Peer Revenue Equity in science

Papers

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

All of arXiv

Wednesday 30 September

Neural and Evolutionary Computing

Large Language Model-Guided Evolutionary Discovery of Native Neural Architectures for Spiking Sequence Modeling

Ruoyu Zhao, Jiaqi Wu, Chenyu Zhu et al. (+1 more)

Spiking neural networks (SNNs) offer low-energy sequence modeling through sparse, event-driven computation. However, interactions among spike encoding, neuronal dynamics, and information propagation complicate architecture design. Existing SNN sequence models often adapt artificial neural network (ANN) architectures designed for real-valued activations, potentially underusing spike-based communication and temporal state updates, motivating automated discovery of native SNN architectures. Most evolutionary neural architecture search (ENAS) methods operate within predefined configuration spaces, limiting discovery to mechanisms expressible within those spaces. We introduce OpenArchEvo, which uses large language models (LLMs) to evolve executable architecture code in an open program space under spiking-projection constraints. In this space, code differences need not reflect architectural novelty, while direct performance evaluation requires costly training. We construct a three-view representation spanning code, design rationale, and a behavioral fingerprint to support novelty estimation and performance prediction. The search treats predicted performance and estimated novelty as two objectives, using surrogate predictions to select candidates for expensive training evaluations. With an estimated candidate-training cost of 132 V100 GPU-days, the search uncovers multiple native SNN architectures, exemplified by three designs featuring mechanisms such as spike-activity-dependent control of state updates and residual pathways. The discovered NeuroGate surpasses the ANN DeltaNet on WikiText-103, and the discovered architectures reduce estimated architecture-level arithmetic energy by up to 50.6x (LoopMem) relative to a common dense Transformer (ANN) baseline. Code and all discovered architectures: https://github.com/Optima-CityU/OpenArchEvo
Read the full abstract

All authors Ruoyu Zhao, Jiaqi Wu, Chenyu Zhu, Zhichao Lu

Algebraic Geometry

Structure of higher-genus open-closed Gromov–Witten theory of $\mathcal{O}_{\mathbb{P}^{1}}(p-1)\oplus\mathcal{O}_{\mathbb{P}^{1}}(-p-1)$

Shuai Guo, Jingyi Xu and Qingsheng Zhang

We study the closed and open Gromov–Witten potentials of the toric Calabi–Yau threefold $$ X_p=\operatorname{Tot}\bigl(\mathcal{O}_{\mathbb{P}^{1}}(p-1)\oplus\mathcal{O}_{\mathbb{P}^{1}}(-p-1)\bigr),\qquad p\geq 2. $$ We prove closed and open mirror symmetry under a nonvanishing condition on the torus weights, relating these potentials to topological recursion on the mirror curves. We establish polynomial structures for both the higher-genus closed potentials and the stable open potentials. We also establish double-scaling limits for topological recursion on the mirror curves. In particular, our results for the closed potentials prove the higher-genus ansatz and the double-scaling conjecture of Caporaso–Griguolo–Mariño–Pasquetti–Seminara.
Read the full abstract

Atomic Physics

Nonrelativistic expansions of Dirac-Coulomb energy and first order relativistic Breit potential corrections for light Atomic and molecular systems

Wanping Zhou, Sanjiang Yang and Chen Tong

The nonrelativistic expansion method is applied to the Dirac-Coulomb energy and first order relativistic Breit potential corrections for light atomic and molecular systems to derive high-order effective energy correction operators for few-body systems. It is found that effective operators at the $m\alpha^8$ order contain incompletely separable divergences. These divergences must be canceled jointly with divergences arising from second- and third-order perturbations of lower-order energy correction operators, and such cancellation can be accomplished automatically by the nonrelativistic numerical expansion approach. Using Gaussian basis sets, we numerically compute contributions to the Dirac-Coulomb energy and relativistic Breit potential corrections up to order $m\alpha^8$ for the ground states of the hydrogen atom, hydrogen molecular ion, helium atom, and hydrogen molecule. These results confirm the reliability of the present method and enable its extension to more complex systems.
Read the full abstract

Quantum Physics

A General Theory of Multi-Resource Theories Involving Finite-Group Asymmetry

Yosuke Mitsuhashi and Hiroyasu Tajima

Resource theories provide a universal framework for quantifying properties of quantum states, such as entanglement, magic, athermality, and asymmetry, as resources under various constraints encountered in quantum science. Since realistic settings typically impose several constraints at once, it is important to study multi-resource theories that combine them. Among these, combinations with asymmetry are especially natural, as symmetry is fundamental and ubiquitous in physics. Here we develop a general theory that treats combinations of finite-group asymmetry with a broad range of resource theories in a unified manner. Our key observation is that the effect of combining the two constraints is governed by the accessibility of symmetry information under the underlying resource theory: whether the information needed for the conversion can be extracted from the input and used to control free operations. If this can be done with asymptotically vanishing error, we show that the conversion rate of the combined theory equals the smaller of the rates of the two constituent theories, so combining them incurs no additional cost. This condition holds for LOCC on arbitrary mixed multipartite states, magic-state conversion of multiqubit systems under Clifford symmetries, and Gibbs-preserving operations. Thermal operations, being time-translation covariant, can access only limited symmetry information, giving an additional compatibility condition between the two symmetries. We also analyze symmetry imposed on elementary operations rather than only on the resulting channel. We establish analogous results for LOCC and thermal operations under suitable assumptions, and for magic when the symmetry is represented by Pauli operators. Finally, we show that finite-group symmetry does not reduce the asymptotic ergotropy per copy and hence introduces no new completely passive states.
Read the full abstract

Quantum Physics

Quantum oblique eigenprojection

Alexander M. Dalzell and Yuan Su

Every square matrix decomposes its underlying Hilbert space into generalized, nonorthogonal eigensubspaces. We show that a quantum computer can perform such an oblique eigenprojection $\Pi$ given block encoding access to the input matrix. Our approach has a query complexity nearly linear in the inverse gap and a normalization factor close to $\lVert\Pi\rVert$ under a spectral-set condition on the input. This covers common assumptions on the numerical range or diagonalizability and matches known results for orthogonal eigenprojections. We achieve this with a two-sided block preconditioning that uses a discrete Fourier transform of the matrix resolvent. We describe applications to: (i) preparing eigenstates of matrices with complex eigenvalues, extending the quantum eigenvalue transformation algorithm of Low and Su beyond real spectra; (ii) solving continuous-time algebraic Riccati equations, cubically speeding up a prior solver of Rodenas-Ruiz, Zhao, and Lee; and (iii) solving ordinary Sylvester equations, quadratically improving a direct augmented method of Wang and Liu. Our result suggests a promising route to applying nonanalytic matrix functions on quantum computers.
Read the full abstract

Computer Vision

ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents

Yong Du, Tongbo Chen, Zhengxi Lu et al. (+5 more)

Online training enables computer-use agents (CUAs) to improve through interaction with executable environments. However, existing methods primarily rely on sparse outcome rewards, which provide no supervision for intermediate actions. On-policy self-distillation (OPSD) offers token-level learning signals through privileged rescoring, but directly applying it to CUA online training presents two challenges: fixed guidance may become misaligned with the student's current state, and guidance-induced probability shifts may conflict with step-level correctness. We introduce ComputerSD, an online self-distillation method for CUAs that converts real-time feedback from executed GUI transitions into guidance for policy learning. A fine-tuned GUI analyzer produces guidance and a step-level value score after each action; the guidance provides privileged context, while the score regulates the resulting OPSD signals. ComputerSD jointly optimizes token-level OPSD and trajectory-level GRPO in a fully asynchronous training framework. On OSWorld-Verified, ComputerSD outperforms outcome-only GRPO by 1.9 and 4.1 percentage points on the general-purpose Qwen3-VL-8B-Thinking and specialized EvoCUA-8B backbones, respectively. Evaluation in out-of-distribution settings further supports the generalizability of ComputerSD. These results demonstrate the effectiveness of learning from real-time feedback through online self-distillation for CUAs.
Read the full abstract

All authors Yong Du, Tongbo Chen, Zhengxi Lu, Yizhou Liu, Bofan Chen, Tao Jiang, Wenhao Xu, Yongliang Shen

Information Theory

From Random Quantum Codes to Explicit qLDPC Codes via Local Properties

Fernando Granha Jeronimo, Xiaojuan Ma and Nikhil Shagrithaya

Constructing explicit codes matching the parameters of random codes has been a central and largely elusive question in coding theory. The quantum setting is even more challenging since it is highly desirable that the quantum code be an LDPC code. Local coordinate-wise linear (LCL) [Levi, Mosheiff, and Shagrithaya, FOCS 2025] witnesses provide a unifying language for many coding-theoretic properties, from distance to list decoding and list recovery. In particular, it provides a framework to study properties of random linear codes, which achieve optimal parameters for many properties of linear codes. For CSS quantum codes, however, a local witness has two distinct ranks: its physical rank before quotienting by stabilizers and its logical rank after quotienting. We develop a quantum version of the LCL framework for nested spaces $S \subseteq C$, in which local constraints are imposed on physical representatives while independence is measured in the logical quotient $C/S$. The resulting theory gives a threshold theorem for random CSS codes, and as a consequence shows that the per-sector rate threshold is equal to the classical rate threshold. We also define a quantum analogue of subspace design [Guruswami and Xing, STOC 2013] and show that they can be described in a natural manner within the quantum-LCL framework. Finally, we give explicit constructions for arbitrary folded quantum-LCL properties, in a manner similar to the LCL derandomization of [Jeronimo and Shagrithaya, STOC 2026]. As a consequence, we obtain the first explicit constructions of quantum list-decodable codes and list-recoverable codes that have optimal list sizes, in addition to explicit quantum subspace design codes. We note that all our explicit constructions are qLDPC codes, an important property for quantum error-correcting codes.
Read the full abstract

Logic

PAC fields and difference fields with generic automorphisms

Özlem Beyarslan and Piotr Kowalski

We study the existence of model companions after adjoining an automorphism to field theories in two complementary settings. For bounded PAC fields of characteristic zero containing all roots of unity and admitting a cyclic Galois quotient of prime order, a suitable choice of countably many constants gives a model-complete field theory which has no model companion after adding an automorphism fixing the named constants. This applies, in particular, when the absolute Galois group is free profinite or free pro-$p$ of finite positive rank. Secondly, for every completion of $\ACFA$ in every characteristic, adjoining a second commuting automorphism has no model companion. The two explicit obstructions use a common ordered product of inertia automorphisms along an infinite orbit, combined respectively with an equivariant PAC completion and an algebraically closed difference-field embedding.
Read the full abstract

Quantum Physics

An (almost) efficient classical algorithm for sampling from typical Gibbs states

Saleh Naghdi and Eric R. Anschuetz

Gibbs state preparation has emerged as a potential avenue for quantum advantage in the study of physical systems. Although there is known advantage for computing local properties of worst-case Gibbs states, recent results suggest a different picture for average-case local systems: at temperatures where their Gibbs states can be efficiently prepared quantumly, quasi-polynomial time classical algorithms can also compute local expectation values. Yet this leaves open the possibility of an advantage under a stronger notion of simulation: sampling. Gibbs sampling underlies Boltzmann machine-based quantum learning algorithms and quantum supremacy proposals, but its classical average-case complexity for physically relevant models remains unresolved. Here, we give strong evidence against an exponential quantum advantage in Gibbs sampling for an ensemble of local systems known as the quantum $p$-spin model. We describe a classical algorithm for sampling from the Gibbs state to any polynomially small error in total variation distance. Our algorithm is based on a meta-algorithm known as algorithmic stochastic localization (ASL). Our main technical result is a generalized convergence theorem for ASL which reduces approximate sampling from any distribution on the hypercube to estimating the mean of a related "tilted distribution" to sufficiently low additive error. We then extend known quasi-polynomial time algorithms for computing local expectation values of a related disordered model to compute these "tilted means" for the quantum $p$-spin model. Finally, we give a non-rigorous physics argument that this mean-estimation algorithm works down to the spin glass transition temperature. As efficient quantum algorithms also fail at this transition, this suggests they cannot Gibbs sample the quantum $p$-spin model at temperatures below that for which classical algorithms are also (almost) efficient.
Read the full abstract

Data Structures and Algorithms

Dynamic Time Warping in the Low-Distance Regime

Itai Boneh, Shay Golan and Tomasz Kociumaka

Dynamic Time Warping (DTW) is a classical similarity measure for strings and time series that allows local stretching. Given non-empty strings $S,T$ over an alphabet $\Sigma$ and a cost function $\delta:\Sigma^2\to\mathbb{R}_{\ge0}$, $DTW_\delta(S,T)$ is the minimum total cost of equal-length expansions of $S$ and $T$ obtained by duplicating characters. For strings of length at most $n$, DTW is computable in $O(n^2)$ time, and this is conditionally optimal under the Orthogonal Vectors Hypothesis (OVH). We study the low-distance regime, where an integer $k$ upper-bounds $DTW_\delta(S,T)$, assuming $\delta(a,a)=0$ and $\delta(a,b)\ge1$ for $a\ne b$. For several classical similarity measures, this regime admits $O(n+\operatorname{poly}(k))$ algorithms, whereas for DTW with metric costs the best known bound is $O(nk)$. We show that this dependence is essentially optimal: assuming OVH, computing DTW requires $n^{1-o(1)}k$ time even for the discrete mismatch-cost function, which assigns cost $1$ to every mismatch. The lower bound applies to the whole spectrum of thresholds $k$ between constant and linear in $n$. Our reduction from Orthogonal Vectors encodes vector coordinates in the lengths of equal-character runs. The resulting instances are very structured: collapsing runs to single characters reveals long substrings with short periods. We complement the lower bound with a $\tilde O(n+\operatorname{poly}(k))$-time algorithm whenever, after collapsing runs in the inputs, every substring with period $O(k)$ has length $\operatorname{poly}(k)$. Finally, we extend this lower bound to DTW pattern matching, which asks whether any non-empty substring of a length-$n$ text has DTW distance at most $k$ from a length-$m$ pattern. We prove that the classic $O(nm)$-time dynamic-programming algorithm is near-optimal under OVH, even when $k=O(\log n)$.
Read the full abstract

Quantum Physics

Optimal Purity Estimation with Incoherent Measurements

Junseo Lee and Chirag Wadhwa

In this work, we consider the fundamental task of estimating the purity of an unknown state to within $\textit{multiplicative}$ error $\varepsilon$. When one can perform general collective measurements, $\Theta\left(\frac{\sqrt{d}}{\varepsilon^2} + \frac{d}{\varepsilon}\right)$ copies are known to be necessary and sufficient for this problem [AISW20]. However, implementing collective measurements on such a large number of copies can be experimentally demanding, and we thus aim to characterize the copy complexity of this problem with incoherent measurements. In this setting, the only non-trivial result is a non-adaptive algorithm that uses $O\left( \frac{d}{\varepsilon^2} + \frac{d^2}{\varepsilon} \right)$ copies [PTTW26], which is polynomially larger than the best-known lower bound. Our first result is a new algorithm performing non-adaptive incoherent measurements that succeeds using $O\left( \frac{d}{\varepsilon^2} + \frac{d^{3/2}}{\varepsilon} \right)$ copies, improving on the latter term in the copy complexity. Moreover, we show that for any algorithm restricted to non-adaptive measurements, the above copy complexity is optimal. We also develop a new adaptive estimator for the purity of a state that improves on the above complexity in the high-precision regime, i.e., for $\varepsilon = o(1/d)$. We also show that the non-adaptive and adaptive estimators together yield the optimal complexity for incoherent purity estimation; in particular, we show that the copy complexity of this problem is $$ \Theta\left(\min\left\{ \frac{d}{\varepsilon^2} + \frac{d^{3/2}}{\varepsilon}, \frac{d^2}{\varepsilon} + \frac{\sqrt{d}}{\varepsilon^2}\right\} \right). $$
Read the full abstract

Operating Systems

Herschel: Continuous Optimization of Production LLM Inference through On-Demand Profiling

Luping Wang, Weigao Chen, Yifei Wu et al. (+11 more)

Model-as-a-service platforms call for continuous optimization as complex serving conditions expose inefficiencies missed before deployment. Detailed always-on profiling can incur substantial overhead, while lightweight collection omits information needed for diagnosis. We present Herschel, a continuous optimization system for production large language model (LLM) inference. Our key insight is that adaptive, on-demand profiling can provide rich full-stack evidence without continuous collection. Herschel safely attaches to and detaches from selected running processes without engine changes or restarts, and adapts coverage as investigations reveal missing evidence. Herschel reconstructs operator executions and cross-process dependencies to identify inefficiency mechanisms and suggest solutions using applicable reference fixes. AI agents implement and test engine and kernel changes under controlled conditions that preserve the triggering workload and dependencies, with expert review before deployment. Controlled tests show active-collection overhead below 0.5% for time to first token and 7% for time per output token. Bounded windows, typically 30 s, avoid the continuous cost of always-on tracing. Over six months, Herschel collected approximately 17,000 traces across over 120 model variants and more than 10 accelerator types, identifying inefficiency patterns in 23% of the traces. Representative findings guide widely deployed optimizations, including restructured synchronization, removal of unused computation, and improved operator implementations.
Read the full abstract

All authors Luping Wang, Weigao Chen, Yifei Wu, Yonghe Zhang, Rui Zhang, Wenchao Wu, Jiyu Luo, Haoran Geng, Xin Yang, Chen Cao, Yuemin Wu, Cheng Huang, Guodong Yang, Liping Zhang

Operator Algebras

Connes-embeddability and twisted group von Neumann algebras

Soham Chakraborty and Felipe Flores

We call a countable discrete group cocycle-hyperlinear if all its twisted group von Neumann algebras are Connes-embeddable. We show that the class of cocycle-hyperlinear groups includes extensions of treeable groups by amenable groups and central extensions of SL(n,Z), GL(n,Z) and PSL(n,Z) for n greater than 1. We also show that this class is closed under taking direct products, free products, amalgamated free products with amenable amalgams, HNN extensions by amenable groups, as well as certain quotients, extensions, and overgroups.
Read the full abstract

Robotics

STARS: From Spatiotemporal Dynamics to Social Representations in Human-Robot Interaction

Nathan Tsoi, Michael J. Munje, Tejas Oberoi et al. (+5 more)

Robot navigation in dynamic, human-centered environments requires socially-compliant decisions grounded in robust scene understanding. Recent Vision-Language Models (VLMs) exhibit promising capabilities such as object recognition, common-sense reasoning, and contextual understanding, capabilities that align with the nuanced requirements of social robot navigation. However, it remains unclear whether VLMs can accurately understand complex social navigation scenes (e.g., inferring the spatial-temporal relations among agents and human intentions), which is essential for safe and socially compliant robot navigation. While some recent works have explored the use of VLMs in social robot navigation, no existing work systematically evaluates their ability to meet these necessary conditions. In this paper, we introduce the Social Navigation Scene Understanding Benchmark (SocialNav-SUB), a Visual Question Answering (VQA) dataset and benchmark designed to evaluate VLMs for scene understanding in real-world social robot navigation scenarios. SocialNav-SUB provides a unified framework for evaluating VLMs against human and rule-based baselines across VQA tasks requiring spatial, spatiotemporal, and social reasoning in social robot navigation. Through experiments with state-of-the-art VLMs, we find that while the best-performing VLM achieves an encouraging probability of agreeing with human answers, it still underperforms simpler rule-based approach and human consensus baselines, indicating critical gaps in social scene understanding of current VLMs. Our benchmark sets the stage for further research on foundation models for social robot navigation, offering a framework to explore how VLMs can be tailored to meet real-world social robot navigation needs. An overview of this paper along with the code and data can be found at https://larg.github.io/stars.
Read the full abstract

All authors Nathan Tsoi, Michael J. Munje, Tejas Oberoi, Rishab Maheshwari, Pengen Zheng, Tanush Chauhan, Peter Stone, Joydeep Biswas

Computer Vision

StreamRig: Exploiting Intra-Rig Geometry for Streaming Multi-Camera Odometry

Yufei Wei, Shuhao Ye, Qi Wang et al. (+4 more)

Mobile robots and vehicles carry synchronized multi-camera rigs, yet many streaming 3D foundation models are designed for monocular input, leaving efficient use of rig geometry a challenge. We present StreamRig, a freeze-and-stream framework that builds causal streaming odometry for calibrated rigs on a frozen multi-view 3D foundation model. The frozen front-end jointly perceives the synchronized views using rig calibration. A Rig-Resampler compresses their features, a CausalBridge applies causal attention with a key-value cache, and a lightweight head regresses rig poses. A periodic re-anchoring protocol supports stable pose estimation over long sequences. Only these modules are trained, 74.6M parameters in total, with relative poses as the sole supervision. Our two-stage training strategy combines group relocalization pretraining with causal rig training to transfer the geometric priors of the frozen front-end and the alignment ability of the pretrained modules to streaming odometry. We evaluate on NCLT, TartanGround, KITTI-360, and our self-collected humanoid-robot dataset ZJH, where training uses only simulation and real-world evaluation is zero-shot. Across all four datasets, StreamRig achieves lower translation and rotation drift than the evaluated non-oracle monocular streaming and rig-aware offline models, while maintaining low inference cost. Ablations and controlled camera-count experiments identify the sources of these gains. We further examine how longer training windows affect inference over longer horizons. Code has been released at https://github.com/WeiYuFei0217/StreamRig.
Read the full abstract

All authors Yufei Wei, Shuhao Ye, Qi Wang, Xin Zheng, Qing Huang, Rong Xiong, Yue Wang

General Relativity and Quantum Cosmology

Flat limit of artificial cosmology for scalar waves in Schwarzschild-de Sitter

An{\i}l Zenginoğlu and Govind Arun Kumar

How do cosmological waveforms approximate waveforms from isolated systems for a small cosmological constant? This question is motivated by the observed small value of the cosmological constant, and also by Misner's idea to regulate the formally singular terms introduced into the Einstein equations by conformal compactification. We investigate this question quantitatively using scalar waves on Schwarzschild–de Sitter backgrounds. A suitable flat limit requires the foliation to remain regular at the future conformal boundary. We compare bridge foliations extending from the black hole horizon to the cosmological horizon with a regular asymptotically flat limit. Comparisons of cosmological and asymptotically flat waveforms on geometrically matched data show that the flat limit is numerically well behaved and recovers the essential features of the asymptotically flat waveform. We extend Misner's artificial cosmology by introducing a cosmological source that leaves the near-zone geometry unchanged and show that it improves agreement of the ringdown waveform but fails to capture the intermediate polynomial tail behavior accurately.
Read the full abstract

Statistics Theory

Riemannian Regression

Oldemar Rodríguez

Classical linear regression assumes that the relevant geometry of the predictor space is Euclidean and that all centered observations contribute to the least-squares fit in the same geometric scale. This paper proposes Riemannian Regression, a regression framework in which the usual vector differences are replaced by locally weighted differences induced by a data-dependent similarity structure. We introduce a generalized framework, termed Riemannian Regression, extending classic regression to any data endowed with a local distance structure. By equipping data tables with local metrics, we adapt regression model to incorporate manifold geometry. Given a similarity matrix $S=(S_{ij})$, obtained from UMAP, ISOMAP, or DBSCAN \cite{mcinnes,isomap,dbscan}, we define the dissimilarity coefficient $\rho_{ij}=1-S_{ij}$ and the induced subtraction $ x_i\ominus x_j=\rho_{ij}(x_i-x_j). $ A Riemannian center $g=x_\lambda$ is selected as a discrete Fréchet mean, and regression is performed on the Riemannian-centered variables $X_R=W X_{c,\lambda}$ and $y_R=W y_{c,\lambda}$, where $W=\operatorname{diag}(\rho_{1\lambda},\ldots,\rho_{n\lambda})$. The resulting estimator has the weighted least-squares form $ \widehat\beta_R=(X_{c,\lambda}^{t}W^2X_{c,\lambda})^{-1}X_{c,\lambda}^{t}W^2y_{c,\lambda}. $ The proposed approach preserves the linear form of the regression model while changing the geometry of the fit. The paper develops three ways to construct the local metric: UMAP-based fuzzy similarities, ISOMAP-based normalized geodesic distances, and DBSCAN-based density similarities. Simulated examples and the Abalone data set illustrate how Riemannian Regression can reduce the influence of locally anomalous observations and adapt to regions with different local densities.
Read the full abstract

Information Retrieval

Decision-Oriented Recommendation Reranking: An Empirical Study of Its Promise and Limitations

Hanjia Lyu and Yinglong Xia

Large language models (LLMs) have shown promise for recommendation reranking, but their use introduces an important trade-off between recommendation quality and serving efficiency. We investigate whether decision-oriented systems provide a useful alternative when the reranking task is fundamentally a structured choice among predefined candidate items. Specifically, we conduct a controlled empirical study of two decision-oriented systems, TypeSafe Jev and OpenAI Decisions, for recommendation reranking. Across three Amazon Reviews domains and four candidate-set sizes, we compare these systems against recommendation-specific models and pointwise and listwise LLM-based rerankers, systematically evaluating recommendation effectiveness, candidate-order robustness, and efficiency and scalability. Our results reveal that Jev frequently achieves a favorable quality–latency trade-off, offering competitive recommendation effectiveness at substantially lower observed latency than stronger LLM-based alternatives. Overall, our findings suggest that specialization for decision-making does not inherently guarantee superior recommendation quality or robustness. Instead, the practical value of decision-oriented inference depends on implementation-specific characteristics and application requirements. These findings motivate further investigation of decision-oriented systems for recommendation and other ranking tasks with structured output spaces.
Read the full abstract

Geometric Topology

Hierarchically Hyperbolic Surface-by-surface Groups

Spencer Dowdall, Matthew Gentry Durham, Chris Leininger et al. (+2 more)

We give many examples of surface bundles over surfaces whose fundamental groups are hierarchically hyperbolic by showing that extensions of surface groups constructed by the third author and Reid (LR surface groups) are hierarchically hyperbolic. This is facilitated by a new combination theorem for bundles of hyperbolic graphs over a hyperbolic base. This result differs from previous combination theorems for graph bundles by being applicable to bundles where the fibers are not properly embedded. We apply our combination theorem to establish the hyperbolicity of certain graph bundles that arise naturally from extensions of parabolically geometrically finite (PGF) subgroups of mapping class groups. These subgroups include LR surface groups, finitely generated Veech groups, free products of multi-twist groups, and many other examples created by a theorem of Udall. When the PGF groups have cyclic peripherals, hyperbolicity of our graph bundle is the key milestone towards showing that the extension groups are hierarchically hyperbolic.
Read the full abstract

All authors Spencer Dowdall, Matthew Gentry Durham, Chris Leininger, Jacob Russell, Alessandro Sisto

Instrumentation and Methods for Astrophysics

Two bits about lossy compression: On the limits of compression in cosmology

Hurum Maksora Tohfa and Matthew McQuinn

Astronomy is in an era of enormous sky surveys and of the large simulation suites needed to interpret them, both costly to store and slow to share. Simulation outputs are stored as 32-bit floats, yet numerical noise and astrophysical uncertainties make better than percent-level pixel accuracy unnecessary. Because cosmological fields are statistically homogeneous with nearly Gaussian mode amplitudes, classic rate-distortion results apply directly, and non-Gaussian structure permits further compression. We use the scientific compression package SZ3, and a neural compressor that fixes the quantization and learns the probability of each quantization bin with an autoregressive transformer. SZ3 beats the Gaussian approach only where the field is smooth on the pixel scale or its values pile up at a single point, while the neural approach matches or beats SZ3 on every field we consider and is 9 to 27% below the Gaussian-optimal coder at fixed distortion, with the largest margin where the field is most non-Gaussian. Because the quantizer, not the network, sets the error, a poorly trained model can waste bits but never cost accuracy. A model trained only on weak lensing convergence maps transfers to $N$-body density fields without retraining, suggesting it has learned generic properties of cosmological structure rather than features of one dataset. Eulerian grids need only 1-4 bits per pixel (bpp) at percent-level accuracy, and particle displacements and velocities 5-6 at the precisions they require. The approach carries over to observational data: on Rubin Observatory Data Preview 1 coadds, with the quantization step set to a quarter of the background noise, the fine-tuned transformer needs 4.0 bpp, 27% fewer than the Gaussian coder.
Read the full abstract

Differential Geometry

Nakai-Moishezon type criteria for Hessian type equations on compact Kähler manifolds

Jixiang Fu, Xuan Li, Dekai Zhang et al. (+1 more)

We establish numerical criteria for polynomial positivity of real $(1,1)$-classes on compact Kähler manifolds for strictly right-Noetherian polynomials. As applications, we obtain existence and uniqueness for the generalized Monge-Ampère equations with a smooth zeroth coefficient, the complex Hessian and Hessian quotient equations, and the critical LYZ equation. We also characterize the cone of $k$-positive classes as a connected component of a numerical positivity cone.
Read the full abstract

All authors Jixiang Fu, Xuan Li, Dekai Zhang, Ziyi Zhang

High Energy Physics (Experiment)

First double-differential cross-section measurements in proton multiplicity and kinematics for mesonless $\nu_{\mu}$ charged-current interactions on argon using the MicroBooNE detector

MicroBooNE collaboration: P. Abratenko, D. Andrade Aldana, J. Asaadi et al. (+186 more)

We present the first double-differential cross-section measurements of muon neutrino charged-current interactions on argon with no mesons and one or more protons ($N_{p} \geq 1$) in the final state as a function of proton multiplicity and kinematics. The analysis uses the full Booster Neutrino Beam dataset collected by the MicroBooNE detector at Fermi National Accelerator Laboratory, corresponding to a total of $1.30 \times 10^{21}$ protons on target. We measure cross sections as a function of the proton multiplicity, the momenta and muon-proton opening angles of the leading and subleading protons up to and including the fourth-leading proton, as well as the total available kinetic energy of the hadronic system. The treatment of the statistical and systematic uncertainties includes correlations between the different extracted distributions. The measured cross sections are compared to predictions from several event generators using different primary and final-state interaction models. The results reveal strong sensitivity to the modeling of final-state interactions, and none of the generators tested provides a fully adequate description of the data. Significant disagreement between data and predictions is observed in the proton multiplicity distributions, particularly at backward opening angles and for subleading proton kinematics.
Read the full abstract

All authors MicroBooNE collaboration: P. Abratenko, D. Andrade Aldana, J. Asaadi, A. Ashkenazi, S. Balasubramanian, B. Baller, A. Barnard, G. Barr, D. Barrow, J. Barrow, V. Basque, J. Bateman, B. Behera, O. Benevides Rodrigues, S. Berkman, A. Bhat, M. Bhattacharya, V. Bhelande, A. Binau, M. Bishai, A. Blake, B. Bogart, T. Bolton, M.B. Brunetti, L. Camilleri, D. Caratelli, F. Cavanna, G. Cerati, A. Chappell, Y. Chen, J.M. Conrad, M. Convery, L. Cooper-Troendle, J.I. Crespo-Anadon, R. Cross, M. Del Tutto, S.R. Dennis, P. Detje, R. Diurba, Z. Djurcic, K. Duffy, S. Dytman, B. Eberly, P. Englezos, A. Ereditato, J.J. Evans, C. Fang, B.T. Fleming, W. Foreman, D. Franco, A.P. Furmanski, F. Gao, D. Garcia-Gamez, S. Gardiner, G. Ge, S. Gollapinni, E. Gramellini, P. Green, H. Greenlee, L. Gu, W. Gu, R. Guenette, L. Hagaman, M. D. Handley, O. Hen, A. Hergenhan, M. Harrison, S. Hawkins, C. Hilgenberg, G.A. Horton-Smith, A. Hussain, B. Irwin, M.S. Ismail, C. James, X. Ji, J.H. Jo, A. Johnson, R.A. Johnson, D. Kalra, G. Karagiorgi, W. Ketchum, A. Kelly, M. Kirby, T. Kobilarcik, K. Kumar, N. Lane, J.-Y. Li, Y. Li, K. Lin, B.R. Littlejohn, L. Liu, S. Liu, W.C. Louis, X. Luo, T. Mahmud, N. Majeed, C. Mariani, J. Marshall, D.A. Martinez Caicedo, F. Martinez Lopez, M. G. Manuel Alves, S. Martynenko, A. Mastbaum, I. Mawby, N. McConkey, B. McConnell, L. Mellet, J. Mendez, J. Micallef, T. Mohayai, A. Mogan, M. Mooney, A.F. Moor, C.D. Moore, L. Mora Lepin, M. A. Hernandez Morquecho, M.M. Moudgalya, S. Mulleria Babu, D. Naples, A. Navrer-Agasson, D. Nawarathne, N. Nayak, M. Nebot-Guinot, C. Nguyen, L. Nguyen, J. Nowak, N. Oza, O. Palamara, N. Pallat, V. Paolone, A. Papadopoulou, V. Papavassiliou, H. Parkinson, S.F. Pate, N. Patel, Z. Pavlovic, E. Piasetzky, K. Pletcher, I. Pophale, X. Qian, J.L. Raaf, V. Radeka, A. Rafique, M. Reggiani-Guzzo, J. Rodriguez Rondon, M. Ross-Lonergan, I. Safa, C. Sauer, D.W. Schmitz, A. Schukraft, W. Seligman, M.H. Shaevitz, R. Sharankova, L. Silva, E.L. Snider, S. Soldner-Rembold, J. Spitz, M. Stancari, J. St. John, T. Strauss, A.M. Szelc, K. Terao, C.Thorpe, D. Torbunov, D. Totani, M. Toups, A. Trettin, Y.-T. Tsai, J. Tyler, M.A. Uchida, T. Usher, B. Viren, M. L. Velazquez Fernandez, L. Wang, M. Weber, H. Wei, A.J. White, S. Wolbers, T. Wongjirad, K. Wresilo, W. Wu, E. Yandel, T. Yang, L.E. Yates, H.W. Yu, G.P. Zeller, J. Zennamo, C. Zhang, Y. Zhang

Computation and Language

Comparison of techniques for fine-tuning open-weight models for entity extraction from radiology reports

Aawez Mansuri, Kush Mehta, Mohammadreza Chavoshi et al. (+10 more)

Converting free-text radiology reports into structured labels supports cohort building, quality assurance, and monitoring of clinical imaging models, but the strongest label extractors are hosted proprietary models whose use raises privacy, cost, and reproducibility concerns. We asked whether a fine-tuned open-weight model (Gemma-3-12B) can match GPT-4o at multi-label intracranial hemorrhage (ICH) acuity extraction from non-contrast head-CT reports, and which ingredients matter. Using a 2x2 design, we crossed two adaptation strategies (a discriminative classification head, CH; generative instruction fine-tuning, IFT) with two training-data sources (distillation of real GPT-4o-labeled reports; synthetic reports generated by GPT-4o from real exemplars), across five training sizes, benchmarked on 100 expert-adjudicated reports against GPT-4o and the un-tuned open-weight base. The distilled instruction-tuned model (DIFT) matched GPT-4o (macro-F1 0.845 vs 0.850; p = 1.000) and exceeded the base model by 0.178. The decisive factor was the training-data source, not the fine-tuning method: both synthetic-data models failed to exceed the un-tuned open-weight base at any training size and underperformed the distilled models across all acuity classes. Fine-tuning and inference fit within the memory envelope of a single 24 GB consumer GPU. For narrow, high-value clinical label-extraction tasks, distilling real reports, rather than generating synthetic ones, is what closes the gap to a hosted model, enabling a private, low-cost, version-stable on-premises alternative.
Read the full abstract

All authors Aawez Mansuri, Kush Mehta, Mohammadreza Chavoshi, Jahanzaib Malik, Theodorus Dapamede, Frank Li, Rohan Isaac, Beatrice Brown-Mulry, Chiratidzo Rudado Sanyika, YoungSeok Jeon, Judy W. Gichoya, Ali Emami, Hari Trivedi

Machine Learning

Distribution Matching Distillation for Continuous Diffusion Language Models

Paul Le Van Kiem, Dario Shariatian, Umut Simsekli et al. (+1 more)

Continuous diffusion language models generate all tokens in parallel, yet high-quality generation can still require hundreds of network evaluations (NFEs). We study how distributional distillation can reduce this cost by exploiting the student's probabilistic token outputs. Our unified formulation connects the student's output parameterization to the resulting gradient estimators and yields two methods with the same student architecture and reverse-KL matching objective: Simplex-DMD uses continuous token relaxations and pathwise gradients, while Reinforce-DMD uses categorical sampling and REINFORCE with a learned density ratio. We develop both methods for multi-step generation and investigate the training and sampling choices associated with each parameterization. On OpenWebText, for sequences of 1,024 tokens, Simplex-DMD achieves a generative perplexity of 45.6 at a unigram entropy of 5.44 nats in just 4 NFEs, a 49% reduction relative to the strongest evaluated diffusion baseline at matched entropy and sampling budget. Reinforce-DMD improves the frontier at larger budgets, reaching a generative perplexity of 14.9 at an entropy of 5.00 nats with 256 NFEs, a 20% reduction under the same comparison protocol.
Read the full abstract

All authors Paul Le Van Kiem, Dario Shariatian, Umut Simsekli, Alain Durmus

High Energy Physics (Phenomenology)

Interpreting the High-Recoil LUX-ZEPLIN Event with Bino-/Singlino-like and Higgsino Dark Matter

Subhojit Roy, Pedro Schwaller and Carlos E. M. Wagner

The LUX-ZEPLIN (LZ) experiment has recently reported a nuclear-recoil candidate at $E_R\simeq248$ keV. We study two supersymmetric interpretations based on distinct neutralino couplings to the $Z$ boson. In the elastic case, a bino- or singlino-like neutralino scatters through a diagonal axial $Z$ coupling set by the Higgsino asymmetry $X=|N_{13}|^2-|N_{14}|^2$ rather than by the total Higgsino fraction. Since the spin-dependent blind spot does not coincide with the spin-independent one, Higgs-mediated scattering can be suppressed while retaining $X\sim10^{-2}$. In the MSSM this favors low $\tan\beta$, in tension with the observed Higgs mass for a few-TeV squark spectrum, a tension relieved by the tree-level quartic of the NMSSM. An explicit NMSSM benchmark with $m_\chi\simeq500$ GeV and $X\simeq0.007$, below current LZ limits, obtains its relic abundance from a singlet-like pseudoscalar resonance just above threshold, which also suppresses present-day annihilation. In the inelastic case, a pseudo-Dirac Higgsino scatters endothermically, with a splitting of a few hundred keV generated by gauginos at the PeV scale which can be lowered by opposite-sign bino-wino cancellation, or by weak singlino mixing; the minimal thermal realization is strongly constrained by solar-neutrino searches. The two mechanisms predict distinct recoil spectra away from the candidate.
Read the full abstract

Astrophysics of Galaxies

An ALMA sub-kiloparsec test of the Kennicutt–Schmidt relation in a strongly lensed dusty star-forming galaxy at $z\sim2.78$

C. Cornil-Baïotto, J. Molina, G. Mahler et al. (+9 more)

[abridged] Star formation is a key process driving galaxy evolution, but understanding how it proceeds within the interstellar medium (ISM) requires resolving cold molecular gas and dust on sub-kpc scales. At z~2-3, near the peak of cosmic star formation, such resolution can generally only be achieved through strong gravitational lensing. We present an ALMA study of SMMJ0658, a 20x-magnified main-sequence star-forming galaxy at z=2.7768 +/- 0.0002, lensed by the Bullet Cluster (z=0.296) into three images. We use 0.2" and 0.6" ALMA Band-3 observations to test the Kennicutt-Schmidt (KS) relation at sub-kpc scales at cosmic noon. Using a JWST-based strong-lensing model of the Bullet Cluster, we reconstruct source-plane CO(3-2) and rest-frame 803um dust-continuum maps down to ~200pc scales. We derive spatially resolved molecular gas and star formation rate surface densities, depletion times, and the scale dependence of the CO-to-dust flux ratio over 200pc-3.2kpc apertures. The 0.2" ALMA data reveal massive, dense star-forming clumps, with Sigma_mol ~1.3-2.3 x 10^3 M_sun pc^-2, Sigma_SFR ~0.5-2.3 M_sun yr^-1 kpc^-2, and tau_dep ~0.82-2.56 Gyr. CO-dust spatial decorrelation produces a strong scale dependence in tau_dep. We find a KS-relation breakdown scale of 0.8 +/- 0.1 kpc, consistent with individual regions tracing distinct stages of the star-formation cycle. The ALMA and JWST data also reveal evidence for a lensed galaxy pair at z~2.78, although whether SMMJ0658 is undergoing an early-stage interaction remains uncertain. Our results show that the apparent universality of the KS relation breaks down once individual star-forming regions are spatially resolved, in line with observations from the local universe to z~1. We extend this picture to z~2.78, providing new insight into the resolved properties of the clumpy cold ISM and the star-formation cycle at cosmic noon.
Read the full abstract

All authors C. Cornil-Baïotto, J. Molina, G. Mahler, G. Rihtaršič, M. Boquien, M. Bradač, E. Ibar, A. H. Gonzalez, T. Verdugo, O. López-Cruz, J. Magaña, V. Motta

Probability

Spectra of Random Polynomial Matrices: the Petaloid Law

Rikhav Shah and Edward Zeng

We study the distribution of the zeros of $\det P_N(z)$ where $P_N(z)$ is a random monic polynomial matrix, i.e., $P_N(z)=z^dI-\sum_{j=0}^{d-1}A_{j,N}z^j$ for possibly coupled random matrices $A_{j,N}$, scaled to have entrywise variance $O(1/N)$. We provide general conditions under which this distribution almost-surely weakly converges as $N\to\infty$ to a deterministic measure, depending on just the variance and covariance of the entries of the $A_j$. This generalizes the circular, elliptic, and semicircle laws, which concern the special case of this question where $d=1$. Unlike those classical laws, these measures can combine nonuniform two-dimensional densities with singular components supported on curves, producing a variety of petal-shaped regions, inspiring our name “the petaloid law”. We give explicit formulas for the limiting densities and supports. Under a Gaussianity assumption, we also show that there are almost surely no eigenvalues outside small neighborhoods of the limiting support.
Read the full abstract

Quantum Physics

A Quantum Algorithm for $st$-Transport on Flat Connection Graphs

Stacey Jeffery, Tobias J. Osborne and Galina Pass

We study a generalization of undirected $st$-connectivity to graphs whose edges carry quantum operations. Let $G=(V,E)$ be an undirected graph on $n$ vertices in which each edge $\{u,v\}$ is labeled by a unitary $U_{uv}\in\mathbb{C}^{k\times k}$, with $U_{vu}=U_{uv}^\dagger$. We assume the labels form a flat connection: the ordered product of labels along any path between a pair of vertices $u$ and $v$ is independent of the path. Equivalently, the connection is pure gauge, i.e., gauge-equivalent to the trivial connection; such graphs are exactly the consistent connection graphs of spectral graph theory and the noiseless instances of group synchronization. Consequently, whenever $s$ and $t$ are connected, transporting a state from $s$ to $t$ defines a unique unitary $U_s(t)$. Given states $|\psi_s\rangle,|\psi_t\rangle\in\mathbb{C}^k$ and an oracle that returns the neighbours of a vertex while coherently applying the corresponding edge unitaries, the \emph{$st$-transport problem} is to decide whether $s$ and $t$ are connected and, if so, to estimate the squared overlap between $U_s(t)|\psi_s\rangle$ and $|\psi_t\rangle$ to additive error $\varepsilon$. When $k=1$ and all labels are trivial, this is exactly undirected $st$-connectivity. We give a bounded-error quantum algorithm for $st$-transport that runs in time $\widetilde{O}(n/\varepsilon)$ and uses $O(\log n+\log k+\log(1/\varepsilon))$ space. We do this by designing a transducer and applying a Metropolis-Hastings reweighting to the input graph. We also prove an $\Omega(n)$ quantum query lower bound that holds even when $s$ and $t$ are promised to be connected, so for constant $\varepsilon$ our algorithm is optimal up to polylogarithmic factors.
Read the full abstract

Computer Vision

EviRover: Reinforcing Agentic Perception Beyond a Glance

Kaixuan Fan, Kaituo Feng, Tianshuo Peng et al. (+4 more)

Visual perception is conventionally formulated as a one-shot prediction from a single glance at the image, under the assumption that the image content and the model's parametric knowledge suffice to resolve the query. This assumption often fails in real-world scenarios that hinge on fine-grained visual details or require knowledge-intensive and up-to-date information. We term such cases perception under insufficient evidence and formulate perception as an agentic process that can obtain information beyond a single glance. To address the absence of data for this setting, we design two dedicated data generation pipelines, yielding EviRover-SFT-5K and EviRover-RL-12K for training. We further construct EviLens, a human-verified benchmark comprising 688 instances across five perception categories. Building on these data, we present EviRover, to our knowledge the first perception agent explicitly trained to resolve perceptual queries through interaction, using supervised fine-tuning followed by agentic reinforcement learning. Experiments show that the 4B EviRover outperforms its backbone by 30 points on average on EviLens, reaching performance comparable to advanced proprietary models. The gains transfer beyond EviLens to WebEyes, conventional perception benchmarks, and general multimodal benchmarks, including a 15-point improvement on BrowseComp-VL. All code, models, and data are released.
Read the full abstract

All authors Kaixuan Fan, Kaituo Feng, Tianshuo Peng, Yilei Jiang, Manyuan Zhang, Junke Wang, Xiangyu Yue

High Energy Physics (Phenomenology)

$\mathcal{O}(\alpha\alpha_s^2)$ corrections to the Top Quark mass

Daniele Gaggero and Ciaran Williams

We calculate the three-loop $\mathcal{O}(\alpha\alpha_s^2)$ mixed QCD-EW corrections to the top quark mass and obtain the relationships between the pole ($M_t$) and $\overline{\rm{MS}}$ ($m_t(\mu)$) masses at this order. In performing our calculation, we express the three-loop self-energies as a function of three-loop master integrals (MIs). These MIs are defined so as to solve a differential equation in $\epsilon$-factorized form, allowing for their solution to be expressed as iterated integrals. We discuss the types of differential forms that we encounter, which include periods of elliptic curves, and their role in the final physical result. Using our results, we calculate the shift in $m_t(M_t)$ arising at this order. For $M_t = 172.3$ GeV, the shift is -216 MeV, which is comparable to the currently quoted experimental uncertainty.
Read the full abstract