Cosmology and Nongalactic Astrophysics
Low-cost neural summaries for cosmological inference: an ensemble of convolutional neural networks trained on approximate simulations
Iñigo Sáez-Casares, Davide Bianchi, Matteo Calabrese et al. (+1 more)
Machine learning can build near-optimal summary statistics for extracting cosmological information from the large-scale structure of the Universe. Training them, however, requires large numbers of costly simulations. We aim to build a neural summary trained on fast, approximate low-resolution simulations that remains informative when applied to higher-resolution ones, reducing the cost of the training set. We train an ensemble of convolutional neural networks with different architectures to compress the matter overdensity fields of low-resolution particle-mesh simulations. The outputs of a subset of networks are concatenated into a summary statistic, which we use to constrain $\Omega_{\rm m}$ and $\sigma_8$ from high-resolution particle-mesh simulations, each about a thousand times more costly than their low-resolution counterparts. Theoretical predictions are provided by a Gaussian process emulator built from the high-resolution simulations. We test three strategies for selecting the smallest informative subset of networks. Constraints are compared with those from a reference power spectrum analysis. We also consider dark matter
halo catalogues from independent high-resolution $N$-body simulations. The neural summary extracts information beyond the binned power spectrum. Using multiple networks is essential: a single network can be less informative than the power spectrum. A greedy search based on posterior volume provides the tightest constraints with the fewest networks. The neural summary also outperforms the power spectrum for dark matter
haloes, at different number densities and
redshifts, despite being trained on low-resolution matter fields at $z=0$. This work demonstrates that an ensemble of networks trained on approximate simulations yields a summary statistic that remains informative on higher-fidelity data, including dark matter
haloes. (Abridged)
Read the full abstract ▾
All authors Iñigo Sáez-Casares, Davide Bianchi, Matteo Calabrese, Luigi Guzzo
Computation and Language
Frustratingly Easy Majority Voting Using Length Conditioning
Fiorenzo Parascandolo, Enver Sangineto, Xiaoyi Yu et al. (+4 more)
Majority Voting (MV) is a simple and widely used strategy which improves the
reasoning accuracy of Large
Language Models (LLMs) by allocating additional inference-time computation. Specifically, in MV multiple Chain-of-Thoughts (CoTs) are independently sampled and the most frequent final answer is selected. In this work, we propose Length-Conditioned Majority Voting (LCMV), a Test-Time Training approach which improves MV leveraging an empirical observation that has recently been established in several studies: longer CoTs tend to be associated with a lower probability of correctness. We formulate this inductive bias as a length-conditioned prior probability, which is modeled by a simple logistic function. Moreover, we adapt this prior at test time using only unlabeled testing data and without fine-tuning the LLM. Finally, the adapted prior is used to weight the vote of each sampled answer. Extensive experiments across multiple LLMs and
reasoning benchmarks show that LCMV consistently improves over both standard MV and existing length-based and confidence-based aggregation strategies, while retaining the simplicity and model-agnostic nature of MV. Our code is available at: https://github.com/aimagelab/LCMV.
Read the full abstract ▾
All authors Fiorenzo Parascandolo, Enver Sangineto, Xiaoyi Yu, Wenhui Tan, Jipei He, Ruihua Song, Rita Cucchiara
Signal Processing
Comparative Analysis and Behavioral Modelling of Power Amplifiers in FR3
Cel Thys, Rodney Martinez Alonso, Ali H Alsarraf et al. (+3 more)
Power amplifier (PA) nonlinearity is a critical transmitter impairment in the upper mid-band (7 GHz to 24 GHz), designated as Frequency Range 3 (FR3) in emerging 6G standardisation. The 3GPP PA model baseline, TR 38.803 Annex A, was derived from 2.1 GHz and 28 GHz measurements and has not been revisited yet for FR3 standards work. This paper addresses that gap with characterisation measurements of multiple PA devices-under-test (DUTs), driven into gain compression in the 7 GHz and 15 GHz bands with 100 MHz and 200 MHz bandwidth. Memoryless
polynomial and Generalized
Memory Polynomial (GMP) behavioral models are extracted by least-squares regression and compared with 3GPP models spanning FR1 and FR2. The GaN PA DUT is modelled best at a higher nonlinear order and shows markedly stronger phase distortion than the wideband PA. The
memory terms of the GMP model measurably improve the Normalized Mean Square Error (NMSE) at all operating points. An orthogonal frequency-division multiplexing (OFDM) ambiguity function analysis from the Integrated Sensing and Communication (
ISAC) literature, using the extracted models, illustrates how PA compression raises the sidelobe floor of the sensing response. These results provide a reference for PA impairment modelling in FR3, directly applicable to
ISAC system design. All model coefficients and measurement data are publicly available [1].
Read the full abstract ▾
All authors Cel Thys, Rodney Martinez Alonso, Ali H Alsarraf, Zhuangzhuang Cui, Dominique Schreurs, Sofie Pollin
Quantitative Methods
Maximum likelihood, neural networks and posterior mean for inferring fractional Brownian motion parameters
Daniel S. Han, Vaishnavi Ananthanarayanan and Matthew J. Simpson
Fractional Brownian motion (fBM) is a model of anomalous
diffusion parameterised by a Hurst exponent $H \in (0,1)$ and a generalised
diffusion coefficient $D_H > 0$. We jointly estimate $H$ and $D_H$ using maximum likelihood (MLE) from density measurements with Poisson counting noise and single-particle trajectory with Gaussian localisation error. We show that one density measurement in time cannot infer both $H$ and $D_H$ and that several time points are needed, which matters most for inference precision. For trajectories, we show that a neural network trained with a squared error loss approximates the posterior mean of $H$, which can be computed exactly at negligible cost. Paired with the profile-likelihood estimate of $D_H$, the posterior mean has the lowest root mean squared error in both $H$ and $D_H$, except near $H=0$ and $H=1$ where MLE estimates $H$ better. In vesicle trajectories from two
cell types, $H$ and $D_H$ form a continuum rather than distinct modes of motion. The posterior probability that each window within a track is subdiffusive or superdiffusive shows that $26\%$ of MRC5 fibroblast and $21\%$ of HeLa vesicle trajectories contain both subdiffusive and superdiffusive windows, against $2\%$ and $7\%$ expected from estimation uncertainty alone. Intracellular cargoes therefore switch between subdiffusion and superdiffusion over time rather than forming populations with fixed $H$.
Read the full abstract ▾
Differential Geometry
A parabolic flow approach to generalized Gauduchon metrics
Han Luo
We give a parabolic proof of the Fu-Wang-Wu
theorem for semilinear elliptic equations associated with generalized Gauduchon metrics. Fixing a Hermitian metric $\omega$, we introduce a flow $\partial_t v-\Delta v=\psi(\vert\nabla v\vert^2)+\langle B,dv\rangle-c(t)-f(x)$ for the conformal factor, whose stationary points solve the semilinear elliptic equation of Fu-Wang-Wu. Under the growth condition $\liminf_{t\to+\infty}\psi(t)/t^\mu>0$ for some $\mu>\tfrac{1}{2}$, we prove that the flow exists uniquely for all time and converges; the limit solves the elliptic equation and is unique up to addition of a constant. Our convergence argument does not rely on a gradient-flow structure or a monotone energy functional. In particular, this recovers, by a flow method rather than the continuity method, the existence and uniqueness
theorem of Fu-Wang-Wu on $k$-th generalized Gauduchon metrics, providing a new parabolic approach to their elliptic theory.
Read the full abstract ▾
Statistical Methodology
Interactive and Robust PCA Graphics with FSDA in MATLAB
Marco Riani, Anthony C. Atkinson, Aldo Corbellini et al. (+3 more)
Principal Component Analysis (PCA) is a standard method for dimension reduction and the exploratory analysis of multivariate data, but a complete applied workflow often requires more than numerical eigenvectors and static plots. This article documents a set of routines in the FSDA toolbox for MATLAB that implements an integrated workflow for interactive and robust PCA graphics. The software contribution consists of coordinating classical SVD-based PCA, dynamic biplot scaling controlled by row and column parameters, score-distance and orthogonal-distance diagnostics, hard-trimming robust PCA through the Minimum Covariance Determinant and the Forward Search, linked brushing, geographical displays, and three-dimensional projection diagnostics. The implementation returns reproducible numerical output while also exposing graphical controls for changing biplot geometry, confidence regions, displayed labels, robust trimming levels, and linked selections. The methodology is illustrated on quality-of-life indicators for Italian provinces, where the software reveals variable groups, atypical provinces, robust and non-robust PCA structures, and geographical patterns. The emphasis is on statistical software: no new PCA
estimator is proposed, but established PCA and robust-PCA methodology is combined in a documented, extensible, and script-reproducible MATLAB environment.
Read the full abstract ▾
All authors Marco Riani, Anthony C. Atkinson, Aldo Corbellini, Domenico Perrotta, Francesca Torti, Gianluca Morelli
Atmospheric and Oceanic Physics
Generative modelling for multivariate spatio-temporal downscaling of monthly climate fields
Yu Huang, Shangshang Yang, Sebastian Bathiany et al. (+3 more)
Inferring
climate fields at higher spatio-temporal resolutions from coarse representations is essential for accessing socioeconomic and environmental impacts. The downscaling of modelled
climate fields faces a trade-off between data storage constraints, simulation efficiency, and resolution requirements. While many methods for spatial downscaling have been proposed, joint spatio-temporal downscaling has received comparable little
attention. When deriving, for example, daily fields from
climate datasets that are available only as monthly averages at coarser spatial resolution, it is highly challenging to faithfully reproduce extreme event statistics and spatiotemporal consistency. To address this task, we introduce a generative machine learning approach, using a
diffusion model to jointly downscale monthly temperature and precipitation fields into temporally consistent daily fields at higher spatial resolution. We demonstrate that our approach can generate high-dimensional multivariate
climate fields at daily temporal and 0.25{\deg} spatial resolution from monthly input fields at 1{\deg} spatial resolution, exhibiting coherent temporal dynamics and physically plausible spatial structures. In particular, by incorporating a conditional guidance technique, we show that a pre-trained
diffusion model can be directly leveraged to efficiently generate daily high-resolution fields conditioned on monthly inputs, without requiring additional fine-tuning or retraining. Our probabilistic approach thus provides an efficient and generalizable framework for spatio-temporal
climate downscaling, enabling flexible
climate fields generation for
climate impact assessments and digital Earth simulator constructions.
Read the full abstract ▾
All authors Yu Huang, Shangshang Yang, Sebastian Bathiany, Michael Aich, Philipp Hess, Niklas Boers
Numerical Analysis
Second-Order Flows for Approaching Stationary Points via the SAV-E Approach
Haifan Chen, Wei Liu, Ziqing Xie et al. (+1 more)
Second-order flows represent an emerging computational framework for approaching stationary points of variational energies, characterized by damped
hyperbolic partial differential equations. In this paper, we develop a scalar auxiliary variable (SAV) method for this class of flows, referred to as SAV-E. The resulting scheme is linear and decoupled, and we prove unconditional modified-energy stability. We further establish subsequential convergence to stationary points for the Ginzburg-Landau and conserved Swift-Hohenberg models under suitable assumptions. Numerical experiments support the analysis and demonstrate robustness and efficiency compared with representative baselines and the corresponding gradient flows.
Read the full abstract ▾
All authors Haifan Chen, Wei Liu, Ziqing Xie, Yifang Yuan
Machine Learning
Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction
Robin Uhrich, Marco Bagatella, Núria Armengol Urpí et al. (+2 more)
The growing interest in behavior foundation models, combined with recent advancements in generative modeling, has led to an appealing recipe for reinforcement learning from unlabeled data: pre-train a generative model to model occupancies, i.e., discounted state visitations, on reward-free trajectories, such that a policy can be easily extracted once a small reward-labeled
dataset becomes available. We decompose this recipe along three axes: (i) the degree to which the multi-modal structure of occupancies must be captured, (ii) the class of generative model, and (iii) the method used for policy extraction. Along each axis, we consider existing choices, propose simpler alternatives, and validate each decision experimentally. This analysis culminates in a novel, streamlined method that leverages drifting, a modern one-step generative sampler, to model the occupancy. We show that this enables policy extraction via direct differentiation through the occupancy model, and bypasses the need for critic
distillation. Overall, we observe that this nimble setup leads to improvements in performance on standard
benchmarks.
Read the full abstract ▾
All authors Robin Uhrich, Marco Bagatella, Núria Armengol Urpí, Pavel Kolev, Georg Martius
Computer Vision
BluePixel: Toward Native 4K Agentic Image Editing
Shanyan Guan, Shiyu Li, Xin Li et al. (+17 more)
Instruction-driven image editing has reached native 4K, yet how such editors are trained remains undisclosed, and pretraining one from scratch is beyond most teams. The task itself compounds the difficulty: a single end-to-end editor cannot make precise parametric adjustments, match specialized experts, or interpret the vague instructions users actually write. We argue that 4K editing is a system problem rather than a single-model problem, and present BluePixel, which couples a native-4K editing model with an
agent harness; the complete training recipe is disclosed in this paper. The model makes 4K training practical: it turns a 1K-pretrained editor into a native-4K one through progressive VAE adaptation that keeps 4K training as affordable as 1K, absorbs a long tail of vertical editing skills through continual learning, and refines the aesthetic-fidelity trade-off with reinforcement learning.
Distillation via source-anchored target projection (SATP) then compresses 50 denoising steps to 4, completing a 4K edit in 8.7 s on a single Kunlunxin P800 XPU while preserving task success rate and background consistency. The harness, BluePixel
Agent, makes editing controllable: it decomposes user instructions into specialist and general operations, verifies every result with the same reward model that guided training, and retries within bounded attempts. On VINS-4KEval++ (3,305 native-4K cases), BluePixel ranks second among seven models, behind only GPT Image 2 in non-edited-region consistency.
Read the full abstract ▾
All authors Shanyan Guan, Shiyu Li, Xin Li, Zhizhou Chen, Yunzhe Xu, Yuehao Liu, Ruixue Shen, Shiju Zhang, Min Jin, Qiang Liu, Junbo Liu, Penggang Shi, Sheng Sun, Xiao Wang, Yong Zhang, Zhongang Qi, Wei Li, Chao Ma, Ying Tai, Yanhao Ge
Astrophysics of Galaxies
PDRs4All XXIII. JWST-MIRI template spectra of PAHs in the Orion Bar obtained with blind signal separation
P. Amiot, O. Berné, J. Roux-Morin et al. (+12 more)
The aromatic infrared bands (AIBs) are observed in a wide variety of astrophysical sources, and now in unprecedented detail by the James Webb space
telescope (JWST). These bands are generally attributed to the emission from polycyclic aromatic hydrocarbon (PAHs). Our aim is to extract mid-infrared AIB template spectra from JWST observations of the Orion Bar photo-dissociation region that can be used to derive PAH properties from observations of other objects. We used the JWST MIRI-MRS observations from the PDRs4All program. We first isolated the AIBs from the data using a multi-component fitting approach. We then applied a non-negative
matrix factorization to the isolated AIB spectra and extracted two templates, named AIBirrad and AIBshield following previous work. The spatial distribution and spectral characteristics of these two spectra suggest that they originate from two chemical distinct populations. We interpret AIBirrad as emission originating from a PAH population strongly exposed to ultra-violet (UV) radiation and mainly composed of large, compact ionized PAHs. Conversely, AIBshield is interpreted as originating from a more shielded population of PAHs, which is less exposed to the UV by being more embedded in the cloud. This population consists of small, asymmetric and mainly neutral PAHs with bonded aliphatic side groups. These templates can serve to analyze PAH emission in a wide variety of objects observed with MIRI-MRS.
Read the full abstract ▾
All authors P. Amiot, O. Berné, J. Roux-Morin, I. Schroetter, C. Boersma, A. Fuente, J. R. Goicoechea, E. Habart, A. Maragkoudakis, T. Onaka, E. Peeters, M. Robberto, A.G.G.M Tielens, D. Van De Putte, S. Vicente
Computer Vision
CoDy: Codec Dynamics as Complementary Priors for Unsupervised Temporal Action Localization
Tianyu Yan, Qingyi Yang, Qin Yang et al. (+5 more)
Unsupervised temporal action localization (UTAL) aims to identify action categories and localize their temporal boundaries without annotation supervision. Existing methods typically derive supervision for both tasks solely from RGB and optical-flow representations, often resulting in noisy category assignments and inaccurate temporal boundaries. In this work, we reveal an overlooked property of
video codecs: group-of-pictures quantization patterns preserve category-discriminative dynamics, while motion-vector and residual variations exhibit pronounced responses around action transitions. Based on this insight, we propose CoDy, a codec dynamics-guided framework that exploits the codec as complementary information to the visual view for UTAL. CoDy integrates a codec-enhanced view with RGB and flow through confidence-aware consensus clustering, producing reliable soft pseudo-labels. It further converts codec dynamics into a class-agnostic boundary prior, which supervises visual boundary learning and guides proposal initialization and iterative refinement. The codec branches are only used during the training stage and introduce no extra codec-side computation at inference. Extensive experiments on THUMOS14 and ActivityNet v1.3 demonstrate that CoDy consistently outperforms existing UTAL methods and achieves state-of-the-art localization performance. These results establish codec signals as effective annotation-free priors for jointly inferring which actions occur and when they occur.
Read the full abstract ▾
All authors Tianyu Yan, Qingyi Yang, Qin Yang, Xufei Guo, Yuan Tang, Rui Wang, Yixue Hao, Long Hu
High Energy Astrophysical Phenomena
Black Hole Binary Mergers in Galactic Nuclei: Combined Effects of Secular Dynamics and Stellar Flybys
Hao Wang, Bin Liu and Johan Samsing
Galactic
nuclei, with the central supermassive black holes (SMBHs) and surrounding dense
star clusters, serve as natural laboratories for studying gravitational dynamics in extreme gravitational environments. In this work, we model the long term
evolution of black hole binaries (BHBs) in these environments by incorporating the tidal influence of the central SMBH, the potential of the nuclear
star cluster, SMBH induced relativistic precession, and
stochastic perturbations from stellar flybys. Our numerical simulations show that the long term
evolution of BHBs is predominantly governed by the SMBH-driven ZLK mechanism. However, once stellar flybys are included, they can modify both the binary angular momentum and energy. The cumulative effect of repeated flyby encounters substantially reshapes the long term dynamical
evolution, eventually leading systems toward either merger or disruption. Based on a population synthesis study, we find that this combined secular and non-secular channel produces a merger rate of approximately $\mathcal{R}\simeq16\mathrm{Gpc^{-3}yr^{-1}}$, which is significantly higher than the rates predicted by dynamical channels that neglect stellar flybys. Furthermore, we predict that a non-negligible fraction of merging systems would retain detectable eccentricities at low frequencies around 5 Hz, thus providing a distinctive observational signature for next generation ground based GW detectors. Our results demonstrate that the
stochastic perturbations from the surrounding environment are essential for modeling BHB
evolution in galactic
nuclei, and that this channel may contribute a major fraction of observed mergers.
Read the full abstract ▾
Machine Learning
Neural Wasserstein Splines: Representation-Invariant Second-Order Priors for Trajectory Inference
Amartya Banerjee, Harlin Lee and Caroline Moosmüller
Population dynamics are often observed only as unpaired snapshots, which leave the
evolution between measurements underdetermined. Second-order regularization can constrain it, but a penalty on the accelerations of sampled particles is a prior on joint particle paths, and different particle representations of the same marginal curve can have different acceleration costs. We therefore introduce Neural Wasserstein Splines (NWSP), a sample-based method that learns a continuous population curve under a Wasserstein smoothness prior. A neural network maps latent variables to the control points of a cubic spline in time, so samples at any time come from the spline basis without numerical integration. We project the generator's velocity onto gradients of a finite dictionary of scalar potentials and differentiate this measure-dependent projection in time, obtaining a finite-dimensional surrogate for the intrinsic Wasserstein acceleration, trained jointly with distributional fit losses. For a fixed dictionary and ridge parameter, we prove that the population penalty depends only on the marginal curve and bound its sensitivity to moment-estimation errors. Learned potentials (NWSP-N) sharpen acceleration estimates. On various synthetic and single-
cell benchmarks, NWSP is highly competitive with transport and flow-based methods, ranking in the top two in 11 of 16 settings. The penalty improves on an unregularized spline in 13 of 16 settings and, while competitive with a particle-acceleration penalty, adds provable representation invariance.
Read the full abstract ▾
High Energy Astrophysical Phenomena
Linear Polarization of Synchrotron Radiation from Very-small-pitch-angle Electrons
Meng Nie and Jirong Mao
For synchrotron radiation with pitch angle $\alpha \ll 1/\gamma$ where $\gamma$ is the electron Lorentz factor, referred to as very-small-pitch-angle synchrotron radiation, the polarization properties differ significantly from those of conventional synchrotron radiation ($\alpha \gg 1/\gamma$). In this work, we investigate the linear polarization characteristics of very-small-pitch-angle synchrotron radiation from relativistic thermal and nonthermal electrons. We calculate the linear polarization by applying power-law and Maxwellian electron energy distributions. Moreover, some synchrotron sources exhibit a hybrid distribution of nonthermal and thermal electrons, and we also calculate the polarization degree for a hybrid electron energy distribution. The calculations are performed in the
radio, optical, X-ray, and $\gamma$-ray bands. Our results show that the linear polarization degree of very-small-pitch-angle synchrotron radiation is generally lower than that of conventional synchrotron radiation. Very-small-pitch-angle synchrotron radiation from both nonthermal and thermal electrons can produce significant linear polarization. However, the polarization exhibits diverse behaviors across different physical parameters. The general scenario provided in this work can be considered for the application to all kinds of synchrotron sources, as we suggest that the very-small-pitch-angle case is crucial for the study of synchrotron polarization.
Read the full abstract ▾
Machine Learning
SPARS: A Stability-Plasticity-Aware Region Selection Framework for Continual Learning
Teif Aldaajani, Huda Hakami and Maha A. Thafar
Continual learning aims to acquire new knowledge while retaining previously learned knowledge. However, updating shared parameters when learning new tasks often overwrites existing knowledge, causing catastrophic forgetting.
Memory replay methods mitigate catastrophic forgetting but rely on previous samples, raising
privacy,
memory, and scalability concerns. On the other hand, parameter isolation-based methods preserve learned knowledge but may limit knowledge transfer, whereas regularization promotes adaptability but remains susceptible to interference between old and new tasks. To maintain the stability-plasticity trade-off, this paper proposes Stability-Plasticity-Aware Region Selection (SPARS), a replay-free hybrid framework combining global parameter isolation with localized parameter regularization. SPARS estimates parameter importance using either the Fisher Information
Matrix (FIM) or a novel Confidence Score (CS)
estimator that leverages logit sensitivity and achieves comparable performance to FIM without requiring labelled data. Based on parameter importance, SPARS partitions parameters into three importance-aware regions namely, winning, neighboring, and non-important parameter regions. Each region receives a different adaptation strategy: selective freezing, soft regularization, or unrestricted learning based on its role in the stability-plasticity trade-off. To mitigate classifier bias in class-incremental learning, we employ Quadratic Discriminant Analysis to improve performance on previously learned classes .Experiments on CIFAR-10 and TinyImageNet demonstrate consistent improvement in average incremental accuracy over existing isolation- and regularization-based methods across task- and class-incremental learning. On TinyImageNet, SPARS achieves the best performance with 6% improvement for class-incremental learning while retaining over 60% of parameters for future tasks.
Read the full abstract ▾
Representation Theory
Uniqueness of twisted linear periods over Archimedean division algebras
Yufeng Li and Junyan Xiao
Let $\DD$ be $\R$, $\C$, or $\HH$, and let $H=\GL_p(\DD)\times\GL_q(\DD)$ be the block diagonal subgroup of $G=\GL_{p+q}(\DD)$. We prove that $\dim\Hom_H(\pi,\eta)\leq1$ for every irreducible Casselman–Wallach representation $\pi$ of $G$ and every smooth character $\eta$ of $H$.
Statistical Methodology
Principal Component Analysis for Zero-Inflated Probability Measures via Lorenz Curves
Giacomo Aletti and Enea G. Bongiorno
This paper develops a principal component analysis framework for probability measures associated with non-negative real-valued random variables with structural zeros. Specifically, leveraging the associated Lorenz Curve (LC) and a suitable perspective transformation of the plane, a bijective
embedding is defined to treat such measures as functional data in a
Hilbert space. Once the random LC is embedded in this space, standard Functional Principal Component Analysis (FPCA) becomes directly applicable, and the diffeomorphic nature of the
embedding allows the main convergence results and their interpretations to be transferred back to the original space of LC. The a.s. and $L^2(\mathbb{P})$ consistency of the estimated eigenfunctions,
eigenvalues, and modes of variation is subsequently investigated, both in the embedded space and, by virtue of the
embedding's diffeomorphism property, in the original LC space. Crucially, this consistency analysis accounts for the double stochasticity of the estimation problem: one level of randomness is tied to the underlying LC process itself, while the other stems from the empirical estimation of the LC from sample data. As a consequence, consistency of the Karhunen–Loève reconstructions is also established, guaranteeing faithful approximation of individual curves in both spaces. Finally, the proposed FPCA methodology is illustrated on a real-world
dataset of global export flows and compared with a naive FPCA approach.
Read the full abstract ▾
Quantum Physics
Acoustic scattering from quantum dynamics
Farhang Loran and Ali Mostafazadeh
Scattering of sound waves propagating in a compressible fluid is linked to the
quantum dynamics generated by a class of non-Hermitian
Hamiltonian operators. The scattering amplitude is expressed in terms of the corresponding
evolution operator. This is used to provide a new proof of the reciprocity
theorem in acoustic scattering, conduct analytic low-frequency acoustic scattering calculations, and study the scattering of sound waves by a finite planar array of delta-function point scatters embedded in an infinite acoustic slab. A detailed discussion of acoustic scattering properties of a homogeneous acoustic slab including a pair of delta-function point scatters on one of its faces is provided.
Read the full abstract ▾
Computation and Language
RCCM: Diagnosing Gains and Harms in Reasoning-Prompt Transfer
Yuqi Zhang and Di Zhang
A
reasoning prompt can improve one deployment and degrade another. RCCM (
Reasoning-Control Capability Mismatch) examines these reversals by pairing answers to the same questions with records of how generation ends. Across five open-weight deployments and nine strategies, aggregate accuracy hides substantial question-level gains and harms; a common-budget MATH500 evaluation examines transfer on a separate sample. On 60 additional questions, concise
reasoning improves Phi's automatic answer matching at a 2,048-
token cap across three collections, but sharply reduces Qwen's. Author-reviewed final-answer commitments give a smaller, statistically unresolved Phi gain. Reserving an answer stage also improves matching, with additional input and call costs. These results show why a
prompt benefit must be interpreted alongside its execution and evaluation: matching a reference answer need not mean delivering it, and an apparent compression benefit may admit another explanation. RCCM records this evidence for reproducible transfer comparisons.
Read the full abstract ▾
Computation and Language
Rationale Distribution Distillation via Flow Matching for Distributional Continuous Latent Reasoning
Mingyeong Kim, Gyusang Cho, Juho Lee et al. (+1 more)
Sampling multiple chain-of-thought rationales from a
language model can uncover correct solutions missed by a single attempt. However, these rationales are typically lengthy, making repeated autoregressive generation costly. Continuous latent
reasoning offers a compact alternative, but point-valued methods condition repeated answer sampling on a single rationale, which can limit exploration of diverse paths and answer coverage. We thus propose rationale distribution
distillation (RDD), training a conditional flow over fixed-length latent sequences by distilling from multiple textual
reasoning paths. Furthermore, we expand the
distillation targets through within-query mixup, interpolating between complete latent rationales for the same query. The flow nonautoregressively generates the entire latent sequence, and a frozen
language model autoregressively decodes answers from sampled rationale contexts. Empirically, the resulting distributional reasoner achieves higher answer coverage than the evaluated point-valued latent
reasoning baselines, particularly on advanced mathematical tasks.
Read the full abstract ▾
All authors Mingyeong Kim, Gyusang Cho, Juho Lee, Giung Nam
Statistical Mechanics
Understanding the Mpemba Effect through a Rule of Thumb
Ran Yaacoby
In the Mpemba effect, a system prepared farther from
equilibrium relaxes faster than one prepared closer to it. Its persistence across diverse systems, and under different dynamics of the same system, suggests that it is largely rooted in
equilibrium properties, with the dynamics deciding only which mode relaxes slowest. Here we propose a simple rule of thumb for anticipating the Mpemba effect: the effect is expected when a single low-order
equilibrium moment is nonmonotonic in temperature. The bath singles out this moment as the slowest to relax, so that its displacement from its bath value sets the amplitude of the slowest relaxation mode. We establish this relation exactly for low-dimensional Ornstein–Uhlenbeck processes. We then extend the rule to mean-field models by approximating their relaxation near the final
equilibrium state by an Ornstein–Uhlenbeck process. For the mean-field Ising antiferromagnet, which we solve exactly in the thermodynamic limit, this approximation reproduces the main structure of the Mpemba phase diagram, including strong effects far from any phase transition. Finally, we give a general prescription for mean-field systems: identify the slowest
symmetry-allowed moment of the normal coordinates at the bath, and follow its
equilibrium value as a function of preparation temperature. This offers a practical route from
equilibrium observables to anomalous relaxation without diagonalizing the dynamics.
Read the full abstract ▾
Sound
SEA-HELM Audio: Evaluating Multilingual ASR-LLM Pipeline Performance Across SEA-HELM Subsets
Yixian Tan, Aulia Adila, Jonathan Heng et al. (+3 more)
We introduce SEA-HELM Audio, a Southeast Asian (SEA) multilingual evaluation framework that extends the text-based SEA-HELM
benchmark to assess performance degradation in cascaded automatic
speech recognition and large
language model (ASR-LLM) pipelines. Using synthesized
speech, we compare
speech-input and text-input performance across five
languages (Chinese, English, Malay, Indonesian, and Tamil) on multiple-choice
reasoning and instruction-following tasks. We evaluate open-weight and commercial ASR-LLM systems using identical verifiers and quantify the
speech gap as the difference between text and
speech accuracy. Results show substantial degradation, with accuracy drops of 9.00 to 18.50 percentage points for multiple-choice
reasoning and 20.00 to 42.86 points for instruction following. Tamil exhibits the largest degradation, exceeding 40 percentage points. Our findings highlight the limitations of text-only
benchmarks in predicting voice-
agent reliability and demonstrate the value of SEA-HELM Audio for evaluating multilingual voice-
agent systems.
Read the full abstract ▾
All authors Yixian Tan, Aulia Adila, Jonathan Heng, Kunat Pipatanakul, Donghang Wu, Anocha Sutaveephamochanon
Artificial Intelligence
Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable Rewards
Tianyu Liu, Davide Belli, Bence Major et al. (+2 more)
Training LLM
agents on long-horizon interactive tasks with reinforcement learning typically relies on human-annotated tasks and reliable verification, both expensive to construct for stateful environments. Self-play offers an alternative, but must generate achievable, compositional tasks and provide feedback beyond binary outcomes. We propose Explorer-Zero, a self-play framework that jointly generates tasks and fine-grained, verifiable rewards through environment interaction. A learned proposer explores the environment, constructs tasks, and records intermediate and final states as verification targets. These checkpoints reward partial completion, while alternating proposer–solver training adapts the generated curriculum. We show theoretically that subtask rewards reduce collapse among unsuccessful rollouts when they complete different numbers of subtasks. Without human-annotated training tasks, Explorer-Zero outperforms the evaluated self-play baselines across ALFWorld, ScienceWorld, and AppWorld; on AppWorld, our 30B
agent reaches 59.67% task success, approaching 62.65% for RL on human-annotated tasks. Further analyses support task-relevant composition and benefits on complex tasks, while self-play initialization improves subsequent human-data RL across all three
benchmarks.
Read the full abstract ▾
All authors Tianyu Liu, Davide Belli, Bence Major, Arash Behboodi, Tribhuvanesh Orekondy
Social and Information Networks
From Content Sharing to Opinion Shaping: A Theoretical Framework for Cascade-Driven Opinion Dynamics
Elisabetta Biondi, Chiara Boldrini, Andrea Passarella et al. (+1 more)
Online social networks (OSNs) tightly couple information
diffusion and opinion formation, yet most existing models treat these processes separately. Building on the Friedkin-Johnsen on Cascades (FJC) model, which integrates opinion updating within
stochastic information cascades, we advance its analytical characterization and systematic validation. We derive an exact conditional layer-wise recursion that reveals the statistical coupling between
diffusion realizations and opinion
evolution, showing that a closed-form recursion in unconditional expectations is not available. To overcome this limitation, we introduce a mean-field decoupling approximation (MFA) that yields a tractable linear recursion for the expected opinion trajectory. We also develop scalable tree-based reductions that drastically reduce computational complexity while preserving structural
diffusion properties, and validate the analytical recursion and its approximations through extensive simulations. Using
Stochastic Block Model networks with controlled initial opinion configurations, we show that, unlike the classical Friedkin-Johnsen model, which typically leads to global homogenization, the FJC dynamics generate richer regimes of bounded assimilation and sustained polarization. Overall, this work bridges
diffusion and influence modeling, providing a theoretical foundation for understanding how cascade-driven interactions reshape collective opinion landscapes in structured social networks.
Read the full abstract ▾
All authors Elisabetta Biondi, Chiara Boldrini, Andrea Passarella, Marco Conti
Data Structures and Algorithms
Optimal Online Stochastic Matching
Paul Duetting, Michal Feldman and Mathieu Molina
We resolve the problem of unweighted online
stochastic matching with a known arrival law by establishing its optimal competitive ratio. For Poisson arrivals, we achieve $1-e/e^e\approx0.8206$, matching the upper bound of Chierichetti, Giacchini, Panconesi, and Vattani [ICALP 2025]. For exactly $n$ i.i.d.~arrivals, we obtain the same guarantee up to an additive $O(n^{-1/2})$ loss. Our key technical contribution is an exact correspondence between Poisson matching and exponential scheduling, preserving both online performance and the expected offline optimum. To prove the latter, we show that the distribution of the maximum of a random supermodular set function is determined by its joint distributions along chains of nested sets, a result of independent interest. This reduces the analysis to chains, where classical Poisson–exponential duality applies. The offline equivalence extends to arbitrary nonnegative edge weights. We then design a posterior allocation policy for exponential scheduling and use the correspondence to transfer its optimal competitive guarantee to online matching. Additionally, we show that the optimal competitive ratio under an unknown arrival law is strictly lower than $1-e/e^e$, establishing a strict separation between the known and unknown arrival models.
Read the full abstract ▾
Earth and Planetary Astrophysics
LiDO: The Large inclination Distant Objects Survey
Mike Alexandersen, Rosemary E. Pike, Ying-Tung Chen et al. (+20 more)
Previous trans-Neptunian objects (TNOs) discovery surveys targeted the ecliptic plane, where the overall highest density is seen. High-inclination objects spend little time near the plane, and are thus under-represented in previous TNO surveys. The Large inclination Distant Objects (LiDO) survey targets these high-inclination populations in 34 sq$.$deg., located in two blocks 14-21 degrees north and south of the ecliptic plane, finding 148 objects, all with i>14 degrees due to survey design. A careful characterization of detection efficiency and other survey parameters results in a "characterized sample" of 125 objects detected within the survey's magnitude and rate limits, thus suitable for model comparison via survey simulations. LiDO detected 62 TNOs classified as resonant, including 21 Plutinos, of which 12 are securely Kozai resonators (always coming to pericenter off-ecliptic), as well as 9 objects in the 5:2 resonance, and several mixed mode inclination-eccentricity resonators. No objects beyond 63 au were detected, despite sensitivity to objects as far as 150 au. We found that the i>14 degree orbital distribution of the previous model from OSSOS was rejectable with the inclusion of the 38 characterized hot classical TNOs from LiDO. We constructed a proof-of-concept orbital distribution model of hot classical TNOs based on long-term stability sculpting, which is statistically acceptable for the LiDO and OSSOS samples, while producing population estimates (21,000$^{+3000}_{-3000}$ hot classicals with H_r<8.3) consistent with OSSOS estimates. The discovery of 125 characterized mid-to-high inclination objects in a variety of orbital classifications provides useful constraints for population modeling and the migration history of the Solar System.
Read the full abstract ▾
All authors Mike Alexandersen, Rosemary E. Pike, Ying-Tung Chen, Samantha M. Lawler, Mark Comte, Lowell Peltier, J. J. Kavelaars, Kathryn Volk, Cameron Semenchuck, Cameron Collyer, Jaidyn Draper, Stephen Gwyn, Hermosillo Ruiz, Matthew J. Holman, Katrina Kianpoor, Cameron J. Morgan, Ruth Murray-Clay, Jonathan Pond, Benjamin Proudfoot, AbdulRahman Rabaa, Darin Ragozzine, Emily Rieger, Mriana Yadkoo
Artificial Intelligence
Euler: Towards Human-Centered Multi-Agent Mathematical Discovery
Qiuyu Chen, Qi Wang, Liang Xu et al. (+3 more)
Mathematical research drives scientific discovery, and recent advances in AI for mathematics have shown that
language-model
agents can solve challenging problems and develop substantial mathematical proofs. Yet much of today's AI-for-mathematics work centers on obtaining a final answer. This emphasis leaves the exploration behind a result difficult for researchers to understand, while long-running investigations face fragmented context and the risk of unsupported claims propagating across
reasoning steps. We present Euler, a human-centered mathematical research system built around a shared, structured representation of the research process. Euler organizes mathematical statements, assumptions, proof dependencies, alternative approaches, failed attempts, and verification evidence into connected research graphs. This representation allows researchers to trace how results are developed and supplies
agents with persistent, task-relevant knowledge for planning, collaboration, and verification across sessions. Researchers can also select a graph node linked to a saved team state, revisit the
agents' earlier
reasoning, and explore alternative directions through continued interaction. We illustrate Euler's capabilities through three mathematical research case studies and describe the implemented system and its software validation.
Read the full abstract ▾
All authors Qiuyu Chen, Qi Wang, Liang Xu, Xin Jin, Baao Xie, Wenjun Zeng
Multiagent Systems
When Extremists Enter LLM Societies: Investigating Extreme Behavior in Multi-Agent Systems
Guo Chen, Huaijin Xie, Ziwen Li et al. (+5 more)
Large
language model (LLM)
agents increasingly serve as computational actors in multi-
agent social simulations. While existing work emphasizes representative personas and average-case behavior, social science shows that persistent tail actors can disproportionately shape collective dynamics. We study this regularity in controlled LLM-based multi-
agent systems (LLM-MAS). In these systems, such actors cannot be reduced to fixed scalar states:
prompts, natural-
language histories, and backbone generation mediate their behavior and influence. We first analyzed three properties of extremists in sociology and model the LLM extremist based on LLM-native perspective. We introduce EXAM, an EXtremist
Agent Modeling framework that capable of configuring extremist
agents for specific tasks and precisely controlling their population proportion, spatial distribution, interaction protocols, and measurement interfaces. Across four scenarios: congressional election, financial crisis, epidemic public opinion, and iterated prisoner's dilemma, we observe that small persistent tail groups produce measurable, backbone- and scenario-dependent collective shifts. EXAM shows how this social-science regularity manifests through
language-mediated mechanisms specific to LLM-based social simulation.
Read the full abstract ▾
All authors Guo Chen, Huaijin Xie, Ziwen Li, Reed Li, Yu Lu, Haibo Shi, Bingbing Xu, Junjie Huang
Cosmology and Nongalactic Astrophysics
specwizard: A Public Python Framework for Creating Synthetic Absorption Spectra from Hydrodynamical Simulations
Andrés Arámburo-García, Tom Theuns, Joop Schaye et al. (+1 more)
We present specwizard, a publicly available modular Python package for generating synthetic absorption spectra from hydrodynamical simulations. The code combines atomic data from the National Institute of Standards and Technology database with ionisation fractions computed using the cloudy spectral synthesis code for multiple ionising backgrounds, enabling consistent modelling of bound–bound and bound–free absorption for resonant transitions of all ions. specwizard employs a column-density–conserving projection scheme, includes natural line broadening for the Lyman series, and offers optional treatments of instrumental resolution and noise for direct comparison with high-resolution
quasar spectra. As example applications of the code, we shoot sightlines through the eagle cosmological simulation to generate a z = 3.2 synthetic
quasar spectrum containing H i Lyman-series transitions, including damping wings and Lyman-limit absorption. Additionally, we compute metal-absorption lines probing the circumgalactic medium of simulated
galaxies, and investigate the correlated absorption of H i and Si iii. We demonstrate that specwizard applied to outputs from eagle accurately reproduces key observational diagnostics of the Lyman-alpha forest, including the
redshift evolution of the mean transmission, the transmission PDF and the transmission power spectrum. We take advantage of the fact that the eagle simulation also reproduces the measured
evolution of the temperature-density relation to infer the
evolution of the amplitude of the ionizing background, finding excellent agreement with recent theoretical models and measurements.
Read the full abstract ▾
All authors Andrés Arámburo-García, Tom Theuns, Joop Schaye, Alexey Boyarsky