Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Wednesday 9 September

Optimization and Control

New Pareto Z-eigenvalue localization theorems for tensors with applications

In this paper, we establish some new localization sets for Pareto Z-eigenvalues of tensor eigenvalue complementarity problems and discuss the relationships among existing ones. As applications, a sufficient condition for the solution existence of the tensor complementarity problem (TCP) and the tensor generalized eigenvalue complementarity problem (TGEiCP) are presented. In addition, this condition is also used to test the vacuum stability of a general scalar potential
Read the full abstract

Probability

Local Anticoncentration for Gaussian Boson Sampling via Conditional Wishart Geometry

Local anticoncentration controls the conversion of additive output probability estimates to relative estimates in Gaussian boson sampling. We prove a center-uniform quadratic small-ball bound for hafnians of complex Gaussian transpose Gram matrices for $2n$ selected modes, with an explicit finite coefficient that is polynomial when the number of Gaussian rows is at least of order $n^2/\log n$. Letting the Gaussian row count tend to infinity yields anticoncentration for the independent complex symmetric Gaussian ensemble. For the transpose Gram hafnian bound, the proof compresses dependent hafnian cofactors by Fourier interpolation and controls their self-coupled variance through a conditional Wishart identity that preserves the transpose Gram coordinate. Independent perturbations satisfy related bounds, and a positive mixture density shows that the local quadratic power is exact. Together with the uniform hiding theorem in the companion paper, the result gives local anticoncentration for finite Haar interferometers.
Read the full abstract

Computer Vision

A Low Grounding Score Is Not an Ungrounded Judge: Identifying the Perceptibility Confound in Multimodal Oversight

Model judges now supervise multimodal systems at scale, filtering training data, selecting outputs, and supplying the reward that shapes multimodal reasoning models. Trusting one means first checking that it uses its evidence, and that check is itself worth scrutinizing, so we ask whether a counterfactual probe of visual grounding measures what it claims to. The probe edits the image so the ground truth flips, holds the reasoning trace fixed, and asks whether the verdict follows. We formalize it as the Verdict Grounding Score and show it cannot be read the way such scores are read. A verdict responds only to an edit that reaches the judge's decision-relevant reading, so the score is capped by how perceptible the edit is, and unless editing makes the attribute easier to read, the error is one-sided: the score can only make a judge look less grounded than it is. The practical failure is therefore a false alarm, an auditor discarding a usable overseer. Under assumptions we state, we show this missing quantity is not merely bounded but identified from three quantities the same audit protocol already collects, which makes the false-alarm rate directly measurable rather than merely a concern. Auditing nine judges, we find the predicted ordering holds strictly across our entire primary pool, and the typical judge there acts on only about half of the edits whose attribute it can otherwise resolve. Applying a conservative rejection threshold certifies several cells as false alarms outright, the clearest being a judge that detects the injected error essentially every time while still scoring as if it had not used the image at all. The rule that follows is that an image-side counterfactual score should never be reported alone: a detection probe on the unedited image upper-bounds it, certifies its false alarms, and costs nothing extra to run.
Read the full abstract

Strongly Correlated Electrons

Emergence of electronic modes and triplet pairing from spin-1 antiferromagnetic insulators in the Kanamori-Hubbard model

The electronic band structures of conventional band insulators generally remain unchanged under doping via chemical-potential shifts or increasing temperature. However, in Mott and Kondo insulators described by models such as the Hubbard and Kondo lattice models, distinct electronic modes with momentum-shifted magnetic dispersion relations have been shown to emerge from band edges into the gap. Here, by investigating the two-orbital Kanamori-Hubbard model, this study elucidates how and why the band structures of spin-1 antiferromagnetic insulators change with doping and temperature, revealing emergent spectral features characteristic of orbitally degenerate systems. The nature of the emergent electronic modes is clarified using weak-hopping effective theory, selection-rule analysis, and numerical calculations. In a typical parameter regime, the doping-induced modes in each orbital reflect distinct types of spin modes, while the temperature-induced modes primarily reflect the low-energy spin mode in both orbitals; in both cases, the emergent modes exhibit momentum-shifted spin-mode dispersion relations. The momentum regimes of these emergent modes can differ between orbitals if signs of the hopping parameters differ. For small inter-orbital repulsion, onsite spin-triplet pairs can form after doping, leading to a pair-breaking gap, while the temperature-induced electronic modes continue to reflect the low-energy spin mode without a pair-breaking gap. Moreover, the emergence and characteristics of electronic modes in nonequilibrium states by spin or charge perturbations are elucidated. These findings provide a comprehensive understanding of how the interplay among the spin, charge, and orbital degrees of freedom in large-spin antiferromagnetic insulators leads to emergent electronic modes, opening avenues for exploring unconventional quantum phenomena in strongly correlated systems.
Read the full abstract

Number Theory

PageRank on Lubotzky–Phillips–Sarnak graphs

We compare return probabilities for random walks stopped at an independent geometric time on Lubotzky–Phillips–Sarnak graphs and on the infinite regular tree, with the McKay measure giving the tree value and the Ihara zeta function describing the additional returns caused by cycles. We estimate this correction when the mean walk length is comparable to the length of the shortest cycle, a regime in which walks long enough to traverse a cycle receive non-negligible weight. Unconditionally, we prove an upper bound with a double logarithmic factor and bounded fixed moments over prime levels after multiplication by the number of vertices. Under the Generalized Riemann Hypothesis for quadratic Dirichlet $L$-functions, the correction is of the same order as the reciprocal of the number of vertices.
Read the full abstract

Combinatorics

Counting Weight-$k$ Italian Dominating Sets on Trees and Cycles

We study the problem of counting Italian dominating functions (IDFs) of each possible weight on paths, trees, and cycles. An Italian dominating function on a graph $G=(V,E)$ is a function $f\colon V\to\{0,1,2\}$ such that every vertex $v$ with $f(v)=0$ satisfies $\sum_{u\in N(v)}f(u)\ge 2$; the Italian domination polynomial $D_I(G,x)=\sum_k d_I(G,k)\,x^k$ records the number $d_I(G,k)$ of such functions of weight~$k$. This polynomial is the Italian analogue of the well-studied domination and Roman domination polynomials, and computing its coefficients on a fixed graph class is, in essence, a constrained integer composition problem: passing to the deficit variables $y(v)=2-f(v)$, each coefficient $d_I(G,2n-j)$ counts the number of compositions of $j$ into $n$ parts, each at most~$2$, that additionally satisfy a local adjacency capacity constraint. We make this connection explicit and use it to derive closed-form expressions for $d_I(C_n,2n-j)$, $j=0,\ldots,5$, as trinomial coefficients minus explicitly enumerated forbidden configurations. On the algorithmic side, we give dynamic programming algorithms that compute all coefficients of $D_I$ on path graphs ($O(n^2)$ time, $O(n)$ space), on general trees ($O(n^2)$ time, $O(n)$ space, via a post-order merge with a deficit parameter), and on cycle graphs ($O(n^2)$ time, $O(n)$ space, via boundary conditioning), together with formal correctness proofs and complexity analyses. The known values $\gamma_I(P_n)=\lfloor n/2\rfloor+1$ and $\gamma_I(C_n)=\lceil n/2\rceil$ are recalled with references rather than reproved. All numerical results are verified against brute-force enumeration, and complete tables are provided.
Read the full abstract

Social and Information Networks

Structural Properties of DomiRank Centrality: Symmetry Axioms, an Entropy Bound, and Spectral Interpolation

DomiRank is a node-importance algorithm for unweighted networks, defined by a dynamical model governed by a competition-strength parameter, a dominance threshold, and a decay rate. We establish three structural results. First, on symmetry. Automorphism-related vertices share the same DomiRank value, and a graph whose values are pairwise distinct must be asymmetric. The orbit count bounds the number of distinct values, and regular and vertex-transitive graphs admit closed forms. Second, on entropy. The maximum of the DomiRank entropy over connected graphs, log_2 N, is attained if and only if the graph is regular. The initial entropy derivative is the degree-log-degree edge covariance, making the decrease rigorous near vanishing competition for structurally disassortative graphs, and an assortative counterexample shows it does not extend to all graphs; a sufficient-condition theorem and a refined conjecture cover the rest. Third, on spectrum. The normalized distribution is sigma-invariant if and only if the degree vector is an adjacency-matrix eigenvector. The parameter sigma interpolates continuously between degree and smallest-eigenvector centrality within the resolvent (Bonacich-type) family, its sensitivity being a second-order suppression operator on the degree vector. Graph transformations are also characterized, including the m-product. Experiments on four real networks and synthetic families up to 10^4 vertices confirm the interpolation and exhibit monotone entropy decrease, continuous ranking interpolation, decaying positive fraction and ranking reversals; where the positive fraction has decayed far from one, the limit describes the solution vector rather than the induced ranking. DomiRank is thus a tunable complement to principal-eigenvector centrality. A companion study quantifies how much of its task-level edge over a degree baseline survives degree correction.
Read the full abstract

Hardware Architecture

SyntheticHLS: Building Diverse Synthetic High-Level Synthesis Datasets using LLMs

Deep learning and large language models (LLMs) are rapidly gaining adoption in semiconductor design, driving demand for training datasets. Most efforts focus on hardware description languages (HDLs) while designs for high-level synthesis (HLS), a popular approach to domain-specific accelerators, remain scarce. HLS dataset efforts emphasize manual curation or design parameterization, seldom addressing high-quality LLM-based generation or diversity in code length, hierarchy, design-space size, latency, resource utilization, and application domain, potentially limiting model generalization. We propose SyntheticHLS, a framework for generating large-scale, complex, diverse synthetic HLS datasets using LLMs. Its two key ideas are: 1) an iterative feedback-guided mutation loop that uses paired HLS source code and design-space specifications to incrementally transform seed designs into more complex, scalable designs; and 2) quantitative metrics of HLS design complexity and design-space scalability that serve as measurable objectives for LLM-guided mutation. We systematically cross-validate an HLS Quality-of-Results (QoR) deep learning model trained and tested across common HLS benchmarks, zero-shot synthetic designs, and iteratively mutated synthetic designs. Synthetic designs transfer well to common benchmark test sets while the reverse does not hold. SyntheticHLS's iteratively mutated designs provide the most generalizable training corpus among the datasets studied. Analysis of the mutation process and dataset shows that metric-guided trajectories consistently improve targeted complexity and scalability objectives without regressing non-target metrics. Mutated designs span a substantially broader, more diverse design space than zero-shot generated designs. Our framework, dataset, and evaluation are open-source: https://github.com/sharc-lab/synthetic-hls.
Read the full abstract

Representation Theory

Self-Orthogonal $\tau$-Tilting Modules and Tilting Modules III: Finitistic Dimension

Let $A$ be an Artin algebra and let $T$ be a $\tau$-tilting right $A$-module. Put $C=A/{\rm Ann}_A(T)$. We prove that finite little finitistic dimension of $C^{\rm op}$, together with a finite range of self-extension vanishing for $T$, forces $T$ to be a classical $1$-tilting module. We also show that if the finitistic conjecture holds for all Artin algebras then the self-orthogonal $\tau$-tilting conjecture holds for all Artin algebras. Moreover, a counterexample to the self-orthogonal $\tau$-tilting conjecture will give a counterexample to the finitistic dimension conjecture. As an application, the self-orthogonal $\tau$-tilting conjecture holds for Artin algebras with radical cube zero.
Read the full abstract

Machine Learning

OMP-MoE: Efficient Expert Pruning for Mixture-of-Experts LLMs via Orthogonal Matching Pursuit

Mixture-of-Experts (MoE) models enable efficient scaling of large language models but face critical deployment challenges due to massive memory requirements. Existing pruning methods either incur prohibitive search costs or neglect the dynamic interdependencies between experts. To address these challenges, we present OMP-MoE, a novel training-free compression framework for reducing expert redundancy in MoE-based LLMs. Based on observations of expert contribution patterns, we reformulate the pruning problem as a sparse signal reconstruction task solved through Orthogonal Matching Pursuit. Specifically, our method first treats individual expert contributions as dictionary atoms and selects experts that greedily minimize reconstruction error with linear computational complexity. Then, we optimize cross-layer expert allocation through a water-filling strategy that accounts for both reconstruction quality and routing stability. Finally, we introduce OMP-MoE{\dag}, an adaptive inference mechanism that dynamically adjusts expert activation based on energy prediction. Comprehensive experiments on Qwen, DeepSeek-V2, GPT-OSS, and Mixtral MoE demonstrate consistent improvements over existing methods at 25-50% pruning ratios. For Qwen3-30B-A3B at 50% compression, we retain 93.3% of original performance, achieving 33$\times$ faster search and 1.55$\times$ inference speedup. Codes will be available after acceptance.
Read the full abstract

All authors Dezhi Li, Lujun Li, Qiyuan Zhu, Hao Gu, Bei Liu, Sirui Han, Yike Guo

Machine Learning

Replay in the Silent Degrees of Freedom: Continual Learning Without an Offline Phase

Replay-based continual learning rehearses past data either in an offline phase, during which the agent stops acting, or interleaved with the live stream, where it perturbs the computation serving the current input. An agent that learns in deployment can afford neither. Motivated by local sleep, the use-dependent off periods of individual cortical circuits in awake animals, we show that replay can instead be written into the degrees of freedom the current input leaves unused. In a network with k-winner-take-all hidden layers, confining replay updates to synapses whose presynaptic unit is silent or whose postsynaptic unit is inactive leaves the hidden computation on the current batch invariant: exactly so for silent and suppressed units, and for all but 0.3% of samples in practice. A refractory rule under which units that have just fired sit out the next competition doubles the width of this channel and carries most of the accuracy. On class-incremental split-MNIST the resulting learner, with no offline phase, matches or exceeds the best offline rehearsal schedule and outperforms experience replay, ER-ACE and unmasked interleaved replay, each re-tuned under the same micro-batch schedule. Against DER++ the comparison splits by protocol: with five epochs per task DER++ leads by 1.5 points once it runs under that schedule, and in a single pass over the stream, the regime closest to the agent deployment setting, the system leads it by 1.6 while offline rehearsal falls 15 points behind. On split CIFAR-10 it again leads offline rehearsal and experience replay, but trails ER-ACE and DER++ by two to three points. The construction is not tied to the local learner: on a backprop network with k-winner-take-all hidden layers under the same schedule, refractory rotation adds half a point in five epochs and three in a single pass, and isolation again costs nothing on top of it.
Read the full abstract

Computation and Language

ChestPheNoT: Deployable, Auditable Label-Status-Evidence Extraction from Radiology Reports

Structured phenotype extraction from radiology reports supports cohort construction, quality auditing, and clinical analytics, but practical deployment requires local inference and auditable predictions, while expert annotations remain scarce. Conventional labelers provide structured findings and assertion states but no supporting evidence, while API-hosted large language models may be unsuitable when clinical text cannot leave institutional infrastructure. We present CHESTPHENOT, a compact 0.5-3B language model that jointly extracts finding labels, three-class status (present/absent/uncertain), and verbatim supporting evidence spans. CHESTPHENOT is trained using hybrid CheXbert+72B silver supervision followed by supervised fine-tuning and lightweight GRPO refinement. Across three human-annotated gold sets spanning in-distribution, cross-taxonomy, and cross-institution evaluation, the 3B model remains below its CheXbert silver teacher in distribution but is competitive under distribution shift, significantly surpassing CheXbert on cross-institution detection (+2.0 F1). Task-specific training also enables the 3B model to match or exceed substantially larger prompted models on most detection and status comparisons. For evidence-grounded extraction, over 99% of final evidence spans are locatable in the source report, and the 3B model achieves 47.5 auditable-F1, outperforming Qwen2.5-7B one-shot prompting by 7.6 points and approaching Qwen2.5-72B. These results demonstrate that locally deployable models can provide competitive and directly auditable radiology-report extraction without relying on external inference APIs. Code and the full extraction/judge prompts will be made available at https://github.com/yukkai/ChestPheNoT.
Read the full abstract

All authors Kai Yu, Chenyu Zhu, Zaifu Zhan, Meijia Song, Min Zeng, Xiaoyi Chen, Mingquan Lin, Rui Zhang

Computation and Language

Auditing and Repairing LLM-as-Judge Failures in a Production Text-to-SQL Pipeline

Production text-to-SQL pipelines often end with an LLM-as-judge whose agreement with human annotators has never actually been measured. When we checked ours, the deployed gpt-4o-mini judge agreed with two-author gold at only Cohen's kappa = 0.04 on a disagreement-enriched set and 0.42 on a uniform-random spot-check, over-flagging 77.1% of the human-FAITHFUL cases in the enriched set. Most of its over-flags trace back to a single mechanism we call GRADE-HALLUCINATION. A self-hosted Qwen3.6-27B replacement (kappa = 0.72) lands in the same range as Claude Opus 4.7 (kappa = 0.71); the head-to-head is underpowered at n = 96, but for the deployment decision that hardly matters, since Qwen costs roughly 1/300 as much per call. Ensembling does not help for free. Pairing the weak judge with a stronger one degrades agreement, whereas three strong judges under unanimity routing reach kappa = 0.79 at 89.7% auto-coverage. Applied out-of-domain, the same audit recipe flags 25.5% of BIRD-financial's expert-authored gold SQLs as candidate gold-SQL issues under our annotation protocol. Code and pre-registration are at https://github.com/JamesL404/synca-audit.
Read the full abstract

Computation and Language

Not All Memories Are Equal: Hierarchical Collaborative Memory for Validity-Aware Retrieval in LLM Agents

In team collaboration scenarios, memory is heterogeneous and continually evolving. Team memories capture collective decisions, protocols, and current consensus, while individual memories preserve member-specific observations, execution traces, and intermediate progress. Existing memory-augmented systems typically retrieve from all stored memories as a flat pool, ranking them by semantic relevance, importance, or recency without modeling hierarchical structure or evolving validity. As a result, they often surface semantically relevant but outdated or conflicting memories, especially individual memories that no longer align with current team consensus, instead of prioritizing currently valid memories. This is particularly problematic when collaborative LLM agents answer user questions, since their responses should be grounded in valid memories. We propose HiCoMER, a framework for hierarchical collaborative memory management and validity-aware retrieval in LLM agents. HiCoMER first maintains the validity of team and individual memories and then retrieves memories that remain valid, rather than retrieving directly from all stored memories. It consists of three components: a Hierarchical Memory Conflict Updater, a Validity-Aware Memory Retriever, and a Memory-Grounded Answer Generator. To evaluate HiCoMER, we construct two new datasets for memory-grounded question answering in collaborative settings. Experiments on both datasets show that HiCoMER consistently outperforms strong baselines by reducing outdated retrieval, preserving current team consensus, and improving downstream QA quality.
Read the full abstract

All authors Yufei Shi, Rujing Yao, Ang Li, Yang Wu, Zhuoren Jiang, Xiaozhong Liu

Computation and Language

Manifold Projection and Iterative Autoencoder Refinement for Masked Language Modeling

In Transformer-based masked language models, attention is the primary mechanism for context mixing, but there are other ways to mix data across tokens. Recent attention-free mixers replace attention with fixed or hypernetwork-generated MLPs, alternating their dynamic, content-dependent weighting for computational simplicity. We build an alternative that gets the same property from a low-rank bottleneck autoencoder. We replace attention with a stack of autoencoder-based mixing modules, one operating over local neighborhoods, one over the full sequence, and one across attention heads, each compressing and reconstructing its input through a bottleneck, and its width is a hyperparameter rather than a training effect. In masked positions, we introduce an iterative refinement procedure that has two distinct steps. A pulling step that pulls an embedding representation toward a weighted average of its neighbors, and a correcting step that projects the result back to the learned manifold via an autoencoder. Our architecture achieves a significant portion of attention's performance at about $1.9 \times$ fewer FLOPs when pretrained on C4 and evaluated with parameter-matched BERT baselines. Our model equals parameter-matched BERT and TinyBERT baselines on the rarest-token frequency bucket using a frequency-aware training schedule that samples rare tokens more than uniformly for the masking tasks.
Read the full abstract

Algebraic Geometry

Non-smoothable Calabi-Yau threefolds from reflexive polytopes

We construct projective Calabi-Yau threefolds with isolated Gorenstein canonical singularities that admit no smoothing, as anticanonical hypersurfaces in Gorenstein Fano toric fourfolds. One example has a unique singularity, the anticanonical cone over the first Hirzebruch surface. Another has a unique singularity whose reduced miniversal base is a smooth curve but whose deformations are all singular. The construction combines Altmann's deformation theory with the smoothing criterion of Corti-Filip-Petracci: a two-dimensional face of the ambient fan polytope with primitive edges obstructs smoothing if it has no Minkowski decomposition into unit segments and unimodular triangles. We also classify the 217 isolated Gorenstein toric threefold singularities whose polygon edge vectors have coordinates in $[-2,2]$ in some lattice basis, recording their deformation types and smoothing components. Subject to completeness and absence of repetitions in the database copy used, a scan of the Kreuzer-Skarke classification finds this obstruction in $39{,}175{,}536$ of its $473{,}800{,}776$ polytopes (approximately $8.27\%$), whose generic anticanonical hypersurfaces therefore admit no smoothing.
Read the full abstract

Algebraic Geometry

Affine Spherical Mass and a Gaussian Kinematic Formula for Tropical Two-Fans

Let $F$ be an effective balanced rational two-dimensional fan in a rank-four lattice. We introduce an affine spherical mass \[ M_*(F)=\inf_{covol_g(N)=1}M_g(F) \] and prove the two-sided comparison \[ \sqrt{\frac83}\,m(F)\leq M_*(F)\leq 3\sqrt{2\,q(F)}. \] Here $q(F)=deg(F\cdot F)$ is stable self-intersection and $m(F)$ is the least degree of $F$ under a primitive rank-two lattice quotient. For a surjection $A$, one has $deg(A_*F)=deg(F\cdot[\ker A_R])$; hence $m(F)$ is equivalently the least stable intersection degree with a complete weight-one rational two-plane. In particular, \[ q(F)\geq\frac4{27}m(F)^2. \] The proof combines an exact Gaussian–kinematic formula for mixed stable intersection with an affine normalization principle for stationary spherical links. In dimension four, the identification $Gr^+(2,4)\simeq S^2\times S^2$ converts ambient isotropy into a sharp second-moment transversality estimate. We prove affine covariance for $M_*$, derive equality and stability conditions for the transversality estimate, and give examples where ambient isotropy coexists with zero mixed intersection, as well as a three-fan in six-space with zero self-intersection. Determining the optimal numerical constant remains open.
Read the full abstract

Instrumentation and Detectors

Real-Time SiPM Pulse Deconvolution for a High-granularity Dual-readout Calorimeter with Neutral Networks

The High-Granularity Dual-Readout Calorimeter (HG-DREAM) designed for FCC-ee aims to achieve unprecedented energy resolution through fine three-dimensional shower imaging, simultaneous measurement of Cherenkov and scintillation signals, and machine learning capabilities. A key enabling technology is longitudinal segmentation via timing measurements, where multiple energy deposits along optical fibers are distinguished by the arrival times of Cherenkov photons. We present quantized one-dimensional convolutional neural networks that predict pulse locations directly from 80-sample, 16 ns waveforms digitized at 200 ps, and evaluate their implementation in front-end electronics. Three architectures, a compact baseline and two variants using dilation and stride to reach the $\sim$ 20-sample receptive field set by the SiPM pulse width, achieve ROC AUC above 0.95 with four to eight filters per layer. Fixed-point quantization to a $<$16,8$>$ baseline costs less than $10^{-3}$ in AUC relative to the floating-point reference. Applied to pairwise deposits, the smallest model separates two pulses with better than 95% probability at longitudinal separations of 20 cm and above. Synthesized at 400 MHz, this model uses 9663 LUT on a Xilinx XCVU13P FPGA and occupies $\sim$ 18500 $\mu$$m^2$ in TSMC 28 nm CMOS with reconfigurable weights. Assuming a triggered readout architecture for FCC-ee, the $\sim$ 500 kHz Z-pole trigger rate will be the most stringent throughput requirement across all FCC-ee operating modes. The ASIC implementation of the model sustains 4.8 MHz, exceeding the 0.5 MHz requirement by an order of magnitude, while the reconfigurable-weight FPGA implementation fails to do so. Sustaining the 40 MHz FCC-ee Z-pole bunch-crossing frequency in the triggerless architecture is not feasible with the models developed and the hardware targeted in this work.
Read the full abstract

All authors Spencer Allen, Matteo Cremonesi, Jordan Damgov, Erdem Yigit Ertorer, Janna Goodman, Giuseppe Di Guglielmo, James Hirschauer, Shuichi Kunori, Peter Meiring

Materials Science

Efficient spectral Galerkin framework for nonlinear transient heat transfer in finite domains

Accurately modelling the temporal evolution of heterogeneous temperature fields requires resolving strong nonlinearities in the heat equation arising from temperature-dependent thermophysical properties, latent heats of transformation and any local heat sources/sinks. In this work, we present a spectral Galerkin (SG) framework to solve the fully nonlinear transient heat equation in finite domains to attain the accuracy of high-fidelity finite element (FE) simulations at considerably lower computational cost. The heat equation is reformulated into a linear reference problem with constant thermophysical properties and residual forcing terms. Solving the reference problem provides a complete three-dimensional orthonormal trigonometric basis, whose Galerkin projection reduces the heat equation to a set of modal ordinary differential equations (ODEs) in time; the reference operator is diagonal in the modal basis, and results in independent modal updates for a fixed nonlinear forcing. These ODEs can be integrated using exponential time differencing and iteratively corrected for nonlinearities. The SG method eliminates the global solve required by FE methods, and its use of structured grids allows efficient GPU parallelization. Applied to rapid laser-metal interactions, the SG solver reproduces high-fidelity FE temperature fields with less than 1% relative error while achieving 227-fold faster GPU runtimes. Applied to a part-scale laser scanning study [Ramani et al., Additive Manufacturing 52 (2022) 102643], the method shows that accounting for evaporation and latent heat more than halves their proposed processing metric. The source code of the SG heat solver and some worked examples are available at https://github.com/manasvupadhyay/spectral_galerkin_heat under the Apache 2.0 license.
Read the full abstract

High Energy Physics (Theory)

Dirac oscillator in a self-gravitating cosmic string: confinement spectra beyond the conical approximation

The gravitational field of a physical cosmic string is regular on the axis, curved across a finite vortex core, and only asymptotically conical. We test whether the bound-state spectrum of a transverse Dirac oscillator resolves this structure beyond the ideal-cone approximation. The fermion is treated as a test field on self-consistent Einstein–Abelian-Higgs vortex backgrounds, and the radial problem is cast as a generalized Hermitian eigenvalue problem. The numerical formulation is checked against the exact Minkowski and ideal-cone spectra and by independent shooting calculations. For a reference vortex family, the finite-core correction grows from the weak-confinement regime toward the percent level. When the normalized asymptotic cone is held fixed, the spectrum still varies with the measured Higgs-core radius, and the sampled backgrounds bracket a change of sign of the ground-state correction. The bound-state spectrum therefore contains information about the resolved core profiles that is not fixed by the asymptotic cone alone.
Read the full abstract

General Mathematics

n-tuple-periodic Solutions of the Navier-Stokes Equations: Existence and Obstructions

We introduce a phase-angle framework for constructing and classifying n-tuple-periodic solutions of the incompressible Navier-Stokes equations, and use it to determine, dimension by dimension, when such solutions exist and why they fail. The velocity field is a difference of two n-fold cyclic trigonometric products, made divergence-free for every phase and every n >= 3 by a telescoping identity. The problem then reduces to two explicit conditions on the n phase angles: a reduction condition, necessary and algebraic, and a pressure-integrability condition, sufficient and differential. Separating them allows a failed construction to be located rather than merely recorded. Unlike the stream-function route on which earlier tri-periodic solutions rest, the framework is not confined to three dimensions, and does not presuppose the Beltrami property. For 3 <= n <= 8 we prove that unforced solutions exist at n=3 and n=4 and nowhere else. In three dimensions the admissible phases form exactly two families, mirror images distinguished by the sign of the helicity; these families are known, and we establish that the construction admits no others, settling a completeness question left open when they were found. In four dimensions integrability reduces a one-parameter curve to a single family, so the quarter-period offset is forced, not assumed. The obstructions divide by parity: arithmetic at n=5,7, where every admissible phase vector annihilates the field; differential at n=6,8, where the field survives but its convective term is not a gradient. Every obstructed dimension nevertheless carries an exact forced solution in closed form. These are not mere self-balancing: the phases are those at which strain and rotation are in exact pointwise balance, so that no pressure is generated at all. What obstructs the unforced problem becomes, in the forced problem, an exactly prescribable input.
Read the full abstract

General Physics

Constraining a model supported in Moiré gravity with a recent data release

Moiré theory is a revolutionary alternative arising from solid-state physics that can be extended to gravity, forming a framework called Moiré gravity. The idea behind it is that Einsteinian gravity has a validity limit and must be extended to Moiré gravity, in which the Friedmann equation changes and a new scalar field emerges. In this vein, in this paper, we propose an analytic solution for the Moiré scalar field coupled with the modified Friedmann equation, studying the viability and constraining the value for the main parameter of the scalar field and cosmology in general, through the use of datasets that contain cosmic chronometers, type Ia supernovae, intermediate-luminosity quasars, and hydrogen II galaxies. Our results allow for an accelerated phase with two peaks, but also predict an extremely decelerated stage, with $q(z=0)=15.376^{+6.403}_{-4.692}$ when the characteristic parameter of the theory is $\jmath=0.677^{+0.001}_{-0.002}$, in contrast to other models that decelerate at $z=0$.
Read the full abstract

All authors J. A. Astorga-Moreno, Miguel A. García-Aspeitia, A. Hernández-Almada, E. A. Mena-Barboza

Geometric Topology

Cubical residual complexes and profinite rigidity of real hyperbolic lattices

We study profinite rigidity of real hyperbolic lattices, with the main application in dimension three. The basic tool is a profinite version of the dual cube complex of a pair of filling curves on a surface, which we call a profinite cubical residue complex. We prove that an equivariant isomorphism between the completions of two such complexes is induced, after conjugation by a single element, by an isomorphism of the discrete groups and complexes, and that the same holds for free cocompact cellular actions of finitely generated residually finite groups on connected, locally finite, finite dimensional regular CW complexes. On a closed orientable surface of genus at least two, a theory of divisible fillings determines the completed cut tree of a nonseparating simple closed curve. Combined with the homological intersection criterion of Boggi and Zalesskii, this determines the completed dual square complex of a nonseparating filling pair, and hence realizes every isomorphism of profinite surface groups which preserves the two marked cyclic subgroups. In dimension three we combine these results with virtual fibering, Liu's profinite invariants of hyperbolic $3$-manifolds and the goodness of $3$-manifold groups. Massey products and virtual domination give cohomological integrality, and pseudo-Anosov dynamics, periodic orbit traces and cross sections of suspension flows give a correspondence of prime periodic orbits together with the fiber curves needed for the realization. Profinite rigidity of closed and of cusped lattices follows, the cusped case through cyclic orbifold fillings, and we obtain $\Out(\Gamma)\cong\Out(\wh\Gamma)$ for every lattice of $\PSL_2(\C)$.
Read the full abstract

Logic

A Note on Lowness for $\Pi^0_1$-Immunity

We show that the only degree which cannot co-enumerate a non-trivial $\Pi^0_1$-immune real is $\emptyset$, resolving a conjecture of the author. The main ingredient is that if a real $A$ has that all $A$-maximal sets are c.e., then $A$ has c.e. degree, which can be proved via a coding mechanism and two classical results of Lachlan.
Read the full abstract

Computer Vision

Harnessing Image Question Dependence for Better VLM Test-time Reinforcement Learning

Test-time reinforcement learning can adapt vision-language models (VLMs) to unlabeled target data, but its effectiveness is fundamentally limited by the reliability of self-generated learning signals. To assess the reliability of consensus-based learning signals, we analyze VLM test-time reinforcement learning across diverse VQA datasets and model sizes, revealing two limitations. First, gains from consensus-based test-time training largely come from answer normalization rather than content correction. Second, many initial VLM responses are incorrect due to the model's limited ability to jointly use the image and the question; consensus rewards derived from these outputs may preserve the resulting grounding errors rather than correct them. Motivated by these, we propose TTIQ, a test-time reinforcement learning framework that harnesses image-question dependence for better vlm adaptation. TTIQ teacher-forces each sampled response under the original image-question pair and its image- and question-ablated variants, using the resulting token-level likelihood changes to estimate dependence on each input. It combines image and question dependence with calibrated confidence to construct a response-level reward that favors jointly grounded responses, and uses the token-level signals to assign greater positive policy credit to tokens supported by both inputs. This design favors responses that are jointly grounded in the image and the question and sufficiently confident, rather than merely popular. Experiments across eight VQA datasets and multiple VLM sizes show that TTIQ achieves the best average performance at every model scale. It further generalizes across VLM families, while models trained on one dataset improve performance on unseen datasets without further training.
Read the full abstract

All authors Xinrui He, Ting-Wei Li, Junting Wang, Mengting Ai, Xinyu He, Hanghang Tong, Jingrui He

Robotics

Ergodic Control and Controlled Diffusion for Robot Learning: Review and Tutorial

Diffusion learning leverages the statistical mechanism of diffusion processes for learning, reasoning, and inferring complex distributions from data. Recent advances in diffusion learning have been transformative, with robot learning emerging as a key opportunity area, with applications spanning perception, control, and decision-making. At the same time, the statistical mechanism of diffusion processes can be controlled to shape the temporal evolution of the state distribution underlying robot trajectories, inducing ergodic behavior in robotic systems. The frameworks of controlled diffusion and ergodic control were developed around the same time as diffusion learning, and their theories and algorithms have increasingly converged. Ergodicity induced by controlled diffusion has several significant implications for robot learning, distinct from applying diffusion learning to robotics problems: it formally enforces statistical properties required to ensure optimality of robot learning, enables non-myopic search over uncertain information landscapes for data collection, and enables behavior specification based on spatial rather than temporal characteristics of trajectories. This survey introduces the intuition behind controlled diffusion for robot learning, explores its connection to diffusion learning, presents theoretical foundations and numerical tutorials for solving controlled diffusion and ergodic control problems, and reviews applications across robotics. Finally, we discuss key challenges and future opportunities in leveraging controlled diffusion for robot learning.
Read the full abstract

All authors Max Muchen Sun, Cem Bilaloglu, Ananya Rao, Stefan Ivic, Guillaume Sartoretti, Kathleen Fitzsimons, Ian Abraham, Sylvain Calinon, Todd Murphey

Computer Vision

VectorHarness: Recovering Editable, Relation-Preserving Structure from Scientific Graphics

Converting scientific graphics into editable representations remains a challenging problem for image-to-code generation because of their heterogeneous elements and complex layouts. Recent multi-agent reconstruction systems have advanced this line of work, but often follow a copy-paste paradigm: the reconstructed image closely resembles the original, while complex regions remain effectively uneditable. We instead formulate a different objective, raster-to-authoring reconstruction, which aims to recover an authoring representation that supports native, customized editing rather than mere visual replication. To this end, we present VectorHarness, a multi-agent framework for raster-to-authoring reconstruction that recovers heterogeneous components using type-appropriate native representations. Text, formulas, shapes, connectors, icons, charts, and tables are reconstructed as natively editable objects, while intrinsically image-based regions remain raster content. To systematically evaluate reconstruction quality, we introduce VectorHarness-Bench, which jointly assesses rendering fidelity, raster fallback coverage, executable object edits, and relation-preserving edits. Experiments show that VectorHarness improves executable edit success and relation preservation, reduces avoidable raster fallback, and maintains high visual fidelity across heterogeneous graphics.
Read the full abstract

Computer Vision

Reliable Egocentric Action Anticipation via Temporal Reliability Suppression and Compositional Graph Decoding

Wearable action anticipation systems must remain reliable despite missing frames, masking, and sensor noise, yet existing egocentric anticipation methods largely assume clean observations. We identify two complementary failure modes under temporal corruption: unreliable temporal evidence during encoding and implausible, low-support verb-noun compositions during decoding. We address them with a lightweight framework combining Temporal Reliability Suppression (TRS) and Robust Verb-Noun Graph (RVG) decoding. TRS predicts a per-frame suppression score from the projected input embedding and uses it as a learned key-side attention penalty at every encoder block and to derive reliability-weighted temporal pooling. RVG re-ranks verb-noun pairs using a PMI-based compatibility graph constructed from training labels. Under corruption-augmented training, TRS+CA+RVG reaches 29.1% average corrupted accuracy and 88.2% relative robustness across six corruptions, including three mechanisms absent during training, while reducing rare verb-noun predictions from 15.4% to 1.1%. Multi-seed and diagnostic experiments show that TRS responds to synthetic masking; shuffled-graph and frequency-only controls further indicate that RVG gains depend on genuine pairwise compatibility rather than marginal-frequency effects alone.
Read the full abstract

Atmospheric and Oceanic Physics

Aries: A Proprietary Medium-Range Weather Prediction Model for the Energy Industry

Medium-range weather forecasting underpins operational and planning decisions across the energy industry. Developing competitive weather models was once the domain of national meteorological centers, but recent advances in machine-learned weather prediction (MLWP) have opened the field to industry. We present Aries, a SwinTransformer-based MLWP model developed at InCommodities. Aries is trained on ERA5 reanalysis data at 0.25\textdegree{} resolution, predicting 74 prognostic and 11 diagnostic atmospheric variables. We evaluate the model on 2025 ECMWF Analysis initializations, ensuring a recent and strictly out-of-sample test period for all models compared. On 10-metre wind speed, Aries outperforms both ECMWF HRES and AIFS in terms of RMSE for lead times up to four days, while on 2-metre temperature it achieves RMSE on par with AIFS operational. These results demonstrate that proprietary development of competitive weather models is technically viable, supporting a broader set of forecasts available for operational and planning applications in the energy industry.
Read the full abstract

All authors Lukas Hedegaard Morsing, Arian Bakhtiarnia, Jonas Lynge Olesen, Tómas Bragi Björnsson Leth, Christian Gøbel Bach

Sound

ScorePrompts: Natural-Language Exploration of Symbolic Music Scores through Analysis

We present ScorePrompts, an interactive system in which users upload a score, receive natural-language descriptions of its musical structure, ask questions about specific passages, and inspect the corresponding analysis results in staff notation. Specialist MIR components first estimate harmony, tonality, cadences, formal boundaries, texture, and note-level roles, organizing their outputs at note, beat, measure, and piece levels. A schema-constrained language model converts these results into descriptions rather than inferring musical structure directly from raw MusicXML. For questions such as "What changes in measures 14-18?", a deterministic router selects the relevant measures and analytical levels and returns a concise response together with the underlying results and caveats. Verovio renders the score and links the returned information to cited measures and note-level attributes. The interface also exposes intermediate tables and disagreements between analytical levels. ScorePrompts is intended for exploratory score analysis and explanation, not score editing. The demo shows how existing analysis models, constrained language generation, Q&A retrieval, and notation-based visualization can provide natural-language access to symbolic music analysis while keeping intermediate results inspectable.
Read the full abstract