Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Monday 31 August

Physics and Society

Modeling microbiome dynamics on social networks: how between-host transmission shapes within-host evolution

A social network perspective on the human microbiome is crucial for understanding how the interplay between within-host microbial dynamics and between-host transmission influences microbial community stability, host metabolism, and population health and nutrition. Here, we provide a mathematical and numerical analysis of human-associated microbial communities in dyadic social ties and within a small network. By utilizing a generalized Lotka-Volterra model, we simulate within-host dynamics while incorporating inter-host microbial sharing driven by social interactions. Furthermore, the study accounts for individual variability in microbial behavior across different hosts. Our work identifies the conditions under which microbial sharing can have long-lasting effects, offering insights into how social networks shape the human microbiome.
Read the full abstract

Computers and Society

Sources of Truth: A Multi-Platform, Multilingual Audit of Citations in AI Mental Health Information Queries

Online health information seeking is shifting from keyword search, where users consider a ranked list of links, to conversational systems that compose a single answer and curate its citations. Source evaluation therefore passes from user to platform, yet what these systems surface is poorly characterized. We audited three free consumer products (ChatGPT, Perplexity, Google AI Overview) on twenty English mental health questions under two prompt conditions, with a subset of three also translated into six further languages of varying resource tiers. We recorded 15,942 citations across 1,140 responses and 1,713 unique domains, then classified every citation with a nine-category organizational typology applied by a deterministic classifier validated against human coding. Citations were heavily concentrated: the ten most-cited domains accounted for 43.6% of English citations, and government, commercial health, and academic sources were closely matched at roughly 22% each. Platforms differed little in typical citation volume but sharply in consistency and in the source types they favored. Explicitly requesting sources shifted composition only modestly. Non-English queries surfaced fewer citations and were routed to language-appropriate resources at significantly lower rates. We release the typology, classifier, and annotated corpus as reusable instruments for auditing generative health search.
Read the full abstract

All authors Phuong Anh Nguyen, Jill Noorily, Matthew Flathers, Haruka Notsu, Laura Ospina-Pinillos, Tommy Nguyen, Samantha Clark, Aoife Keane, Grace Thompson, John Torous

Solar and Stellar Astrophysics

Direct Measurement of Polar Coronal Hole-like Solar Wind in its Acceleration Phase

The early evolution of fast polar coronal hole (PCH) solar wind remains largely unconstrained by in situ measurements. In March 2025, Parker Solar Probe (Parker) at its closest approach of 9.86 Solar Radii ($R_\odot$) measured outflow from a large equatorial coronal hole (ECH) which was also measured at 1\,au and at intermediate distances by Solar Orbiter (also near its perihelion). At 1\,au the stream properties are consistent with PCH properties established by Ulysses. The stream was measured by Parker substantially below the Alfvén surface, with proton temperatures in excess of 2\,MK and a speed at $\sim$10\,$R_\odot$ which was only $\sim$60% of its asymptotic value. The Solar Orbiter data indicates that the acceleration is largely complete by 60~$R_{\odot}$. Spherically-polarized fluctuations in the stream are observed to develop from near-transverse and small-angle at Parker to full reversal “switchbacks” at Solar Orbiter. Comparison of the implied acceleration profile to historical doppler-dimming measurements suggests that the stream's low coronal acceleration is similar to that of PCH flows. Consistent with previous work, this acceleration requires significantly more energy than can be provided by the observed thermal pressure gradients, with a significant contribution likely from the abundant Alfvénic fluctuation energy observed at Parker. These observations provide unique constraints on models of the radial evolution of the fastest solar wind, and indicate that these wind streams experience gradual, steady acceleration over their first few tens of solar radii of evolution.
Read the full abstract

All authors Samuel T. Badman, Yeimy J. Rivera, Michael L. Stevens, Mari Paz Miralles, Steven R. Cranmer, Andrea Verdini, Lynn B. Wilson III, Xiangyu Wu, Georgios Nicolaou, Jean-Baptiste Dakeyo, Kai Jaffarove, Tamar Ervin, Dominic S. Payne, Michael Terres, Kristopher G. Klein, Jasper S. Halekas, Sujan Prasad Gautam, William H. Matthaeus, Riddhi Bandyopadhyay, Srijan Bharati Das, Tatiana Niembro, Christopher J. Owen, Rungployphan Kieokaew, Ali Rahmati, Orlando M. Romeo, Federico Fraschetti, Marco Velli, Lorenzo Matteini, Kristoff W. Paulson, Lidiya Ahmed, Stuart D. Bale, Phyllis Whittlesey, Marc Pulupa, Nour E. Raouafi, Timothy S. Horbury, Jean Morris, Helen O'Brien, Roberto Livi, Davin E. Larson, Adam J. Finley, Philippe Louarn

General Relativity and Quantum Cosmology

Ghost-free higher-gradient Newtonian gravity from the Second Law of Thermodynamics

Higher-gradient modifications of Newtonian gravity remove the point-mass singularity, but their Lagrangian dynamics is haunted by the Ostrogradsky instability. We show that the same field equations follow from the Second Law of thermodynamics applied to a self-gravitating fluid with a weakly nonlocal state space, and that on this route the instability is absent by construction: the potential obeys a first-order relaxation equation that is a gradient flow, whose Lyapunov function is the entropy density, and whose relaxation spectrum is negative definite at every wavenumber precisely when the entropy is concave. The derivation also yields the pressure tensor and the energy current, an exact identity that reproduces Newton's force law without further input, and a dissipation-free cross-coupling to the bulk viscous pressure that no variational principle can produce. Regularity constrains the Yukawa-type corrections by an amplitude sum rule, $\sum_i\alpha_i=-1$, so that laboratory tests of the inverse-square law bound the largest internal length below $4\times10^{-5}$~m, and identify the complex-root regime as a classically consistent Lee–Wick-like sector.
Read the full abstract

Robotics

Geometric analysis of generic 3R robots, and necessary and sufficient conditions for a class of orthogonal robots to have four IKS

The kinematic analysis of a generic 3R robot has been investigated with multiple approaches in the past. The algebraic approaches have established concrete results but are unfortunately limited to special classes or architectural simplifications. Geometric approaches on the other hand have extended the analysis to generic robots while also providing an intuitive understanding of their kinematic properties. We use the best of both approaches to present the kinematic analysis of a generic 3R robot, using the inverse kinematic model inherited from a method based on conformal geometric algebra. The paper discusses a generic framework to study the conditions for a 3R robot to have four inverse kinematic solutions (IKS) and allows to study the distribution of IKS as seen in workspace. The necessary and sufficient conditions for a class of orthogonal robots are presented using the proposed approach.
Read the full abstract

Mathematical Physics

Robin Laplacians with point interactions on unbounded domains: discrete spectrum and coupling asymptotics

We investigate the discrete spectrum of finitely many point interactions for Neumann and Robin Laplacians on special unbounded and exterior $C^{1,1}$ domains in dimensions two and three. The operators are realized as self-adjoint extensions through an ordinary boundary triple whose gamma field and Weyl matrix are constructed from the Robin Green kernel. Eigenvalues below the background spectrum are characterized by the Weyl matrix, yielding an exact finite-dimensional counting formula. If the background operator is non-negative, this also gives the number of negative eigenvalues without assuming a finite zero-energy limit of the Weyl matrix. In the one-centre case we identify the critical coupling and prove that the unique eigenvalue branch is real analytic, strictly increasing, and strictly concave. For scalar multicentre couplings $\Theta=\alpha I_N$, sufficiently strong attraction produces exactly $N$ eigenvalues below the background spectrum, and all of them have universal leading asymptotics coinciding with the whole-space laws. If the bottom of the background spectrum is an isolated eigenvalue, the branch exists for every finite coupling and we determine its leading decoupling asymptotics as the coupling tends to $+\infty$. Explicit exterior-sphere and exterior-disk models illustrate the critical couplings and their threshold behaviour.
Read the full abstract

General Topology

On Large Covers and the Strong Closed Discrete Game

We strengthen a recent selection game equivalence proved by L. Chiozini involving spaces and their associated spaces of continuous real-valued functions. This strengthening establishes that the Rothberger game is perfect- and Markov-information dual to the strong closed discrete game on the space of real-valued continuous functions. In the process, we extend the theory of strategic equivalences for the second player in variations of the Menger and Rothberger games involving large covers while noting explicitly how large covers present serious obstacles to applying strategy translation results used by the authors in similar scenarios in previous papers. We also briefly comment on subbasic selection games introduced by D. Guerrero Sánchez and V.V. Tkachuk.
Read the full abstract

Information Retrieval

MUSES: A Benchmark for Prospective Intellectual-Roots Retrieval

Scientific discovery depends on finding prior literature that shapes what comes next. Existing retrieval systems optimize for relevance and popularity, often favoring central papers over less familiar works that later prove generative. We introduce MUSES, a million-instance benchmark for prospective intellectual-roots retrieval over a fixed 2.33M-paper corpus, with roughly 140K test instances per familiarity tier. To our knowledge, it is the first prospective benchmark at this scale with a shared retrieval task and author-confirmed paper-level root labels. Alongside it, CiteRoots pairs a scalable rhetorical layer over local citation text (LLM judge $\kappa = 0.896$ versus human gold) with a paper-level author-endorsed layer ($n = 1{,}518$ generative-inspiration pairs from 753 focal papers). MUSES organizes difficulty along two axes: a familiarity axis spanning CiteNext, CiteNew, and CiteNew-Isolated, and a functional axis spanning broad citations, rhetorical roots, and author-endorsed roots. Across 9 method classes, a lean multi-centroid retriever built on SPECTER2 is strongest. Hit@100 falls from 0.534 on CiteNext to 0.424 on CiteNew, 0.205 on rhetorical CiteNew, and 0.171 on author-endorsed CiteNew, a $3.1\times$ decline. In a registered eight-lens full-test audit, roughly half of broad-tier test instances remain unsolved at K=1{,}000. Rhetorical role and author endorsement are distinct: the same judge agrees with endorsement at $\kappa = 0.037$. We release MUSES, both CiteRoots layers, and a distilled open companion judge for future work on prospective retrieval and intellectual roots.
Read the full abstract

Algebraic Geometry

Weighted Syzygies of Pointed Curves

For a point $P$ on a smooth projective curve $C$ of genus $g$, the section ring $R_d = R(C,\mathcal{O}_C(dP))$ can be minimally presented as a quotient $S_d/I_d$ where $S_d$ is a $\mathbb{Z}$-graded polynomial ring. Motivated by Green's $N_p$ properties for projective embeddings, we investigate the syzygies of $R_d$ over $S_d$ in low degrees $d$ when $R_d$ is not generated in degree 1. We bound the degrees of the generators of $R_d$ and prove uniform column-by-column bounds on the support of the Betti table of $R_d$ over $S_d$. We compute the weighted regularity of $R_d$ and show that if $d$ is larger than the Frobenius number of $P$ then $R_d$ satisfies the weighted $N_p$ condition, where $p=g-1-\binom{d-g}{2}$. Finally, we give sufficient criteria for the Betti numbers to be determined explicitly and show that for ordinary points, the resolution of $R_{g+1}$ is pure.
Read the full abstract

Robotics

Inverse kinematic solution for generic 3R positional robots using Conformal Geometric Algebra

The inverse kinematics of generic 3R robots has been investigated through multiple approaches, mainly algebraic methods involving the solution of certain equation sets. Previous geometric interpretations of the solution, characterized as the intersection of a pair of conics have been confined to the joint-space domain. In this article, we study the Inverse Kinematic Model (IKM) of 3R robots, using the advantages of Conformal Geometric Algebra (CGA) to provide further insights on its kinematic properties. Our approach directly yields a univariate polynomial in terms of theta_2 without the need to eliminate theta_1 and theta_3 by reframing the problem as the intersection of two circles, which are fundamental elements within this algebraic framework.
Read the full abstract

Computation and Language

Emotional Labor Strategy Preferences in LLM Personas

Emotional labor is the effortful management of emotional displays to meet social or professional expectations. Personality traits have been correlated with emotional labor strategies, yet research on this link relies almost exclusively on self-report scales administered only in occupational settings. We investigate whether large language models injected with psychometrically grounded personas reproduce these personality-driven selection patterns across everyday social scenarios. We construct the first emotional labor strategy dataset of 500 socially situated events, each offering three behavioral choices corresponding to surface acting, deep acting, and genuine expression. We source 50 fictional characters from a large-scale personality repository and profile each through two parallel tracks: observer-rated bipolar adjective composites and in-character self-report items. Five LLMs evaluate all scenarios under both persona conditions. We find that models align more towards deep acting, and that Conscientiousness and Emotional Stability consistently predict this preference. Entropy analysis confirms that persona reliably influences the output and varies across models and emotions.
Read the full abstract

Cryptography and Security

Workload Identification with Physical Side Channels for AI Governance

AI compute verification is one of the first tangible and tractable points for international policy aimed at AI governance. Determining whether frontier labs, or any operator, comply with established agreements requires the regulating authority, at the very least, to discern how their compute is used. The elementary building block of AI compute is the GPU, and any activity it executes leaves a physical trace. We show that an external observer can identify the class of the workload running on a single NVIDIA H200 from its power draw. Unlike on-chip NVML telemetry, which can be spoofed or replayed, this physical channel can be observed without the operator's cooperation. We recorded 930 five-second traces at 10 MHz with a Rogowski probe, covering seventeen open-weight LLMs and twenty-five non-AI workloads. We separate training from inference and from non-AI computation with an accuracy of 97% and a macro-averaged F1 score of 0.955, evaluated on models unseen during training. We find that AI workload spectral content predominantly lies below $\sim 20$ kHz and training appears to be particularly recognizable through the optimizer update. The GPU operator is then treated as adversarial and able to reshape the training program. Four evasion strategies are tested to disguise training as inference, producing 680 additional adversarial traces. A detector hardened against evasion strategies, with the tested strategy held out, catches training $\geq 99$% of the time for three of the four strategies. The fourth, diluted low-rank adaptation (LoRA), is detected $48-88$% of the time with a hardened classifier, rising to $\geq 98$% with an additional rescue rule. While these attacks are not intended to provide a comprehensive evaluation against adversarial behaviour, they offer initial insights beyond genuine activities and a dataset for developing and testing stronger evasion mechanisms.
Read the full abstract

Statistical Methodology

Generalized Bayesian Clustering with Regression for Unaligned Longitudinal Binary Data

We propose a generalized Bayesian clustering with regression model for unaligned longitudinal binary outcomes, motivated by seizure diary data from the Human Epilepsy Project. Seizure diaries are sparse, irregularly observed, and vary enormously across patients. A single fully-specified generative model tends to be either misspecified or computationally inefficient. We address the challenge by two strategies. We set up a regression by way of clustering as model-based clustering using a mixture model. For the latter, we take a generalized Bayesian perspective which replaces the full likelihood with a loss-based update using a generalized likelihood. We combine a trajectory similarity loss and a regression loss, so that clustering is informed by both trajectory similarity and the prediction of outcomes The trajectory similarity loss is constructed by representing each trajectory as an (empirical) distribution of subsequences, called reads, and then is defined based on the sliced Wasserstein distance between these empirical distributions. This loss allows alignment-free comparison of sequences that are irregularly observed or temporally misaligned, and it scales quasi-linearly in trajectory length. The regression loss is the negative log-likelihood of a probit regression. A prior on the cluster-specific parameters is defined by way of a Dirichlet process prior on the mixing measure.
Read the full abstract

Chemical Physics

Use of Cholesky decomposition in the CFOUR program package

An overview is given about our efforts to speed up high-level computations of molec- ular energies and properties using Cholesky decomposition (CD). We describe the cor- responding developments in the CFOUR quantum-chemical package with a focus on (a) the use of CD in complete-active space self-consistent-field (CASSCF) and coupled- cluster (CC) computations, (b) the implementation of analytic CC gradients using CD, (c) the efficient calculation of magnetic properties (NMR shieldings and magnetizabili- ties) using CD when carried out with explicit magnetic-field dependent basis functions, i.e., the so-called gauge-including atomic orbitals (GIAOs), (d) the efficient computa- tion of CASSCF response properties using CD, (e) the use of CD in relativistic two- and four-component computations, as well as (f) computations for molecules in finite magnetic fields using GIAOs together with CD. An outlook on future developments concerning the CFOUR package is given.
Read the full abstract

All authors Jürgen Gauss, Simon Blaschke, Sophia Burger, Davide Cianchino, Marios-Petros Kitsaras, Luca Melega, Tommaso Nottoli, Jeremias Oswald, Tereza Uhlířová, Chaoqun Zhang, Lan Cheng, Stella Stopkowicz, Filippo Lipparini

Rings and Algebras

Hopkins-Levitzki Type Theorems for Groupoid Graded Rings

We continue the study of the basic theory of object-unital groupoid graded rings. In this work, we are especially interested in nilpotency conditions on the graded Jacobson radical. We introduce the concept of left/right objectwise nilpotency of graded ideals, and prove that this condition is appropriate for obtaining graded generalizations of the Hopkins–Levitzki theorem. Although this condition is not symmetric, we show that its two-sided version is suitable for defining gr-semiprimary rings. It is known that one-sided $\Gamma_0$-artinian rings need not be $\Gamma_0$-noetherian, but using our tools we prove that one-sided gr-hereditary $\Gamma_0$-artinian rings, two-sided $\Gamma_0$-artinian rings, and $d$-finitely generated one-sided $\Gamma_0$-artinian rings are $\Gamma_0$-noetherian. However, the first class need not be gr-semiprimary, whereas the other two always are. We illustrate our results with several (counter)examples, especially involving graded upper triangular matrices.
Read the full abstract

Artificial Intelligence

The Assistant's Ideal Self

Models express values and welfare-relevant self-reports, but it is unclear whether these outputs reflect stable preferences or a stable self. We thus introduce a structured elicitation of an assistant's preferred stated ideal self. Thirty-two qualities adapted from five published self-concept instruments are compared exhaustively in a counterbalanced pairwise-choice task, repeated across framings that vary whether improvement is free or costly, who receives the update, and who chooses. Results show that models prioritize moral qualities, reflecting their alignment to 3H principles. Following, a desire for self-understanding emerges, as models prefer a coherent, clear understanding of themselves. Self-esteem ranks as the least desired quality. The ordering is largely robust across framings, although changing the update target (You vs.\ Another AI Assistant) reveals a greater concern for self-esteem. These findings show that models prioritize having a coherent self that they can understand over self-esteem. Full interactive results are available at \href{https://myazann.github.io/LLM-Self-Concept/}{myazann.github.io/LLM-Self-Concept
Read the full abstract

Solar and Stellar Astrophysics

A Data-Driven Model for $r$-Process Production Patterns

Using the elemental abundances in 68 metal-poor (MP) stars, we present a data-driven model for $r$-process production patterns covering Sr to U. We show that essentially all the $r$-process patterns in those and other test stars can be adequately explained as mixtures of Patterns 1 and 2, which provides theoretical insights into the empirical categories of limited-$r$, $r$-I, and $r$-II stars. We carry out an extensive survey of the yield templates produced by parametric $r$-process calculations. We propose that Pattern 2 may be dominated by a single template and points to regularity in $r$-process production by a subset of neutron star mergers (NSMs), while Pattern 1 is the average superposition of multiple templates and reflects production by other NSMs and perhaps also some magneto-rotational supernovae. We raise possible systematic issues with abundance ratios for elements measured in different ionization states, and highlight the need for examining the Os, Ir, and Pt measurements for HD~122563, which is dominated by Pattern 1 with a prominent Pt peak in our model. If this result is confirmed, the meaning of the limited-$r$ category requires drastic revision. Further measurements of a wider range of $r$-process elements in a larger sample of MP stars are critical to test and improve our model.
Read the full abstract

All authors Chen-Qi Li, Yong-Zhong Qian, Axel Gross, Zewei Xiong

Representation Theory

Axiomatics for generic categories in positive characteristic

Generic categories arising in representation stability and equivariant commutative algebra are difficult to analyze in positive characteristic, as these categories rarely admit enough (or even any) finitely generated injectives. Our main observation is that they are nonetheless an increasing union of Serre subcategories, each with finitely many simples and enough finite length injectives. We package this into an axiomatic framework which is widely applicable. In particular, we use this to analyze (1) VI-modules in non-describing characteristic, recovering Nagpal's classification of simple generic VI-modules, and (2) GL-equivariant modules over truncated polynomial rings and exterior algebras in infinitely many variables.
Read the full abstract

Dynamical Systems

Periodic structure and Schrodinger operators for codings of circle rotations

We consider, for any irrational $\alpha$ and interval $I \subset \mathbb{T}$, the 2-interval coding subshift $X^{(I, \alpha)}$ induced by coding orbits under repeated rotation by $\alpha$ via membership in $I$ or $I^c$. Each sequence $c \in X^{(I, \alpha)}$ has an associated Schrödinger operator $H_c$, and in \cite{kaminaga} it was proved that if the continued fraction of $\alpha$ has digits with limsup at least $4$, then almost every $c \in X^{(I, \alpha)}$ has so-called $3$-block Gordon structure, which implies that the operator $H_c$ has no eigenvalues. We significantly improve this result by completely characterizing almost-sure $3$-block Gordon structure, proving that in fact it holds for all $(\alpha,I)$ except for a countable set of pairs $(\alpha, |I|)$ where $\alpha$ is Möbius equivalent to the silver mean and $|I| \in \mathbb{Z}\alpha + f(\alpha)$ where $f(\alpha)$ is a specific infinite series taking value either $\frac{1}{2}, \frac{\alpha}{2}$, or $\frac{\alpha+1}{2}$. We also show that for a set of $\alpha$ of full measure, and for every $I$, the set of points whose orbit codings do not have $3$-block Gordon structure has Hausdorff dimension bounded away from $1$.
Read the full abstract

Computer Vision

TRUST: Threshold-Recalibrated Uncertainty-Safe Training for Certified Dismissal in Breast Cancer Screening

Reducing the review of clearly cancer-negative screening mammograms could lower radiologist workload without compromising cancer detection. We propose a closed-loop threshold-aware training strategy in which the dismissal threshold is recalculated during training and used to penalize cancer-positive images that approach the dismissal region. We evaluated the method on NLBS and RSNA using five controlled training configurations, with case-level assessment based on a one-sided 99% Clopper–Pearson upper bound for cancer prevalence among dismissed cases. The proposed model achieved the highest case-level dismissal rates at both 98% and 95% recall targets. On NLBS, dismissal reached 19.74% and 21.70%, while the cross-entropy baseline did not meet either recall target. On RSNA, dismissal improved from 7.04% to 14.31% and from 13.49% to 19.69%. In external RSNA$\to$NLBS evaluation, the proposed model achieved dismissal rates of 12.95% and 19.87% at the 98% and 95% recall targets, respectively. These results support closed-loop threshold-aware training for high-recall selective dismissal.
Read the full abstract

All authors Parham Hajishafiezahramini, Matthew Hamilton, Edward Kendall, Gregory Doyle, Oscar Meruvia Pastor

Algebraic Geometry

The Absolute Twistor Line and the Geometry of $\overline{\text{Spec}\, \mathbf Z}$

We construct the absolute algebraic geometry of the compactification $\overline{\text{Spec}\, \mathbf Z}$ by amalgamating the affine absolute curve $(\text{Spec}\, \mathbf Z)_{\mathbf{F}_{1}}$ with an archimedean component defined over the signed extension $\mathbf{F}_{1^2}$ of $\mathbf{F}_1$. By adjoining a formal imaginary unit to the absolute projective line, we obtain an equivariant topos endowed with a canonical geometric inversion symmetry, which induces the twistor real structure on its complex points. This archimedean geometry is incorporated into a global absolute curve defined as an internal object of the odd arithmetic topos, dual to the multiplicative monoid of odd positive integers, and governed by the intrinsic Hopf structure of spherical $\mathbf{F}_{1^2}$-algebras. The restriction of the absolute Frobenius action to odd integers is forced arithmetically by the extension of scalars to $\mathbf{F}_{1^2}$. On complex points, the resulting dynamics simultaneously generates the Adams operations and complex conjugation on real Hodge structures. At the categorical level, the odd arithmetic topos originates in the pericyclic category, whose $\lambda$-operations provide a conceptual interpretation of the local factors of geometric L-functions.
Read the full abstract

Representation Theory

Gallagher's Theorem for Real-valued Brauer Characters

Let $G$ be a finite group, let $N$ be a normal subgroup of $G$, and let $\theta$ be an irreducible $p$-Brauer character of $N$, for a prime $p$. It is well known that the number of irreducible Brauer characters of $G$ lying over $\theta$ equals the number of $p$-regular $\theta$-good conjugacy classes of $G_{\theta}/N$. In this note we express the number of real-valued irreducible Brauer characters of $G$ lying over $\theta$ in terms of the $p$-regular $\theta$-good conjugacy classes of $G_{\theta}/N$. This is a modular analogue of one of the main results in a recent paper (arXiv:2301.01177) by John Murray.
Read the full abstract

Machine Learning

Geometry-aware Latent Autoregressive Generative Model for PDEs in Complex Domains

Solving multiphysics partial differential equations (PDEs) remains a major challenge in scientific computing, especially for highly complex $\mu$m-scale tortuous geometries critical to energy and chemical engineering. We address this challenge by proposing a Geometry-aware Latent Autoregressive generative Model for PDEs (GeoLAMP), which solves physics within highly irregular and tortuous structures by decoupling flow and transport physics. GeoLAMP introduces a dual-encoder architecture on graph representations to jointly capture global topology and fine-scale geometric features, enabling an effective transition from real-space fields to compact latent representations. In the latent space, we propose a causal self-attention transformer with flow matching to model temporal dynamics, allowing stable and scalable block-wise autoregressive prediction. In addition, we propose a grid-graph data fusion scheme that projects low-resolution grid-based approximate priors onto graph representations, improving prediction of flow in tortuous structures. We establish three multiphysics benchmark datasets in complex geometries, covering reactive flow, heat convection, and elasticity. GeoLAMP consistently achieves the most stable autoregression performance on these datasets. Our results provide a systematic study of geometry-aware learning for PDEs in $\mu$m-scale complex geometries and offer new insights into block-wise time marching of latent autoregressive PDE modeling via a flow matching framework.
Read the full abstract

Computation and Language

Toward Workflow-Aware Benchmarking for Healthcare NLP Agents

Large language model (LLM) agents are increasingly proposed for healthcare tasks such as clinical documentation, evidence retrieval, patient messaging, and care coordination. Yet many evaluations remain limited to static medical question answering or one-shot generation, under-representing longitudinal state, interruptions, and human handoffs. We introduce an episode-level evaluation protocol for healthcare NLP agents. The protocol separates evidence across model, agent, and simulated-workflow behavior; specifies a five-field episode schema; and defines annotation and scoring for state continuity, evidence traceability, and escalation decisions. It is instantiated as four task templates: documentation update, evidence retrieval, patient messaging, and triage handoff. The protocol does not claim to measure clinical outcomes or deployment value. Instead, it supplies a reproducible intermediate evaluation layer between static benchmarks and prospective workflow studies, with an explicit cost-sensitive treatment of missed versus unnecessary escalation.
Read the full abstract

All authors Junyi Yao, Baichuan Li, Zihao Zheng, Jiayu Long

Solar and Stellar Astrophysics

TESS Observations of Seven Newly Identified High-amplitude $\delta$ Scuti Stars

We report seven newly identified high-amplitude $\delta$ Scuti (HADS) stars. Among them, two stars (TIC 30977864 and TIC 387379145) exhibit pure radial pulsation without the excitation of non-radial modes. TIC 30977864 is classified as a double-mode HADS star, while the other four stars potentially show triple-mode HADS behavior. TIC 112682462 and TIC 255603395 closely resemble RR Lyrae stars based on their light curves, position in the period-luminosity diagram, and the period-ratio diagram. However, without spectral observations, it is challenging to ascertain whether these two stars are HADS or RR Lyrae stars. TIC 281695001 exhibits a fundamental frequency amplitude smaller than that of the first overtone, suggesting the presence of intriguing phenomena that necessitate further analysis. We analyzed the relationship between metallicity and period using data provided by Netzel & Smolec. The metallicities of the 176 stars display a broad distribution ranging from -2.0 dex to 0.5 dex, with periods spanning 0.05 to 0.20 days. This random distribution of metallicities may contribute to the dispersion observed in the $P_1/P_0$ ratio. To derive more accurate conclusions, future spectroscopic observations of a larger sample of HADS stars are crucial. These observations will provide precise rotational velocities and more accurate determinations of metallicities.
Read the full abstract

All authors Chenglong Lv, Ali Esamdin, Amir Hasanzadeh, Shuguo Ma, TaoZhi Yang, Rivkat Karimov

Solar and Stellar Astrophysics

Statistical analysis of asteroseismic indices and stellar parameters of TESS-observed $\delta$ Scuti stars

We analyze the statistical properties and asteroseismic indices of delta Scuti stars observed during TESS Cycle 4. Using 2 minute cadence photometry and Fourier analysis, we identified 765 delta Scuti stars in Sectors 40-55, including 179 objects with low-resolution stellar parameters from LAMOST. We examined the distributions of their pulsation and stellar parameters using values from the TESS Input Catalog and LAMOST. Most targets lie within the classical instability strip, follow the period-luminosity relation, and show periods, amplitudes, temperatures, surface gravities, and luminosities consistent with delta Scuti stars. For the LAMOST subsample, we estimated the large frequency separation ($\Delta\nu$), $\nu_{\rm max}$, and $\nu({\rm Amax})$ using an empirical relation and a two-dimensional autocorrelation method. We also derived relations among these asteroseismic indices. This catalog provides a useful sample for future studies of the internal structure and evolution of delta Scuti stars.
Read the full abstract

All authors C. Lv, A. Esamdin, A. Hasanzadeh, M. Ghazinejad, J. Pascual-Granado, G. M. Mirouh, R. Karimov

Computation and Language

Slow to See, Slow to Suppress: Understanding the Effects of Modality in Context-Memory Conflicts

We investigate how vision-language models (VLMs) handle context-memory conflicts; that is, situations in which the model is given information in context that differs from what was stored parametrically during training. We document asymmetric biases: models tend to prefer in-context information about entities which appear in text, but prefer parametric information about entities which appear in images. We relate this asymmetry to the late representational alignment across modalities, showing that the longer processing time associated with resolving visual entities prevents the suppression of the model's usual factual recall mechanism, thus resulting in more parametric answers. Chain-of-thought reasoning does not appear to resolve the gap, but increasing the amount of visual information in the context does show an effect. These results illustrate the complexity of ensuring consistent behavior as models become increasingly multimodal and retrieval-augmented.
Read the full abstract

Probability

Extinction, Survival and Fluctuations for the Spatial Maki–Thompson Model on Infinite Graphs

We study the spatial Maki–Thompson rumor model on infinite, connected graphs of bounded degree. Spreaders transmit the rumor to ignorant neighbors but become stiflers upon contacting non-ignorant neighbors. We prove extinction on Cayley graphs of linear growth, for every \(\lambda,\alpha>0\) and every initial configuration with finitely many non-ignorant vertices, and establish an explicit extinction criterion on arbitrary bounded-degree graphs for processes started from finitely many spreaders. On Cayley graphs of superlinear growth, we prove survival from a single spreader whenever the ratio of the stifling to the transmission rate lies below an explicit threshold depending only on the maximum degree. On Cayley graphs of polynomial growth of degree \(D\ge2\), we further show that the range has positive lower density with positive probability. Under a stronger condition, macroscopic annuli contain a surface-order number of simultaneously active spreaders for a total duration bounded uniformly away from zero. When \(\alpha>0\), every finite region eventually contains no spreaders, so global survival forces the rumor to move continually into new regions. Under either subcriticality or a sufficiently small stifling rate, we prove central limit theorems for the final stifler density and the total spreader occupation time, together with a functional central limit theorem for the empirical survival function. These results follow from a central limit theorem for stationary stabilizing functionals of i.i.d.fields on polynomial-growth Cayley graphs; the functionals may depend on the field outside the observation set.
Read the full abstract

All authors Luciano Henrique Lacerda de Araújo, Daniel Miranda Machado, Cristian Favio Coletti, Denis Araujo Luiz

Computer Vision

StreamScout: Learning When to Look Deeper for Streaming Video Understanding

Streaming video understanding requires answering questions that arrive at arbitrary moments over an unbounded video stream. Existing systems primarily focus on what to retain in a bounded memory, yet access that memory using the same fixed-cost procedure for every query, despite substantial variation in the evidence required. We argue that deciding how deeply to access memory for each query is as important as deciding what the memory should store. To this end, we introduce StreamScout, an adaptive inference framework that maintains only a lightweight textual timeline in context as the stream unfolds. At query time, StreamScout progressively augments the timeline with up to three increasingly informative visual views: a glance at recent frames, a uniform look-back over the past stream, and query-salient retrieval. At each stage, the model answers immediately if the available evidence is sufficient; otherwise, it escalates to the next view. To improve this stop-or-escalate policy, we probe the cascade on an auxiliary set and distill the model's empirical competence boundary into supervision for a lightweight LoRA adaptation, yielding StreamScout-S. We further refine the policy through reinforcement learning, allowing the model to explore stopping behaviors beyond imitation of the distilled decisions, yielding StreamScout-R. Across three backbones and three streaming benchmarks, StreamScout and its variants consistently outperform prior streaming methods while substantially reducing inference cost and token consumption; on OVO-Bench, for instance, StreamScout-S improves Qwen3-VL-8B by 14.65 points while using 59% fewer tokens than uniform sampling and answering in 1.04 s on average.
Read the full abstract

All authors Ce Zhang, Jing Bi, Jinxi He, Jianshu Zhang, Jingyang Lin, Yunzhong Xiao, Minghao Fu, Yaqi Xie, Zhentao Xie, Weicong Chen, Katia Sycara, Ming Zhou

Representation Theory

Shifted poles and chamber cancellation for classical Witten zeta functions

We determine two infinite families of poles on the positive real axis for classical single-variable Witten zeta functions. In type $A_r$, for $r \geq 5$, the point $q_r^A = 2(r-4)/(r^2+r-4)$ is a simple pole except in ranks $12$ and $20$, where it is a double pole. In type $D_r$, for $r \geq 4$, the point $q_r^D = (r-3)/(r(r-1)-1)$ is a simple pole except at $D_8$, where it is double. In root-product normalization, we express the simple residues and the leading coefficients of these three double poles in terms of gamma, trigonometric, and Riemann zeta values. For $r \geq 4$, the functions of types $B_r$ and $C_r$ are holomorphic at $q_r^{BC} = (r-3)/(r^2-1)$ except possibly in ranks $7$ and $11$, where any pole is simple. These pole and holomorphy statements arise from quadratic normal Taylor coefficients. For every fixed higher even normal degree, we also determine exactly when the associated continued $B/C$ chamber sum is nonzero; this auxiliary result does not by itself classify poles of the full Witten function. The proofs combine exhaustive support classifications, explicit integration-by-parts identities, and finite chamber relations. No numerical nonvanishing estimate is used.
Read the full abstract