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

WireSeg-32K: A Physics-Grounded Synthetic Dataset for Wire Instance Segmentation

Deformable linear objects such as wires and cables are difficult to segment because they are thin, highly deformable, and frequently self-occluded, while large-scale instance-level annotations are expensive to obtain in real scenes. Existing resources either focus on cable tracing or semantic segmentation under constrained settings, or generate visually plausible images without physically grounded wire deformation. We present WireSeg-32k, a synthetic dataset for wire instance segmentation with 32,000 RGB images, instance masks, depth maps, and a complementary real-world test set with annotations. To generate this dataset, we develop DeformX, a co-simulation pipeline that couples Cosserat-rod dynamics with photorealistic Isaac Sim rendering, enabling physically plausible, contact-consistent wire shapes, CAD-based wire assets, and diverse visually grounded scenes. As a simple baseline, LoRA fine-tuning SAM3 on WireSeg-32k alone improves real-world mAP@75 by 10.2% over the off-the-shelf model, showing that physically grounded synthetic data can transfer to real wire perception.
Read the full abstract

All authors Zilin Dai, Lehong Wang, Yi Yang, Xiang Fei

Analysis of PDEs

The emergence of the Cauchy horizon from the crease for $3D$ compressible Euler flow

We study open sets of nearly plane symmetric data for the $3D$ compressible Euler equations with dynamic entropy and non-trivial vorticity. In our prior work, we proved that the solutions develop a gradient singularity along a singular boundary, which is a hypersurface that emanates from a co-dimension 2, spacelike submanifold called the crease. In the present paper, we prove that a null hypersurface, called a Cauchy horizon, emanates from the crease, and propagates in a direction transverse to the singular boundary. The solution remains smooth up to the Cauchy horizon, even though it “feels” the influence of the gradient singularity at the crease. The union of the Cauchy horizon, the crease, and the singular boundary, make up a connected portion of the boundary of a maximal globally hyperbolic development (MGHD) of the data. Roughly, an MGHD is the “largest” way that smooth data can evolve into a classical solution. Known examples show that the only way one can guarantee uniqueness of an MGHD is by proving that certain structural properties hold along its entire boundary. We prove that a local version of the needed properties are satisfied along the Cauchy horizon and singular boundary. Our work provides the first description of the formation, structure, and stability of an $O(1)$-size portion of the Cauchy horizon in $3D$ without symmetry assumptions. The presence of dynamic entropy and vorticity stretching, both of which are absent in simpler settings such as $2D$ isentropic Euler, introduces significant difficulties that we resolve with novel techniques. Our approach relies on new foliations of spacetime dynamically adapted to the shape of the crease, and a PDE framework based on double-null foliations and the energy identities we derived in earlier work. Collectively, our techniques allow us to handle a challenging new solution regime where the crease lacks strict convexity.
Read the full abstract

Machine Learning

Distilling deep optical flow stereo methods to retrieve dense three-dimensional wind fields

Geostationary atmospheric motion vectors (AMVs) provide the dense horizontal wind vectors (u,v) and heights ingested into data assimilation systems. Traditional AMVs track features using window-based cross-correlation and estimate heights via infrared brightness temperatures paired with numerical weather prediction (NWP) background states, creating a circular dependency that yields inaccurate heights, high computational cost, and sparse retrievals. Stereo winds from GEO-GEO and GEO-LEO geometrically resolve heights from parallax shifts across different poses, eliminating NWP dependence and improving accuracy, but they remain computationally heavy with limited coverage. In this work, we replace window-based tracking in stereo matching with deep optical flow for efficient, improved retrieval. Fine-tuning balances a self-supervised geometric residual loss with supervised radiosonde reconstruction. To eliminate multi-satellite overlap requirements, we distill the stereo teacher into a single-satellite student model. Chi-square and height uncertainties from the teacher are emulated by the student for quality assurance. The student generates winds across full-disk GEO imagery globally. Validation compares stereo and student models against radiosondes, operational AMVs, ERA5 reanalysis, and EarthCARE cloud profiles. Results through triple collocation show that stereo winds improve performance beyond operational AMVs for water vapor bands (6.2, 6.9, and 7.3 {\mu}m), wit degradation in the long-wave infrared (11.2 {\mu}m) band.
Read the full abstract

All authors Thomas J. Vandal, Dong L. Wu, James L. Carr, Derek J. Posselt, Elise Penn, Tristan Ballard, August Posch, Kate Duffy

Signal Processing

Enhancing UAV Trajectory and Communications Through Vision-Inertial Tracking

In this paper, we propose an energy-efficient and reliable communication system for non-terrestrial networks deployed in dynamic GPS-denied wireless environments, enabled by a Vision–Inertial Tracking-Assisted UAV Communication (VIT-UAVCom) system. To the best of our knowledge, this is the first work to exploit onboard UAV cameras and IMU sensors for UAV-assisted communications. We consider a complete VIT-UAVCom system that incorporates the key design parameters while explicitly accounting for system noise and residual tracking inaccuracies. Building on this framework, we formulate an optimization problem for jointly designing the UAV trajectory and communication performance to improve propulsion energy efficiency, reduce outage probability, and enhance physical-layer security. We then develop a dedicated solution framework to efficiently compute near-optimal trajectory and communication control actions in dynamic scenarios. Furthermore, to enable real-time implementation, we propose and evaluate three optimizers, namely linear search (LS), binary search (BS), and genetic search. Our numerical results demonstrate that our proposed VIT-UAVCom framework significantly outperforms the K-means benchmark in terms of energy consumption while maintaining robust secrecy performance and reliable user coverage. Specifically, our proposed framework improves the energy efficiency by 144% compared to the benchmark. Interestingly, our results also show that, compared with the LS, the BS reduces the computational time by approximately 50%.
Read the full abstract

All authors Abdallah S. Ghazy, Hussein A. Ammar, James Bayes, Francois Chan

Analysis of PDEs

Standing waves for Schrodinger equations with Kato class potentials and $L^\infty$-bounded nonlinearities

We establish the existence of standing waves for a nonlinear Schrodinger equation with potentials belonging to the Kato class and an $L^\infty$-bounded nonlinearity, whose Lipschitz constant is smaller than the distance from zero to the essential spectrum of the linear part. We consider both the nonresonant and resonant cases. Our approach is based on the Conley index theory applied to study invariant sets of the associated parabolic semiflow. Using properties of the Schrodinger semigroup with Kato-class potentials, which follow from its Feynman-Kac representation, we derive a priori estimates for bounded solutions in the $L^\infty$ and $L^2$ norms, as well as regularity bounds in Sobolev spaces. As a consequence, we obtain conditions ensuring the existence of connecting orbits between stationary solutions, which in turn yield the existence of nontrivial standing waves.
Read the full abstract

Quantum Physics

Optimal control theory for measured quantum Schrödinger bridges

Schrödinger bridges and entropic optimal transport are usually formulated as stochastic interpolation problems between initial and final probability distributions. In their computational form, the bridge potentials are obtained by Sinkhorn or iterative proportional fitting, and are often regarded as auxiliary scaling functions. We show that for continuously monitored quantum systems, these potentials acquire a direct measurement-theoretic meaning. The mathematical structure of conditional quantum trajectory theory induces a Fokker–Planck equation on quantum state space. Conditioning this diffusion on a terminal distribution or a terminal measurement effect produces a Doob/Sinkhorn potential whose directional derivative along a unitary control vector field is the imaginary part of a generalized weak value. The same construction connects the Schrödinger-bridge viewpoint to the optimal-path framework for continuously monitored trajectories: the backward bridge potential plays the role of an effect-like costate, and its weak-value directional derivative gives the local control signal. By specifying the desired endpoint distribution, the induced drift produced by the Schrödinger bridge solution is the control solution that minimizes the quadratic cost feedback law, guiding the distribution to its desired endpoint. We quantify the control score of the available control Hamiltonian as a logarithmic directional derivative of the bridge potential, or an imaginary weak value. Three explicit examples identify weak measurement as a natural entropic regularization mechanism for quantum state transport and gives a route from Sinkhorn scaling to quantum feedback and Hamiltonian control synthesis for practical optimal control.
Read the full abstract

Cryptography and Security

A Bayesian Correlated Equilibrium for Early Insider-Threat Detection

We model insider threat detection as a dynamic Bayesian game in which a platform coordinates a committee of strategic certifiers to sustain equilibrium among honest users and detect malicious deviations before exfiltration. Certifiers and users operate under a Bayesian Temporal Correlated Equilibrium (BTCE), where a sealed-envelope correlating device issues private recommendations over time and obedience is verified at every on-path information state. Unlike Stackelberg formulations, BTCE coordinates heterogeneous certifiers without requiring commitment power. We incorporate present bias and loss aversion to capture impulsive escalation dynamics, enabling 1.7–4.5 days earlier detection than rational baselines. We prove three guarantees: (1) calibrated intervention losses make recommended behavior a current-self best response despite behavioral biases, (2) controlled evidence accumulation guarantees intervention in bounded expected time before exfiltration, and (3) median aggregation confines implemented actions to the honest recommendation range when fewer than half of certifiers are Byzantine. On CERT r6.2 our mechanism achieves up to 28.3% pre-exfiltration detection with false positives below 1.6%, while both a transformer baseline and a streaming provenance approximation (HOLMESLite) achieve near-zero pre-exfiltration detection under comparable constraints.
Read the full abstract

Image and Video Processing

Beyond Blur: A Semantic Tri-view Pipeline for Teledermatology Gradability via Skin Micro-relief

Smartphone skin photographs are indispensable to teledermatology, yet assessing the diagnostic suitability of submitted cases (gradability) remains a critical bottleneck in mobile care workflows. Dermatologists routinely review multiple photographic views (regional, angled, and close-up) to identify consistent textural detail rather than relying on a single image. We present the Semantic Tri-view Pipeline, an interpretable architecture for automated teledermatology gradability screening that formalizes epidermal micro-relief as a computable biomarker of image quality. Using an expert-annotated subset of the public SCIN dataset, we train a lightweight DeepLabV3+ model to segment micro-relief fidelity. These spatial masks are then aggregated across up to three case views with a logistic regression classifier, leveraging viewpoint redundancy to support robustness under uncontrolled smartphone acquisition. This approach learns context-aware, clinically intelligible heuristics, such as penalizing high-fidelity texture in regional distance views. Evaluated at a predefined 90% sensitivity operating point, the system's apparent errors largely reflect subjective clinical variance on borderline cases where clinicians rely on non-visual metadata. On SCIN, performance improves from an AUC of 0.81 (80.6% PPV) on variance-heavy majority-consensus cases to 0.96 (97.7% PPV) on optically unambiguous unanimous cases. Overall, this work delivers an interpretable, privacy-by-design, edge-ready system that can provide real-time feedback during case submission to filter ungradable photo sets before review.
Read the full abstract

Statistics Theory

Discrete Gromov-Wasserstein Duality: Algorithms and Isomorphism Testing

The Gromov-Wasserstein (GW) distance provides a principled framework for aligning metric measure (mm) spaces based solely on their intrinsic structure. Its ability to identify isomorphic representations of distributions across spaces renders it valuable for comparing data where equality up to isomorphism occurs naturally such as in graphs or, more generally, distributions on graphs. Recently, a type of dual form for the GW distance between Euclidean distributions with the squared Euclidean or inner product costs was derived, spurring the development of new statistical and algorithmic results for this setting. This work furnishes a novel duality result for GW distances with and without entropic regularization that is applicable to all finitely supported mm spaces. Leveraging this result, we derive the sample complexity of empirical GW distances between finite mm spaces, as well as limit distributions under proper centering and scaling. Furthermore, we propose new algorithms for solving the regularized GW problem which are subject to formal convergence guarantees. These statistical and algorithmic advancements give rise to a principled and efficient framework for testing whether two distributions on the set of graphs with a fixed number of nodes are isomorphic based on samples.
Read the full abstract

All authors Gabriel Rioux, Joanna Marks, Riccardo Passeggeri, Ziv Goldfeld

Metric Geometry

Local stability for the truncated octahedral conjecture

This note proves that the truncated octahedron of Archimedes is a strict local minimizer of the isoperimetric quotient among parallelohedra in three dimensions. This extends the corresponding result from the five-dimensional lattice-Voronoi subclass to the full ten-dimensional local parameter space. A six-zone parametrization and Schur's lemma reduce the Hessian to two positive definite $2\times 2$ matrices.
Read the full abstract

Quantum Physics

Non-Kolmogorov-Arnold-Moser Quantum Sensors for Quantum Parameter Estimation

Non-KAM (Kolmogorov-Arnold-Moser) systems, when subjected to weak time-dependent perturbations, exhibit an abrupt transition to classical chaos through the breakdown of invariant phase-space tori. We showcase the utilization of non-KAM systems in the quantum regime as quantum sensors, leveraging their sensitivity at resonances. Quantum Fisher information (QFI) is a central quantity in quantum parameter estimation theory that measures how much information a quantum state contains about an unknown parameter that is encoded into it. In other words, it quantifies the sensitivity of a quantum state to small changes in that parameter. In this work, through numerical analysis in conjunction with analytical results, we study the performance of the non-KAM systems for quantum sensing applications by computing the QFI. We find that the growth of the QFI is remarkably enhanced when the resonance condition is satisfied. For frequency estimation under Floquet unitary encodings, we derive a transport bound: if the mean excitation number grows as $\langle\hat n(t)\rangle\sim t^{\alpha}$, the QFI obeys $I(t)\lesssim t^{2\alpha+2}$. The quantum kicked harmonic oscillator, a paradigmatic non-KAM system, realizes the full hierarchy: localized dynamics ($\alpha=0$) yield quadratic growth, delocalized diffusion along stochastic webs ($\alpha=1$) yields quartic growth, and translationally invariant resonances ($\alpha=2$) saturate the bound with anomalous hexic growth, $I(t)\sim t^{6}$, established analytically at resonance $R=2$ and numerically at $R=4$. The enhancement stems from resonance-induced translational symmetry rather than exponential instability, identifying non-KAM resonances as a metrological resource distinct from chaos-assisted and criticality-based sensing.
Read the full abstract

All authors Naga Dileep Varikuti, Sourav Manna, Athreya Shankar, Arul Lakshminarayan, Vaibhav Madhok

Number Theory

Perfect numbers, Wieferich primes and the solutions of $\binom{2n}{n}\equiv 2^n \bmod n$

In this article we focus on the solutions of a congruence equation: $\binom{2n}{n}\equiv 2^n \bmod n$. Using the main result of this article and the SageMath software, we improve largely the number of known solutions. Furthermore, we prove that some famous numbers like even perfect numbers and Wieferich primes are connected to solutions of this equation.
Read the full abstract

Machine Learning

The Gradient Does Not See Rank: Rank-Indifference in Matrix-CODI on ProsQA

Continuous chain-of-thought models compress reasoning into latent tokens. Matrix-valued variants, which route each latent token through a d x d matrix bottleneck, introduce rank as a single-sample structural observable on the latent matrix Z. If matrix latents carry parallel reasoning paths via superposition, rank should track them, and truncating Z to low rank should hurt accuracy on tasks whose solutions plausibly require multiple components. Across four training regimes of a matrix-CODI model (three on ProsQA, one on GSM8K-Aug below the learning threshold), the rank-k projection ablation curve is flat to within 0.6 percentage points. A three-seed replication yields 81.0 +/- 2.0 percentage points accuracy while the final effective rank of Z spans {4, 12, 13}; the loss does not reward any particular rank. To test whether rank-blindness arises from the flatten-then-project readout alone, we trained four readouts: a bilinear reparametrization, a bilinear-plus-GELU readout nonlinear in Z, an SVD-augmented readout feeding singular values through an MLP, and a quadratic readout in Z Z^T. All four rank-k curves remain flat (Spearman p-values 0.63, 0.14, 0.82, 0.46). The flat curves persist for readouts nonlinear in Z. A linear probe on Z underperforms a raw pretrained hidden state at target prediction (AUC 0.673 vs. 0.846). A negative control on vanilla GPT-2 SFT (no matrix bottleneck, no Z, three seeds, n=500) reproduces a flat rank-k curve under the same intervention paradigm with pooled-mean range 0.20pp, and a random-h sensitivity floor lands at the same accuracy: the rank-k ablation alone conflates rank-blindness with position-irrelevance.
Read the full abstract

2609.03090 · 2 Sep · Accepted at the ICML 2026 Mechanistic Interpretability Workshop, https://openreview.net/forum?id=Spof4PusVI. 9 pages. Corrects a data-entry error in the workshop version: the seed-1337 accuracy in the three-seed replication was reported as 80.47% (a control run); the archived value is 78.91%, so the three-seed mean is 81.0 +/- 2.0pp (was 81.5 +/- 1.2pp). All other results are unchanged

Cosmology and Nongalactic Astrophysics

TDCOSMO XXIX: JWST/NIRSpec IFU Spatially Resolved Kinematics of Three Time-delay Lenses

Spatially resolved stellar kinematics are critical to the high precision achieved by cosmological probes utilizing time delays of strongly lensed quasars. Combined with high-resolution imaging and lens modeling, dynamical models of the 2D resolved kinematics of the deflector galaxy tightly constrain the mass profile and break the mass-sheet degeneracy, in turn providing tight constraints on the Hubble constant when the time delays are included. We extract stellar kinematics of the deflector galaxies in the quadruply lensed quasar systems HE0435-1223, PG1115+080, and WFI2033-4723 from James Webb Space Telescope Near-Infrared Spectrograph (JWST-NIRSpec) integral field spectroscopy. The kinematic maps reach average per-bin total (statistical and systematic) uncertainties of 6-11%, with average bin-to-bin correlated errors of only $\sim 1.2\%$. The aperture-integrated velocity dispersions are statistically consistent with the values used in the previous TDCOSMO analysis (all within $0.9\sigma$), with their average uncertainty reduced from 3.7% to 2.4% owing to improvements in the data reduction and kinematic-extraction methodology. We classify PG1115+080 as a fast rotator, HE0435-1223 and WFI2033-4723 as slow rotators within the probed radii, and we refine the source redshifts from the kinematics of the lensed host galaxies. Kinematic maps will be combined with time delays, lens models, and line-of-sight convergence estimates to measure cosmological parameters in the upcoming TDCOSMO 2026 milestone publication.
Read the full abstract

All authors Shawn Knabel, Pritom Mozumdar, Anowar J. Shajib, Tommaso Treu, Devon M. Williams, Michele Cappellari, David Law, Simon Birrer, Takahiro Morishita, William Sheu, Massimo Stiavelli

Space Physics

Direct Four-Spacecraft Measurement of Reconnection Exhaust Thickness in the Solar Wind

The thickness of a reconnecting solar wind current sheet is normally inferred from a single spacecraft as the product of the boundary-normal speed and the crossing duration. Multipoint measurements have tested this method for generic solar wind current sheets and constrained the geometry of reconnection exhausts, but a direct comparison with single-spacecraft thickness estimates for the same reconnection-associated crossings is still lacking. We analyze four current-sheet crossings observed by Magnetospheric Multiscale in the pristine solar wind on 2017 November 10 the two boundaries of a confirmed reconnection exhaust and two nearby current sheets. The tetrahedron separation was $\approx 16$ km. Cross-correlation of the magnetic-field ramps gives millisecond-level lag uncertainties, boundary normals to $4^\circ-8^\circ$ and speeds to $5\%-8\%$, while single-threshold timing at these separations is noise dominated. The exhaust edges are 29-31 ion inertial lengths ($d_i$) thick. The two edge normals differ by $22^\circ$ (68% interval [$19^\circ,26^\circ$]), resolving a nonparallel exhaust geometry. At the exhaust edges, the measured boundary speeds agree with three standard single-spacecraft estimates to 4%-15% and remain within about one Alfvén speed of the plasma motion. Estimating the normal from one spacecraft is the larger source of error, though the thickness is still recovered to within tens of percent when a degenerate minimum-variance solution is rejected. At the actively reconnecting sheet, the boundary speed exceeds the local plasma speed along the normal by $29\% \pm 10\%$.
Read the full abstract

Software Engineering

Large Language Models and Language Server Protocol: a match made in context

This article introduces Eiffel-tools, a language server protocol (LSP) implementation for the Eiffel programming language that uses Large Language Models (LLMs) to aid the development of statically verified software. The tool provides various interactive and non-interactive commands to produce code and specifications. It uses language and project specific knowledge to precisely direct the LLM and verifies the output using a static verifier. It crafts rich programmatic prompts for the input and corrects or rejects the output. Furthermore, it handles the retries until the program passes verification. The tool's bug fixing capability is evaluated on 2 public datasets using 3 models. The tool can fix 76% to 95% of bugs by combining LLMs and a formal verifier depending on the model and prompts used. The results show the trade-off between the number of fixing attempts and the success rate.
Read the full abstract

Computer Vision

Solving the Needle-in-a-Haystack Problem in Mammography Vision-Language Model with Differentiable Subset Sampling

There is growing interest in adopting CLIP-style vision–language model (VLM) pretraining for mammography. However, models that directly employ the standard CLIP architecture and training objective exhibit limited zero-shot performance in clinically important tasks such as cancer, finding-type, and BI-RADS predictions. We argue that this underwhelming performance is due to neglecting two characteristics of mammography data: (1) its high-res nature, and (2) homogeneity of radiology reports, largely driven by a predominance of negative/benign findings on examinations. We propose TopKSigLIP, a VLM designed to address these two limitations through a novel architecture and learning objectives. Instead of downscaling high-res mammography images to satisfy GPU memory constraints, TopKSigLIP introduces TopK-Patch module that learns to sample a sparse set of high-res patches likely to contain lesions, sidestepping the resolution–batch size tradeoff of VLM training. The sampled patch locations additionally serve as a built-in localization tool. To address report homogeneity, we replace the contrastive loss, which falsely repels semantically similar pairs, with a Sup-sigmoid loss. Sup-sigmoid loss extends the sigmoid loss from SigLIP with soft labels derived from structured data. TopKSigLIP outperforms existing open-source mammography and general medical VLMs on both internal and external benchmarks on density assessment, BI-RADS classification, finding subtyping, and cancer prediction under zero-shot evaluation. TopKSigLIP remains competitive under linear probing despite using a significantly smaller vision encoder and smaller training batches than baselines. The TopK-Patch module additionally achieves superior lesion localization over post-hoc Grad-CAM. Code and weights are made public:https://github.com/Youngseok0001/TopKSigLIP.
Read the full abstract

All authors Young Seok Jeon, Beatrice Brown-Mulry, Rohan Satya Isaac, Anjana Dissanayaka, Theo Dapamede, Mohammadreza Chavoshi, Judy Gichoya, Hari Trivedi

Analysis of PDEs

Convergence rates to traveling waves for an FKPP model with nonlinear advection

We establish convergence rates of solutions to traveling waves for a reaction diffusion model with nonlinear advection. The convergence rate is proven using the shape defect function. We introduce a novel energy methods-based approach that uses a weighted Nash inequality to obtain $L^\infty$-estimates on the shape defect function. In determining the convergence rates, we also establish the front location for the model.
Read the full abstract

Astrophysics of Galaxies

Same galaxy, different CGM: how the metal loading of galactic winds regulates the baryon cycle in Milky Way-mass galaxies

The circumgalactic medium (CGM) is both the reservoir of gas that fuels star formation and galaxy growth, and the repository for the mass, energy, and metals expelled through stellar evolution and feedback. We present a controlled experiment using a suite of five cosmological magnetohydrodynamical simulations of a Milky Way-mass halo from the Auriga suite, where we vary the metal content and energy loading of galactic winds driven by stellar feedback. We chose these parameters in combination such that all five runs obtain very similar (within $\approx$10%) $z=0$ stellar masses and morphologies and thus produce similar amounts of metals throughout their lifetimes. Our simulations differ substantially in where these metals end up, spatially within the halo and across the different baryonic components (stars, gas disc, CGM, and the intergalactic medium). The metal content affects the cooling efficiency of the halo gas, regulating the CGM cool gas fraction and the accretion rates onto the galaxy. Between our two extreme models, we report an order of magnitude difference in the median gas metallicity at vertical heights $|z| \gtrsim 30$ kpc above the disc plane. Current constraints on the metallicity of the Milky Way halo gas from absorption-line measurements of intermediate- and high-velocity clouds at $|z| \lesssim 15$ kpc are broadly consistent with our simulations, though a direct comparison is limited by differences in the adopted methods. Our findings show that galaxies with nearly identical stellar content and morphology can differ substantially in their CGM, which offers a promising avenue for constraining feedback physics within galaxy formation models.
Read the full abstract

All authors Prachi Khatri, Freeke van de Voort, Rebekka Bieri, Rüdiger Pakmor, Robert J. J. Grand, Thomas A. Rintoul, Maria Werhahn, Rosie Y. Talbot

General Relativity and Quantum Cosmology

Gravitational Waves and Matter Constraints in Gravity from Entropy

We show that Gravity from Entropy (GfE) exhibits a tensor-propagation pathology on generic FLRW backgrounds. In vacuum, in the symmetry-inheriting single-multiplet sector, and for the algebraic perfect-fluid constitutive embedding, the strict-principal TT system has a nonzero real characteristic only when \(\dot H=K/a^2\). Away from this locus the temporal principal roots become complex, signaling a high-frequency hyperbolicity obstruction. Under the stated isotropy assumptions, a rank theorem further shows that one foundational multiplet cannot provide the isotropic bivector response needed to remove this problem.
Read the full abstract

Combinatorics

Non-attacking rook placements on crossword grids

We introduce the notion of a non-attacking rook placement on a crossword grid. A crossword grid is a collection of white squares (which comprise across and down words) and black squares (which separate the words), and a complete non-attacking rook placement on such a grid is a subset of white squares which intersects every across and every down word exactly once. We prove an upper bound on the number of rook placements that a general grid can admit. We then study sparse grids in which no two black squares share an edge and show that rook placements on certain sparse grids correspond bijectively to alternating sign matrices with prescribed $-1$ entries. Specializing further to permutation grids, we prove that every permutation grid admits at least one rook placement, and characterize the permutations whose grids admit exactly one placement in terms of the Robinson–Schensted correspondence. Throughout, we pose a variety of conjectures and open questions.
Read the full abstract

Computer Vision

Exemplar: Few-Shot Biomedical Segmentation That Fits in Seconds

Segmenting a new microscopy dataset usually means training a deep network on its annotations, and training again for every added image. Exemplar solves a segmentation readout in closed form and updates it exactly with each new mask in seconds, as accurately as networks trained for minutes. The updated readout is identical to a fit from all masks at once. Its linear readout combines frozen DINOv3 features, read at a magnification chosen from the masks, with classical filter responses at native resolution. Its regularisation and its pixel weighting are chosen exactly by leaving out one annotated image at a time, so none is set aside for validation. On ten biomedical datasets from cells to vessels and filaments, Exemplar is more accurate than every few-shot method that only applies a pretrained model, and on average than nnU-Net trained on the same masks. A fine-tuned Cellpose-SAM, the strongest baseline, is as accurate on average, but each of its fits takes a median of 25 minutes against 8.5 seconds, and each new mask means fine-tuning it again.
Read the full abstract

Machine Learning

LeanStream: A Speculate-and-Refine Streaming Framework for Efficient on-Device LLM Inference

On-device LLM inference is attractive for privacy and responsiveness, but remains challenging on mobile and embedded devices because model weights far exceed available DRAM. Prior systems exploit activation sparsity and offload weights to SSD or flash storage, but face a fundamental systems trade-off: accurate sparse execution decisions require the latest context, whereas efficient computation-I/O overlap requires early prediction. As a result, existing designs either serialize execution or incur redundant weight fetches, extra computation, and large cache overheads. We present LeanStream, a streaming speculate-and-refine framework for efficient on-device LLM inference. LeanStream progressively refines computation, loading, and cache-retention priorities using partial GPU results, enabling fine-grained overlap between GPU execution and storage I/O. We implement LeanStream on both mobile and embedded platforms. Compared with prior on-device LLM inference systems, LeanStream reduces memory usage by 4.8$\times$ to 7.5$\times$ at the best throughput achieved by prior work, while further improving token generation throughput by 1.6$\times$ to 2.1$\times$. Source code is available at https://github.com/RenyuanLIU/LeanStream.
Read the full abstract

All authors Renyuan Liu, Yuyang Leng, Kaiyan Liu, Yuzhou Zhong, Shaohan Hu, Chun-Fu Chen, Peijun Zhao, Heechul Yun, Shuochao Yao

Databases

Practical Threshold-based Tree Edit Distance Lower-Bounds

Threshold-based similarity search over tree-structured data using tree edit distance (TED) is computationally intensive. Given a query tree and a database of trees, the goal is to retrieve all trees within a predefined TED threshold $\tau$. Because exact TED computation is expensive, practical methods employ lower-bounds to prune dissimilar candidates before verification. Existing lower-bounds exhibit a fundamental trade-off: inexpensive statistical and structural bounds provide limited pruning power, whereas the more precise traversal-based string edit distance (SED) bound is expensive to compute using standard quadratic dynamic programming. Moreover, previous comparative studies do not cover recent structural filters or threshold-aware SED implementations, leaving their practical trade-offs unclear. In this article, we first provide a comprehensive experimental comparison of state-of-the-art TED lower-bounds in terms of pruning precision and computational cost. We then accelerate the SED lower-bound using Ukkonen's bounded string edit distance algorithm, substantially reducing its runtime without affecting its pruning power. Finally, we introduce the SED-struct threshold filter, which strengthens SED with axes-aware constraints capturing structural relationships among tree nodes. Experiments on synthetic and real-world datasets show that SED-struct consistently achieves the highest filtering precision while retaining practical filtering costs. The results suggest that SED-struct is particularly beneficial for heterogeneous tree collections in which the standard SED lower-bound achieves relatively low precision.
Read the full abstract

Computer Vision

Position: Unlabeled IS NOT Equal to No Human Supervision in Visual Learning

This position paper argues that the absence of labels does not imply the absence of human supervision in visual learning, and urges the research community to identify sources of supervision more explicitly. Many recent methods in computer vision build upon representations learned from large-scale unlabeled data, and are therefore grouped under the same umbrella term “unsupervised.” However, different data curation schemes and training objectives embed substantially different human priors on which models rely, and we argue that one “unsupervised” umbrella term is no longer capturing these distinctions. This ambiguity makes it harder to compare unsupervised learning research conducted under different assumptions, coinciding with a sharp decline in papers titled with “unsupervised” in flagship computer vision conferences since 2021, despite continued growth of the field. While we fully embrace pre-training as a strong foundation for modern computer vision, we advocate for a community-level effort toward greater conceptual clarity: authors are encouraged to disclose priors in data selection and learning objectives, and to specify which components of a learning pipeline depend on which assumptions. Standardized disclosure practices can improve academic communication, ensure fairer comparisons, and preserve methodological diversity in unsupervised learning.
Read the full abstract

Mesoscale and Nanoscale Physics

Deterministic nanofabrication for engineering nanowire quantum dot devices

Semiconductor quantum dots (QDs) are a leading platform for realising bright, wavelength-tunable sources of single and entangled photon pairs for photonic quantum technologies. Site-selected nanowire quantum dots (NWQDs) are a promising platform for fabricating such photonic devices in a scalable manner. However, implementing additional structures around the photonic nanowire while maintaining its vertical growth geometry has remained a challenge. In this work, we develop a deterministic pick-and-place technique to conduct a vertical-to-vertical transfer of NWQDs from the growth substrate to arbitrary templates. Using this transfer technique, we enhance the photon extraction efficiency to 75% by implementing a bottom gold mirror and tune the emission wavelength by 3.6 GHz via implementing electrostatic gates around the QD. Importantly, we measure low-multiphoton probability (g^(2)(0) = 0.002) and high indistinguishability (>80% for +/-100 ps) of the QD emission after the transfer process, yielding high-quality devices. These results demonstrate the repeatability and versatility of the developed transfer technique, which is an enabling step towards scalable single and entangled photon sources.
Read the full abstract

All authors Tarun Patel, Matteo Pennacchietti, Greg Holloway, Stephen R. Harrigan, Sayan Gangopadhyay, Anthony Drouin, Megha Jain, Dan Dalacu, Philip J. Poole, Sasan Vosoogh-Grayli, Michael E. Reimer

Strongly Correlated Electrons

Spin-current rectifier and linear magnon Edelstein effect in $p$-wave antialtermagnets

We show that $p$-wave magnetic insulators efficiently generate magnetization via a linear thermal magnon Edelstein effect. Furthermore, they can generate perfectly nonreciprocal spin currents through a purely even-order nonlinear response. This makes them promising candidates for spin-current rectification. More generally, we find that odd-parity-wave magnets with coplanar ground states host magnons with zero magnetic moment in the plane. Instead the magnons carry a colinear, out-of-plane magnetic moment, placing coplanar odd-parity-wave magnets within the class of antialtermagnets. We establish these general results using symmetry arguments and demonstrate them explicitly for a two-dimensional model with a coplanar ground state containing the minimal number of sublattices, inspired by the ground state of CeNiAsO. Our conclusions hold both in the absence and in the presence of spin-orbit coupling.
Read the full abstract

Systems and Control

Finite-Sample Limits of Entropy-Based Structure Identification in Discretized Nonlinear Systems

Discretization fundamentally limits structure identification in stochastic systems. When system stochasticity exceeds the discretization resolution, entropy-based methods lose their ability to distinguish which input drives the output. We study this in Fuzzy Inductive Reasoning (FIR), a nonparametric framework for learning dynamical systems from discretized measurements, where the choice of input variables determines both predictive accuracy and the interpretability of the learned input–output relationships. Entropy-based selection targets explainability, i.e., identifying which variables causally drive the output, while mean-squared-error-based selection targets prediction. We introduce a resolution-stochasticity ratio that governs when entropy-based selection is reliable. Three results follow. First, entropy-based selection is consistent below this threshold but loses discriminative power above it, regardless of sample size. Second, using the entropy-selected variables for prediction instead of the MSE-selected ones incurs a closed-form excess prediction risk that grows with input complexity and shrinks with sample size. Third, reliable identification of the causally relevant inputs requires data that scales with the number of input combinations and inversely with the strength of the entropy signal. The theory is validated on a two-state Markov model and demonstrated on a distribution grid reliability dataset analyzing the impact of infrastructure investment, where the goal is to explain which investments drive reliability improvements rather than merely predict outcomes.
Read the full abstract

Quantum Physics

Optimized Matrix-Product State Simulations of Quantum Error Correction Circuits

Simulating quantum error correction (QEC) circuits including non-Clifford gates at scale is important to accelerate progress toward fault-tolerant quantum computing. Here we demonstrate that matrix product state (MPS) techniques can handle many QEC circuits exactly and without restriction on gate types. Crucially, we find that MPS efficiency depends sensitively on implementation choices, and we introduce a series of targeted optimizations that reduce bond dimensions and simulation time by several orders of magnitude compared to naive approaches. We illustrate this with examples including: (a) a rotated surface code quantum memory up to distance 11, (b) logical Bell-state preparation up to distance 9, (c) a 15-to-1 magic-state distillation circuit including hundreds of QEC rounds that we optimize to be simulated with only 11 logical qubits (187 physical qubits) and a maximal bond dimension of 64 in under 40 seconds, and (d) a narrow, deep random circuit that scales linearly with the number of T gates. These results demonstrate the importance of circuit-level optimizations and position MPS as a valuable complement to near-Clifford simulators for QEC circuits.
Read the full abstract

All authors Asier Piñeiro Orioli, Chen Zhao, Guido Masella, Tommaso Macrì, Hengyun Zhou, Shannon Whitlock