling constant for bosonic quadratic Hamiltonians. We prove general theorems establishing this holomorphy, thereby providing a rigorous justification for formal perturbative expansions. This theory is applied to several concrete models, including the single pair interaction model and the Pauli-Fierz model in the dipole approximation. Importantly, for the Pauli-Fierz model with a realistic ultraviolet cutoff, the perturbation expansions in the elementary charge converge at the physical charge value, even when mass renormalization is taken into account. Furthermore, we explicitly determine the radii of convergence of the ground states and their energies for both the single pair interaction model and the fibers of the translation-invariant Pauli-Fierz model.
Read the full abstract

All authors Kota Imura, Shinnosuke Izumi, Yasumichi Matsuzawa, Itaru Sasaki

Computational Physics

Vectorized Symmetric and Fermionic Tensor Network Implementations for GPU-Accelerated Variational Monte Carlo

Variational Monte Carlo (VMC) calculations based on tensor networks (TN) have recently achieved competitive accuracy in ground-state calculations of strongly correlated spin and fermionic systems. However, existing tensor network VMC (TN-VMC) algorithms have not been formulated in a manner that can fully utilize GPU acceleration. We tackle the key missing ingredient, namely, the vectorized evaluation of tensor network amplitudes and tensor network operations. This ensures high GPU utilization by batching over computations with identical structure. In particular, we show how to achieve vectorization for the practically relevant case of abelian symmetric tensor networks (which includes fermionic tensor networks) by developing a “flat” tensor network formalism for block-sparse tensor representation and contraction. Using this, we construct a GPU-adapted symmetric TN-VMC workflow with batched tensor network computation. In the two-dimensional Fermi–Hubbard model, we demonstrate a GPU speedup of up to $300 \times$ over single core CPU implementations, for fermionic TN variational wavefunctions.
Read the full abstract

All authors Si-Jing Du, Johnnie Gray, Ao Chen, Garnet Kin-Lic Chan

Computer Vision

From Perspective to Fisheye Depth Estimation and Open-Vocabulary Segmentation

Vision foundation models are capable of generalizing across 3-dimensional (3D) scenes with high-fidelity estimates; their empirical success can be attributed to training on large-scale datasets of perspective images. However, when transferred to wide field-of-view (FoV) images, such as those captured by fisheye cameras, they return erroneous outputs due to a covariate shift stemming from the radial distortion on the image pixels. We propose a method to generalize vision foundation models to fisheye cameras. The crux of our method lies in a set of learnable parameters, termed Distortion Extenders (DEX), that model the fisheye distortion coefficients and the distributional shift between fisheye and perspective images encoded in the latent space. By minimizing a self-supervised alignment loss, DEX transforms the latent embeddings of fisheye images to resemble those of perspective images to recover high-fidelity estimates. DEX is architecture- and task-agnostic: We demonstrate DEX on monocular depth estimation and open-vocabulary segmentation for convolution- and Transformer-based architectures, where we consistently improve over baselines across indoor and outdoor fisheye datasets. As a byproduct, the activations of DEX can also be decoded to distortion coefficients to support camera calibration. Code available at: https://github.com/Suchisrit/DEX.
Read the full abstract

Computational Complexity

Improved Subexponential Upper Bounds for $3$-Restricted Matching Vector Families

Matching Vector families (MVFs) are defined by two ordered lists of vectors in $\mathbb{Z}_m^n$ whose inner products satisfy specific residue patterns modulo an integer $m$. Most famously, restricted MVFs are used to construct the best-known constant-query Locally Decodable codes (LDCs). We prove an upper bound of $2^{O\left(\sqrt{n\log n \log m}\right)}$ on the size of $3$-restricted MVFs in $\mathbb{Z}_m^n$ for $m \leq \sqrt{n}$, substantially improving on the previous best bound of $2^{O(n/\log n)}$ by Bhowmick, Dvir and Lovett (STOC'13, SICOMP'14). Our proof relies on a new polynomial method argument that controls collisions in sumsets of matching vectors.
Read the full abstract

Applied Physics

Extending the operating window of scanning electron microscopy through an integrated electron-optical architecture for high-temperature and near-ambient-pressure environments

Scanning electron microscopy (SEM) under simultaneously high-temperature, near-ambient-pressure (NAP), and reactive-gas environments requires coordinated control of vacuum isolation, electron-beam transmission, signal generation, and thermal management, constraints that have long limited the operating window of environmental scanning electron microscopy (ESEM). Here we establish an integrated electron-optical architecture that combines a multistage differential-pressure pathway, front-stage pressure transition, detector optimization, thermal management, and a gas-focusing sampling architecture into a unified ESEM platform. Pressure distribution and electron-beam transmission are quantitatively validated through computational fluid dynamics (CFD), Monte Carlo electron-gas scattering analysis, and direct beam-current measurements, while detector optimization and thermionic-electron suppression preserve stable imaging under elevated pressure and temperature. The resulting system enables stable SEM imaging at pressures up to 20,000 Pa and high-temperature imaging up to 1,400 degrees C, continuous observation of hydrated biological specimens, and synchronized SEM-QMS operando characterization using local gas sampling. These developments establish a general electron-optical framework for extending ESEM toward realistic operando environments where elevated temperature, reactive gases, structural evolution, and gas-phase chemistry can be investigated simultaneously.
Read the full abstract

All authors Yue Chai, Honglong Zhao, Xinning Tian, Chao Ang, Zhu-Jun Wang

Artificial Intelligence

SpikeOPD: Stable On-Policy Distillation for Autoregressive Spiking Language Models

Spiking neural networks (SNNs) offer a path to energy-efficient language modeling through sparse encoding and event-driven computation, but training capable spiking language models from scratch remains difficult. A practical alternative is ANN-to-SNN migration through knowledge distillation (KD), where a pretrained artificial neural network (ANN) teacher supervises an SNN student. Existing migration approaches distill on fixed corpus prefixes, whereas autoregressive inference conditions on self-generated prefixes, creating prefix-source mismatch. It manifests as output-policy mismatch with the ANN teacher and internal spiking-dynamics drift between self-generated and matched corpus prefixes. On-policy distillation (OPD) offers a natural way to mitigate both manifestations by continuing teacher supervision on self-generated prefixes. We evaluate a teacher-only full-KL variant, Vanilla OPD, via a controlled stress test and observe it may suffer from delayed rollout-feedback collapse. This result shows that on-policy coverage alone does not ensure stable adaptation. Motivated by these findings, we propose SpikeOPD, a stable on-policy distillation framework for autoregressive SNNs that learns from self-generated prefixes while maintaining rollout stability. It applies full-KL teacher correction to reduce output-policy mismatch, while matched-prefix policy anchoring constrains policy departure from the frozen reference SNN on the same prefixes. Layerwise spike regularization further limits firing-rate deviations during on-policy adaptation. Across three model scales, SpikeOPD improves average accuracy over the corresponding KD SNNs by 0.8, 1.7, and 2.9 points at 0.125B, 0.35B, and 1.3B, respectively, while preserving their sparse-compute profiles.
Read the full abstract

All authors Enqiao Lu, Xingrui Yu, Yiwei Fu, Zhenglin Wan, Pengfei Zhou, Wangbo Zhao, Muqing Jian, Xueyi Zhang, Yang You, Ivor Tsang

Machine Learning

FedEHR-Agents: Federated Agentic Optimization for Automated EHR Modeling

Recent advances in large language models are enabling autonomous clinical agents to perform increasingly complex electronic health record (EHR) modeling workflows. However, agents deployed at individual hospitals remain constrained by institution-specific data and modeling environments, while direct cross-hospital collaboration is restricted by the sensitivity of patient-level EHR data. Although federated learning (FL) provides a natural foundation for privacy-preserving collaboration, existing approaches remain predominantly model-centric, limiting federation to prediction models or their updates while overlooking the richer modeling experience accumulated by autonomous agents. To address this limitation, we propose FedEHR-Agents, an experience-centric federated agentic optimization framework for automated EHR modeling. Each hospital deploys an autonomous clinical EHR agent that performs data preprocessing and model development while refining local clinical modeling experience through historical memory, task-specific evaluation, and TextGrad-based prompt refinement. The federated server performs evidence-guided experience aggregation to integrate reliable and complementary modeling experience across heterogeneous hospitals and distills the aggregated experience into global meta-prompts for subsequent local refinement. Extensive experiments on real-world multi-hospital EHR benchmarks demonstrate that FedEHR-Agents consistently outperforms local and federated baselines across diverse clinical prediction tasks and remains robust across different federation scales and LLM backbones. These results establish clinical modeling experience as a promising collaborative object beyond conventional parameter-centric FL and point toward federated autonomous clinical intelligence.
Read the full abstract

Computation and Language

AI Writers Have a Consistent Stylometric Footprint, but AI Editors Do Not

Text generated by large language models (LLMs) has been shown to be stylometrically distinct from human-written text (Andre et al., 2023; Shah et al., 2023; Opara, 2024; Soto et al., 2024; Li and Zhang, 2025; Selvioglu et al., 2025). But LLMs are increasingly used not only to generate text but also to edit human writing, and it is unclear whether the two leave the same trace. We show that AI generation leaves a consistent "stylometric footprint": a small subset of features, primarily entropy and lexical diversity, consistently separates AI-generated text from human writing across 8 LLMs and 5 domains, while the remaining features depend heavily on the domain and generator. AI editing, however, does not reproduce the same footprint. Relative to their human- written sources, AI-edited texts show only a small increase in lexical diversity and a decrease in entropy, rather than the joint increase that characterizes AI generation. Lexical density, which contributes little to generation, instead becomes the dominant editing-associated signal. Stylometric features therefore separate AI-edited text from AI-generated text but are substantially less effective at separating it from human-written text. Our results suggest that "AI text" is not a single phenomenon: generation and editing leave qualitatively different stylometric traces and should be studied separately.
Read the full abstract

Functional Analysis

Improved $\ell_0$-Isoperimetry for Convex Bodies via Mass Transport

We study $\ell_0$ isoperimetry for a convex body $K\subset \mathbb{R}^n$, $n\ge2$. For a Borel set $S\subset K$, let $\partial_0^K S$ be the set of points in $K \setminus S$ that can be reached from $S$ by changing at most one coordinate (i.e. the $\ell_0$ boundary of $S$). Suppose that, for some unconditional convex body $Q \subset \mathbb{R}^n$, numbers $r,R>0$, and possibly different centers $x_0,y_0$, \[ x_0+rQ \subset K\subset y_0+RQ. \] Writing $s=\text{vol}(S)/\text{vol}(K)$, we prove that whenever $0<s \le 1/2$, \[ \frac{\text{vol}(\partial_0^K S)}{\text{vol}(S)} \ge \frac{cr}{nR} \min\left\{1,\frac{\log(e/s)}{n}\right\}, \] where $c > 0$ is an absolute constant. Consequently, the associated $\ell_0$-isoperimetric coefficient is at least $cr/(n^2R)$. Previous direct lower bounds were only known for $\ell_2$ and $\ell_\infty$ regularity whereas our lower bound holds directly for any $Q$-regularity, where $Q$ is an unconditional convex body. Compared to $\ell_2$ and $\ell_\infty$ regularity, our lower bound result improves upon the previously best known lower bounds, for any $s$, by a factor of $n$. As an application of our result, we give improved mixing time bounds for the Coordinate Hit and Run walk (CHAR). Our proof of the lower bound is based on a modification of the method of canonical paths applied to a continuous Hamming graph over our convex body. Our construction of canonical paths can be viewed as a suitable coordinate discretization of certain mass transport maps from $S$ to $S^c$. We also give complementary upper-bounds for any $Q$-regularity, with an overall factor of $n$ gap between the two.
Read the full abstract

Plasma Physics

Counter-rotating density and current structures in a partially magnetized $\mathbf{E}\times\mathbf{B}$ plasma

The first 3D kinetic simulation of a magnetron discharge reproduces the measured rotating spoke and reveals that density and current rotate in opposite directions: the $m=1$ spoke turns in the $\mathbf{E}\times\mathbf{B}$ direction at 90~kHz while $m\simeq16$ electron-cyclotron-drift-type filaments turn at 1.0~MHz, with no net propagation along $\mathbf{B}$. Magnetic drifts exchange far more energy than they deposit, yet their spoke-front heating sustains the ionization. Following helical paths, the anode-directed current is relayed by de-trapping turbulence at the sheath edge (50%) and by the spoke channel in the bulk (97%).
Read the full abstract

All authors Pengze Xiao, Ya Zhang, Hongyu Wang, Wei Jiang, Jean-Pierre Boeuf

Analysis of PDEs

On the Well-Posedness of Periodic Strong Transonic Shocks in Divergent Nozzles

This paper studies the $C^1$ existence and dynamical stability of temporal periodic solutions involving strong transonic shocks for the quasi–one–dimensional full compressible Euler equations in diverging nozzles. We reveal the physical structural dissipation mechanism of strong transonic shocks hidden in the Rankine–Hugoniot conditions. This allows us to estimate the resonance at the shock boundary caused by the left acoustic waves in the subsonic downstream region. With this dissipation, we complete the proof by developing a fraction–step linearized iterative method.
Read the full abstract

All authors Peng Qu, Jiahui Wang, Huimin Yu, Xiaomin Zhang

Systems and Control

Graphon Design for Human-Machine Coordination under Bounded Rationality: Optimality of Stochastic Block Models

Coordination is a desirable feature in multi-agent systems, ranging from robotic swarms to socioeconomic networks. This paper is concerned with promoting coordination among heterogeneous agents, e.g., machines and humans, interacting in a stag-hunt game. In our model the agents exhibit bounded rationality at different levels, which leads to uncertainty and a propensity for errors during learning and decision-making processes. This paper addresses the problem of designing a network topology that maximizes a global metric of coordination under such constraints. While optimizing over the discrete space of finite graphs is generally computationally intractable, we employ a mean-field approach to lift the problem into the space of graphons. Within this framework, we analyze agents following a logit learning dynamics. Using calculus of variations, we show that for systems with a bimodal rationality profile, it suffices to search for optimal graphons in the ensemble of stochastic block models. We then propose a water-filling algorithm to find a locally optimal graphon. Finite graphs can then be sampled from the optimized graphon, bypassing the inherent combinatorial complexities of discrete graph optimization.
Read the full abstract

Chemical Physics

Mode-Specific Dynamics of $\text{CO}_2$ Hydrogenation on Copper: The Hidden Role of Molecular Rotation

Catalytic hydrogenation of $\text{CO}_2$ to formate on copper is a key elementary step for $\text{CO}_2$ utilization. Previous experimental and theoretical studies suggested an Eley-Rideal mechanism for this reaction, promoted by bending vibrational excitation, yet direct state resolved evidence remains lacking. Here, we present first-principles dynamical predictions for $\text{CO}_2$ hydrogenation on Cu(111) based on an accurate full-dimensional neural network potential energy surface. Our calculations near-quantitatively reproduce the measured reaction probabilities, including their nozzle-temperature and incidence-energy dependence. Our state-resolved results indicate that while vibrational excitation of the bending mode enhances reactivity, it alone cannot account for the observed reactivity increase with nozzle temperature. Instead, rotational excitation plays a dominant role, mainly attributable to the significant change in anisotropy of the molecular polar orientation as $\text{CO}_2$ accesses the transition state. This mode-specific insight reinforces the hidden role of rotation in surface reactivity, opening new avenues for state-selective control of $\text{CO}_2$ hydrogenation on heterogenous catalysts.
Read the full abstract

All authors Junfan Xia, Zhikai Jiang, Yaolong Zhang, Bo Peng, Hua Guo, Bin Jiang

Dynamical Systems

Many-Revolution Low-Thrust Transfers within Periodic Orbit Families

This paper explores the use of a reduced manifold space initialization method for the generation of both time and fuel optimal many-revolution, low-thrust transfers between three-body periodic orbits within a family. Libration point orbit families are approximated by multi-segment Chebyshev polynomials and 1-dimensional Fourier series. Control is mapped into the family space, and a 1-dimensional targeting scheme in the family space dynamics is used to generate an initial guess for the phase space, directly accounting for winding. An integral collocation-based direct method is then used to first continue a feasible trajectory into the phase space, and then optimize for a particular cost function. The methodology is first applied to the distant retrograde orbit family. Modifications for time-regularized dynamics are then discussed, with an ensuing example of many-revolution transfers in the $L_2$ halo family.
Read the full abstract