Quantum Physics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 3 September
We develop a quantum sensing framework for the phase independent detection of weak coherent optical displacements based on SU(1,1) interferometry. Unlike conventional quantum measurement protocols that require prior knowledge of the signal phase and coherent homodyne detection, the proposed approach estimates the displacement magnitude independently of its phase. We show that, under ideal lossless conditions, a conventional SU(1,1) interferometer employing only total intensity detection saturates the quantum Cramer Rao bound for displacement magnitude estimation. We further derive the analytical expression of the quantum Cramer Rao bound and the sensitivity of the conventional SU(1,1) interferometer with total intensity detection and systematically investigate its performance in the presence of optical loss. The proposed phase-independent intensity detection scheme achieves comparable performance over experimentally relevant operating regimes while eliminating the need for local oscillators, phase locking, and quadrature tracking. These results establish SU(1,1) based intensity detection as a practical platform for phase independent quantum sensing.
Sound
Knowing When Not to Answer: Pseudo-Ensembles for Abstention in Music Audio-Language Models
Music audio-language models are evaluated almost entirely by accuracy on multiple-choice questions. This protocol forces the model to commit to an option, so a lucky guess looks the same as real musical understanding. What is missing is a way to tell when the model does not know the answer, so that it can abstain instead of guessing. The usual solution, an ensemble of independently trained models, is far too expensive here, which leaves the entropy of a single predictive distribution as the only available confidence signal. We instead build pseudo-ensembles from one pretrained model by perturbing its input in ways that cannot change the correct answer, then averaging the resulting distributions over the options. Our main construction simply shuffles the order in which the candidate answers are presented; we also study ensembles built from corrupted audio and from swapped option labels. A pseudo-ensemble gives several predictive distributions per question, so it supports the full family of ensemble-based uncertainty measures (entropy of the expected distribution, expected entropy, and their difference, the mutual information) rather than entropy alone. Evaluating TinyMU on MuChoMusic, we find that averaging over four option orderings raises accuracy from 55.7% to 59.2%, and that the resulting uncertainty measures rank the model's errors better than the single-pass entropy baseline, reducing the area under the error retention curve from 0.293 to 0.261. All of this costs a few extra forward passes and no retraining, which makes abstention practical for compact music audio-language models.
Read the full abstract
Astrophysics of Galaxies
Hint of bimodal Mg-Al anticorrelation in the metal-poor Globular Cluster NGC 4372
We present a detailed chemical abundance analysis of seven red giant branch stars in the very metal-poor globular cluster NGC~4372, based on high-resolution UVES spectra. Our study aims to characterize the chemical evolution of NGC~4372 and investigate the presence of multiple populations. We derived abundances for 17 elements, including light, iron-peak, $\alpha$, and heavy elements, using both equivalent width and spectral synthesis methods. We find a mean cluster metallicity of [Fe/H]=$-$2.36 $\pm$ 0.03 dex, with no evidence of an intrinsic iron spread. The cluster exhibits a typical $\alpha$-enhancement of [$\alpha$/Fe]= 0.33 $\pm$ 0.05 dex and a well defined Na-O anticorrelation, confirming that NGC 4372 hosts at least two distinct stellar populations. Furthermore, the measured ratio of [Ba/Eu] =$-0.47 \pm 0.04$ indicates that the synthesis of heavy elements was dominated by $r$-process, consistent with other very metal-poor environments. We also identify a potential bimodal Mg-Al anticorrelation, which aligns with the patterns observed in GCs of similar mass and metallicity. These results reinforce the status of NGC 4372 as a cornerstone for understanding the early chemical enrichment of the Galactic Halo.
Read the full abstract
Algebraic Geometry
On the torsion in the Chow motive of an Enriques surface
The integral Chow motive of an Enriques surface $S$ is determined by a torsion summand $\mathfrak{t}(S)$, which contributes the torsion in the cohomology of $S$. We study isomorphisms between and endomorphisms of motives of the form $\mathfrak{t}(S)$. This is closely related to the integral Hodge conjecture for products of two Enriques surfaces.
Read the full abstract
Mesoscale and Nanoscale Physics
Quantum oscillations of helical edge states of periodically deformed 2D topological insulator in magnetic field
We study edge-state transport in a two-dimensional topological insulator with a periodically deformed edge subjected to a uniform magnetic field. Zeeman coupling breaks time-reversal symmetry and enables elastic backscattering, producing oscillations of the forbidden-band widths. In the strong-field regime, the gaps can close completely at discrete field values. In the weak-field regime, we identify an important class of periodic deformations for which the dominant semiclassical scattering is controlled by complex infinity rather than by the nearest turning points. We develop a semiclassical treatment of this process and establish its agreement with perturbation theory and direct numerical calculations. The gap modulation should produce observable oscillations of the edge conductance. Unlike conventional magnetic quantum oscillations, which are periodic in inverse field, the predicted oscillations are periodic in the magnetic field itself, with a period determined by the Fermi velocity and effective g-factor
Read the full abstract
Quantum Physics
p-Adic Dirac Equations, Continuous-Time Quantum Walks, and Quantum Networks
We introduce a new class of p-adic Dirac equations, formulated in the standard axiomatic framework of quantum mechanics, in which the ordinary momentum operators are replaced by bounded non-local operators built from arbitrary integrable radial kernels. We diagonalize the free Dirac Hamiltonian in momentum space, construct its plane-wave solutions, and determine its spectrum, which is pure point, with eigenvalues of infinite multiplicity; in the massless case we describe the zero-energy subspace explicitly. We identify the charge-conjugation symmetry of the model: it is given by an antiunitary operator whose square is minus the identity, it exchanges the positive- and negative-energy subspaces, and in an external electromagnetic field it reverses the sign of both the charge and the vector potential. We then discretize the free equation in two ways, both giving genuine continuous-time quantum walks rather than discrete-time, coined walks: one on a countable graph and one on a finite, tree-like graph. In both cases the discretization is exact: the walk is the restriction of the p-adic Dirac evolution to an invariant subspace, and its Hamiltonian is a Dirac-type combination of weighted graph Laplacians with ultrametric weights. We prove that the Born-rule transition probabilities, summed over the internal degrees of freedom and averaged over the initial internal state, form a doubly stochastic matrix at every instant, and that, as for any non-trivial continuous-time quantum walk, this family of matrices is not a Markov semigroup. We discuss how the construction can serve as the foundation of a quantum network with four-component spinor degrees of freedom, complementing earlier, non-relativistic p-adic quantum neural networks, and prove global well-posedness of the associated nonlinear network equation. We close with a discussion of open problems.
Read the full abstract
Image and Video Processing
Cross-modal triage network: a multimodal deep learning framework for severity-based triage and visual explainability in chest radiographs
Purpose: Increased number of chest radiograph (CXR) scans create a triage bottleneck, queueing urgent examinations behind routine ones. Existing AI tools are predominantly unimodal binary classifiers lacking severity awareness, and multimodal systems are rarely benchmarked against expert radiologists. To this end, we developed a multimodal deep learning framework for joint severity triage, pathology detection, and native visual explanation. Approach: We propose the cross-modal triage network (CMTN), fusing a Swin Transformer V2 visual encoder with a PubMedBERT text encoder via gated cross-attention. The CMTN was trained on 34,639 image-text pairs (12,489 patients) from MIMIC-CXR-JPG, optimizing an ordinal focal loss for four-tier severity triage and binary cross-entropy for 14 pathologies. Beyond quantitative benchmarking, attention heatmaps were evaluated against a blinded expert radiologist in a two-phase clinical audit comparing model triage output to expert severity assessment (100 cases) and grading spatial-semantic concordance (116 heatmaps). Results: The CMTN achieved strong ordinal agreement with reference labels (quadratic weighted kappa [QWK] = 0.9341, 95% CI: 0.9219 to 0.9449) and macro-AUROC of 0.9970 across 14 pathologies, with 34~ms latency, outperforming the state-of-the-art BioViL multimodal baseline (QWK = 0.7679). However, the blinded Phase I clinical audit revealed substantially lower agreement with genuine radiologist judgment (QWK = 0.1399). Phase II found 54.3% of heatmaps achieved clinically acceptable spatial localization. Conclusions: The CMTN demonstrated an efficient multimodal architecture for CXR triage. The divergence between algorithmic and radiologist agreement demonstrates that benchmark performance against NLP-derived labels is insufficient, highlighting the need for radiologist-labeled ground truth before clinical deployment.
Read the full abstract
Cryptography and Security
Blockchain-Enabled Secure Logging for Fiscal Electronic Mechanisms: Evaluation of the Greek eSEND and myDATA Tax Systems
This paper analyzes the implementation of blockchain-based integrity mechanisms in Greek Fiscal Electronic Mechanisms (FEMs) and the central tax information system eSEND. The study examines the cryptographic architecture of fiscal devices, including Electronic Cash Registers, Fiscal Printers, Fiscal Signing Machines, and FEMAS devices, which implement double or triple hash-chain structures to ensure transaction immutability. The transmission protocol between fiscal devices and the central database is also evaluated with respect to encryption, sequential validation, and blockchain verification. In contrast, the architecture of Electronic Invoicing Provider Services and the myDATA central platform is analyzed, highlighting the absence of blockchain-based integrity guarantees. The comparison demonstrates that hardware-based fiscal mechanisms provide stronger guarantees for transaction completeness and tamper resistance than purely software-based invoicing infrastructures. The findings highlight architectural weaknesses in the current e-invoicing framework and propose improvements for ensuring transaction integrity in digital tax ecosystems.
Read the full abstract
Robotics
VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models
Pretrained vision-language-action (VLA) models enable broad manipulation but remain unreliable in tasks demanding precision and repeatability. Applying real-world online reinforcement learning (RL) to VLA post-training enables autonomous trial-and-error improvement beyond demonstrations alone, but exposes two bottlenecks: 1) unreliable value signals can induce policy drift; 2) large-VLA overhead constrains throughput and sample efficiency. To address these challenges, we present VLA-Precision, an efficient real-world online RL framework featuring the Asymmetric Co-Bootstrapping (ACoB) algorithm and the ACoB-Stream architecture. Specifically, ACoB establishes asymmetric co-bootstrapping across timescales: early intervention-guided behavioral learning rapidly improves policy performance while enhancing online experience quality. As autonomous experience accumulates, global return propagation and local preference ranking progressively calibrate value estimates, yielding relative action advantages for reference-regularized policy improvement while suppressing drift. To enable ACoB on large VLAs, we develop ACoB-Stream, a closed-loop experience–policy architecture that establishes invariant-state decoupling and on-demand streaming as design principles, delivering up to 10.9$\times$ improvements in throughput and computational efficiency. Extensive evaluations on nine high-precision chemistry tasks across four categories and four robot embodiments show that VLA-Precision achieves 98.3% mean success rate in 45.8 min/task, with 27.6 s episodes running at 1.2$\times$ and 1.8$\times$ the speeds of VLA and RL baselines. Resources are available at https://vla-precision.github.io.
Read the full abstract
Machine Learning
A Quantum Variational Approach to Prototypical Recurrent Unit
We introduce a lightweight Quantum Prototypical Recurrent Unit (QPRU) that requires significantly fewer parameters than both classical recurrent architectures, such as Long Short- Term Memory (LSTM) and Gated Recurrent Unit (GRU), and quantum variants, including Quantum LSTM (QLSTM) and Quantum GRU (QGRU). Despite its compact design, the QPRU achieves competitive forecasting performance, matching state-of-the-art baselines while offering important structural and practical advantages, including enhanced scalability and a reduced number of trainable parameters.
Read the full abstract
Dynamical Systems
Medvedev degrees of SFTs on cocompact Fuchsian groups
We prove that for any cocompact Fuchsian group the class of Medvedev degrees attained by subshifts of finite type is the class of $\Pi_1^0$ degrees. This result relies on the construction of a rigid hierarchical structure in a graph model of the hyperbolic plane, and on the fact that every such group admits a representation in $\operatorname{PSL}_2(\mathbb{R})$ such that the matrices in the image have coefficients that are computable real numbers.
Read the full abstract
Combinatorics
Enumerating Pattern-Avoiding Involutions using Combinatorial Exploration
The enumeration of pattern-avoiding permutations has been a popular area of study over the past several decades, but comparatively little attention has been given to the topic of pattern-avoiding involutions. In this paper, we derive the algebraic generating functions of two Wilf-equivalence classes of involutions avoiding a single pattern of length $4$, $\operatorname{Av^I}(2431)$ and $\operatorname{Av^I}(3421)$. We then adapt the Mosaic method, a fast counting algorithm for permutations, to count involutions and apply it to substantially extend the known initial terms of the counting sequences for the remaining two Wilf-equivalence classes avoiding a pattern of length $4$, $\operatorname{Av^I}(1324)$ and $\operatorname{Av^I}(4231)$. Based on these extended sequences, we empirically analyze the asymptotic behavior of the counting sequences of these two classes.
Read the full abstract
Astrophysics of Galaxies
Extending the Stellar-to-Halo Mass Relation to Dwarf Galaxies with DESI DR1
Constraining the dark matter halos of the smallest galaxies offers fundamental insights into the nature of dark matter and stellar feedback. Using the Dark Energy Spectroscopic Instrument (DESI) Data Release 1, we infer the stellar-to-halo mass relation (SHMR) down to the dwarf scale ($M_\star < 10^9\,M_\odot$), without extrapolation from the higher mass range. Leveraging the unprecedented depth of the DESI Bright Galaxy Survey at $0.01 < z < 0.2$, we construct 12 samples spanning nearly four orders of magnitude in stellar mass, and measure their projected clustering $w_p$, galaxy-galaxy lensing $\Delta\Sigma$, as well as a novel observable: satellite occupation number $N_{\rm sat}$. The addition of $N_{\rm sat}$ enables robust subtraction of satellite contributions to both $w_p$ and $\Delta\Sigma$ across the 12 individual halo occupation distribution analyses, yielding an average halo-to-stellar mass relation (HSMR) of $\log \langle M_h(M_\star) \rangle = 12.06 + 0.58\log(M_\star/10^{11}) + (M_\star/10^{11})^{0.73}$. Combining this HSMR with an observed stellar mass function, we constrain the SHMR across five orders of magnitude in halo mass, with the power-law slope steepening from $0.32 \pm 0.06$ above the Milky Way mass to $2.08 \pm 0.21$ in the dwarf regime. Interestingly, the scatter about the SHMR grows from $0.17 \pm 0.02$ dex at Milky Way-like scales to $0.68_{-0.33}^{+0.21}$ dex for systems comparable to the Large Magellanic Cloud, suggesting that smaller galaxies follow increasingly diverse evolutionary paths. Our work highlights the power of DESI in probing the galaxy-halo connection within the dwarf regime, offering an exciting avenue to bridge the gap between large-scale and near-field cosmologies in the future.
Read the full abstract
Computation and Language
Adapting from Downturns: Prediction of Long-Term Conversational-Skill Development in Mental-Health Crisis Counselors
How do people learn to become better conversationalists? This question is especially important in the context of mental-health counseling, where conversational skills are essential, yet volunteer counselors often have limited access to supervision and structured feedback. Understanding how counselors develop their ability to steer conversations toward positive outcomes – and identifying early which counselors are (not) on track to improve – can help prioritize support for the counselors who need it most. In this work, we introduce the task of predicting, early in a conversationalist's career, whether they will eventually improve at steering conversations toward positive outcomes, and demonstrate the feasibility of this task in the case of volunteer mental-health crisis counselors. Our central insight is that people may struggle with particular kinds of moments in a conversation, and that what is especially revealing of their likelihood of future improvement is how they learn to handle those moments over time. We operationalize this insight by designing a method that identifies the types of moments a counselor initially struggles with, captures how they adapt their response when they re-encounter similar moments in subsequent conversations, and learns which early adaptations predict improvement months or even years later. While this future-prediction task is challenging, our counselor-adaptation approach yields better results than baselines that learn directly from the conversation transcript.
Read the full abstract
Astrophysics of Galaxies
H I Absorbers as Beacons of Hidden Structure at $z \sim 3$: Multi-Component, Metal-Rich Absorption System near a Protocluster
H I gas traces the large-scale structure and provides the primary fuel for star formation. High-$z$ protoclusters are ideal laboratories to study how H I gas is accreted and consumed during the formation of the most massive structures in the Universe. However, much remains unknown about the distribution and physical state of their H I gas. We examine a rare configuration in which a protocluster candidate is located in front of a quasar at $z=3.09$. Our spectroscopic campaign confirms a protocluster at $z=3.079$: however, no corresponding strong H I absorption is found in the background quasar spectrum. Instead, we serendipitously discover a prominent H I absorption feature at $z\sim 3.01$, offset by $\sim 60$ cMpc from the centre of the protocluster. Spanning an exceptionally broad velocity range of $\sim 2000$ km s$^{-1}$ ($\sim 40$ cMpc), this absorption is decoupled from the confirmed member galaxies. Detailed kinematic modelling reveals this absorption comprises five distinct components rather than a single cloud. Moreover, one of these components exhibits a super-solar metallicity ($[\mathrm{O/H}] = +1.19^{+0.91}_{-0.78}$). We propose two physical scenarios for this unique system: (1) an additional, hidden massive protocluster along the line of sight, and/or (2) metal-rich outflows and metal-poor inflows driven by a single massive galaxy. The discovery highlights that while protoclusters are not universally associated with strong H I absorption, targeting the strong H I absorbers may serve as a beacon for uncovering massive, metal-rich protoclusters or complex gas kinematics in the early Universe.
Read the full abstract
Computer Vision
Object Concepts Emerge from Motion
Object-centric visual representations are important for physical-world perception, but existing visual pretraining methods often capture semantic categories without preserving the identity and coherence of individual instances. We present a biologically inspired framework that learns object-centric representations for single images from raw videos. Our approach uses motion boundaries as a source of object-level grouping: off-the-shelf optical flow and clustering produce pseudo-instance masks, which supervise a single-image encoder with pixel-level pairwise metric learning. The framework requires neither human annotations nor camera calibration. We first obtain 195 million pseudo-labeled frames from 7,163 hours of driving and web videos, then expand the supervision to 421 million frames with Motion-Verified Self-Training, which combines model proposals with motion evidence. We train encoders up to Swin-H and distill the learned representations into a family of Swin backbones. Across monocular depth estimation, 3D object detection, 3D occupancy prediction, and end-to-end planning, the resulting models achieve competitive or superior performance relative to supervised and self-supervised pretraining baselines, with particularly strong transfer on geometry- and instance-sensitive tasks. These results show that motion-derived supervision can teach static image encoders to represent visual instances, providing a complementary direction for scalable visual pretraining.
Read the full abstract
High Energy Physics (Theory)
UV/IR scale dependence in a four-derivative scalar field theory
An analog of the standard renormalization procedure may be useful for four-derivative theories, and we illustrate it with the four-derivative scalar field theory in the $m\to0$ limit. The renormalization scale $\mu$ appears in momentum-dependent logarithms associated with both UV and IR divergences. At one loop, the on-shell scattering amplitude satisfies an analog Callan–Symanzik equation involving the analog beta functions. These functions are determined from the combined UV and IR pole parts of the 1PI functions.
Read the full abstract
Formal Languages and Automata
Languages and Recognition in a Category with Factorisation
Language recognition by homomorphisms is a central construction of algebraic language theory. Initially studied for monoids and semigroups, it has subsequently been expanded to other algebraic structures. Our new categorical account is based on fibrations, which have already seen other applications in automata theory. Languages and surjective homomorphisms give indeed rise to two fibrations, and the notion of language recognition is stable under reindexing. We develop this framework in a category with a factorisation system and address two main technical questions in the fibrational setting. First, we provide sufficient conditions under which languages have syntactic quotients (which is a generalisation of syntactic congruences) and we show how such quotients can be described in some concrete cases using a result by Slomiński. Second, we introduce sufficient conditions under which (regular) languages are closed under certain J-limits and J-colimits.
Read the full abstract
Systems and Control
A Constraint-Aware Generative Framework for Synthetic Origin-Destination Demand in Logistics Networks
Large-scale logistics networks require synthetic data generation capabilities to support scenario-based planning under novel conditions-such as network reconfiguration and demand shocks. Existing approaches, which rely primarily on historical observations, lack the ability to generate demand patterns that adapt to changes in network topology while respecting operational constraints. We propose a constraint-aware conditional generative framework for synthetic origin-destination demand generation in hierarchical logistics networks. The framework models demand as a conditional distribution over destinations given each origin, enabling topology-aware synthesis that is both topologically realistic and operationally feasible. Operational guidance is incorporated directly into the generative objective via differentiable constraints, while a flexible conditioning mechanism supports various operational contexts and adaptation to evolving network configurations. We instantiate the proposed framework based on a conditional generative model. Experimental validation on industrial real fulfillment and transportation network demonstrates 16% improvement over graph neural network baselines, 87% operational compliance, and efficient cold-start adaptation, enabling applications in capacity planning, network design evaluation, and routing optimization.
Read the full abstract
Machine Learning
SharedSAE: One Feature Dictionary Across Language Models
Sparse autoencoders (SAEs) are widely used to interpret language model activations, but SAE training and latent labelling are typically repeated for every model. Here, we show that a single shared SAE can replace a collection of dedicated per-model SAEs. Our method, SharedSAE, combines a shared dictionary with model-specific encoder-decoder pairs. Unlike the closest prior method, which discards activation magnitudes and requires all models at inference, SharedSAE instead normalizes only selection scores, preserving magnitudes, and uses model dropout for single-model inference. We train SharedSAE on four 1B-scale base language models spanning distinct families and tokenizers. Despite sharing its latents across models, SharedSAE retains 96.6% of dedicated SAEs' mean explained variance; its latent activations exhibit cross-model correlations 1.8 times as high as separate SAEs aligned post-hoc, and its latent descriptions transfer across models. After the dictionary is frozen, new models can be efficiently adapted to it, achieving near-dedicated-SAE reconstruction quality while reusing the shared latent descriptions.
Read the full abstract
Artificial Intelligence
A Removal Based Approach to Improve LLM Faithfulness at Test-Time
Large language models (LLMs) are increasingly used for consequential decisions, making their explanations an important tool for auditing model behavior. Unfortunately, these explanations can be unfaithful, failing to reflect the actual reasoning underlying the model's decisions. We consider a setting in which an LLM provides both an answer and an explanation in response to a question. We identify two distinct dimensions of unfaithful explanations: incompleteness, meaning that the explanation omits factors that influence the answer, and unsoundness, meaning that the explanation cites factors that did not influence the model's answer. Existing approaches to improving LLM faithfulness include training-time methods, which require access to model weights and extensive computational resources, and test-time methods that largely focus on addressing unsoundness. We introduce a test-time approach that directly targets incompleteness. We remove from the input the concepts not credited in the model's explanation and re-query the model on the reduced input. This eliminates unmentioned influences while preserving the influence of mentioned concepts. Across two datasets, multiple model families, and two independent faithfulness metrics, our approach improves explanation faithfulness compared to both standard prompting and prompting to encourage faithfulness. Our method is model-agnostic and can be applied at inference time without modifying model parameters, providing a flexible mechanism for reducing hidden influences and improving the reliability and safety of LLM-assisted decision making.
Read the full abstract
High Energy Physics (Theory)
Superselected ghost theory: entangled pairs
The superselection-rule approach to ghost theories is extended to theories with a nonreal spectrum containing a complex-conjugate pair of poles on the physical sheet. The two excitations are labeled by their respective complex masses, $M$ and $M^*$, and the superselection sectors are labeled by $n$, the difference between the numbers of $M$ and $M^*$ excitations. The corresponding generalized ghost parity $Q$ assigns a phase $e^{in\alpha}$ to each sector. Under $Q$ superselection, nonvanishing norms occur only in the $n=0$ sector, and the physical state condition $Q^{2}|s\rangle=|s\rangle$ projects onto this sector. For a $MM^*$ pair carrying real total energy and momentum, we define a swap operation $R$. The $R$-eigenstates are entangled pairs, and an $R$ superselection rule ensures positive probabilities. We discuss how the optical theorem describes physical cuts through these entangled pairs.
Read the full abstract
Representation Theory
Schatten norms and determinants of linear combinations of matrix tensor powers via virtual representations
Let $$X_n=\sum_{i=1}^s t_i A_i^{\otimes n},$$ where $A_1,\ldots,A_s\in M_d(\mathbb C)$ and $t_1,\ldots,t_s\in\mathbb C$ are fixed, while $n$ grows. Direct computation of determinants or Schatten norms of $X_n$ is exponential in $n$. For a single tensor power these quantities are elementary, and even the determinant of a generic two-term combination admits a reduction to polynomially many scalar factors; however, no analogous elementary reduction is available for three or more terms. We give an exact representation-theoretic method which, for fixed $d$ and $s$, computes $\|X_n\|_p$, $0<p<\infty$, and determinants in polynomial time in $n$. Schur–Weyl duality yields a simultaneous block decomposition, while Jacobi–Trudi identities in the Grothendieck ring replace Schur modules by signed combinations of tensor products of symmetric powers. For $d=3$, each irreducible contribution reduces to the difference of two explicitly computable symmetric-power terms, leading to an open-source implementation. In a single-thread CPU benchmark, a genuine three-term $3\times3$ trace-norm problem with $n=18$ is evaluated in about $47$ seconds, whereas just storing the unreduced matrix would require approximately $2.4\times10^{18}$ bytes. Direct and reduced computations agree to relative error below $3.4\times10^{-15}$ throughout their common range $n\leq9$.
Read the full abstract
General Relativity and Quantum Cosmology
Waveform models for the gravitational-wave memory effect: III. Phenomenological frequency-domain model for nonspinning binaries
We present a phenomenological frequency-domain model for the gravitational-wave (GW) memory signal from nonspinning binary-black-hole mergers on quasicircular orbits. We develop separate amplitude and phase models for the dominant $(l,m)=(2,0)$ spherical-harmonic mode of the GW memory signal. The amplitude and phase models are built from superpositions of elementary and transcendental functions, which can be evaluated efficiently. Both portions of the model are calibrated against a numerical-relativity surrogate model over mass ratios from one to eight. Their accuracy is assessed by computing their mismatch with the numerical-relativity models using the Advanced LIGO sensitivity curve from the fourth observing run. The mismatches are of the order $10^{-4}\unicode{x2013}10^{-3}$ over the parameter space of total mass covered by LIGO. The resulting frequency-domain model is a more computationally efficient waveform model for the GW memory signal than a related time-domain model earlier produced by the authors. An open-access implementation of both the time-domain and frequency-domain models is provided in the Python package GWMemoryModel, which can be used for relevant analyses of nonspinning binary black holes.
Read the full abstract
Machine Learning
Modular Deep Recurrent Neural Network: Application to Quadrotors
A modular deep Recurrent Neural Network (RNN) is introduced to facilitate the process of deploying various architectures of RNNs, and to automatically compute derivatives for gradient-based learning methods. The modularity leads to a set of new architectures, one of which includes feedforward inter-layer connections. By adding feedforward inter-layer connections in a multi-layer RNN, it is observed that the capability of the RNN to learn and model high-order dynamics and nonlinearities is significantly improved. The problem of vanishing/exploding gradient in space for a multilayer RNN is also alleviated using feedforward connections. These results are demonstrated using a quadrotor case study, for which a model of the altitude dynamics is learned with our particular network structure, while existing methods are unable to generalize as quickly or at all.
Read the full abstract
Data Structures and Algorithms
Computing Lewis Weights to High Precision by Fixed-Point Iteration
The $\ell_p$-Lewis weights of a matrix are defined by a fixed-point equation. For $p<4$, Cohen and Peng [CP15] showed that iterating an equivalent rearrangement of this equation computes Lewis weights to high precision; for $p\geq4$, prior high-precision methods instead use optimization-based approaches. We show that the direct Lewis fixed-point iteration, appearing in the thesis of Lee [Lee16], computes Lewis weights to high precision for every $p>2$. For a matrix $\mathbf{A}\in\mathbb{R}^{m\times n}$ partitioned into row blocks $\mathbf{A}_{[1]},\ldots,\mathbf{A}_{[k]}$, we compute, for $p>2$, coordinatewise $\varepsilon$-approximate $\ell_p$ block Lewis weights in $O\left(p\log\frac{p\sqrt{\sum_{i=1}^k\operatorname{rank}(\mathbf{A}_{[i]})}}{\varepsilon}\right)$ rounds of exact leverage-score-vector computations. For ordinary Lewis weights, this becomes $O\left(p\log\frac{p\sqrt{m}}{\varepsilon}\right)$, improving the $O\left(p^2\log(m/\varepsilon)\right)$ bound of Gribling, Sidford, and Zhang [GSZ26] for $p\geq4$. Our guarantees also extend to approximate leverage-score computations. Our main observation is that each direct Lewis update contracts the KL divergence to the true weights by a factor of $1-\frac{2}{p}$. We also give an alternate explanation of this contraction through volume sampling and entropic independence. Synthetic experiments closely match our predicted local contraction rates and iteration counts grow approximately linearly with $p$; real-data experiments illustrate the information-concentration tradeoff of finite-$p$ block Lewis designs.
Read the full abstract
Nuclear Experiment
Heavy Flavors and Quarkonia at RHIC
After the discovery of the strongly coupled quark-gluon plasma (QGP), a nearly perfect fluid, in 200 GeV Au+Au collisions at RHIC, understanding its microscopic structure and transport properties has become a central goal of relativistic heavy-ion physics. Heavy-flavor particles, containing charm or bottom quarks, provide unique sensitivity to the QGP because they are produced predominantly in the initial hard scatterings and interact with the medium throughout its evolution. This review summarizes measurements of open heavy flavor and quarkonia by the PHENIX and STAR experiments, focusing primarily on 200 GeV collisions recorded during the first two decades of RHIC operations. We discuss key results on charm, bottom, and quarkonium production cross sections; cold nuclear matter effects in small collision systems; nuclear modification factors and elliptic flow of open heavy-flavor hadrons; charm baryon-to-meson production ratios; and the suppression patterns of quarkonium states in Au+Au collisions. We further highlight the resulting insights into the properties of the QGP, including heavy-quark transport, hadronization mechanisms, and quarkonium dissociation and regeneration in the medium. Finally, we discuss the future prospects of the RHIC heavy-flavor program, enabled by the large data sets collected by the STAR and sPHENIX experiments, and its strong synergy with the future Electron-Ion Collider, where precision measurements of heavy-flavor production in electron–proton and electron–nucleus collisions will provide complementary constraints on the structure of QCD matter.
Read the full abstract
Computation and Language
MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering
Medical visual question answering (Med-VQA) is often assumed to require medical fine-tuning, large models, or complex multi-agent pipelines. We revisit this assumption with MedProb, a lightweight probing framework that predicts multiple-choice Med-VQA answers from frozen VLM representations without free-text generation. Across PATH-VQA, SLAKE, and VQA-RAD, MedProb recovers substantially more answer-relevant signal than prompting and performs stronger than medical VLMs and agentic systems. Probing also reduces the apparent gap between small and large models compared to prompting, suggesting that smaller VLMs contain more recoverable Med-VQA signal than generation-based evaluation reveals. Across 14 matched general-purpose and medical VLM pairs, medical adaptation does not consistently improve this linear decodability. Finally, free-text generation exhibits an answer-position bias of up to 10 percentage points, whereas MedProb also has positional bias, however, it is impacted differently than prompting. Our main results target the multiple-choice/multiclass Med-VQA setting; we additionally show the probe can be extended to open-ended generation via a rejection-sampling scoring procedure.
Read the full abstract
Complex Variables
Positivity of Smooth Currents on Singular Spaces
We study two notions of positivity for smooth currents on singular spaces. We show that they are not equivalent by establishing a necessary condition on the fourth Whitney cones. We also construct an explicit compact normal projective variety $X$ and a real smooth $(1,1)$-form $T$ on $X$ such that $T$ has local smooth ddc-potentials and is a Kähler current, but $T$ is not a positive form (hence not Hermitian).
Read the full abstract
Cosmology and Nongalactic Astrophysics
Model Predictive Scoring Shows Specific BAO Observations (not SNIa) Drives $w_0w_a$ Tension
The recent analyses of DESI DR2 BAO, Planck CMB and various Supernovae datasets have shown preferences for evolving dark energy at various levels of significance, under frequentist model comparison tests. However, the same analysis done with a pure Bayesian model comparison test can lead to the opposite conclusion due to the impact of $w_0w_a$ priors. In contrast to these in-sample model comparison route, we approach the problem along a third axis – which model better predicts unseen data. We use the Expected Log Predictive Density (ELPD) metric to score the predictiveness of $\Lambda$CDM and $w_0w_a$CDM models, using the leave-one-redshift-block-out cross-validation estimator. The comparison metric $\Delta$ELPD is out-of-sample, unlike $\Delta \chi^2_{\rm MAP}$ and Bayes Factor, and insensitive to prior width, unlike the Bayes factor. Aggregated, we find modest preferences for $w_0 w_a$CDM, primarily driven by the BAO. Decomposing the $\Delta$ELPD score by redshift we find that a single BAO block, LRG2 ($z = 0.706$), supplies the entire BAO preference. Interestingly, the biggest outlier point, LRG1, contributes little to the model predictive scoring because both $\Lambda$CDM and $w_0w_a$CDM fail to predict the observation by equal amount. On the other hand, all SNIa datasets' $\Delta$ELPD scores are consistent with $0$, indicating that at an out-of-distribution predictive level, the $w_0w_a$CDM tension is entirely driven by one BAO point, not SNIa.