Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Wednesday 2 September

Computer Vision

Doppio: A Dataset for Contactless Weight Estimation of Falling Particles

Measuring the mass of powder, including falling particles, is a common task in industrial applications. While scales are effective for static measurements, many applications require contactless sensing, where existing solutions are often costly, application-specific, and technically complex. In this work, we investigate computer vision as a practical alternative for contactless mass estimation. As an accessible real-world case study, we focus on coffee grinding and introduce Doppio, a novel video dataset capturing videos of falling ground coffee, paired with precise, per-frame ground-truth weight measurements. To demonstrate contactless measuring, we evaluate deep learning-based approaches ranging from purely spatial feed-forward networks to recurrent spatio-temporal models. These models are analyzed with respect to their predictive accuracy and computational trade-offs. We demonstrate that deep learning-based computer vision models accurately estimate the cumulative weight of falling particles, establishing a solid foundation for future vision-based contactless measurement solutions.
Read the full abstract

All authors Simon Kiefhaber, Jan-Martin O. Steitz, Julia Grabinski, Christoph Reich, Paul Wagner, Max Zimmermann, Simone Schaub-Meyer, Stefan Roth

Statistics Applications

Did Mary Shelley Write Frankenstein? A Stylometric Analysis

The novel Frankenstein was published anonymously in 1818, and was first credited to Mary Shelley in a French translation of 1821. Since its publication, several claims - both contemporaneous and recent - have been made suggesting that Frankenstein was actually written by Mary's husband, Percy Bysshe Shelley. We review the background of this controversy and then apply modern techniques from computational stylometry to determine who the true author is. Based on our analysis, we find extremely substantial evidence that Mary Shelley is indeed the true author of Frankenstein, and that it is very improbable that Percy Bysshe Shelley played a heavy role in composing the text. While our finding confirms mainstream scholarly opinion regarding Frankenstein, our analysis is the first application of stylometric techniques to this question and provides strong objective grounds for favouring Shelley by freeing the question from some of the politics which have traditionally accompanied it.
Read the full abstract

Computation and Language

When Persona Attributes Improve Population Alignment in Large Language Models

Large Language Models (LLMs) are increasingly used to predict the responses of human participants in survey panels. Towards that goal, persona prompting has recently emerged as a technique to inform and align large pretrained language models. Persona prompting refers to the practice of using short textual descriptions of 'personas' in prompts to steer the LLM's generations. Personas describe individuals through different attributes such as their socio-demographics, attitudes, or behaviors, with the aim of aligning LLMs to produce responses that correlate with the corresponding human responses. Yet, recent work has produced mixed and partly conflicting results of persona prompting without clear patterns of success and failure. Among the few consistent findings is that the selection of persona attributes matters, and that using more attributes does not necessarily lead to better performance. It remains unclear how different attribute selection methods perform and how to choose among them. In this paper, we propose that observed human response variation of a survey question is a potential explanation for the mixed performance observed so far. In addition, we compare the performance of persona prompting associated with different methods for selecting persona attributes. We evaluate these methods on four different (general) social surveys across two countries, six LLMs, and twenty prediction tasks per survey. Our work helps to identify when persona prompting can be expected to be useful in survey prediction tasks, and provides new insights on the effectiveness of different attribute selection methods for LLM-based survey prediction using persona prompting.
Read the full abstract

All authors Leon Fröhling, Jens Rupprecht, Markus Strohmaier, Claudia Wagner

Portfolio Management

Switching Frictions, Heterogeneous Trading Horizons, and Long-Memory Order Flow

This paper develops a mechanism through which costly changes in the representations used for portfolio choice can contribute to persistent signed order flow. Heterogeneous switching thresholds and opportunity volatility generate heterogeneous residence times, and renewal aggregation maps their execution-weighted tail into the decay of aggregate flow covariance. Under common execution weights, the same tail determines the exponent of representation-spell durations, the order-flow memory exponent, and the horizon at which finite-market scaling must end. First-passage renewal analysis establishes these joint restrictions. Structural simulations recover them from realized paths, quantify the distortion created by mismatched weights, and show how finite cross sections shorten the usable inference horizon. The resulting empirical protocol converts an aggregate persistence fit into cross-dataset restrictions that can determine whether a duration-based kernel is suitable for a separate execution-cost model.
Read the full abstract

Materials Science

A Wannier-first approach for extended chiral systems

We present a real-space formulation of DFT for extended systems in which localized Wannier-like functions are constructed directly from localized Gaussian basis functions without explicitly computing canonical Bloch-like states during the self-consistent cycle. Building on the formalism of Pederson and Lin [Phys. Rev. B 35, 2273 (1987)], a variational set of Wannier-like functions is generated self-consistently within a finite Wannier domain and used to construct the charge density, electrostatic potential, and per cell total energy of the extended system. The occupied space can be determined entirely from the Wannier-like functions. Electronic band structures can be recovered in a post-processing step by solving the full Hamiltonian in a Bloch-like basis constructed from Gaussian orbitals. A key feature of the method is that it can incorporate combined translation–rotation, or screw, symmetries, enabling efficient simulations of chiral and helical systems with finite twist angles at essentially the same computational cost as systems described by pure translational symmetry. The approach is validated through calculations on linear and twisted $\mathrm{{-}C{\equiv}C{-}}$ and $\mathrm{{-}Li{-}F{-}}$ chains, as well as graphene, where total energies and band structures show excellent agreement with reference periodic calculations. To illustrate the ability of the method to treat three-dimensional systems, it is further applied to AA graphite, in which carbon atoms in adjacent graphene layers are aligned directly above one another, as well as helically stacked AA graphite structures. The Wannier-first framework provides a practical route for treating extended systems with nontrivial translational, rotational, and screw symmetries, and provides a natural foundation for the implementation of orbital-dependent functionals such as the Perdew–Zunger self-interaction correction.
Read the full abstract

All authors Yashpal Singh, Juan E. Peralta, Koblar A. Jackson, Mark R. Pederson

Analysis of PDEs

Strong Ill-Posedness in critical and subcritical regimes for the Hunter-Saxton equation

We study the Hunter–Saxton equation on the real line and its global dissipative solution in the energy space $ L^\infty(\mathbb R)\cap \dot H^1(\mathbb R). $ We prove strong ill-posedness through instantaneous failure of Sobolev regularity at and below the Lipschitz threshold. More precisely, for every $s\in(1,\nicefrac32]$, we construct $ u_0\in L^\infty(\mathbb R)\cap\dot H^1(\mathbb R)\cap\dot H^s(\mathbb R) $ whose unique global dissipative solution satisfies $u\notin C([0,T]; \dot H^s (\mathbb R))$, for every $T>0$. The constructions differ substantially in the subcritical and critical regimes. For $1<s<\nicefrac32$, we superpose rescaled, localized bubbles with increasingly negative slopes; subcritical scaling preserves $\dot H^s$-summability, while the explicit characteristic formula produces norm inflation. At $s=\nicefrac32$, where scaling yields no smallness, we use logarithmically distributed compactly supported multiscale profiles whose breaking times converge to zero. A one-sided localization principle and an almost-orthogonality estimate then transfer the inflation of individual profiles to the full dissipative solution.
Read the full abstract

Lattice Field Theory

Probing Criticality Using GMM-Based Potentials

Spin models with a given symmetry are easier to sample than scalar theories with the same symmetry on a lattice, as the constrained nature of spin variables enables cheap heat-bath updates. However, this constraint suppresses radial fluctuations, and consequently, spin models cannot be used to study the phenomena of spontaneous symmetry breaking, such as the Higgs phenomenon. To address this, we propose a class of scalar potentials based on Gaussian Mixture Models (GMMs) that are as easy to sample as spin models with a given symmetry. These potentials can be designed to belong to the same universality class as the theory of interest, thereby reproducing its critical properties while enabling efficient sampling. We construct such models for global $\mathbb{Z}_2$ symmetry, $U(1)$ gauge symmetry, and disordered systems. We also verify by numerical experiment that the case of $\mathbb{Z}_2$ symmetry in two dimensions lies in the two-dimensional Ising universality class.
Read the full abstract

High Energy Physics (Theory)

Universal fingerprint of topological defect cores

Topological defects are ubiquitous in physics, arising from condensed matter physics to the early universe. Although there exist many universal scaling laws for correlations between topological defects, such as Porod scaling, the fingerprint of topological defect cores has remained largely unexplored. Here, we discover a universal scaling law in the region $k>1/\xi$, where $\xi$ is the healing length of the topological defects, taking the scaling of the form factor $S_f \propto k^{-(d+p+2)}$, where $d$ is the spatial dimension and $p$ is the defect codimension. We analytically prove that this exponent originates from a universal V-shaped cusp at the defect core and is independent of the underlying system and dynamics. Numerical simulations verify this scaling law in four typical frameworks: the time-dependent Ginzburg-Landau and Gross-Pitaevskii equations in the weak-coupling regime, the gauge/gravity duality model in the strong-coupling regime, and the Klein-Gordon equation in the Friedmann-Robertson-Walker background in cosmology. Our work provides a new probe for studying topological defects in systems ranging from superconductors to cosmological phase transitions.
Read the full abstract

All authors Zi-Qiang Zhao, Nayun Jia, Zhang-Yu Nie, Hua-Bi Zeng, Jing-Fei Zhang, Xin Zhang

Statistical Mechanics

Branching stochastic mechanics: Relative localization and collective poles from Bohm/Fisher feedback

Branching stochastic mechanics (BSM) provides a reciprocal branching representation of the Schrödinger–Nagasawa pair. Its centered forward–backward kernel $C_{\rm FB}=\mathbb E_\omega[\psi_F\psi_B]$ resolves an organized connected sector, with weight $\rho_{\rm BSM}=-C_{\rm FB}(x,x)$ on the anticorrelated branch. Here we investigate how this sector forms, localizes, and propagates under Bohm/Fisher feedback. We retain the branching covariance on the prescribed background as the bare noise kernel and truncate the nonlinear interaction to Bohm/Fisher drift vertices. A Martin–Siggia–Rose–Janssen–de Dominicis formulation and a causal two-loop two-particle-irreducible (2PI) closure determine response and correlation functions self-consistently. While the free connected theory exhibits secular growth and ultraviolet accumulation, the dressed theory develops a finite relative screening length. A reduced numerical evolution shows bounded formation of the localized sector, and a self-similar Fisher construction defines a saturated information velocity $c_\star$. A Born–Oppenheimer separation connects this internal organization to collective propagation. The instantaneous adiabatic kernel has two pole families at fixed internal momentum: a gapless difference branch and a gapped sum branch. Projecting both inverse response kernels onto the same localized internal profile defines their collective coefficients. If both propagation speeds match $c_\star$ and the projected gap matches $mc_\star^2/\hbar$, the gapped branch takes the infrared Klein–Gordon form. These matching conditions define a candidate relativistic fixed point whose dynamical realization remains to be tested.
Read the full abstract

Machine Learning

Spectral Initialization and Scheduled Graph Smoothness for Uncertain Knowledge Graph Completion

Uncertain knowledge graphs (UKGs) extend knowledge graphs by assigning each triple a continuous confidence score. Since most possible triples lack observed confidences, recent methods rely on semi-supervised learning to generate pseudo-labels. These methods initialize entity embeddings without using the confidence-weighted graph, discarding its global community and hub structure. We introduce QUEST, which adds no trainable parameters to the standard confidence-distribution learning pipeline. First, QUEST initializes entity embeddings using the smallest non-trivial eigenvectors of the confidence-weighted graph Laplacian, incorporating community and hub structure before training. Second, QUEST applies an unbiased mini-batch Dirichlet energy regularizer to enforce early-stage structural consistency. On two UKG datasets, QUEST improves confidence prediction and link prediction on six of eight metric-dataset pairs over prior methods and matches the previous best on the remaining two, while removing the instability spike observed on dense graphs. These results indicate that spectral structural priors combined with a graph Dirichlet energy regularizer improve accuracy, training stability, and checkpoint reliability in UKG completion.
Read the full abstract

All authors Md Abrar Jahin, Taufikur Rahman Fuad, Jay Pujara, Craig A. Knoblock

Quantum Physics

Efficient Simulation of Nonreciprocal Many-body Physics via Quantum Feedback

We present a scheme to efficiently simulate nonreciprocal many-body spin models using a non-Markovian open system. Our scheme utilizes a single quantum emitter coupled to a waveguide mode that is fed back to the same emitter after a time delay. With this coherent delayed feedback, an effective nonreciprocal interaction can be engineered, wherein the emitter at earlier times affects itself at later times, creating a scalable spin chain. We show that all the steady-state quantities of such spin models can be accessed through measurement of the emitter and the output field. Furthermore, using classical feedback that resets the emitter, quench dynamics from arbitrary product states can be simulated. We demonstrate striking features of nonreciprocal models, such as the Liouvillian skin effect, anomalous relaxation dynamics, and quasi-long-range order of output photons. Finally, we show that the proposal is amenable to experimental realization and robust to imperfections. Our work establishes a feasible way to scale up a nonreciprocal many-body system, and provides a new route towards generation of exotic states of light.
Read the full abstract

All authors Kseniia Vodenkova, Andrew Pocklington, Fan Yang, Andrew Lingenfelter, Aashish A. Clerk, Hannes Pichler

Optimization and Control

Projected Subgradient Methods for a Class of Nonsmooth and Nonconvex Optimization Problems

We investigate the optimization problem of minimizing a nonsmooth function that satisfies a nonsmooth version of the descent lemma over a nonempty and closed but not necessarily convex set. The objective function belongs to the class of upper-$\mathcal{C}^2$ functions, whereas the constraints may promote a sparse or low-rank structure. We propose a projected subgradient method with two different globalization strategies: (a) a nonmonotone linesearch and, under additional assumptions, (b) an auto-conditioned method, where the stepsize is given by a formula depending on data from past iterations. We show that both methods converge to solutions that satisfy a stronger stationarity concept than one would expect from the subdifferential sum-rule, which is particularly important since the optimization problems of interest are inherently nonconvex. Finally, we present promising numerical results when applying the algorithm to an MPEC-style problem as well as the matrix optimization problems MAXCUT and Robust PCA.
Read the full abstract

Analysis of PDEs

Power convexity of the torsion function on horo-convex domains in hyperbolic space

In this paper, we study the torsion problem on bounded, smooth, strictly horo-convex domains in hyperbolic space. We prove that if the diameter of the domain is sufficiently small, then $-\sqrt{u}$ is strictly convex, where $u$ denotes the torsion function. The main ingredients of our proof are the constant rank theorem, a boundary convexity estimate, and a deformation argument for the domain. We also show that the exponent $1/2$ is optimal, even among strictly horo-convex domains of arbitrarily small diameter.
Read the full abstract

Fluid Dynamics

Numerical Investigation of Mach 10 Kerosene-Fueled Oblique Detonation Waves at Different Flight Altitudes

The initiation and propagation mechanisms of kerosene-fueled oblique detonation waves are key issues in the development of oblique detonation wave engines. In this study, numerical simulations of kerosene-fueled oblique detonation at a flight Mach number of 10 were conducted using the two-dimensional conservative Euler equations and a second-order two-step global chemical reaction model. The objective was to compare wedge-induced initiation with bump-forced initiation. The results show that, at different flight altitudes, the oblique detonation flow field follows the h̊o}L binary scaling law. Wedge-induced initiation requires highly precise matching of parameters such as the inlet-exit conditions, wedge angle and length, and equivalence ratio. In contrast, bump-forced initiation exploits the high total temperature and total pressure at the stagnation point and therefore does not require precise matching of these parameters. Therefore, the bump-forced initiation method provides reliable oblique detonation initiation and a stable oblique detonation flow field.
Read the full abstract

Distributed Computing

AceSpec: An Asymmetric Edge-Cloud Collaborative Framework for Communication-Efficient LLM Inference

Deploying Large Language Models (LLMs) on edge devices typically relies on model compression or split inference. However, compression degrades reasoning capabilities, while split inference suffers from severe Wide Area Network (WAN) communication bottlenecks. Edge-cloud speculative decoding emerges as a promising alternative, leveraging an edge small model to draft tokens for cloud verification. Yet, over volatile WANs, inevitable prediction rejections trigger catastrophic pipeline stalls and network-wide rollbacks, neutralizing collaborative gains. To overcome this, we propose AceSpec, an asymmetric edge-cloud collaborative framework. AceSpec utilizes un-saturated edge compute to proactively construct a probabilistic state cache, effectively transforming network-wide pipeline flushes into $\mathcal{O}(1)$ local memory lookups. To preserve bandwidth, it employs an asymmetric communication protocol that transmits minimal main-chain indices uplink and compact sparse distributions downlink. Furthermore, we introduce a network-aware, Lagrangian-optimized resource allocation strategy that dynamically maximizes the local cache hit rate. Evaluations demonstrate that AceSpec achieves up to a 3.52$\times$ throughput speedup and exhibits exceptional bandwidth immunity, sustaining near-peak inference performance even under severely constrained 50 Kbps WAN conditions.
Read the full abstract

All authors Yida Zhang, Zhiyong Gao, Shuaibing Yue, Jie Li, Rui Wang

Complex Variables

Prescribed Lelong Numbers for One-Pole Green Functions on Complex Projective Space

Let $\omega_{\mathrm{FS}}$ be the normalized Fubini–Study form on $\mathbf P^n$, with $n\geq2$. We prove that the one-pole Lelong-number range in $DMA(\mathbf P^n,\omega_{\mathrm{FS}})$ is exactly $[0,1]$: for every $\lambda$ in this interval there is a Green function with a single pole, Monge–Ampère measure equal to the Dirac mass at that pole, and Lelong number $\lambda$. This answers Question~9 in the survey of Dinew–Guedj–Zeriahi, where the problem is attributed to Coman and Guedj. The construction adapts Li and Xia's local zero-Lelong-number scheme through variable degrees and homogeneous finite stages, while also establishing the exact Lelong number and membership in the global $DMA$ class. We then study the relation between the one-pole range $\mathcal R_\alpha(x)$ and the Seshadri interval $[0,\varepsilon(\alpha,x)]$. An application of Koike's equivalence theorem gives a point on a degree-one del Pezzo surface where the Seshadri endpoint is not attained. Conversely, a finite-pullback criterion and an explicit finite morphism show that every ample rational class on a product of projective spaces realizes its full Seshadri interval at every point.
Read the full abstract

Computer Vision

Beauty is in the AI of the beholder: MLLMs systematically overrate facial attractiveness

Beauty assessments from Multimodal Large Language Models (MLLMs) are increasingly popular amongst users, companies, and aestheticians. This raises the question of whether these AI models can accurately reflect human judgments of attractiveness. In a pre- registered exploratory study, we compared the attractiveness ratings of 2,513 human participants to four widely used commercial AI models: Claude, Gemini, GPT, and Grok. Results showed that MLLMs systematically rate faces more favourably and within a narrower range than humans and, at the time of study, do not reproduce human ratings in absolute terms. However, MLLMs exhibit strong correlations with human attractiveness judgments, accurately tracking the rank-ordering of faces. MLLMs may judge faces by different cues than humans; only face age was a predictor of facial attractiveness in both humans and MLLMs, with inconsistent patterns across models for ethnicity and gender. AI models strongly agree with one another, except for Grok, which also showed the lowest agreement with humans. Our findings suggest that while they may be able to approximate rank-orderings of human attractiveness, current off-the-shelf commercial MLLMs systematically overrate the beauty of human faces.
Read the full abstract

All authors Santiago Grandas, Juan Sebastian Cely-Acosta, Mohit Mendiratta, Shafee Hassan, Macken Murphy

Graphics

Telligram: Text-Driven Calligram Generation via Diffusion-Guided Skeleton Optimization

Compact calligram generation aims to form a semantic shape while keeping letters recognizable. Most existing methods are shape-conditioned and mainly solve downstream letter layout inside a given contour. We study text-only calligram generation without an input contour. This setting is difficult because semantic shape formation and letter readability strongly interfere with each other when optimized in a single stage. Pushing the word toward a clear figure can easily damage glyph structure, while preserving readable letters can weaken the target shape. To address this difficulty, we present Telligram, a training-free, low-tuning, two-stage framework composed of Semantic Occupancy Prior Formation and Readability-Constrained Glyph Realization. The first stage uses Variational Score Distillation (VSD) with structured skeleton optimization and hierarchical gradient projection to produce a semantic occupancy prior. The second stage converts this occupancy prior into per-letter regions and reconstructs readable glyph layouts through lightweight geometric processing. The framework generates coherent and creative word-level semantic calligrams directly from text prompts.
Read the full abstract

Computer Vision

Orthogonal Ensembles and Tested Explanations for Performer-Independent Body-Motion Emotion Recognition

We study body-only, 12-class acted-emotion classification from skeleton motion under leave-performer-out (LPO) evaluation, a hard, underdetermined setting: chance is 8.3%, and a protocol-matched reproduced STGCN++ baseline reaches only 25.73 +/- 4.03% Macro-F1. We show that reliable gains come not from a new architecture but from combining eleven models with orthogonal error modes: under 10-fold LPO cross-validation on the labeled training performers, an equal-weight logit-mean ensemble reaches 36.80 +/- 4.00% per-fold Macro-F1, a protocol-matched +11.07 pp (+43% relative) over the same-split reproduced baseline. Our central contribution is a tested explanation suite: for a strong ensemble member, part-masking and counterfactual edits show (rather than assert) that its decisions depend on motion-grounded body-region evidence, and this region saliency aligns with rule-based Laban Movement Analysis (LMA) attributes far more than with classical kinematics: region-level saliency-LMA Spearman rho = +0.500 versus +0.033, roughly 15x, and the alignment holds for the submitted 11-way ensemble itself at rho = +0.517; the audit is post hoc and needs no retraining. The same suite faithfully reports a negative: within-window temporal saliency is diffuse rather than localized. On the hidden challenge test set the submitted ensemble scored 37.23 % Macro-F1 and received the Best Performance Award of the MMAC Challenge 2026 (Human score is 39 %). Code is available at https://github.com/nawta/diema-challenge and the presentation at https://nawta.github.io/mmac2026/.
Read the full abstract

Representation Theory

Algebraic study of quantum configuration spaces of decorated flags

Let $G$ be a connected, simply connected complex simple algebraic group and $\mathscr{A}_G=G/U^+$ its base affine space, whose elements are called decorated flags. We introduce the quantum configuration space of decorated flags $\mathcal{O}_q(\mathrm{Conf}_K \mathscr{A}_G)$ and initiate its algebraic study, based on the representation theory of quantized enveloping algebras. Our algebra $\mathcal{O}_q(\mathrm{Conf}_K \mathscr{A}_G)$ gives a quantum analogue of the configuration space $\mathrm{Conf}_K \mathscr{A}_G$ of $K$ decorated flags, which provides local building blocks for the Fock–Goncharov moduli space $\mathscr{A}_{G,\Sigma}$ of decorated twisted $G$-local systems on a marked surface $\Sigma$. We establish basic algebraic properties of $\mathcal{O}_q(\mathrm{Conf}_K \mathscr{A}_G)$ such as the domain property, quantum normalization of representatives, the quantum cyclic shifts, the quantum Wilson lines, whose classical counterparts have been fundamental in the study of $\mathscr{A}_{G,\Sigma}$. Moreover, we construct quantum seeds for $\mathcal{O}_q(\mathrm{Conf}_4 \mathscr{A}_G)$ by transporting the Berenstein–Zelevinsky quantum cluster structure on $\mathcal{O}_q(G)$ via quantum Wilson lines, and prove that $\mathcal{O}_q(\mathrm{Conf}_4 \mathscr{A}_G)$ coincides with the corresponding quantum cluster algebra and its upper counterpart after the localization at frozen variables. The exchange matrices for our quantum seeds agree with the Goncharov–Shen exchange matrices. We also show that quantum seeds for $\mathcal{O}_q(\mathrm{Conf}_4 \mathscr{A}_G)$ restrict to those for $\mathcal{O}_q(\mathrm{Conf}_3 \mathscr{A}_G)$. Finally, we construct quantum seeds for $\mathcal{O}_q(\mathrm{Conf}_K \mathscr{A}_G)$, $K\geq 5$, and prove that the corresponding quantum cluster algebras contain the quantum configuration space $\mathcal{O}_q(\mathrm{Conf}_K \mathscr{A}_G)$.
Read the full abstract

Group Theory

The binary actions of sporadic groups

An action of a group is binary if it induces the group of automorphisms of some homogeneous edge-coloured directed graph. In this paper we study the quasisimple groups $G$ with $G/Z(G)$ a sporadic simple group—we produce a complete classification of their binary actions.

Machine Learning

Rethinking the Teacher-Student Framework for Test-Time Adaptation

Test-Time Adaptation (TTA) has recently emerged as a promising strategy that allows the adaptation of pre-trained models to changing data distributions at deployment time, without access to any labels. To mitigate error accumulation, researchers have widely adopted the teacher-student framework, though its long-term stability is often taken for granted. In this work, we challenge the common strategy of setting the teacher weights to an exponential moving average of the student by showing that error accumulation still occurs, although it is mostly apparent on longer sequences compared to those commonly utilized. We analyze the stability-plasticity trade-off within the teacher-student framework and propose to use an intransigent teacher that does not update its weights. Surprisingly, we show that this simple change allows TTA methods to significantly improve their performance on multiple datasets with longer scenarios and result in increased robustness to changes in hyperparameters. Finally, we show that those changes can be seamlessly and effectively applied to various architectures and experimental setups, including semantic segmentation. The code is available at https://github.com/dmn-sjk/intransigent_teacher.
Read the full abstract

All authors Damian Sójka, Marc Masana, Bartłomiej Twardowski, Sebastian Cygert

Strongly Correlated Electrons

Disorder-driven competing magnetic interactions and glassy magnetic behavior in quaternary Heusler alloy FeRuMnGe

In this combined experimental and theoretical study, we investigate the role of disorder in governing the magnetic ground state of the quaternary Heusler alloy FeRuMnGe. In the FeRuMnZ (Z = Ga, Si) series, chemical substitution modifies atomic ordering and electronic structure, resulting in distinct magnetic ground states. Motivated by this, we extend the series to FeRuMnGe. X-ray diffraction reveals B2-type antisite disorder, where Fe–Ru and Mn–Ge intermix. Theoretical calculations show that such disorder modifies exchange interactions, leading to competing ferromagnetic and antiferromagnetic couplings, and drives the system from half-metallic to metallic. Magnetic measurements reveal competing interactions, giving rise to a cluster-glass state coexisting with long-range magnetic order. The absence of a thermodynamic signature at $T_f$, together with ac susceptibility and relaxation measurements, supports the presence of short-range magnetic interactions and a glassy magnetic state. The compound exhibits an enhanced magnetic response below $\sim 161$ K and a maximum magnetization of $\sim 1.64~\mu_B$/f.u. at 5 T. Overall, this work establishes a direct correlation between antisite disorder, competing exchange interactions, and glassy magnetism in quaternary Heusler alloys. Combined experimental results and theoretical calculations reveal that disorder drives the system from an AFM-dominated state in FeRuMnSi to an FM-dominated state in FeRuMnGe, providing deeper insight into the role of the extent of disorder in governing the magnetic properties of QHAs.
Read the full abstract

All authors Manikantha Panda, Prabuddha Kant Mishra, Sonali S. Pradhan, Aarti Gautam, Bhagyashree Pol, P D Babu, Ashok Kumar Ganguli, V. Kanchana, Tapas Paramanik

High Energy Physics (Phenomenology)

Dark Matter as the Z_2 Partner of the Standard Model Higgs Boson

A possible gamma-ray excess from the Galactic halo favors dark matter with a mass of several hundred GeV. In previous work, we found that an electroweak scalar doublet is a particularly economical, and nearly uniquely selected, interpretation among simple renormalizable thermal dark-matter models. We point out that its unbroken Z_2 symmetry is exactly equivalent to an interchange symmetry between two Higgs doublets, making dark matter the Z_2-odd partner of the Standard Model Higgs boson and suggesting that the two doublets may share a common weak-scale origin. We connect this model to the recently reported 248 keV nuclear-recoil event in the LUX-ZEPLIN experiment. For dark-matter masses of 440–600 GeV, a technically natural, mass-dependent neutral-state splitting of order 350 keV defines a narrow strip near the kinematic boundary where the fixed off-diagonal Z interaction can produce the observed recoil. Because the scattering probes the extreme high-velocity tail of the Galactic distribution, this interpretation predicts a large annual modulation, with future events preferentially occurring around early summer. Neither observation is yet conclusive, but their common interpretation is sharply testable.
Read the full abstract

Computer Vision

SR-Edit: Region-Aware Image Editing via Self-Refinement

With the recent rapid progress in generative models, image editing has made remarkable advances, yet achieving faithful edits that precisely modify only the target regions while strictly preserving all other regions remains challenging. Since externally provided region annotations are often difficult to obtain in practice, a growing body of work seeks to improve preservation by automatically inferring edit and non-edit regions, and then enforcing consistency on the latter. However, these approaches still suffer from inaccurate region estimation and heuristic correction strategies that distort the native inference process, making methods designed for fidelity themselves a new source of artifacts. We propose SR-Edit, an image editing framework that overcomes these issues via iterative self-refinement. Specifically, at each iteration, SR-Edit first (i) extracts progressively precise and self-consistent region separation from the model's own predictions by lightweight post-processing, and then (ii) enforces preservation in non-edit areas through correction updates that remain aligned with the original sampling dynamics. Extensive experiments demonstrate that SR-Edit achieves superior preservation and overall image quality compared to existing editing techniques.
Read the full abstract

All authors Andong Wang, Zehua Chen, Yuxuan Jiang, Jun Zhu

Quantum Physics

Single-Shot Fidelity Reveals Hard and Soft Limits: A Universal Yardstick for Photon-Number-Resolving Detectors

Photon-number-resolving (PNR) detectors are essential for photonic quantum computing, where a single measurement outcome must reliably herald a specific quantum state. However, detector fidelity is conventionally evaluated using ensemble-averaged statistics obtained from many measurements, which can remain high even when individual photon-number assignments are frequently misidentified. Here we introduce a universal single-shot fidelity that directly quantifies the probability of correctly identifying a photon number in a single measurement. The framework combines an efficiency-based POVM with a resolution-driven confusion matrix derived from the detector response, allowing photon loss and photon-number misidentification to be treated separately and then recombined into a single operational metric. This distinction reveals two fundamentally different limitations. Detection-efficiency loss represents an unrecoverable hardware constraint, whereas resolution-driven misidentification can be reduced by introducing a rejection region, trading generation rate for confidence. Because the metric is defined independently of detector architecture, it enables direct comparison between energy-resolving detectors such as transition-edge sensors and multiplexed click-based detectors on the same footing. Applying the framework to calibration data from three distinct detector architectures, we demonstrate quantitative comparison across photon-number regimes relevant to both discrete-variable and continuous-variable photonic quantum computing. The resulting benchmark provides a common operational metric for evaluating photon-number-resolving detectors and connecting detector performance to photonic quantum-computing requirements.
Read the full abstract

Computer Vision

Blending Concepts: Benchmarking Visual Metaphor Generation in Text-to-Image Models

Text-to-image (T2I) models have achieved remarkable success at faithfully rendering specified objects and attributes, yet their ability to produce visual metaphors, images that convey abstract ideas by combining elements from two distinct domains, remains largely unexamined. To bridge this gap, we introduce VMetaphor-Bench, the first benchmark for evaluating visual metaphor generation in T2I models. It comprises 1,500 visual metaphors curated from real-world creative imagery, organized into three levels and ten categories, with each sample paired with two prompts of differing specificity. For evaluation, we develop a hybrid framework within an MLLM-as-judge paradigm, combining a multiple-choice question (MCQ) based protocol of 9,594 questions across four levels of metaphorical fidelity with a dimension-based scoring protocol along three perceptual dimensions. Extensive evaluation of 11 representative T2I models reveals that even the strongest proprietary models struggle with compositional structuring and cross-domain mapping, key aspects of metaphorical expression, highlighting visual metaphor generation as an important frontier for future T2I research.
Read the full abstract

All authors Chuer Chen, Zichen Wang, Yi He, Zhengxi Yu, Nan Cao

High Energy Physics (Phenomenology)

Scalar Portal Verifiable Light Dark Matter and Correlated Gravitational Wave Signatures

The lack of signals in direct detection experiments has placed the canonical Weakly Interacting Massive Particle (WIMP) paradigm under severe tension, motivating a shift toward the sub-GeV Light Dark Matter (LDM) regime. However, realizing detectable LDM interaction rates typically requires large couplings to the visible sector, which leads to a severe thermal underabundance of the dark matter relic density within standard cosmology. Furthermore, LDM models featuring vector mediators face stringent constraints from the Cosmic Microwave Background (CMB) due to late-time energy injection. In this work, we propose a minimal scalar portal extension featuring a vector-like fermion dark matter candidate, which naturally evades CMB bounds via inherent p-wave annihilation suppression. To simultaneously achieve the correct relic density and large direct-detection couplings, we invoke a pre-Big Bang Nucleosynthesis (BBN) non-standard cosmology dominated by a stiff fluid ($w > 1/3$). The enhanced Hubble expansion during this epoch triggers an early dark matter freeze-out, successfully rescuing the asymptotic relic abundance. Crucially, this stiff pre-BBN phase heavily blue-shifts inflationary gravitational waves that re-enter the horizon prior to BBN, imprinting a distinct high-frequency tilt on the stochastic gravitational wave background. We establish a robust correlation between the non-standard expansion history, the particle physics parameters verifiable in future terrestrial direct detection experiments, and the unique gravitational wave signatures observable by forthcoming space-based interferometers like LISA and DECIGO. This framework highlights how multi-messenger observations can concurrently probe the dark sector and the pre-BBN thermal history of the Universe.
Read the full abstract

Algebraic Geometry

Flexibility of affine cones over blow-ups of weighted projective planes

Let $S_m^n$ be the surface obtained by blowing up $\mathbb{P}(1,1,m)$ at $n$ general smooth points, where $m\geq2$. For every very ample Cartier divisor $H$ on $S_m^n$, we prove that $\mathrm{Affcone}_H(S_m^n)$ is flexible when $n\leq m+1$. For $n=m+2$ and $n=m+3$, we prove that $\mathrm{Affcone}_H(S_m^n)$ is generically flexible.
Read the full abstract

Information Retrieval

Training seeds and model-selection stability in recommender-system evaluation

Recommender-system experiments often rely on a single random training seed, assuming that run-to-run stochasticity has limited impact on evaluation conclusions. This assumption is risky, as a training seed may influence several algorithm-dependent mechanisms, including parameter initialization, mini-batch ordering, dropout, masking, latent sampling, and training-time negative sampling. We examine this assumption by fixing the data partition and varying the training seed across hyperparameter configurations. We analyze seed effects at three levels: user-level metric sensitivity, validation-based model selection and recommendation-list agreement. Results show that seed variation is often detectable. Its impact depends on whether configurations are clearly separated, whether validation results transfer to test, and whether similar scores lead to similar top-$k$ lists. Findings suggest that reporting single-seed results can overstate the stability of recommender system evaluation, and that training seeds should be treated as part of the evaluation protocol rather than as incidental implementation noise.
Read the full abstract