Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Systems and Control

Towards Passivity-informed Data-driven Predictive Control

We propose a passivity-informed data-driven predictive control framework based on passivity-preserving sampling of unknown continuous-time linear systems. We incorporate the passive supply rate into the objective functional alongside an output-regulation cost, thereby explicitly accounting for energy extraction. Despite being practically motivated, the resulting objective is not always convex; its convexity depends on the collected input-output training data. We exploit passivity-induced positive semidefiniteness to derive an equivalent reformulation that remains a convex quadratic program independently of the data. Furthermore, we establish recursive feasibility and asymptotic output convergence using passivity and a terminal equality constraint. Numerical simulations illustrate the closed-loop performance and computational benefits of the proposed approach.
Read the full abstract

All authors Guanru Pan, Oleksii Molodchyk, Mohammad Itani, Timm Faulwasser

Machine Learning

K-prop: Deep Online Learning by Backpropagating Temporal Kernels

Self-recurrence represents an important component for temporal processing in neural networks, both artificial and biological, where it appears naturally in the form of leaky integration. However, due to the complex temporal dependencies of the gradient, an efficient and effective method of online training is still lacking. Here we propose K-prop, a novel online learning method for leaky neural networks. It dramatically reduces the computational and memory cost of real time recurrent learning (RTRL) by propagating kernel gains. K-prop also outperforms existing RTRL approximations and achieves performance comparable to backpropagation through time in sequence learning, classification, and reinforcement learning tasks. We further analyze how its sole approximation affects gradient quality and identify conditions under which K-prop remains accurate and effective.
Read the full abstract

All authors Bingkun Liu, Paul Haider, Federico Benitez, Mihai A. Petrovici, Walter Senn

Human-Computer Interaction

Perceived Agency During Agent-Avatar Control Transitions in Virtual Reality

Prior beliefs about whether a virtual human (VH) is controlled by a human or a computer can shape how users interpret and respond to its behaviour. As emerging hybrid VH systems allow control to transfer between human operators and computer programs during an ongoing interaction, it remains unclear whether perceived agency updates with the actual controller or continues to reflect users' prior beliefs and first impressions. We investigated this question in a within-subject VR study with 24 participants across four conditions: always computer-controlled, computer-to-human, human-to-computer, and always human-controlled. We designed an interaction task in which participants acted as interviewers, asking the VHs a fixed set of questions. Before each interaction, participants were informed whether the VH was human- or computer-controlled, matching the actual controller at the beginning, but were unaware of possible control transitions. We found that perceived agency was strongly influenced by the controller presented at the beginning of the interaction, with no evidence of updating following the control transitions. Social evaluations including likeability and co-presence were generally higher in conditions where human control occurred later and covered a greater proportion of the interaction. The results suggest that prior beliefs and first impressions continued to shape perceived agency, whereas social evaluations were more sensitive to the VH's actual behaviour over the interaction.
Read the full abstract

All authors Jingyi Zhang, Ruijun Sun, Klara Brandstatter, Nels Numan, Anthony Steed

Number Theory

Ternary quadratic genera and generalized Hurwitz class numbers via quaternion orders

We classify the genera of primitive positive definite ternary quadratic forms in the extremal cases of Lehman's level and discriminant conditions. The genera are parametrized by their level, discriminant, and genus conductor, which also distinguishes local forms having the same isotropy. Using the correspondence with quaternion orders of level $(N_1,N_2)$, we express their weighted representation numbers and class numbers in terms of generalized modified Hurwitz class numbers. The formulas allow arbitrary admissible prime power exponents and include the additional genera obtained by Watson's transformation. As an application, we determine all $255$ one-class genera in these families and give an explicit representation formula for each. Of these formulas, $110$ concern nonsquarefree $N_1N_2$.
Read the full abstract

Sound

WaveGen: Truly End-to-End Waveform Generation via Internal Spectral Trajectory Learning

This work proposes WaveGen, a fully end-to-end diffusion-based text-to-waveform framework based on Spectral Forcing, which jointly learns an internal spectral trajectory and the target waveform trajectory within a single model, enabling direct end-to-end text-to-waveform generation. Unlike recent TTS pipelines, WaveGen does not rely on separately trained or externally decoded intermediate representations, such as neural audio codecs, Mel-spectrogram vocoders, or VAE-based latent models. In addition, WaveGen performs internal text-speech alignment within a single model, eliminating the need for external alignment modules such as duration predictors. To preserve a truly end-to-end formulation, our framework further avoids dependence on self-supervised representations, such as BERT-based text embeddings and Wav2Vec 2.0-based semantic representations, as well as semantic distillation methods for training optimization. To achieve this, we carefully design Spectral Forcing, a model architecture with disentangled diffusion heads, and a training objective based on a multi-scale spectral $v$-loss. We further validate the scalability of end-to-end TTS models by scaling the model size from 0.1B to 0.6B parameters. Using a single architecture, open-source data, and one-stage training, without external modules such as duration predictor or REPA, WaveGen demonstrates fully end-to-end waveform generation achieving state-of-the-art performance on the Emilia training dataset. In particular, WaveGen achieves a WER of 1.47 and a SIM of 0.75 on the Seed-en benchmark. Source code and checkpoints are available at https://github.com/sh-lee-prml/WaveGen
Read the full abstract

Robotics

Galilean Robots, Aristotelian Sensors: Joint Spatiotemporal Trajectory Estimation and Extrinsic Calibration

Multi-sensor state estimation requires accurate spatial and temporal calibration between heterogeneous sensors. Existing spatiotemporal calibration methods typically represent sensor extrinsic parameters as a rigid-body transformation together with a scalar temporal offset, causing the spatial and temporal parameters to enter the estimation problem through separate mechanisms. While the spatial transformation composes directly with the robot pose, accounting for the temporal offset requires evaluating the trajectory at a shifted time, typically through motion compensation, interpolation, or a continuous-time trajectory representation. In this work, we formulate joint trajectory estimation and spatiotemporal calibration using the special Galilean group for the robot trajectory and the Aristotelian subgroup for sensor extrinsics. This representation incorporates the coupling between motion and temporal offset directly through the group structure, avoiding the need for a separately parameterized continuous-time trajectory. We develop twist- and IMU-driven motion priors, accommodate both ideal and non-ideal clocks, and evaluate the framework with camera, position, and range measurements. Simulated and real experiments demonstrate calibration and trajectory-estimation performance comparable to established methods across heterogeneous sensor configurations. Our simplified framework therefore offers a quality-of-life improvement over existing methods without compromising performance.
Read the full abstract

Probability

Probabilistic counting of labelled trees with prescribed vertical edge profile

The vertical edge profile of a labelled plane tree $t$ is the collection of integers counting the number of oriented edges from label $i$ to label $j$ for every possible pairs of labels $(i, j)$. In a previous work by the second author, it was shown that a process constructed from the vertical edge profile forms a Markov chain for a wide class of models of random labelled plane trees where edge label increments are constrained to lie in $\{-1, 0, 1\}$. The purpose of this paper is twofold. First, we continue the study of Markov properties of the vertical edge profile. In particular, we derive general formulas for the transition kernel and the initial distribution of a class of Markov chains related to the vertical edge profile. Second, we apply these results to the probabilistic enumeration of trees with a prescribed vertical edge profile. These applications yield a unified probabilistic derivation of several enumeration formulas. In particular, we recover some results established Bousquet-Mélou and Chapuy for binary trees and Cayley trees. These enumeration formulas admit a remarkable factorized form, already noted in previous works. The probabilistic methods developed here explain this phenomenon by showing that these factorizations arise naturally from the Markov structure underlying the vertical edge profile.
Read the full abstract

Computer Vision

EditJudge-Bench: Auditing VLM Image-Edit Judges with Synthetic Ground Truth

Vision-language models (VLMs) are increasingly used as automated judges for image-editing benchmarks, yet their reliability is poorly understood. Auditing these judges is difficult because real edited images lack controlled ground truth and entangle the requested changes with model artefacts, unintended side effects, and ambiguous degrees of success. In this paper, we introduce EditJudge-Bench, a diagnostic benchmark of 1,382 synthetic before/after image pairs rendered in Blender, with edits parameterised across nine categories derived from real user requests and exact ground truth from known 3D scene parameters. Because every edit is generated by a known intervention on a controlled scene, EditJudge-Bench evaluates VLM judges as binary verifiers of whether an instruction was correctly executed, isolating judge perception from editor quality. Across 25 judge configurations, we find substantial reliability gaps: (a) the widely adopted VIEScore template reaches only 69.0% AUROC with GPT-4o; (b) fine-tuned judges score 51.7-66.3%; spatial edits such as rotation and shape remain challenging; (c) prompt-formatting choices like image-text ordering shift performance by up to 14.1 points. Despite using only synthetic data, judge rankings on EditJudge-Bench correlate strongly with two real-image human-preference benchmarks ($\rho$=0.95 and $\rho$=0.71), indicating that controlled audits surface failure modes relevant to real evaluation.
Read the full abstract

Analysis of PDEs

Existence of a non-zero steady-state solution to a population density equation subject to horizontal transfer

From plasmid dynamics and efflux pump expression in bacteria to risk-taking behaviour in human groups, horizontal interactions can drive the transmission of deleterious traits that spread despite carrying no selective advantage, thereby burdening the population. How can such detrimental traits persist, and under what conditions can the population avoid extinction? To address these questions, we model a population structured by a continuous trait representing individual burden within a non-linear, non-local integro-differential framework. From an analytical perspective, mutations are modeled via a jump operator rather than a diffusion term; consequently, the mutation operator lacks regularizing properties and is generally non-self-adjoint. Moreover, working in a weighted L 1 setting precludes the direct application of standard Krein-Rutman theory. To overcome these hurdles, we construct a custom fixed-point operator by first freezing the non-linearities to solve an auxiliary eigenvalue problem, and then tuning the resulting principal eigenvalue. Applying the Leray-Schauder fixed-point theorem to this operator rigorously yields the existence of a non-trivial steady state. Finally, numerical simulations illustrate the qualitative dynamics of the model and delineate the parameter regimes governing population persistence and extinction.
Read the full abstract

Geometric Topology

Rationally Null-homologous Floer Simple Knot

We prove that every rationally null-homologous Floer simple knot in a closed, oriented, irreducible three-manifold with nonzero Thurston norm is unknotted. The proof combines a cut-and-reglue comparison for extremal Floer groups with Binns's theorem on Floer-minimal knots in balanced sutured manifolds.
Read the full abstract

Machine Learning

Can the Confidence and Decisions of System 1 Models Be Trusted? An Explainability and Perturbation Audit of Jev-Style Decision Models

"System 1" decision models (pioneered by TypeSafe AI's Jev) replace autoregressive text generation with a single forward pass returning typed probability distributions over declared schemas (Choice, Score, Noul). While deployed for low-latency routing and automated triage, their statistical calibration and post-hoc explainability remain uncharacterized. We introduce an open, deterministic audit benchmark evaluating System 1 decision models across calibration under label-preserving perturbations, cross-question interference, and post-hoc attribution faithfulness. Evaluating five open-weight models (0.8B to 27B) alongside generative System 2 baselines – both single-token logprob scoring and structured JSON/CoT decoding – across NLI (e-SNLI), adversarial syntactic shortcuts (HANS), synthetic clinical triage (N=900), prompt injection defense (Tensor Trust, N=250), and an 86-case logic-gate trigger suite yields five findings: (1) provider confidence is an exact closed-form dispersion transform (1 - H(p)/ln K on sub-10B models; normalized margin on Clef-27B) rather than empirical accuracy; (2) bundling multiple questions in one request induces directional cross-question interference (11.7% to 22.4% binary flips on Tev1-0.8B, vs. 0.1% to 0.2% on Tev1-4B); (3) System 1 models with >= 4B parameters are faster and better calibrated under syntactic shifts (ECE_w = 0.030 to 0.087) than logprob (0.144) and JSON (0.378) baselines; (4) for prompt-injection defense (N=250), >= 4B models achieve 0.988 to 0.998 AUROC with up to 95.2% recall and 0.0% false alarms at 48 to 60 ms; and (5) single-pass inference makes 512-coalition XAI interactive (1.3 to 3.2 s/input), while KernelSHAP with a logit link (S1-SHAP) mitigates softmax saturation on disjunctive OR gates (89.3% to 100.0% gate recovery vs. 57.1% to 82.1% in prob space; p_adj = 0.004).
Read the full abstract

Lattice Field Theory

Symanzik Effective Hamiltonian for Lattice Quantum Field Theories

We propose a framework to construct the Symanzik effective Hamiltonian for lattice quantum field theories directly based on dimensional analysis and symmetries in the Hamiltonian formulation. This allows to classify all operators which can contribute at a given order in the spatial lattice spacing. This list can be reduced using the equations of motion and other identities. We exemplify the framework for $\phi^4$ and pure gauge theory, obtaining results consistent with earlier publications for the case of classical improvement. It provides the means to develop a non-perturbative improvement programme in the Hamiltonian formulation of lattice quantum field theories in the future.
Read the full abstract

Combinatorics

Chvátal's Toughness Conjecture for Dense Almost Regular Graphs

Chvátal's toughness conjecture asserts that there is a constant $t_0$ such that every $t_0$-tough graph on at least three vertices is Hamiltonian. Most partial results on this conjecture concern graphs with a restricted structure, such as chordal graphs and $H$-free graphs for small linear forests $H$. We confirm the conjecture with $t_0=4$ for dense almost regular graphs, without any structural assumption: for every $\alpha>0$, there exist $\varepsilon>0$ and $n_0$ such that every $4$-tough graph $G$ on $n\ge n_0$ vertices with $\delta(G)\ge\alpha n$ and $\Delta(G)-\delta(G)\le\varepsilon n$ is Hamiltonian. This result also bears on a conjecture of Bollobás and, independently, Häggkvist from the 1970s that every $k$-connected $D$-regular graph on $n$ vertices with $D\ge n/(k+1)$ is Hamiltonian. This conjecture is false for every $k\ge4$, even for large $n$. Our result shows that its conclusion holds for all large $n$ when $k$-connectivity is replaced by $4$-toughness, and that the degree condition can then be weakened to $D\ge\alpha n$ for an arbitrarily small fixed $\alpha>0$. We also construct $3/2$-tough regular non-Hamiltonian graphs of linear degree, showing that the constant $4$ cannot be replaced by $3/2$.
Read the full abstract

Operator Algebras

The space of Dimension functions on a $C^*$-algebra

In this paper, we prove that the space of normalized dimension functions on a unital $C^*$-algebra is either empty or a Choquet simplex. As a byproduct, we show that $W(A)$ is hereditary in $\operatorname{Cu}(A)$ for every $C^*$-algebra $A$. The argument combines an exact factorization for Cuntz comparison with Birkhoff's decomposition of lattice valuations. Moreover, we show that the space of normalized rank functions on a unital ring is either empty or a Choquet simplex.
Read the full abstract

High Energy Astrophysical Phenomena

Binary pulsar studies with multi-wavelength sky surveys. II. Companion star follow ups

The identification of the stellar companions to binary pulsars and the extraction of the physical information encoded in their optical spectra are key to study the evolution of the binary system and how this is influenced by the interactions between the two stars. Obviously, extending the characterisation of the companion star spectra over an as large as possible wavelength range allows one to maximise the amount of information. In Paper I, we used 11 source catalogues built upon multi-wavelength (optical, infrared, ultraviolet) imaging sky surveys to search for the stellar companions of 144 field binary pulsars (i.e. not in Globular Clusters) selected from the Australia Telescope National Facility (ATNF) pulsar catalogue. For nine of these pulsars, for which the companion star had been already identified, we obtained new spectral information from imaging photometry in the surveys' bands. In particular, for some of these pulsars we provided the first possible detections of the companion star in either the ultraviolet or mid-infrared. Here, we serve this additional flux information as a base to update the spectral energy distribution (SED) of the companion stars and the multi-epoch catalogue photometry to look for flux modulation at the orbital period of the binary system.
Read the full abstract

All authors R. P. Mignani, C. Pallanca, V. Testa, A. Corongiu

Statistics Theory

Linear Functional Estimation under Multiplicative Measurement Error with Unknown Error Distribution

In this work we consider i.i.d. observations from a multiplicative deconvolution model, i.e. generated by $Y= X\varepsilon$, where $X$ and $\varepsilon$ are independent strictly positive random variables. The goal is to statistically infer the unknown signal $X$. Starting from the scenario where the probability density $f_\varepsilon$ of the error is known, we revisit local estimation strategies for several quantities of interest including the probability density function f and the survival function $S^X$ of the signal $X$. We use these results to propose a thresholded plug-in estimation technique for linear functionals in situations where the distribution of the error $\varepsilon$ is unknown, while having some additional data at our disposal. This estimation procedure combines the reconstruction of the unknown Mellin transform of $f$ as well as its regularized inversion by a spectral cut-off. We establish oracle-type upper bounds and discuss rates of convergence in explicit scenarios including estimating $f$ and $S^X$ pointwise. In addition, we develop a fully data-driven choice of the cut-off parameter via model selection, which is able to obtain the oracle risk up to a constant without any previous knowledge of the error distribution. The results are illustrated in a simulation experiment, where we compare different classes of distributions in settings with known and unknown $f_\varepsilon$.
Read the full abstract

Atmospheric and Oceanic Physics

Is AMOC collapse too unstable in climate models?

It has been a long-held paradigm that the Atlantic Meridional Overturning Circulation in coupled climate models is too stable, underestimating the possibility of a collapse. A recent analysis of extended simulations to year 2300 in the Coupled Model Intercomparison Project, CMIP6, demonstrated a partial AMOC collapse where the deep circulation completely vanishes. This collapse was simulated by at least one model in all emission scenarios, and it occurred in every model forced by the highest emission scenario when extended to year 2300. Given the modelling community's previous emphasis on stability in climate models, in this perspective, we explore the alternate possible mechanisms by which the AMOC in CMIP6 might have become too unstable. This is motivated by the lack of evidence of downward trends in the AMOC in the observational records, and a subset of CMIP6 models displaying too high climate sensitivity and too strong sensitivity to aerosol removal, both of which could predispose them toward a spurious collapse. For these reasons, it is critical to evaluate the mechanisms by which CMIP6 models simulate AMOC collapse, to be able to develop programs further investigating the trustworthiness of simulated AMOC collapse scenarios.
Read the full abstract

Statistical Mechanics

Holonomy, Spectral Topology, and Detailed Balance in Classical and Open Quantum Systems

Band topology and non-Hermitian spectral theory are now applied to mechanical metamaterials, chiral active fluids, stochastic networks, and open quantum systems. We give a tutorial synthesis organized by holonomy and by one question: what can the spectral data of a linear generator reveal about detailed balance? Three holonomies recur: the Berry phase of a band, the Burgers vector of a dislocation (the integrated torsion of the defect geometry), and the cycle affinity of a Markov network (the holonomy of the connection $\ln(W_{ij}/W_{ji})$). We rederive the Chern numbers $C_m=m(\mathrm{sgn}\,\Omega+\mathrm{sgn}\,\nu_o)$ of a compressible odd-viscous fluid and restate the dislocation zero-mode count of Paulose, Chen, and Vitelli as a pairing of integrated torsion with topological polarization. Our central message is that spectral topology can obstruct detailed balance but does not determine the magnitude of dissipation. Detailed balance is self-adjointness in the inner product fixed by the stationary state, so a non-real eigenvalue, an exceptional point (EP), or a nonzero point-gap winding number of a Markov generator rules it out and forces entropy production $\sigma>0$, with $\sigma\ge k_B j^2/D$ for any current, whereas a real, diagonalizable spectrum does not in general certify it. In the Hatano-Nelson ring, which is normal yet has complex spectrum, $\sigma$ sweeps $(0,\infty)$ at fixed winding number. EPs need a separate assessment of the dynamics: with generalized detailed balance including time-reversal parity, an EP is compatible with equilibrium only through time-reversal-odd dynamics, as at critical damping of a Langevin oscillator, whereas an EP of a system with only even variables, such as an odd elastic medium, certifies a nonequilibrium steady state; in both cases $\sigma$ is analytic across the EP. We propose benchtop tests for each statement.
Read the full abstract

Statistical Methodology

Robust Posterior Projection for Bayesian Inference Under Order Restrictions

In archaeology and geochronology, stratigraphic information constrains the relative chronological order of dated samples. Bayesian chronological models incorporate these constraints either through the prior, or by projecting an unconstrained posterior distribution onto the set of ordered chronologies. Dating measurements may nevertheless appear inconsistent with their stratigraphic position, owing to disturbances that do not justify discarding them. We introduce different robust projection maps based on the Huber loss and characterize the non-uniqueness of the considered maps which justifies a Tikhonov regularization. The robust maps also decompose each sample into a monotone baseline and a spike, which flags questionable samples. We adapt the Sequential Block Merging algorithm for the Huber-type losses, and propose a data-driven heuristic for selecting the hyperparameter on which each loss depends on. Simulation studies and geochronological case studies illustrate the gains in robustness and their limits.
Read the full abstract

Databases

Certifying the Error-Detection Value of Functional Dependencies without Ground Truth

Functional dependencies are used to flag erroneous cells, but without error labels the value of these flags is unknown. Dependency strength cannot settle it, because identical data and flags can have positive value in one clean state and negative value in another. We treat this value as partially identified and certify a fixed detector when its value is positive in every clean state compatible with the observation under an error model and budget. The largest such budget is the certification radius. For a detector that flags departures from the unique mode of each class, the worst case reduces to a 0/1 knapsack over class flips, yielding sharp bounds on true positives and the exact radius in polynomial time. Allowing bounded legitimate exceptions in the clean data, we derive an exact threshold for the existence of a positive certificate and compute the radius with the same knapsack. Across five relations and 21,981 candidate dependencies, rules with nearly equal $g_3$ and flag rate can have certification margins differing tenfold. Error injection confirms sound certification and shows that, at equal error counts, changing the organization of errors reverses certification outcomes for every tested rule.
Read the full abstract

Fluid Dynamics

Efficient high-order Discontinuous Galerkin simulations of multi-species high-enthalpy flows

Numerical simulation of hypersonic vehicles is an attractive approach to investigate the complex phenomena involved in such technologies. For this, second-order finite volume methods (FVM) are the established method in computational fluid dynamics, with the cost of high dissipation in affordable meshes, thus needing refined ones or adaptive mesh refinements techniques to obtain accurate results. An alternative is to make use of high-order discontinuous Galerkin methods, in particular discontinuous Galerkin spectral element method (DGSEM), once it achieves higher formal order of accuracy on element-local polynomial bases. Therefore the goal of this work is to present the numerical simulations of a two-dimensional scramjet configuration to test and verify the applicability of the DGSEM approach for a complex hypersonic flow where shock-shock interactions occur. Results revealed to be in agreement with reference FVM solvers, thus verifying the code implementation.
Read the full abstract

All authors Beatriz Fagá, Georgii Oblapenko, Yiyang Zhu, Daniel Doehring, Manuel Torrilhon

Mathematical Physics

Real-Time Path Integral Resurgence: Which Classical Path Contributes?

Which complex classical paths contribute to a real-time Feynman path integral? Picard-Lefschetz theory answers this question with the intersection numbers of the thimbles, which are notoriously hard to compute in infinite dimensions. We show that resurgence provides a practical alternative. We compute the asymptotic series of the real and complex classical paths to high order, with Feynman diagrams and with a transport recursion for the cumulants, and read off the adjacency of the classical paths from the divergent tails of the series with Borel-Padé methods. The South-East rule then turns adjacency into relevance. For a particle meeting a Rosen-Morse barrier, we identify the relevant complex classical paths in front of and behind the barrier, including the complex path that carries quantum tunnelling, follow ordinary and higher-order Stokes phenomena as the final point moves, and verify the predictions with the exact propagator. The resummed series reproduce the propagator with exponential accuracy. Finally, we outline how the expansion extends to relativistic path integrals governed by a Hamiltonian constraint.
Read the full abstract

Machine Learning

UniHam: A Large-Scale SOC-Complete Dataset and Benchmark for Hamiltonian Learning in Materials

Accurate prediction of electronic Hamiltonians would enable broad property inference while avoiding the high computational cost of Density Functional Theory (DFT). However, progress toward general-purpose materials foundation models is limited by a data bottleneck: existing Hamiltonian datasets are typically small, lack structural diversity, and often omit essential relativistic physics such as spin–orbit coupling (SOC). We therefore construct UniHam, a large-scale Hamiltonian dataset and benchmark suite comprising 100,000+ DFT-computed complex-valued Hermitian Hamiltonians with full SOC, covering 72 elements and a wide range of crystal geometries and symmetries (spanning diverse lattice types and space-group families). Building on UniHam, we benchmark two representative state-of-the-art models under a standardized protocol and introduce complementary evaluation metrics that jointly assess three dimensions: (i) Hamiltonian reconstruction accuracy, (ii) out-of-distribution (OOD) generalization across composition/symmetry shifts, and (iii) the ability to support downstream property prediction from the predicted Hamiltonians. Experiments on UniHam demonstrate that the proposed benchmark and metrics effectively differentiate model capabilities, revealing intrinsic SOC- and element-dependent failure modes, large variations in compositional OOD robustness, and the necessity of spectral-level evaluation to assess whether Hamiltonian predictions reliably support downstream electronic-structure properties. Overall, UniHam provides a reproducible, SOC-complete benchmark that can sharpen model comparisons and accelerate the development of next-generation foundation models for quantum materials.
Read the full abstract

Computer Vision

MARS: Multiscale Appearance-Residual Streams for Video Forgery Detection

Generalizing deepfake detection across facial manipulation and fully generated video remains challenging due to forgery traces varying with the manipulation method and scene content. Explicit spatial and temporal feature differences offer a way to capture this variation, but also respond to natural texture and motion, motivating their joint interpretation with appearance features. We investigate whether combining multiscale feature differences with adapted appearance representations improves detection across both forgery categories. We introduce Multiscale Appearance-Residual Streams (MARS), a two-path video detector. The residual path applies fixed spatial and temporal difference-of-Gaussians operators to frozen intermediate features and models the resulting differences alongside appearance context across frames. The semantic path adapts the encoder's normalization parameters for frame classification, and a clip-dependent gate combines their predictions. Under a shared protocol across seven datasets, MARS achieves the highest average AUROC of 91.43%, outperforming seven compared detectors. When trained on facial manipulations alone, MARS achieves 93.99% AUROC on GenVideo and 75.50% on DF26, demonstrating transfer to generated videos without generated-video training data. MARS also provides signed patch and frame contributions, whose faithfulness we evaluate through controlled spatial and temporal removal against random controls. MARS is publicly available for research: Project page: https://vectorinstitute.github.io/mars/ Github: https://github.com/VectorInstitute/mars
Read the full abstract

Differential Geometry

Non-connected minimal clusters of small volume in a compact Finsler manifold

We prove that isoperimetric clusters in general Finsler manifolds can be disconnected even when the volume constraint is small. This is in contrast with the Riemannian case and the “fixed-norm” case, where connectedness for arbitrarily small volume has recently been proved.

Numerical Analysis

An approximation for solutions to Helmholtz equation in polygons and its application in learning based numerical method

A learning based method for Helmholtz equations was proposed in [J. Comput. Phys. 520:113478,2025]. Based on Runge's approximation, it can be proved that when the solution can be continued outside the computational domain, using fundamental solutions as learning samples is sufficient to obtain the solution mapping with spectral accuracy. However, for complex geometries, such as domains with corners, approximation by fundamental solutions may only have logarithmic convergence due to the regularity issue, which is computationally impracticable. Moreover, approximation of a solution to the elliptic equation with those on larger domains is an ill-posed problem. In this work, we investigate how to feed additional learning samples to essentially get a practicable convergence rate and a stable algorithm, so as to ensure the generalizability of the learning based method. Error analysis and numerical examples illustrate flexibility and high-accuracy features.
Read the full abstract

Logic

Projections between Ramsey quasi-orders and topological Ramsey spaces

A Ramsey quasi-order is a quasi-order $(\cal{R},\leq)$ such that $\cal{R}$ is a Polish space, $\leq$ is analytic and every analytic subset of $\cal{R}$ is Ramsey. In this paper, we study the transfer of the analytic Ramsey property via monotone quasi-open maps, which are closely related to a family of forcing projections between quasi-orders (ranging from weak projections to strong projections). We apply this framework to the class of topological Ramsey spaces by providing a detailed list of examples of strong projections between topological Ramsey spaces, and show that, for many of these examples, there is no weak projection in the reverse direction. We also exhibit pairs of topological Ramsey spaces that do not weakly project into each other. Finally, partially confirming a conjecture of Dobrinen, we show that every space in a large class of topological Ramsey spaces either strongly projects into the Ellentuck space or contains a dense set of atoms.
Read the full abstract

Earth and Planetary Astrophysics

ALMA detection of continuum emission from GQ Lup B

GQ Lup B is a massive (10-40 M$_{\text{Jup}}$) companion orbiting the young T-Tauri star GQ Lup A at a projected distance of $\sim$110 au (0.7”). It is outside the circumprimary disk in the submm continuum, but its ${\sim}\,63^\circ$ inclined orbit intersects the primary's gas disk twice per period. We report a 5.8$\sigma$ detection of continuum emission at the location of GQ Lup B, obtained with the Atacama Large Millimeter/submillimeter Array (ALMA) at 0.89 mm (band 7). We associate this emission with a dusty disk around the companion, already detected by VLT/NACO and JWST/MIRI in the mid-Infrared. To date, this is the third disk observed at mm wavelengths around a companion spanning the brown dwarf and planetary mass regimes, following PDS 70 c, a gas giant embedded in the parent circumstellar disk, and SR 12 c, a planetary-mass companion at 1200 au from the primary. We retrieve a point source with peak intensity $54.5\pm9.4\,\mu\rm Jy\, beam^{-1}$, co-spatial with GRAVITY astrometry. The emission is spatially unresolved, placing an upper limit on the source radius of 2.3 au (from a synthesized beam of $\sim\,29\times25$ mas). For comparison, this is smaller than 1/3 of the Hill's radius (3-5 au for GQ Lup B, depending on the assumed mass), the theoretical upper limit for circum-planetary disks. Assuming optically thin thermal emission from mm-size grains in a disk-like structure, we infer a mass of 0.0032 M$_\oplus$, while the optically thick limit yields an emission radius of 0.2 au. Assuming a gas to dust ratio of 100, our results indicate a very low disk mass to accretion rate ratio, of the order of $\sim10^4$ years, analogously to the highest accretors in the stellar mass regime. In the case of GQ Lup B, this can be explained by its periodic crossings of the circumprimary disk, which could deliver a substantial amount of gas to the companion's disk over a long timescale.
Read the full abstract

All authors Pietro Pirola, Stefano Facchini, Gabriele Cugno, Myriam Benisty, Tomas Stolker, Jaehan Bae, Ruobing Dong, Brendan P. Bowler, Ya-Lin Wu, Richard Teague, Pietro Curone, Felipe Alarcon

Optimization and Control

Preference-Based Constraint Handling for Multi-Objective Design-Decision Optimisation: Acceptability Relaxation and Feasibility Guidance

Multi-objective design-decision optimisation (MODO) problems contain negotiable acceptability limits and non-negotiable feasibility constraints. Hard constraints define the feasible design space, whereas actors may reconsider acceptability limits through deliberation and compromise. Relaxing such limits is a preference judgement: penalties or violation measures do not necessarily represent how actors value departures from initial requirements. This paper introduces a preference-based constraint-handling framework. For negotiable limits, preference functions model how actor preference changes with relaxation, allowing the preferred design and acceptable degree of relaxation to be determined through the same aggregation process. Approach 1 represents relaxation as an additional objective and search variable; Approach 2 incorporates it into the preference function of the affected criterion. When constructed consistently, both yield the same best-fit-for-common-purpose solution but differ in complexity and computational effort. The framework addresses stalemate problems, where relaxation is required to recover an acceptable solution, and non-stalemate problems, where modest relaxation may reveal a better design. Hard feasibility constraints remain fixed, and infeasible designs remain inadmissible. A feasibility-guidance function supplies IMAP-IGS with graded information on scaled violations, guiding the search towards difficult-to-reach feasible regions. Every feasible design takes precedence over every infeasible design, and the final recommendation is selected exclusively from the feasible set through A-FINE preference aggregation. The framework integrates acceptability negotiation, feasibility guidance, and design optimisation without conflating their meanings, while preserving preference consistency and decision validity.
Read the full abstract

Optimization and Control

Vanishing Control Times for Single-Input Bilinear Quantum Systems under Rapid Drift Spectral Growth

This paper investigates the asymptotic behavior, as the Hilbert space dimension tends to infinity, of the state-to-state control time of single-input bilinear quantum systems whose control operator is drawn from an isotropic (unitarily invariant) random ensemble. Relying on deterministic graph-theoretic bounds, namely star cuts of the interaction graph for the lower bound and spanning trees for the upper bound, and on integration over the Haar measure, we derive high-probability lower and upper bounds on the control time, expressed in terms of the empirical spectral variance and of the relative concentration of the drift spectrum. Whenever the spectrum grows fast enough (for instance polynomially with an exponent larger than 3/2, geometrically, or super-exponentially), the upper bound vanishes and the control time tends to zero in probability: in these cases, enlarging the Hilbert space constructively speeds up control.
Read the full abstract