Peer Revenue Equity in science

Papers

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

All of arXiv

Friday 6 November

Differential Geometry

Differentiation of tepui groupoids and the holonomy groupoid of singular foliations

Tepui groupoids are a generalization of Lie groupoids which allow the arrow space to be singular, as is the case for holonomy groupoids of singular foliations. We construct a Lie differentiation functor for a large class of tepui groupoids. When restricted to effective and source-connected tepui groupoids, this differentiation recovers the correspondence between holonomy groupoids and their underlying singular foliations.
Read the full abstract

Solar and Stellar Astrophysics

Interferometric survey of stellar parameters. Reliable methodology for a direct measurement of the angular diameter in exoplanet host stars

Since their pioneer observations more than thirty years ago, exoplanets have been a major topic among the astrophysics community. Despite their ubiquity, our knowledge of planetary parameters and internal compositions still mostly relies on stellar quantities estimated through evolutionary models, which suffer from a calibration bias, preventing the community to have accurate discussions on the diversity of planets across the Universe. We use our sample to demonstrate the current best reachable precision of the CHARA Array to be $\sim$ 2% on radius determination, with an error budget dominated by the uncertainty on the angular diameter, while systematics are shown to be of the order of 1% thanks to our methodology. While the precision matches values found in the literature, we demonstrate the uniqueness of the present study in the relevance of the results put forward. Finally, we show that our methods naturally translate into an improved accuracy of the planetary parameters, along with an improvement in precision by $\sim$ 50% in radius and $\sim$ 30% in mass.
Read the full abstract

All authors H. Nowacki, D. Mourard, N. Ebrahimkutty, R.V. Ibañez Bustos, J. Jonák, M. Vrard, G. Schaefer, P. Berio, J. Dejonghe, F. Morand, N. Nardetto, D. Salabert, M. Bailleul, O. Creevey, S. Deheuvels, A. Domiciano de Souza, R. Ligi, A. Meilland, K. Perraut, M. Wittkowski, J.D. Monnier, S. Kraus, N. Anugu, M. Gutierrez, N. Ibrahim

Mesoscale and Nanoscale Physics

Growth and fabrication of advanced PbTe-Pb quantum devices

We report recent progress in the growth and fabrication of advanced PbTe-Pb Majorana devices for three-terminal correlation experiments. A critical prerequisite for these devices is a continuous Pb film spanning from the PbTe nanowire to the substrate. In previous studies, PbTe nanowires were spatially separated from the substrate mask, making them unable to fulfill this requirement. To address this issue, we have developed a new growth and fabrication approach. Key fabrication steps are introduced between the PbTe growth and Pb deposition. The top facet of the PbTe nanowire is covered with a Te capping layer before breaking the ultra-high vacuum, enabling ex situ processing for subsequent steps. The mask is then removed, followed by the fabrication of shadow walls. After reloading the chip into the growth chamber, the Te protection layer is removed through thermal decomposition, and the Pb film is subsequently deposited. Notably, a pristine PbTe-Pb interface with an in-situ-equivalent quality can be achieved despite the ex-situ processing. Our technique facilitates the develop of advanced PbTe-Pb quantum devices with complex designs, paving the way toward the realization of topological qubits.
Read the full abstract

All authors Zehao Yu, Wenyu Song, Lining Yang, Zeyu Yan, Yuhao Wang, Weizhao Wang, Xinchen He, Ruidong Li, Jiaye Xu, Shuai Yang, Zonglin Li, Shan Zhang, Xiao Feng, Tiantian Wang, Yunyi Zang, Lin Li, Runan Shang, Qi-Kun Xue, Ke He, Hao Zhang

Optics

Differential chiral sensing at exceptional points in intrinsically chiral metastructures

Discriminating opposite enantiomers through their weak chiroptical responses remains a fundamental challenge in optical sensing. Exceptional points (EPs), non-Hermitian singularities at which eigenvalues and eigenvectors coalesce, offer a route to enhanced spectral responses through their characteristic square-root eigenvalue splitting. However, their potential for chiral discrimination remains largely unexplored. We demonstrate exceptional points in an intrinsically chiral hybrid metamaterial and reveal strongly asymmetric responses to enantiomeric perturbations. The interplay between intrinsic chirality and non-Hermitian degeneracy is realized through the coupling of a chiral plasmonic nanoantenna array to a photonic crystal slab. We observe characteristic complex-square-root scaling, with both resonance frequencies and loss-rates exhibiting pronounced handedness dependence. Remarkably, opposite enantiomers produce complex mode splittings that differ by more than an order of magnitude. Our findings establish intrinsic chirality as a new degree of freedom in non-Hermitian physics and reveal a fundamental route toward topological singularities enhanced discrimination of chiral matter.
Read the full abstract

All authors Suzit Hasan Nayem, Zhanni Wu, Mariusz Drong, Kamyar Behrouzi, Boubacar Kanté

Quantum Physics

Neutral-atom quantum features as complementary structural encodings for graph learning

Structural descriptors provide graph neural networks (GNNs) with positional and topological inductive biases. However, widely used classical encodings based on random walks or Laplacian spectra emphasize specific structural statistics of a graph, leaving room for alternative descriptors. Neutral-atom quantum processing units (QPUs) can realize graph-dependent Rydberg Hamiltonians, allowing graph topology to be probed through dynamical correlation measurements. In this work, we use these measurements as quantum graph features (QGFs), providing a complementary, hardware-derived family of structural encoding that is not accurately described by classical features derived from Laplacian and adjacency matrices, even when supplemented with an extended set of random-walk, geometric descriptors and classical physical models. By appropriately selecting the quench parameters of the Hamiltonian, we access two characteristic correlation regimes with distinct structural signatures: antiferromagnetic-like and ferromagnetic-like. On an intrinsic information benchmark of 15 topological and spectral probing tasks, quantum and hybrid quantum-classical features match or surpass classical descriptors on local topological properties. When supplied to message-passing, attention, and graph-transformer backbones, hybrid quantum-classical features yield improvements that are within cross-fold noise on most cells but sign-consistent across architectures. When comparing noiseless, shot-sampled emulation-derived quantum features with a neutral-atom QPU implementation, the downstream hybrid results with compressed hardware features match their emulated counterparts within $0.01$ F$_1$ on all backbones. On a synthetic family of ladder graphs, QGFs distinguish two specific classes, outperforming the strongest classical structural descriptors by a margin exceeding cross-fold variability.
Read the full abstract

All authors Mathieu Garrigues, Igor O. Sokolov, Matthias Hecker, Lorenzo Moro, Shaheen Acheche

Medical Physics

Dual-energy cardiac DSA imaging using a dual-layer flat-panel detector: in vivo validation study

Background: The temporal misregistration in conventional X-ray digital subtraction angiography (DSA) remains a key challenge for its practical application in cardiac catheterization, leading to incomplete separation of cardiac vessels from background anatomical structures. Purpose: In this study, the X-ray spectral information is unveiled for DSA imaging to generate better visualization of cardiac vessels with negligible anatomical background. Methods: A dedicated dual-energy DSA imaging model is proposed based on the subtraction residuals between pre-contrast and post-contrast spectral projections, enabling accurate material decomposition without requiring precise scatter correction. Specifically, a dual-layer flat-panel detector (DL-FPD) is assumed to acquire the needed temporally registered low-energy and high-energy projections. Moreover, a penalized likelihood material decomposition algorithm is proposed to enhance the material decomposition performance. To validate, in vivo swine experiments were conducted for the first time on a clinical DSA system equipped with a DL-FPD. Results: Monte Carlo simulation results demonstrated that the impact of scatter on the subtraction residuals can be reduced by more than 80%, substantially improving material decomposition accuracy. Compared with analytical material decomposition, the penalized-likelihood material decomposition algorithm significantly improves image contrast-to-noise ratio (CNR), enabling the visualization of small vessels and even the myocardial blush effect. Furthermore, compared with conventional dual-energy subtraction, the penalized-likelihood approach provides more quantitative material maps with fewer overlapping anatomical background artifacts. Conclusion: This preliminary study verifies that dual-energy DSA imaging method is able to generate high-quality cardiovascular imaging results.
Read the full abstract

All authors Yuhang Tan, Jiongtao Zhu, Xin Zhang, Haodi Wu, Xiao Xuan, Mang Feng, Mingzhi Cui, Hairong Zheng, Dong Liang, Yongshuai Ge

Combinatorics

Reflexive polytopes can have doubly exponentially many vertices

Reflexive polytopes are lattice polytopes containing the origin in their interior such that also the dual polytope is a lattice polytope. On the basis of low-dimensional observations and of bounds known for special classes of reflexive polytopes, the second author conjectured in 2004 that the number of vertices of $d$-dimensional reflexive polytopes is bounded from above by $6^{d/2}$. In this note, we present counterexamples showing that this conjecture is false in every dimension $d \ge 8$. In fact, rather surprisingly, in sufficiently large dimensions $d$ the number of vertices (equivalently, by duality, the number of facets) can be as large as theoretically possible, namely doubly exponential in $d$. These results were found using GPT-6.1 Sol.
Read the full abstract

Sound

Fuzzy-FedAvg: Server-Aware Client Selection for Federated Speech Emotion Recognition in Heterogeneous Environments

Federated speech emotion recognition (SER) is challenged by speaker-dependent, class-imbalanced data and by heterogeneous client devices. This paper proposes Fuzzy-FedAvg, a server-aware framework in which the server learns Client Selection Weights (CSWs) through fuzzy Q-learning from global-model feedback. The same CSWs weight client accuracy, training time, and \divkpi{} in both client selection and aggregation, while a dedicated PCAS node manages stragglers through staleness-discounted deferred updates. On IEMOCAP and CREMA-D, Fuzzy-FedAvg achieves higher unweighted average recall (UAR) than FedAvg in all configurations under severe non-IID conditions, and in 22 of 24 participation-level and 10 of 12 training-horizon configurations overall. It also outperforms FedProx and Oort, while SCAFFOLD and Power-of-Choice remain competitive, and it reduces cumulative training time by 81.2% in a FedScale-based evaluation. The benefit is greatest under strong statistical heterogeneity and narrows as data become more balanced.
Read the full abstract

All authors Farrukh Pervez, Siddique Latif, Moid Sandhu, Raja Jurdak

Numerical Analysis

Multilevel kernel approximation of log-Euclidean manifold-valued functions

We study kernel approximation and multilevel residual correction for functions with values in a manifold that admits a global logarithm onto a finite-dimensional Hilbert space, such as positive definite matrices, exponential Lie groups, and Hadamard manifolds. Approximating in logarithmic coordinates keeps the approximant on the manifold and transfers Hilbert-space error estimates verbatim. To express these errors in the natural geometry of the target, we prove metric comparisons. In particular, we obtain a sharp, scale-invariant bound between the log-Euclidean and affine-invariant distances on positive definite matrices, with a constant that depends only on condition numbers. For multilevel approximation, compactly supported kernels scaled to the fill distance give algebraic convergence with sparse, uniformly well-conditioned systems. These guarantees require appropriate sample locations and approximation scales. Compact groups are treated through a principal logarithm, where a bi-invariant metric gives error bounds with constant one. Finally, the log-Euclidean and Karcher means differ by $\mathcal O(r^3)$ near the identity and by $\mathcal O(r^2)$ around a general point, where $r$ is the logarithmic data radius and the second bound is sharp.
Read the full abstract

Number Theory

Burgess' bounds with prime moduli

Let $S_{\chi}(N)= \sum_{n\leq N} \chi(n)$, where $\chi$ is a Dirichlet character modulo a prime $p$. For integral $r\geq 2$, Burgess gave bounds of the shape $S_{\chi}(N) \ll N^{1-1/r} p^{(r+1)/(4r^{2})} \log p$. The factor of $\log p$ was reduced by several authors. We show that it can be removed completely.
Read the full abstract

Differential Geometry

Left-Invariant Kernels of the Hodge-de Rham Operator on Three-Dimensional Lorentzian Lie Groups

We determine the kernel of the Hodge–de Rham operator on left-invariant differential forms for the seven customary families of three-dimensional Lorentzian metric Lie algebras. Every parameter stratum is treated, including zero structure constants, singular bracket matrices, and null modular directions. If $\theta(X)=\mathrm{tr}\,\mathrm{ad}_X$ is the modular character, the full kernel is the direct sum of constants, the space $K=\mathrm{Ann}([\mathfrak{g},\mathfrak{g}])\cap\theta^\perp$, its Hodge dual, and the volume forms. We give explicit real bases in all seven families and compute the exterior differential and geometric codifferential on every basis form. The non-unimodular calculation remains valid without the customary extra normalization conditions, and hence also covers metric representatives omitted by those conditions. The resulting dimensions are $2,4,6$, or $8$; the two middle-degree dimensions agree. Explicit examples distinguish the kernel of the first-order operator from the kernel of its square. All assertions concern smooth left-invariant forms, with no square-integrability condition.
Read the full abstract

Robotics

It Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for Assembly

Geometric assembly requires fine spatiotemporal coordination to place parts without causing collisions or deadlocks under tight clearances. To address this, we present an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the pieces, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component. Per-agent proximal policy optimization with a centralized graph critic directly learns motion and temporal coordination within a single feedback policy. We study this formulation through planar Tangram assembly, where pieces must slide and rotate into a given configuration while remaining on the assembly surface. Procedural generation of target configurations and randomized initial arrangements support training and evaluation across diverse assembly problems. Our experiments isolate the roles of relational structure and multi-agent optimization, and compare learned control with sampling-based planning. We demonstrate concurrent assembly in simulation and transfer the policy to sequential execution on a real robot, as our object-centric formulation decouples learned assembly coordination from low-level physical control. Project page: https://tangrambench.org/
Read the full abstract

All authors Paul Knaußt, Kay Pompetzki, Nico Bohlinger, Niklas Funk, Cristiana de Farias, Rolf Findeisen, Jan Peters

Signal Processing

Energy-Aware Vertically Layered Pinching-Antenna System Design with Probabilistic LoS Availability

This paper investigates a vertically layered pinching antenna (PA) system, in which dielectric waveguides deployed at different heights provide additional vertical spatial degrees of freedom for line-of-sight (LoS)-favorable PA placement. We consider a practical energy model that captures in-waveguide attenuation, base station-to-feed-point optical-fiber loss, and PA repositioning energy, along with a statistical LoS-availability model that depends on elevation and distance. We then jointly optimize the transmit beamforming, PA positions, time allocation, and transmission rates to maximize the expected successfully delivered frame-effective sum rate under per-user outage and quality-of-service constraints. To avoid enumerating the exponentially many LoS states, we derive a certified conditional-moment outage envelope for outage-guaranteed rate selection and a differentiable safe approximation of the chance constraints to facilitate tractable optimization. Building on these results, a block coordinate descent algorithm with a penalty-aided majorization-minimization procedure attains a high-quality suboptimal solution. Simulation results confirm the gains of vertical layering and reveal a fundamental energy-LoS tradeoff: LoS-favorable PA placement generally enhances both performance and reliability, although it incurs higher attenuation, path loss, and energy costs.
Read the full abstract

Social and Information Networks

Dynamic signatures of collective knowledge through navigation in a network with semantic structure

What distinguishes how humans navigate an information network like Wikipedia from what the underlying graph structure and language statistics alone predict? To study human collective navigation, we analyze 70,770 human trajectories through Wikispeedia, an online game where players navigate between articles to reach a target. To compare groups of trajectories rather than individual ones, we merge all trajectories leaving or entering a given article into a Flow Tree. Flow Trees compress collections of trajectories into hierarchies of group splits. We build forward and reverse Flow Trees for 815 of the most commonly traversed articles, representing divergence and convergence, respectively. We compare humans against four random-walker null models, either unweighted or weighted by graph structure, word frequency, or semantic similarity to target. Human Flow Trees are deeper with longer trunks and later splits, indicating players stay together longer than random walkers, whose Flow Trees tend to fan out immediately. An article's degree predicts 51.3-70.9% of the variation in Flow Tree branching factor for random walkers, but only 24.1-36.2% for humans. Taking what graph structure fails to predict as a two-dimensional coordinate system, we define four article types (beacons, funnels, forks, and corridors), distinguished by how humans enter and exit them. These human-defined types are not explained by out-degree or title-word frequency are, and are not recovered by any random walker. A fifth null model, built from first-order human click preferences, partially reproduces human Flow Tree shape but differs significantly in how much of that shape degree explains. We consider collective knowledge to be the structure shared by players who never communicate, beyond what graph structure imposes. It comprises first-order click preferences and a significant component that extends across clicks.
Read the full abstract

Artificial Intelligence

I Think, Therefore I Attack: Evaluating Thought Injection Attacks on Reasoning Agents

Large Reasoning Models (LRMs) are one of the key components of current agentic systems. LRMs augment their output with an explicit chain-of-thought reasoning stage, introducing a new attack surface for thought injection, in which untrusted text resembling the model's own deliberation is fed back to the model. We present a two-part study of this attack, performed against the prevailing large, production class open-weight models (e.g., GLM-5.3, MiniMax-M2.5, Qwen3.8-Flash-Next). We first evaluate the effectiveness of static thought-injection prompts under different styles as well as tagged and untagged conditions and find that the effects of thinking tags and payload styles are all properties of the specific target model. In the second study, we exploit this model-specificity in a containerized agentic cyber range, using indirect injection prompts styled in the form of the target model's own reasoning traces to achieve attack success rates (ASR) up to 50% on operational objectives such as reverse-shell exfiltration and repository backdooring. We also find that the same malicious objective in standard prompt injection user voice never succeeds, and thinking tags suppress the agent's tendency to disclose the injection. Our results show that imitating a reasoning model's own thought style is effective for this style of attack, and that tag stripping is an unreliable defense.
Read the full abstract

All authors Parth Illendula, Garik Livingston, Zain Shamsi, Ravi Srinivasan, Alex Liu

Algebraic Geometry

Galois lines for the maximal curve with the second largest genus in characteristic two

We determine all Galois lines for the embedding of the maximal curve with the second largest genus in characteristic two into $\mathbb{P}^{3}$ induced by its Frobenius linear series. We also determine the associated Galois groups.

Information Theory

Sensing-Based Communication: A Non-Orthogonal Passive Reflection Framework for ISAC

While integrating uplink without disrupting the primary downlink remains a critical challenge for existing reflective communication systems, this paper proposes a non-orthogonal passive integrated sensing and communication (ISAC) mechanism termed Sensing-based Communication. Originating from and validated within an optical retro-reflective system, this mechanism possesses an underlying wave-agnostic architecture with the potential to be generalized across other physical modalities. It leverages incident downlink signals for parasitic modulation via passive tags. By relying solely on fundamental on-off keying (OOK) reflection, the tag unifies the reflective sensing channel and the uplink communication channel, and simultaneously achieves both sensing and duplex communication. Consequently, our approach introduces uplink capability without sacrificing the downlink data rate or consuming additional spectrum resources. At the receiver, we design a joint decoding architecture combining cross-correlation computation and successive interference cancelation (SIC) to fully exploit the degrees of freedom in broadband signals for multi-user interference suppression. We derive closed-form expressions for the interference and bit error rate (BER) assuming zero-mean Gaussian signals. Using an experimental testbed, we validate the pivotal theoretical finding, the scaling relationship of SINR with signal bandwidth and cross-correlation time–where linear regression yields a strong goodness-of-fit with $R^2 = 0.9688$. Furthermore, evaluations under noise-like orthogonal frequency division multiplexing (OFDM) waveforms exhibit high consistency between theoretical derivations and simulation data ($R^2 = 0.9372$, mean absolute percentage error, $\text{MAPE} = 14\%$), confirming the model's accuracy and waveform generalizability.
Read the full abstract

All authors Xin Xiong, Mathew Salas, Sihua Shao, Chen Chen, Mostafa Hassanalian, Rafael Fierro

Computer Vision

Impact of color in computer and human-based object recognition under occlusion

Partial occlusions challenge object recognition in both human and computer vision (CV), but the impact is largely dependent on the occlusion and model design. Properties relating to color are particularly important in biological recognition and frequently leveraged for camouflage, however there are still open questions about what makes them effective and how they translate to CV. This work studies partial occlusions that create camouflage, including those that blend in with target objects or with scenes, which alter object representations in distinct ways. Experiments evaluate recognition performance on both human participants and models ranging from convolutional neural networks (CNNs) to transformers, and additional analysis is performed on models to uncover trends in processing and object representations under multiple occlusion types. Novel metrics offer explainability of model behavior under occlusion: non-occlusion precision (NOP) and non-occlusion recall (NOR) measure alignment of model activations with and without occlusion, and activation in occlusion (AIO) measures the percentage of activations that occur in occluded regions. Findings show that human recognition, and most transformers, are negatively impacted by generated camouflaging occluders, with scene-based camouflage performing worse than object-based. Classical CNNs show a strong opposite trend, performing the worst on neutral-colored occlusions, suggesting that contextually-related colors, particularly scene-based, benefit object recognition rather than disrupting shape or boundary cues. Additionally, generated camouflaging occluders for training augmentation improves robustness to unseen camouflage for selected CNN model, and occlusion metrics suggest trends about learned mechanisms. Findings have interesting implications for alignment in human and computer vision, model vulnerabilities, and explainability.
Read the full abstract

Thursday 5 November

High Energy Physics (Phenomenology)

Geometric Formulation of Covariant Spin Operator

In relativistic quantum theory, Lorentz boosts induce momentum-dependent Wigner rotations, complicating the coupling of particle spins in arbitrary frames. We describe particle spin spaces as fibers of an associate bundle over the velocity manifold, and different particle's spins residing at distinct fibers. Analogy to general relativity, we introduce parallel transport to bring spins at different velocities into a common frame, where they can be coupled. Using spinor-helicity variables, we construct the parallel transport operator along geodesics and show how it is uniquely determined by little group covariance. This construction gives explicit form of the covariant spin and orbital operators, and yields a covariant 3-point amplitude basis with definite spin and orbital angular momentum. Higher-point amplitudes can be constructively obtained and applied to the covariant partial wave expansion for amplitudes, form factors, and correlation functions, etc.
Read the full abstract

General Relativity and Quantum Cosmology

2-dimensional Metric-Affine Geometries and Dynamical Gravity

We study non-Riemannian metric-affine geometries in 2-dimensions. We explicitly show that if the connection of the space is not metric compatible (i.e. the non-metricity is non-vanishing) then the underlying 2-dimensional manifold is not an Einstein manifold, in contrast to the Riemannian case. Furthermore we prove that in general the curvature in such spaces is specified completely by the Ricci tensor which, however, is not symmetric now. An enlarged symmetry of the scalar curvature under vectorial connection shifts is also found. We then apply our results to the study of 2-dimensional Metric-Affine Gravity Theories and discuss some special cases.
Read the full abstract

General Mathematics

How random are affine maps over finite cyclic groups and generalized cyclotomic permutations over finite fields?

We analyze the dynamical behavior of affine permutations over the ring $\mathbb{Z}_{p^e}$ and generalized cyclotomic permutations (GCPs) over the finite field $\mathbb{F}_q$. We use the description of their functional graphs to determine two statistical parameters: the expected number of cycles and the expected length of a cycle that contains a random element. We obtain explicit closed-form expressions for both expectations and an exact reduction from GCPs to affine permutations. We compare these parameters with the corresponding expectations for random permutations, to determine how closely the cycle statistics of each family resemble those of random permutations.
Read the full abstract

Combinatorics

Longitudinal Subdivision Invariance of Periodic Directed Strips and Its Application to Elongated Octagonal Strips

We prove that longitudinal refinements of finite-width periodic directed strips preserve subdivision containment. The refinement lengths may be arbitrary positive integers, without a uniform upper bound, and no matching condition is imposed on the horizontal arcs. We construct subdivision models in both directions that preserve each designated rail as a whole. The equivalence also holds when the designated rail images are required to lie in a prescribed end. A nonperiodic example shows that reverse containment may fail without periodicity. As an application, we obtain an elongated octagonal strip $O_n$ by subdividing selected longitudinal arcs of the directed hexagonal strip $H_n$. We show that every bounded face of $O_n$ is an octagon for $n\geq 2$ and that $O_n$ and $H_n$ are subdivision-equivalent. Consequently, the known exact end in-degree thresholds and forced-width formula for $H_n$ also hold for $O_n$.
Read the full abstract

General Physics

Gravitational memory, spin memory and S-duality at null infinity of the heterotic string

We determine the asymptotic structure at future null infinity of heterotic string theory compactified on $T^6$, in the consistent truncation to the Einstein-frame metric, the axion–dilaton $\tau=a+ie^{-2\phi}$ and the gauge fields descending from ten dimensions, through first order in $\alpha'$. Three results stand out. First, the radiative data, balance laws and memory observables organize under the S-duality group of the four-dimensional theory: the scalar flux in the Bondi mass loss is the invariant metric of $SL(2,\mathbb R)/U(1)$ evaluated on the scalar news, the gauge flux is the norm of a self-dual radiation field, and at fixed vacuum the residual duality rotation acts with charges $0$, $1$ and $2$ on gravitational, electromagnetic and dilaton–axion radiation, so that S-dual processes share their gravitational waveform and memories and differ in their scalar memory. Second, the spin memory receives a gauge-field contribution whose ordinary part is fixed by the Dirac–Schwinger–Zwanziger product of the total charge with the change of the charge aspect; it is generated by mutually nonlocal dyons, and its dipole part agrees in magnitude and sign with the angular momentum radiated by a charge scattered off a magnetic monopole. Third, within the four-derivative effective action, the Gauss–Bonnet and Pontryagin couplings fixed by the Green–Schwarz mechanism leave the mass-loss and angular-momentum balance laws unchanged at first order in $\alpha'$ and first enter the evolution of the spin-two curvature coefficient, $\partial_u\Psi_0^0\supset\tfrac{3i\alpha'}{2}\tau_1\bar\Psi_4^0$, which combines dilaton and axion into the complex scalar charge aspect and is not invariant under the residual duality rotation. For black holes with fixed physical charges $\tau_1$ is of order one, so this is a first-order effect in $\alpha'/M^2$.
Read the full abstract

General Relativity and Quantum Cosmology

Optical Appearances and Multiple Lensed Rings of Self-Interacting Boson Stars with a Double-Well Potential

As horizon-scale imaging opens a new era of strong-field gravity tests, self-interacting boson stars have emerged as compelling dark matter candidates and natural black hole mimickers. We systematically investigate the optical appearances and multiple lensed rings of such horizonless objects illuminated by a geometrically thin accretion disk. To overcome the severe numerical stiffness caused by the slowly decaying scalar tail, we introduce a robust hybrid metric reconstruction framework—seamlessly bridging an interior 6-parameter Padé approximant to the exact exterior Schwarzschild spacetime via a globally $C^2$-continuous cubic spline. This ensures high-precision ray-tracing across the entire domain. For the adopted double-well potential, we identify a strict upper bound ($\phi_0 \lesssim 0.317$) for the existence of regular static solutions. Crucially, we uncover a striking non-monotonic evolution in the shadow size, driven by a macroscopic mass-amplitude turnover that is fundamentally rooted in the microphysical competition between attractive and repulsive self-interactions. Within the most compact configurations, extreme gravitational lensing generates prominent nested higher-order photon rings and distinctly separated secondary images. Furthermore, exploring the scalar potential parameter space ($\Lambda, \Gamma$) reveals a distinct diagonal gradient in the shadow size. Our findings demonstrate that next-generation VLBI observations can move beyond merely identifying horizonless objects, offering a novel geometric avenue to directly probe the microphysical couplings of dark matter scalar fields.
Read the full abstract

Computer Vision

Where Does an On-Device Prescription Reader Lose Medications? A Frozen-Replay Study, from a 270M Language Model to a 2B Vision-Language Model

On-device models let a phone read a prescription without sending the photograph anywhere, but small models fail in ways that are hard to attribute: to the camera, the OCR, the model, or the rules around it. We study a phone-based pipeline that reads a prescription photograph with PaddleOCR, extracts medications with a 270M-parameter language model, and filters the result with a deterministic backstop before showing suggestions. Replaying identical frozen inputs through each configuration, with deterministic generation verified, we compare configurations with paired document-level intervals on 49 captures of 15 synthetic prescriptions (6 typed, 9 handwritten by one writer). Production recall is 8%, and 0 of 45 on handwriting. The backstop removes no correct medication and cuts unsafe outputs from 45 to 1. On handwriting nearly all medications never reach the model in readable form; on typed sheets the model mostly picks the wrong text. We trace much of the handwriting floor to a defect in our own crop step, then replace the pipeline: rotated-box crops, a 2B-parameter vision-language model that re-reads prescription lines and extracts the medications, and per-line numeric grounding. On the same captures recall rises from 8% to 67% (paired +59 points, 95% CI +40 to +78): typed 19% to 100%, handwritten 0 to 20 of 45, with no unsafe output. Rotated crops account for most of the reading gain and the model for most of the recall gain, and the OCR confidence gate is now the main limit. The cost is 61 s per photograph against about 12 s, and about 1.9 GB of memory. An earlier claim that better OCR lowered recall does not survive the controlled comparison, and we withdraw it. All results are in-sample, from one writer and one device.
Read the full abstract

General Relativity and Quantum Cosmology

Metric perturbations in Einstein-Hilbert gravity: An alternative perturbative formulation

We derive the ordinary gauge-fixed quadratic operators of the four-dimensional Einstein–Hilbert action in exponential and inverse-metric-density variables on an arbitrary reference background. With a single nonlinear background-harmonic gauge transported through the exact field map, the density operator is the trace-reversed Lichnerowicz operator, while the exponential operator contains an additional algebraic Ricci-curvature term. This difference arises from the nonlinear transformation of the background tadpole and vanishes on Ricci-flat backgrounds; a cosmological-constant check gives the corresponding cancellation at Einstein saddles. The construction retains the exact determinant contribution and specifies the associated Faddeev–Popov operator. At the first interacting order on a flat background, we show that the two gauge-fixed cubic actions differ by a term proportional to the free field equation, and independently evaluate the same nonzero three-point coefficient in both variables, with its normalization matched to the Einstein-gravity interaction. An exact Schwarzschild configuration satisfies the same background-harmonic gauge and is stationary under unrestricted exterior variations of the gauge-fixed action. Its density and exponential reconstructions are checked together with the bulk and boundary terms; a separate radial variation illustrates the loss of an equation after imposing a restricted ansatz. These results provide an explicit comparison of established field coordinates at the classical and tree levels; quantum loop calculations additionally require a common measure, regulator and source prescription.
Read the full abstract

General Relativity and Quantum Cosmology

Kalb-Ramond black bounces: throat-blind shadows and throat-sensitive neutrinos

Regularizing a black hole removes its singularity, but which observations could detect the regularized core? We address this question for a Simpson-Visser black bounce built on the power-law hairy black hole that a Lorentz-violating Kalb-Ramond background produces. Measured in the areal radius, the bounce leaves the metric function untouched and changes only the radial measure. This single fact sorts the observables into two classes. The shadow, the innermost stable orbit, orbital frequencies and gravitational redshifts are blind to the throat. The instability of the photon ring, the brightness of its subrings, the oscillation phase of neutrinos and the energy deposited by neutrino pair annihilation respond to it through one universal factor. We derive an exact law for the subring demagnification that holds for any metric function. We also show that the throat enters the weak-field neutrino phase and the light deflection with different weights, so that combined photon and neutrino lensing can separate the mass from the throat. Ray-traced polarized images of a turbulent thin disk confirm that the direct image hides the throat while the first subring reveals it. For neutrinos emitted near a stellar-mass compact object the throat reshapes the flavor survival pattern and enhances the annihilation energy deposition. Two independent computations reproduce each headline number.
Read the full abstract

Computer Vision

ARC-VLN: Adaptive Recursive Computing-based Visual-Language Navigation

Efficient aerial visual-language navigation (VLN) requires reducing model storage and inference costs while maintaining instruction-conditioned action prediction. We present ARC-VLN, a token-type-aware recursive computation framework built on a language backbone with shared parameters. ARC-VLN retains full-depth computation for text and action tokens while adaptively selecting exit depths for visual tokens. Representations of early-exiting visual tokens are retained, and active-token compaction shortens the sequences processed by subsequent recursive blocks. To mitigate routing collapse to a single exit depth, we design a dedicated learning strategy that encourages the model to distribute visual tokens across multiple exit depths.On OpenFly's Test-Seen and Test-Unseen subsets, ARC-VLN achieves the lowest navigation error and the highest success rate and success weighted by path length among all compared methods. Meanwhile, it reduces the average number of decision steps per trajectory, total floating-point operations, mean model inference time per trajectory, and average GPU memory usage during inference, indicating its great potential as a VLN backbone for lightweight applications.
Read the full abstract

All authors Xiang Zhang, Xiangyu Dong, Yang Lu, Xiaoguang Ma

General Relativity and Quantum Cosmology

Transient Information in Moving-Mirror Quantum Correlations

We study how a finite dynamical history remains encoded in a quantum field after a moving mirror returns to a common thermal trajectory. Two compact deformations of the Carlitz–Willey ray map are related by reversal of a control parameter and share the same asymptotic null line and Hawking temperature. The pairing induces an exact even–odd grading of the peeling field and Bogoliubov transformation. The complex continuum covariance retains the transient sign at first order, while the leading monochromatic quadratic response is even under reversal. Finite-bandwidth wave packets convert this distinction into signed interference with the thermal reference covariance. In linear response, the signed squared correlation is quartic in bandwidth for the selected right-moving packets and cubic for matched Rindler/Milne packets. The relative cross-covariance phase has a linear bandwidth dependence, with its leading coefficient maximized in magnitude by the flat window among normalized real symmetric windows. Finite-amplitude corrections introduce an additional linear bandwidth term in the squared Rindler/Milne response. The distinguishing packets sample the transient history in retarded time, while both trajectories share the same asymptotic Hawking temperature.
Read the full abstract

All authors M. Baran Ökten, Asl{\i} S. Turan, Michael R. R. Good, Kuan-Nan Lin, Pisin Chen

Instrumentation and Detectors

Observation of Non-Equilibrium Phonon Loss Across a Solid-Superfluid Interface

We measure the performance of an athermal phonon calorimeter in vacuum and in thermal contact with superfluid $^4$He. Using a cryogenic photon calibration system, we show that the superfluid suppresses the collection of the phonon signal by over 99%. We conclude that submerged sensors will not benefit TESSERACT's superfluid helium dark matter target, but that superfluid helium may be useful to superconducting qubits as a sink for athermal phonons generated by radiation and other spontaneous sources. Our measurement is the first probe of solid-to-superfluid phonon transmission at energies higher than those stably supportable by the superfluid. We discuss thermal flow inside the calorimeter, as well as systematics from an alternative calibration scheme.
Read the full abstract

All authors TESSERACT Collaboration: A. Armatol, C. Augier, L. Bergé, J. Billard, H.J. Birch, J. Blé, C.L. Chang, Y.-Y. Chang, L. Chaplinsky, G. Cline, A. Cochard, I. Cojocari, J. Colas, M. De Jesus, P. de Marcillac, K. Dwinger, R. Faure, S. Fiorucci, M. Garcia-Sciveres, J. Gascon, C. Girard-Carillo, W. Guo, L. Haegel, S.J. Haselschwardt, S.A. Hertel, K. Hunter, L. Juigne, A. Juillard, A. Kavner, J. Lamblin, T. Le-Bellec, X. Li, J. Lin, R. Mahapatra, S. Marnieros, C.A. Marrache-Kikuchi, N. Martini, W. Matava, D.N. McKinsey, J. Menu, K. Moraa, V. Novati, E. Olivieri, B. Penning, H. Pinckney, M. Platt, M. Pyle, D. Poda, Y. Qi, M. Reed, R.K. Romani, I. Rydstrom, B. Sadoulet, S. Scorza, B. Serfass, P. Sorensen, S. Steinfeld, H. Su, A. Suzuki, R.L. Vaughn II, C. Veihmeyer, V. Velan, G. Wang, P. Vittaz, Y. Wang, M.R. Williams, J. Wuko