Chaotic Dynamics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 28 August
Nonlinear dynamical systems often evolve on low-dimensional manifolds with intricate geometry, making them difficult to approximate with a single global reduced-order model (ROM). To address this, we previously introduced quantized local reduced-order models (ql-ROMs), which partition the solution manifold into $K$ clusters and construct local ROMs around the corresponding centroids. In the sharp formulation, discontinuous switching between clusters can induce discontinuities, spurious oscillations, and loss of accuracy near cluster boundaries. Here, we propose a fuzzy extension of ql-ROMs based on fuzzy c-means clustering. The resulting fuzzy local ROM (fl-ROM) advances all local reduced states in parallel and combines their predictions through continuous membership weights, yielding a smooth partition-of-unity coupling of the local dynamics. We also introduce a pseudo-Bayesian information criterion to guide the selection of the number of clusters in the fuzzy setting. The methodology is assessed on the two-dimensional Kuramoto-Sivashinsky equation in periodic, travelling-wave-like, quasi-periodic, and chaotic regimes. In all cases, the local ROMs outperform the global ROM, while the fl-ROM provides smoother and generally more accurate predictions than the sharp formulation. In the chaotic regime, where long-time trajectory tracking is inherently limited, the fl-ROM achieves the best compromise between short-time predictive accuracy and long-term statistics. These results show that fl-ROM provides an effective, robust, and interpretable framework for reduced-order modeling of nonlinear dynamical systems with regime transitions and complex attractor geometry.
General Physics
Bianchi Type I Space -Time Geometry of the Universe with Time Dependent G and $\Lambda$ Within the Framework of General Relativity: Observational Aspects
Inspired by the latest progress in the hunt of acceptable cosmological model of the universe, present paper is devoted to explore a mathematical model of the universe having initial anisotropy and with accelerated evolution that attains isotropic character. We have considered a widely accepted anisotropic Bianchi type I space-time geometry of the universe to explore a physically viable model. To find an acceptable model, we adopted a hyperbolic form for the scale factor as $ a(t)= \left( sinh bt \right)^{\frac{1}{n}} $ and a relation $ \Lambda=\Lambda_0 \frac{\ddot{a}}{a}$. The model parameters are constrained by $ \chi^2$ minimization techniques. Using 35 CC measurements we obtain present day expansion rate $H_0= 65.715 Km/S/Mpc$ and the value of $ n= 1.3024$. Further to estimate the local uncertainties of the best fit parameter, we calculate covariance from the Jacobin of the normalized residuals. The corresponding 1 $\sigma$ uncertainties are $ H_0=65.715 \pm 2.392 Km/S/Mpc$ and $n= 1.3024 \pm 0.0842$. The physical and dynamical behaviour of present model has been discussed by a graphical representation of cosmological parameters. The observational constraints on the expansion history are found to be consistent with a present accelerated phase and a transition from deceleration to acceleration at an intermediate redshift. The corresponding evolution of the derived cosmological quantities is also investigated within the adopted model.
Read the full abstract
General Physics
Influence of Generalized Ghost Dark Energy on Wormhole Geometry
This article presents a wormhole solution constructed from a generalized ghost dark energy (GGDE) source in general relativity. At first, a brief review of GGDE is presented along with the necessary mathematical frameworks. We employed the Markov Chain Monte Carlo (MCMC) technique to constrain the free parameters of the model using the CC+BAO, $Pantheon^+$, and their combined datasets. Furthermore, the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to statistically assess the model's performance and determine its level of acceptance. Next, the basics of wormhole geometries along with the thin-shell formalism are explained in detail. After that, three different wormhole solutions associated with three different choices of the redshift function are presented ,and their various geometric properties as well as energy conditions are studied graphically. Finally, the stability of these thin–shell structures is examined analytically as well as graphically by studying the associated effective potential.
Read the full abstract
Mesoscale and Nanoscale Physics
Coherent phononic frequency combs in ferroelectric CMOS oxides
Modern electronic systems require tens of clock and carrier frequencies, each synthesized by a dedicated phase-locked loop from a shared reference, imposing routing, power and synchronization burdens that grow with every domain. Optical frequency combs solved this problem in photonics, whereas electronics has lacked an equivalent source in its native radiofrequency domain. Here we report broadband phononic frequency combs in ferroelectric hafnia-zirconia nanoelectromechanical resonators built from complementary metal-oxide-semiconductor (CMOS) oxides. Lithographically defined detuning of a 2:1 internal resonance selects the generation mechanism: two-tone-seeded wave mixing yields more than 170 lines distributed over two octaves, with mutual coherence verified for representative pump and generated lines, whereas an autonomous Hopf route yields hierarchical combs of more than 200 lines through torus and period-doubling dynamics, in agreement with slow-flow bifurcation theory. Geometric scaling extends comb generation across 0.44 GHz to 2.1 GHz. Heterodyne measurements, analyzed using the modified Allan deviation (MDEV), establish a two-timescale law. The mechanism governs short-term stability: seeded combs inherit the white-phase-noise scaling of their pumps, whereas autonomous combs acquire the phase diffusion of a free-running oscillator, with the one-second MDEV increasing from 1e-11 to 1e-8. The material governs long-term stability: in air and without active thermal control, the temperature-compensated stack suppresses the random-walk drift that dominates uncompensated resonators. These results establish mechanism- and material-level design rules for operating a single resonator as chip-scale frequency infrastructure, from multi-clock generation to radiofrequency parallel processing.
Read the full abstract
Social and Information Networks
Algorithmic Simplification for Million-Vertex Diffusion History Reconstruction
Diffusion history reconstruction infers latent node states between sparse observations of SI or SIR processes. HERMES combines parameter fitting, a learned graph-neural proposal, and feasibility-aware Markov chain Monte Carlo. We remove these stages one at a time and evaluate each version on all 12 canonical datasets. The final method uses deterministic mean-field forward-backward inference, threshold decoding, and fixed rates. This fixed-rate variant, Battus-Z, achieves mean macro-F1 of 0.8726 and NRMSE of 0.1010, compared with published HERMES aggregates of 0.8692 and 0.1483. The benchmark pins the final observed frame before scoring, so we also exclude all observed frames. Under this metric, Battus-Z obtains macro-F1 0.8431 and NRMSE 0.1181. Thus, the learned proposal, MCMC, and fitting stages can be removed while retaining the published aggregate quality on the evaluated HERMES benchmark and scoring protocol. A CUDA implementation processes generated histories with up to 4.84M vertices on LiveJournal and 117M edges on Orkut. On the same CUDA backend, Battus-Z reduces the geometric-mean algorithm interval relative to fitted Battus by 5.1x for SI and 20.3x for SIR. Its event-weighted causal-violation rates are 7.50% for SI and 8.77% for SIR; graph-constrained decoding remains future work.
Read the full abstract
Analysis of PDEs
Analysis validation of a new continuum model for the evolution of grain boundaries in polycrystalline materials
To describe the evolution of grain boundaries based on the underlying microscopic mechanisms of line defects (disconnections) and the integrated effects of a diverse range of thermodynamic driving forces, Zhang, et al. in 2017 formulated a continuum equation. We prove the global-in-time existence, uniqueness, and regularity of the weak solution to an initial-boundary value problem for this model. The existence, uniqueness of the stationary solution are also established. Finally, we investigate the large-time behavior of the weak solution of the evolution problem, and show that the solution converges to the stationary solution in a suitable sense. Numerical simulations are carried out to validate the analysis results. The main difficulties in the proof of main theorems are due to a non-local term with singularity, a non-smooth coefficient of the highest derivative associated with the gradient of the unknown, and the special form of free energy which is not uniformly bounded from below. The key ingredients in the proof are the energy method, an estimate for a singular integral of the Hilbert type, Fourier transform of Hilbert transformations, and an estimate with a weight, for time-derivative of the unknown.
Read the full abstract
Materials Science
Giant staggered Dzyaloshinskii-Moriya vectors emerged in synthetic antiferromagnets with spatially alternating stress
Dzyaloshinskii-Moriya (DM) interaction is a source of chiral magnetic physics, and it manifests microstructural symmetry which could link to exotic magnetic and electronic properties such as altermagnetism. For a new paradigm in magnetism and spintronics, it is crucial to have control on the DM interaction as well as ordering of the DM vectors which characterize the DM interaction. Here, we report on a strong interlayer DM interaction emerged in Co/Ru magnetic superlattices whose structural symmetry is unambiguously broken by alternating mechanical strains developed using the peculiar technique of our stress mechanism. Moreover, theoretical analyses reveal that our magnetic superlattices host a staggered order of the DM vectors. Our results indicate that the spatially varying strains with our breakthrough technique are quite effective in breaking the symmetry of the system and prove that it can control the DM interaction strength as well as the DM vector order.
Read the full abstract
Numerical Analysis
Minimizing the Arithmetic and Communication Complexity of Jacobi's Method for Eigenvalues and Singular Values: Part Two – Parallel Algorithms
This paper presents several parallel versions of Jacobi's method for the symmetric eigenvalue problem and the SVD. A continuation of [Demmel, Luo, Schneider, & Wang 2025], we develop parallel Jacobi algorithms whose arithmetic cost is optimal and whose bandwidth or latency can match the corresponding lower bounds of parallel matrix multiplication. Our focus is a standard distributed-memory setting with variable processor layouts, including both 2D and 2.5D processor grids. In the 2D case, we demonstrate that a standard implementation of parallel Jacobi achieves a perfect speedup in arithmetic cost – i.e., complexity $O(n^3/P)$ when done with $P$ processors – while hitting the 2D matrix-multiplication lower bound for bandwidth and (nearly) the lower bound for latency. By employing a 2.5D processor grid and leveraging 2.5D matrix multiplication, equivalently by increasing the memory per processor, we demonstrate that parallel Jacobi can achieve even lower bandwidth/latency, though we also prove that these costs cannot simultaneously match the best-known bounds for parallel matrix multiplication in any Jacobi algorithm. Finally, we extend our results to one-sided Jacobi SVD.
Read the full abstract
High Energy Astrophysical Phenomena
Anomalous horizontal track-like events at the HAWC observatory: candidate neutrino-induced charged leptons from the Pico de Orizaba volcano
Two track-like events with exceptionally large charge deposits, pointing back to the region of maximum overburden ($>18$ km.w.e.) of the Pico de Orizaba volcano, were reported by the HAWC Collaboration in a search for Earth-skimming neutrinos. Using exclusively published information, including the published Monte Carlo charge distributions and the scattered-muon background model, we show that these events are kinematically inconsistent with all identified backgrounds. A conservative energy-conservation bound places the in-tank energy deposit of the candidates above $\approx$190–210~GeV, with a worst-case minimum of 122–136~GeV, in either case exceeding the 100~GeV maximum of the scattered-muon background model. Calibration systematics derived from the published simulations act unidirectionally, pushing the implied deposits into the multi-TeV regime. Direct atmospheric muons require $E_\mu \gtrsim 0.67$~PeV to penetrate the overburden and contribute $4.5\times10^{-5}$ expected events ($\lesssim 10^{-3}$ under the largest possible prompt-charm flux); collinear muon bundles are excluded by five independent arguments. Conversely, the neutrino hypothesis predicts identifiable events precisely in the observed charge window ($10^{3}$–$10^{4}$ photoelectrons) and in the highest-overburden analysis region, both observed. The two track-like events imply a rate that exceeds the conventional atmospheric neutrino flux expectation at the 2.1–2.5$\sigma$ level. The charges of the observed candidates require an emerging lepton above about 20~TeV; the published calibration does not bound the energy from above. These would constitute the first neutrino candidates detected in Mexico and the first obtained with the Earth-skimming volcano technique.
Read the full abstract
Human-Computer Interaction
The Web-CLI: Verifiable Privacy for Tools, Models, and Inference Engines in the Browser
We introduce the Web-CLI, a novel application architecture deploying powerful computational capabilities (command-line tools compiled to WebAssembly, models run through client-side inference runtimes, and GPU-accelerated engines) as zero-install, offline-capable browser applications that preserve full underlying capability. Unlike web-based alternatives that require server-side processing and expose user data to third parties, Web-CLI applications execute entirely on the client, providing a verifiable privacy guarantee by architecture rather than policy. We define the pattern and its four properties: fidelity, progressive disclosure, offline-first, and zero egress. We present four reference implementations across distinct domains: ffmpeg-webCLI, a browser-based video editor built on FFmpeg; whisper-webCLI, speech transcription via Transformers.js; chat-webCLI, WebLLM-based language model inference; and 3mf-webCLI, a deterministic tool segmenting 3D models into multi-material files for physical 3D printing. Together they demonstrate that the pattern generalizes across deterministic media processing, neural speech recognition, LLM inference, and geometry processing with a physical output, and we outline how it extends to AI-native interfaces in which a local language model becomes the command surface itself. We further report early, anecdotal signs of independent reuse by third-party tools, suggesting the pattern generalizes beyond its reference implementations. We evaluate the primary implementation against native FFmpeg on performance and feature parity, and argue that progressive disclosure lowers the barrier for non-technical users. We argue that for applications processing sensitive user data (medical, legal, journalistic, or personal), the Web-CLI should be the default architecture, as it makes data locality an independently verifiable technical property rather than a policy promise.
Read the full abstract
Machine Learning (Statistics)
The information geometry of product-reference discrete diffusion: Interaction growth complexity and optimal scheduling
We study a class of product-reference diffusion algorithms for sampling from a discrete distribution. We show that their sampling performance can be characterized using a path-based measure of data geometry that we call the interaction growth complexity (IGC). We show that a bivariate IGC kernel gives an exact representation of both the KL discretization error and a simple one-step upper bound. The simpler univariate IGC density can be used to study the effect of stepsize choices on the iteration complexity required to obtain $\epsilon$-accurate samples in KL divergence. Samplers that traverse the path with equi-spaced steps in log-squared-reliability-odds have performance that depends on the aggregate IGC mass, whereas refined choices of stepsizes have a lower complexity depending on a square-root functional. In the fine-grid limit, both of these characterizations become sharp. We also allow general product reference distributions and show that the reference law can substantially reshape the IGC profile and the resulting sampling complexity; in particular, references far from both the uniform and the data marginals can yield dimension-dependent improvements. Finally, the aggregate IGC mass admits bounds in terms of total correlation and dual total correlation, thereby connecting the pathwise geometry to classical measures of multivariate dependence.
Read the full abstract
Machine Learning
From the Loss Landscape to Diverse Feature Learning in Neural Networks
Over the course of the last decade, neural networks have grown from an academic curiosity to moving the markets of nations. Despite this explosion in both research and deployment, relatively little is understood about how they achieve the solutions they do. This is both scientifically relevant, and pressing for society. When neural networks make decisions across self-driving, construction, law, hiring and health, there have been and will continue to be unintended consequences. However, attempting to generalize the failures of the largest and most important production systems makes for a very difficult task. Yet signs of these failures exist at all scales of neural networks, so we should be able to study a much more tractable setting. All neural networks must undergo an optimization process, called training, to be useful. To a great degree, understanding neural networks is understanding their optimization: through what process and exposure to which data did they arrive at their results. Yet our knowledge on this topic as a field is quite imprecise. In particular, a curious phenomenon called mode connectivity, the ability to connect neural networks in the loss surface, defies explanation entirely. This dissertation elucidates, explains and exploits this special structure in the loss landscape...
Read the full abstract
High Energy Astrophysical Phenomena
A Self-Sustaining Black Hole Engine Powered by the Tidal Disruptions of Stars
Tidal disruption events (TDEs) strongly prefer host galaxies undergoing, or recovering from, a burst of star formation, implying that a rare minority of galaxy types produces most events. The required per-galaxy rates, as high as $10^{-2}$ gal$^{-1}$ yr$^{-1}$, are hard to achieve through stellar relaxation alone. Molecular clouds surrounding nuclear clusters help set the relaxation rate, offering a path to higher rates. We show that (post-)starburst galaxies, whose nearby prototypes contain large molecular gas reservoirs, are likely in a self-sustaining cycle: an enhanced disruption rate compresses the surrounding clouds through momentum injected by the unbound debris, the denser clouds compress the cluster, and the cluster disrupts stars faster still. The runaway is arrested only when stars begin to collide. Solving for the steady state, the rate saturates at $1.4\times10^{-2}(M_{\rm h}/10^{6}M_\odot)^{-0.84}$ yr$^{-1}$, two orders of magnitude above the canonical rate at $10^{6}M_\odot$, with a normalization uncertain by a further two orders of magnitude through the stellar collision rate. Three independent requirements — that tidal debris outweigh AGN feedback, that the molecular clouds fit within the disk that holds them, and that the cusp be no denser than observed nuclei — bound the flattening of the stellar cusp to $0.18 \lesssim f_\ast \lesssim 0.25$. Because no engine can run below a threshold black hole mass, holes seeded beneath it stay dark until accretion carries them across, switching on after $\sim 1$ Gyr and offering a natural explanation for the late peak recently measured in the TDE delay time distribution. The nucleus is buried under $A_V \simeq 50$ whatever its geometry, so most such disruptions should be hidden from optical surveys and emerge instead in the infrared.
Read the full abstract
Quantum Physics
On the entropy of a pseudo-density matrix
The pseudo-density matrix (PDM) formalism naturally extends the notion of density operator to the spatiotemporal domain. While PDMs are Hermitian and of unit trace, they admit negative eigenvalues when encoding temporal correlations unattainable for spacelike separated systems. Consequently, there have been various approaches to extending the von~Neumann entropy—which is only defined for positive operators—to PDMs. Here, we prove that there exists a unique extension of the von~Neumann entropy functional to Hermitian matrices of unit trace satisfying two simple assumptions: unitary invariance and strong additivity with respect to affine combinations within the interval $[-1,1]$, which we prove contains the eigenvalues of a PDM. We also prove that this unique extension of von~Neumann entropy to PDMs is subadditive for single qubit dynamics, and we analyze its behavior in a number of examples.
Read the full abstract
Artificial Intelligence
Automated Researchers Can Mitigate Well-characterized Alignment Failures
Automating alignment research may accelerate progress toward aligned AI, but whether it does is hard to measure. Luckily, many alignment failures, such as deception, sycophancy, and jailbreaks, are already measurable by public benchmarks. We study whether automated alignment researchers (AARs) can post-train to mitigate alignment failures by proposing training methods and data to simultaneously optimize multiple safety benchmarks, while largely preserving general capability. Across 10 alignment failures, the strongest AAR methods significantly reduce the targeted alignment failures and generalize to a held-out benchmark, multi-turn behavioral audits, and models up to 4.7x larger than the target model. As a human baseline, 28 experienced researchers receive up to eight hours to develop one-shot methods for the same benchmarks, but their methods underperform the best AAR methods. Using human ideas as the AARs' initial research direction does not improve performance, suggesting current AARs may not need guidance from experienced researchers. These results suggest that automating alignment research on well-characterized failures may be practical in the near term.
Read the full abstract
Artificial Intelligence
Oculi: A Conversational Agentic Platform for Automated Credit Risk Analysis
Credit risk analysis in financial institutions traditionally requires analysts to manually write SQL queries, run statistical computations, and build visualization dashboards. This is a time-consuming workflow that limits exploration to familiar segments. We introduce Oculi, a conversational platform that transforms natural language questions into comprehensive credit risk analyses, complete with data queries, statistical testing, and interactive visualizations. Oculi employs a three-layer architecture that separates reasoning (LLM-powered agent), execution (Model Context Protocol tool servers), and presentation (agentic UI), enabling analysts to discover high-risk portfolio segments. Within Oculi, a new segment discovery pipeline is proposed that combines deterministic statistical methods with LLM-guided feature selection, leveraging LLM semantic domain knowledge alongside data-driven metrics to identify meaningful, actionable portfolio segments. Evaluated on a mortgage portfolio with 200+ features, Oculi demonstrates effectiveness in discovering material risk segments previously intractable through manual exploration, reducing time-to-insight significantly while maintaining auditability and statistical rigor.
Read the full abstract
Statistical Methodology
Saturation in G: simple & robust causal inference in cluster randomized trials with informative cluster sizes
Cluster randomized trials (CRTs) can exhibit informative cluster sizes (ICS) where cluster size is associated with outcomes and/or treatment effects. Under ICS, the individual and cluster-average treatment effects (iATE, cATE) can diverge, and the conventional linear mixed-effects model (LMM) and generalized estimating equation (GEE) with an exchangeable working correlation can produce data-dependent weighted contrasts that are not consistent for either estimand. In these settings with ICS, we propose easy to implement "cluster-size saturated models with g-computation" (CS-g), which employ a simple two-step adjustment to standard practice: (1.) augment the appropriately weighted working LMM or GEE with a saturated continuous cluster-size main effect and treatment x cluster-size interaction, and (2.) apply g-computation to target an interpretable marginal estimand. We prove that the appropriately weighted cluster-size saturated LMM with g-computation and more general cluster-size saturated GEE with g-computation can consistently target the iATE and cATE, among a broad class of interpretable estimands, while allowing for ICS. Crucially, this consistency holds under arbitrary misspecification of other model components, including the functional form of the saturated cluster-size terms. Furthermore, we demonstrate exact finite-sample equivalence between these consistent CS-g estimators and their model-robust standardization counterparts. Across simulations with continuous and binary outcomes, the proposed CS-g estimators were unbiased, more efficient than other consistent estimators, and returned greater power to detect ICS. A re-analysis of the PPACT P-CRT further illustrates the approach. Altogether, CS-g offers a simple, robust, and efficient route to target interpretable marginal effects in P-CRTs with ICS.
Read the full abstract
High Energy Physics (Theory)
Action-angle variables and phase space formulation of Hermitian matrix models
We develop a phase space description of large $N$ Hermitian one matrix models directly from the recursions satisfied by the orthogonal polynomials. In the one-cut phase, the recursion lattice naturally gives rise to a semiclassical action-angle pair, which can be mapped canonically to eigenvalue and momentum variables. We show that the resulting momentum profile is determined by the density of zeros of the orthogonal polynomials, and that the same phase space structure is reproduced from the Wigner transform of the Christoffel-Darboux projector and from the planar spectral curve. The Gaussian model provides a simple consistency check of the proposal. We then extend the construction to the symmetric quartic two-cut phase, where a period-two Jacobi recursion produces two Bloch bands and two disconnected phase space components with actions given by the partial 't Hooft couplings. Finally, we outline the finite-gap generalization appropriate to multicut phases.
Read the full abstract
Numerical Analysis
A GPU-Accelerated Blocked Adaptive Randomized Range Finder Based on an Implicit Householder QR Decomposition
Low-rank methods can reduce the memory and computational requirements of deep neural network training in approaches such as GaLore. Randomized range finders offer an attractive alternative to singular value decompositions, particularly when the required rank is determined adaptively from a prescribed approximation tolerance. We introduce a blocked adaptive randomized range finder based on a Householder QR decomposition and an optimized hybrid CPU–GPU implementation. Numerical experiments show that it preserves orthogonality and approximation accuracy in regimes where block Gram–Schmidt without reorthogonalization becomes unstable. The blocked formulation exposes matrix–matrix operations and enables overlap of CPU panel factorization with GPU updates. On an NVIDIA GH200, the overlapped implementation reduces the runtime for the largest tested matrix from 9.91 s on the CPU to 0.407 s. The method provides a stable and efficient building block for low-rank approximation on heterogeneous systems with applications in computational science and engineering.
Read the full abstract
Optics
First-return statistics and the point-spread function of a Henyey-Greenstein random walk - a far-field stable law
A particle launched normally into a half space, scattering by the Henyey-Greenstein (HG) phase function, exits after a first-return time whose count of flights n approaches the distribution-free Sparre Andersen exponent, f(n) ~ n^{-3/2}, with an amplitude fixed by the launch rather than by the theorem. Conditioned on n, and past a correlation-set transient, the lateral exit position is Gaussian with variance proportional to n, a scaling we test directly. Subordinating that Gaussian displacement to the one-sided 1/2-stable first-return time gives, in the diffusion limit, an isotropic 1-stable scaling limit: the far tail of the point-spread function (PSF) for a normally incident pencil beam is the two-dimensional Cauchy tail p(r) ~ r^{-3}, the large-radius limit of the half-space Poisson kernel. The index alpha = 1 is the product of the central limit theorem and the distribution-free first-return exponent, so no phase function admitting the same diffusion limit - finite step variance, short-range angular correlation - can move it. Absorption weights each path by exp(-mu_a L) in its total length L, truncating the subordinator and producing the screened radial structure (mu_eff + 1/r) exp(-mu_eff r)/r^2 carried by each source term of the Farrell-Patterson-Wilson (1992) diffusion dipole; the finite-depth two-source dipole is not derived here. Keeping the launch-depth factor that the tail calculation discards continues the same kernel into a finite core of radius 1.7 l*, with no additional fitted parameter. A scale-dependent index alpha_eff from the empirical characteristic function tracks the similarity parameter gamma, separating phase functions matched in g_1 by 0.15-0.22 near one transport mean free path. Results are Monte Carlo at g = 0 to 0.9 with up to 10^6 walkers.
Read the full abstract
Information Theory
Authentication over Arbitrarily Varying Channels with Causal Adversaries
We study authentication over a discrete memoryless arbitrarily varying channel (AVC) in which the adversary selects the channel state causally based on past channel outputs. We prove that the authentication capacity is positive exactly when the channel is not distribution-overwritable under stochastic encoding, and not I-overwritable under deterministic encoding. In the stochastic case, whenever the authentication capacity is positive, it coincides with the no-adversary Shannon capacity. Both converses follow from a novel wait-and-overwrite attack: the adversary tracks the no-adversary posterior over the message via the posterior guessing probability, and once it concentrates past a threshold, it samples a guess and a decoy from the posterior and overwrites the channel to mimic a no-adversary transmission of the decoy. The positivity results are obtained by constructing positive-rate codes with controlled overlap between codewords, together with a martingale concentration argument that handles adaptive adversarial strategies. For stochastic encoding, the full-capacity achievability further combines a capacity-achieving channel code with an authentication tag. These characterizations separate the causal stochastic-code and causal deterministic-code settings from each other and from the oblivious-adversary setting studied by Kosut and Kliewer.
Read the full abstract
Combinatorics
Braess' Paradox in Uniform Affine Grid Networks
Braess' Paradox is the phenomenon in which adding an edge to a congestion network increases total travel time. We study the paradox in directed rectangular grids where every edge shares the latency function $\ell(x)=ax+b$ and an added chord has latency $\ell_{*}(x) = cx+d$, where $a > 0$ and $b,c,d \ge 0$. Using an analogy with electrical networks, we bound the change in total travel time. We then give a necessary and sufficient condition for a chord to induce the paradox for some choice of nonnegative coefficients and compute the exact proportion of such chords in all grids with dimensions at most $100$. Such chords are scarce, and the fraction is maximized near an aspect ratio of $2:1$. We then improve the established $4/3$ upper bound on the Braess Ratio to one depending only on the grid dimensions, approaching $1.207$ on squares and $4/3$ on thin grids. Finally, we prove any ratio-maximizing chord must have zero latency.
Read the full abstract
Image and Video Processing
Multimodal Deep Learning for Uncertainty-Aware Radiation Pneumonitis Risk Prediction
Radiation pneumonitis (RP) is a potentially serious complication of thoracic radiotherapy, making reliable risk prediction important for identifying high-risk patients and guiding effective toxicity management. Although deep learning models can predict RP risk, uncertainty quantification remains underdeveloped in radiation toxicity assessment. We developed a Multimodal Bayesian Diffusion Transformer (MM-DiT) that integrates pretreatment computed tomography (CT) and three-dimensional planning dose distributions for RP risk prediction and uncertainty quantification. MM-DiT combines self-supervised multimodal representation learning with Bayesian classification to estimate individual risk and characterize aleatoric, epistemic, and label uncertainty. In independent Mount Sinai Health System (MSHS) and RTOG cohorts, MM-DiT achieved areas under the receiver operating characteristic curve (AUCs) of $0.89$ and $0.86$, respectively, with expected calibration errors of $0.02$ and $0.03$. Incorporating post-treatment CT increased AUC to $0.95$ in MSHS and reduced aleatoric and epistemic uncertainty by approximately $40-45\%$ and $40\%$, respectively. These findings establish a framework for uncertainty-aware radiation toxicity prediction that complements risk estimates with information about predictive reliability. By integrating multimodal learning and uncertainty quantification, MM-DiT may advance more interpretable and trustworthy AI-assisted toxicity assessment and support individualized radiotherapy planning.
Read the full abstract
Earth and Planetary Astrophysics
Magnitude of stable iron isotope fractionation limited by multiple stages of terrestrial core formation
One proposed mechanism for generating iron isotopic differences between planetary mantles and chondrites is metal-silicate equilibration during terrestrial core formation. Prior studies of this isotopic fractionation effect employ single-stage core formation models that are inconsistent with reproducing the siderophile element budget of the Earth's mantle. Here, we model iron isotopic evolution of the Earth's mantle in a multistage core formation scenario that is consistent with dynamic models of planet formation and reproduces the geochemistry of the bulk silicate Earth, specifically the refractory moderately siderophile elements. We find that multiple stages of core formation rebalance the fractionating effect of metal-silicate equilibration for the iron isotopic system, comparable to or smaller than typical analytical uncertainties in mantle rocks. Ultimately, this rebalancing effect - coupled with stochastic accretion and differentiation histories which further complicate planetary formation processes - makes it unlikely that isotopic fractionation during metal-silicate equilibration is the primary mechanism responsible for the iron isotopic composition of the Earth's mantle.
Read the full abstract
Fluid Dynamics
NeuralFlowNet: Towards Data-Free Physics-Informed Neural Network Solutions of Navier-Stokes Equations Across Low and High Reynolds Numbers
Physics-informed neural networks (PINNs) have emerged as a compelling pathway toward trustworthy artificial-intelligence-based computational fluid dynamics (CFD) by embedding governing equations directly into the learning process. Many existing AI flow models require large simulation or experimental datasets and often remain problem-specific, limiting their generalization and physical reliability. Data-free PINN frameworks offer an alternative by learning flow solutions from the Navier-Stokes equations and prescribed boundary conditions, potentially reducing dependence on expensive CFD datasets while retaining physical consistency. However, traditional PINN frameworks have struggled at high Reynolds numbers, limiting their application in CFD. In this work, we present NeuralFlowNet, a data-free, physics-informed proof-of-concept framework designed to solve steady Navier-Stokes problems across low- to high-Reynolds-number conditions. We describe the proposed methodological framework and demonstrate its applicability using benchmark problems with increasing physical and geometric complexity. The results demonstrate that NeuralFlowNet can accurately recover steady flow fields across a broad Reynolds-number range, including cases with strong pressure gradients, without training on external flow-field data and while maintaining good agreement with reference numerical solutions. These findings establish NeuralFlowNet as a reliable framework for future testing of unsteady and more complex simulations and could provide a basis for developing trustworthy and efficient AI solvers for high-Reynolds-number fluid dynamics.
Read the full abstract
Machine Learning
Revisiting the Provable-Auditable Privacy Gap of DP-SGD
Differential privacy (DP) has traditionally been used to provide theoretical upper bounds on an algorithm's stability to changing its training data. In modern private machine learning applications, achieving strong tradeoffs between utility and theoretical privacy is challenging, and thus one may optimistically hope that existing theoretical privacy analyses are loose. Recent work on privacy auditing has adopted a dual viewpoint, instead lower bounding the true privacy of an algorithm by constructing empirical distinguishing events. The auditing literature has thus far yielded a pessimistic outlook on the looseness of theoretical privacy bounds for DP-SGD, the de facto private training method in modern ML, as nearly-matching empirical lower bounds have been achieved under various threat models [NHSBTJCT23, AC24, CBP25]. In this work, we propose the empirical privacy lower bound of an algorithm as a concrete metric to optimize for, complementary to the theoretical upper bound. We give a lightweight defense framework that generically augments optimization methods in the ML pipeline to have significantly-improved empirical privacy on standard benchmarks. Moreover, we show that our framework comes at no theoretical privacy cost when augmenting DP-SGD, unlike previously-proposed defenses against membership inference attacks. We evaluate our defense against a broad range of audit constructions, models, and datasets to demonstrate its flexibility.
Read the full abstract
Computer Vision
NBS: No Bias Stereo
Stereo reconstruction is one of the last remaining Computer Vision tasks where all state-of-the-art methods employ a heavy architectural inductive bias. Even though it has been demonstrated that the task can be solved using general-purpose methods, it is widely believed that inductive biases in stereo are strictly necessary for both high-quality results and computational efficiency. We challenge this paradigm. In this paper, we demonstrate that both state-of-the-art accuracy and superior runtime efficiency are achievable with a model completely devoid of architectural inductive biases, relying instead on a simple, end-to-end Vision Transformer. By training on massive synthetic datasets, we show that pure data-driven learning can surpass explicitly engineered geometry. This work proves that explicit inductive biases are no longer a prerequisite for stereo matching, ultimately unlocking true scaling laws for continuous improvement in 3D reconstruction.
Read the full abstract
Computation and Language
VocalAffectBench: Evaluating Vocal Emotion Recognition in AI Audio Models
Voice products increasingly need affective cues that are present in speech but absent from transcripts. We introduce VocalAffectBench, a public, test-only benchmark for evaluating whether AI audio models can identify expressed vocal emotion from raw audio. The benchmark contains 273 human-recorded English WAV clips from 51 speaker accounts totaling 1.95 hours across seven labels: angry, disgusted, fearful, happy, neutral, sad, and surprised, with 39 clips per class. All baselines are evaluated from audio alone, without transcripts or contextual metadata. Across six released baselines, average accuracy is 35.5%. The strongest baseline, gemini_3_5_flash, reaches 46.5% on the seven-way task, above the 14.3% random baseline but far from robust emotion recognition. A secondary valence-bucket analysis maps labels into positive, neutral, and negative classes, excluding surprised because its valence is ambiguous. Aggregate accuracy under this coarser view is 50.9%. Performance is highly uneven across classes. By recall, neutral is identified most reliably at 75.6% averaged across baselines, while surprised and fearful reach only 10.7% and 15.4%, respectively. These results show that the evaluated baselines can extract some affective signal from speech, but discrete expressed-emotion recognition remains fragile, especially for non-neutral emotions that are often most important in voice agent workflows.
Read the full abstract
Information Retrieval
MERIT: Mitigating Exposure Bias in Generative XMC for User-Interest Propensity Modeling
Matching users to interest categories at scale is central to personalized shopping, but the task is challenging in large e-commerce platforms, where label spaces continually evolve and user-interest signals are sparse and long-tailed. Autoregressive language models are appealing because their world knowledge and semantic priors over descriptors generalize across extreme label spaces and accommodate multiple valid label assignments. Yet under teacher-forced fine-tuning, inference-time predictions become part of the conditioning context: early errors steer later outputs toward co-occurring labels, over-generating near-correlates and missing unrelated true interests. We present MERIT, a framework for user-interest propensity modeling that mitigates this exposure bias through a self-correction objective. A permutation-invariant multi-target loss over shuffled mixtures of gold and mined hard-negative labels exposes the generator to erroneous prefixes while preserving the efficiency of teacher-forced training. This training objective concentrates supervision at classification positions, yielding propensity-aligned hidden states powering a lightweight scorer for bidirectional retrieval (interests for users and users for interests). On a proprietary e-commerce dataset with 250k+ interest categories, MERIT improves global recall by at least 11.9% and average Hit@k by 6.1%. In production A/B tests, it achieves +0.26% gain in user conversion.
Read the full abstract
Computation and Language
The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice
Hidden-state probes effectively detect LLM hallucinations, but the geometry of the signal remains poorly characterized, driving increasingly complex probe architectures. Across three 7B-scale models and three datasets in a paired-example paradigm, we find the signal overwhelmingly dominated by a single mean-shift component, and removing this direction collapses detection to chance. Shrinkage linear discriminant analysis closes about 73% of the gap between 1D and full-dimensional classifiers, so apparent architectural complexity largely reflects high-dimensional covariance estimation difficulty rather than exploitable non-linearity. A simple L2-regularized logistic regression (0.952 AUROC) bounds or outperforms twelve controlled architectural alternatives, and our multi-layer aggregation exceeds CLAP cross-layer attention probing under matched paradigm. Because the signal spans a contiguous layer band, LayerMix aggregates it to match oracle-layer performance without oracle access. Our claims characterize the geometry within the controlled paired-example paradigm. Our code is available at https://github.com/js-lee-AI/LayerMix.