Peer Revenue Now accepting positions

Papers

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

All of arXiv

Saturday 7 November

Numerical Analysis

DMP-Oriented Anisotropic Mesh Adaptation for SUPG Finite Element Methods for Time-Dependent Convection-Diffusion Problems

We study discrete bound preservation for fully discrete piecewise-linear streamline upwind Petrov–Galerkin (SUPG) finite element approximations of time-dependent convection–diffusion equations on anisotropic triangular meshes. Although the continuous homogeneous-source problem satisfies a parabolic maximum principle, SUPG approximations can exhibit nonphysical undershoots and overshoots. We derive sufficient nodal discrete maximum principle (DMP) conditions that account for the stabilized mass matrix, the combined effects of physical and streamline diffusion, time-step restrictions, and transfer between successive meshes. The analysis identifies a SUPG-effective diffusion tensor and motivates an anisotropic metric aligned with its inverse. We construct a combined metric that retains the effective-tensor geometry while introducing solution-dependent mesh density through recovered Hessian information. The resulting DMP conditions provide a theoretical basis for anisotropic mesh design. Numerical experiments for time-dependent transport problems with isotropic and moderately anisotropic diffusion demonstrate that, under mass-compatible limited SUPG stabilization, the combined metric substantially reduces measured undershoots relative to other methods at matched nominal mesh budgets, although improved bound behavior can carry an accuracy penalty. A controlled tensor–density comparison in the anisotropic transport test supports a contribution from the effective tensor beyond the prescribed mesh density.
Read the full abstract

Statistical Methodology

Overlap-Excluded Estimation of Participation Probabilities for Non-Probability Samples

Correction of selection bias is crucial for making valid statistical inferences about a finite population using non-probability datasets. Various approaches have been proposed to correct this bias and obtain estimates of finite population quantities such as means and totals. The quasi-randomisation approaches, in particular, aim to estimate a `participation probability' for units in the non-probability dataset. These probabilities are then applied in an inverse probability weighting scheme to produce population estimates. In this paper, we propose to estimate the participation probabilities for the non-probability sample by combining the non-probability sample with an available probability sample, while excluding the overlapping units in the probability sample. When the true overlap cannot be determined, we show how the Expectation-Maximization method can be used to simultaneously estimate the overlap and the participation probabilities. We compare the performance of our approach against several alternative methods theoretically and in an empirical context, and demonstrate the efficiency gains that are possible when the overlapping probability sample is excluded.
Read the full abstract

All authors Lyndon Ang, Robert Clark, Bronwyn Loong, Anders Holmberg

Artificial Intelligence

Improving Analytic Constants via Interpretable AI-Guided Lean Proof Refinement

AI agents are increasingly capable of discovering new mathematical results, yet their reasoning and generated proofs can be difficult to understand and verify. We introduce an interpretable framework for AI-guided mathematical discovery, focused on improving quantitative bounds on analytic constants. Our key insight is that existing proofs often contain slack in intermediate lemmas, which can be tightened without reconstructing the entire argument. To exploit this structure, we represent proofs as partially-formalized Lean blueprints: human-interpretable dependency graphs of definitions, lemmas, and theorems. An AI agent then iteratively refines the Lean code while preserving the blueprint's dependency structure, allowing local improvements to automatically propagate to a verified final bound. This interpretable structure allows humans to inspect the agent's refinements and understand how local improvements propagate to the final theorem. For Grothendieck's constant ($\mathcal{K}_G$), we obtain $$ \mathcal{K}_G \leq \frac{\pi}{2\log(1+\sqrt{2})} - 10^{-17}, $$ an explicit numerical improvement over Krivine's 1977 bound. For Korenblum's constant ($\mathcal{K}_K$), we obtain $$ \mathcal{K}_K \geq 0.356565, $$ improving Wang's 2025 bound of $0.3554$. Our experiments demonstrate the potential of both partial formalization and iterative agentic refinement for interpretable mathematical discovery.
Read the full abstract

All authors Rahul Saha, Alan Li, Anton Xue, Swarat Chaudhuri, Adam Klivans, Pravesh K. Kothari, Raghu Meka

Computer Vision

Anatomy-Activated Mixture-of-Experts for 3D Medical Vision-Language Pre-training

Medical Vision-Language Pre-training (VLP) leverages the semantic richness of radiology reports to learn 3D representations, yet current architectures remain fundamentally anatomy-agnostic. While clinical diagnostics rely on organ-specific context, standard encoders apply a uniform, dense parameterization to all volumetric patches, and existing Mixture-of-Experts (MoE) routers gate on isolated local features without explicit access to the organ composition of the scan or crop. We propose Anatomy-Activated Mixture-of-Experts (A$^{2}$MoE), a framework that incorporates anatomical inductive biases within the model's computational path. A$^{2}$MoE introduces two key components: (1) a Histogram Representation Router (HRR) that conditions expert selection on global visual context and a predicted anatomical composition histogram, encouraging anatomy-dependent expert weighting; and (2) an Organ-Query Multi-Scale Projector (OQMSP) that utilizes learnable queries to extract organ-level embeddings for fine-grained alignment without requiring manual annotations at inference. By routing at the patch level rather than the token level, A$^{2}$MoE reduces routing complexity by orders of magnitude while satisfying load balancing by construction. Pre-trained on over 60,000 multi-phase CT scans, A$^{2}$MoE achieves state-of-the-art performance across segmentation, classification, and detection benchmarks. Furthermore, we demonstrate that our anatomically grounded inductive biases support robust generalization to MRI and ultrasound, establishing a scalable foundation for medical representation learning. The code is available at https://github.com/gmum/A2MoE.
Read the full abstract

All authors Szymon Płotka, Gizem Mert, Pedro R. A. S. Bassi, Wenxuan Li, Zongwei Zhou, Anna Marcinkiewicz, Wiktoria Romańczyk, Jarosław B. Ćwikła, Łukasz Struski, Ewa Szczurek, Jacek Tabor, Arkadiusz Sitek

Solar and Stellar Astrophysics

Radial Velocity Confirmation of Two Binary Systems Detected with the Pulsation Timing Method

We present follow-up radial velocity (RV) results for the subdwarf B$+$white dwarf binaries TIC 202354658 and TIC 69298924, which were originally identified using the pulsation-timing and Fourier-modulation methods applied to ultra-short-cadence photometric data from the Transiting Exoplanet Survey Satellite ($TESS$). Using spectra obtained with the Gemini-North telescope and the Nordic Optical Telescopes, we derived the radial velocity curves for both targets. The orbital solutions we obtained are consistent with the predictions based on pulsation timing, and they provide more precise orbital information compared with the previously obtained results. We therefore confirm that both TIC 202354658 and TIC 69298924 are binary systems with white dwarf companions. The RV solution of TIC 202354658 indicates a relatively large eccentricity ($e$ = 0.313 $\pm$ 0.013), which would be uncommon for a short orbital period sdB binary.
Read the full abstract

All authors Tomomi Otani, A. S. Baran, Kaydence Boyle, Julia M. Clark

Software Engineering

Hybrid Source Code Change Impact Analysis

Change impact analysis (CIA) helps developers identify code entities impacted by their initial changes. While several approaches for CIA were proposed, they still struggle with either low precision or low recall, thereby increasing manual effort for developers to verify the results. Furthermore, limitations in the used datasets and experimental setups can bias the results reported by many CIA approaches. In this paper, we first address the dataset limitations by manually labeling 1,931 changed code entities stemming from 263 commits of 22 open-source Java projects. Next, we introduce two variants of katIA, our code change impact analysis approach that combines static impact analysis with the reasoning capabilities of large language models (LLMs). The first variant uses a single prompt (katIA-SP) for refining an initial static impact set while the second variant uses three specialized agents (katIA-MA). The evaluation with our curated dataset shows that katIA-SP with Gemini 2.5 Flash achieved an F1-score of 22.4% when matching impacted entities by their fully qualified names (FQN) and 27.0% when matching them by their class and entity names (CE). The latter shows an improvement of 2.0 percentage points over the state-of-the-art. This score is further improved by katIA-MA which, using Gemini 2.5 Flash, yields an FQN-matching F1-score of 29.1% and a CE-matching F1-score of 33.3%. While these improvements come at the cost of increased median generation time from 0.570 s to 11.214 s and median token usage from 1,235 to 6,462 tokens, the improved performance justifies the additional cost. The comparison of three LLMs using katIA-SP finds that Gemini 2.5 Flash is the most consistent model. These results indicate that combining static impact analysis with LLM-based refinement, particularly with the multi-agent variant, can improve code change impact analysis.
Read the full abstract

Statistical Methodology

A Unified Framework for Robust Blinded Sample Size Re-estimation Based on Nuisance Parameters

Blinded sample size re-estimation aims to mitigate uncertainty in nuisance parameters at the design stage while preserving the integrity of the trial. It allows the required sample size to be updated without revealing treatment assignments and thereby reduces the risk of an inappropriate sample size. Although numerous procedures have been proposed for different outcome types, a unified framework is still lacking. We therefore develop a general framework for blinded sample size re-estimation based on nuisance parameters. Within this general framework, we develop robust methods to reduce the influence of extreme observations. The proposed framework combines asymptotically linear estimators for the parameter of interest with robust Z-estimators for nuisance parameters, thereby unifying a broad class of existing procedures through different choices of estimating functions. We also propose a general principle for constructing nuisance parameters that can be estimated from blinded data. The proposed framework is supported by asymptotic theory showing that the type I error rate converges to the nominal level under regularity conditions. Simulation studies demonstrate that the proposed method maintains reliable type I error control across a range of settings, including contaminated data and small pilot sample sizes. Compared with existing blinded and unblinded methods, the proposed framework is more flexible and less sensitive to outliers. A clinical trial application illustrates its practical utility.
Read the full abstract

Risk Management

Risk-Sensitive Reinforcement Learning for an Options Market Maker: An Empirical Study with Real S&P 500 Options

Options market makers absorb client order flow and must hedge the resulting inventory risk under trading frictions. We study whether an option hedging reinforcement learning (RL) agent trained on historical market data outperforms three alternatives in tail-risk protection: (i) traditional delta hedging, (ii) simple hedge-ratio rules, and (iii) the same agent trained in simulated markets. Most RL hedging agents are trained in model-generated markets that do not capture the crashes and volatility regimes present in real data. We introduce an empirical replay environment using S&P 500 option data from 1996 to 2017 and compare against a matched geometric Brownian motion (GBM) environment, a stochastic-volatility (SABR) simulator, and a hybrid that values historical paths with Black-Scholes, all sharing the same portfolio mechanics, stochastic arrivals, costs, and test period. We train a distributional Twin Delayed Deep Deterministic Policy Gradient (TD3) agent and evaluate it out of sample over 2018-2023. The agent trained on empirical data reduces Conditional Value-at-Risk (CVaR) tail loss by 3.1% compared to delta hedging and outperforms every hedge-ratio rule chosen during the training period and every alternative training environment. The reduction reaches 8.9% in volatile and falling markets, but reverses to a cost of 7-10% in rising markets. Empirical training induces a state-dependent short net delta: the policy under-hedges negative-delta books, most strongly after prices have fallen. This tilt disappears with GBM training and is weaker with SABR or hybrid training. These findings suggest that exposure to historical market dynamics can materially change a tail-risk hedging policy, although the evidence rests on a single test path and the edge is regime-conditional.
Read the full abstract

All authors Jacky Chen, Ing-Haw Cheng, John Hull, Landis Li, Zissis Poulos, Justin Shi, Andreas Veneris, Zeyu Wang

Statistical Methodology

Causal Outcome Learning with Noisy Proxies

Modern observational biomedical datasets contain increasingly extensive baseline and outcome measurements, yet detecting a clinically meaningful treatment effect remains difficult because investigators must identify an outcome that captures treatment response while adjusting for confounding distributed across numerous noisy baseline variables. We introduce SCOUT (Score-based Causal Outcome Learning Using Treatment signals), a framework that learns a clinically interpretable, treatment-responsive outcome while adjusting for confounding throughout the learning process. SCOUT constructs low-dimensional adjustment representations of high-dimensional baseline measurements to estimate the average treatment effect (ATE) and conditional average treatment effect (CATE) for the learned outcome. Importantly, the final adjustment space is at most two-dimensional, allowing treatment-effect heterogeneity to be estimated in this space rather than in the full high-dimensional covariate space. SCOUT also introduces a fourth-order statistic for testing treatment-effect heterogeneity that emphasizes unusually large deviations from the ATE. In synthetic and semi-synthetic experiments, the algorithm achieved competitive ATE estimation accuracy and substantially lower CATE estimation error. In an observational study of patients with bipolar disorder, SCOUT identified a clinically recognizable lithium-associated adverse-effect profile, detecting both a significant ATE (Holm-adjusted $p$ = 0.0026) and significant treatment-effect heterogeneity (Holm-adjusted $p$ = 0.033).These results demonstrate how jointly learning outcomes and controlling for confounding can facilitate the discovery of treatment effects in high-dimensional observational data.
Read the full abstract

Signal Processing

MPSi: Neural Extraction of Site-Specific Multipath Propagation Properties for Massive MIMO

Accurate extraction of site-specific multipath propagation properties enables a wide range of wireless applications, including initial access, wideband channel reconstruction, and the generation of channel datasets for learning-based and continual-learning systems. Classical parametric channel-sounding algorithms such as RIMAX provide physically interpretable estimates, but their data-dependent model-order selection and iterative nonlinear refinement introduce substantial delay and computational cost. This paper proposes MPSi, a physics-guided neural framework that replaces these costly stages with a compact path-count network and a fixed-depth, deep-unfolded Levenberg-Marquardt refinement. MPSi estimates the complex gain, angle of arrival, and propagation delay of each multipath component while preserving the underlying physical channel model. We evaluate MPSi in two site-specific ray-traced environments, a small cell and a dense urban area, and show that it achieves initial-access and full-channel-estimation performance comparable to an enhanced RIMAX baseline while requiring significantly fewer optimization iterations. The reconstructed channels also yield bit-error-rate performance close to the perfect-channel knowledge reference. These results position MPSi as a promising low-overhead framework for extracting multipath propagation properties in massive-MIMO systems and for generating cell-specific channel pseudo-labels for training when direct ground-truth measurements are unavailable.
Read the full abstract

All authors Maryam Rezvani, Raviraj Adve, Amr El-Keyi, Akram bin Sediq

Combinatorics

Dominant permutations, positive mixed-variable formulas, and product-splitting equivalence for (double) Schubert polynomials

We present new positive combinatorial formulas for triple Schubert coefficients in mixed-variable products of double Schubert polynomials, $\mathfrak{S}_u(x;y)\mathfrak{S}_v(x;z)$, expressed in the $\mathfrak{S}_w(x;y)$ basis, in the case where $u$ is dominant. This result complements earlier results where $v$ was assumed dominant. To generalize further, we introduce a relation on permutations that we call “domination,” under which dominant permutations dominate all others. When $u$ dominates $w$, we obtain an explicit positive formula for $c^w_{u,v}(y;z)$, which in particular yields positivity when $u$ and $v$ have separated descents. In addition to these new positivity results, we establish a product–splitting equivalence for double Schubert polynomials, relating triple Schubert coefficients to splitting coefficients. This equivalence highlights structural symmetries and provides reduction methods for computing coefficients that are otherwise difficult to access. Taken together, these results extend the range of double Schubert polynomial products for which positivity can be established combinatorially and yield explicit formulas that can be implemented computationally.
Read the full abstract

Information Theory

Optimal Sampling and Transmission Policy for Information Freshness in Gossip Networks

We study the optimal sampling and transmission policies for a transmitter communicating with n gossiping receivers with the objective of tracking a source's status over time using a semantic freshness metric: Version Age of Information (VAoI). The transmitter observes the version changes in the source, and can acquire a new information version by paying a sampling cost. The receivers gossip, allowing information exchange among each other without direct transmission. Gossiping can save on transmission cost, but it may propagate stale version information, necessitating transmitter intervention. With imperfect communication links, we formulate an infinite horizon average cost Markov Decision Process (MDP) in order to jointly optimize receivers' VAoIs, sampling and transmission costs in a time slotted fully connected (FC) gossip network. We first solve the MDP using Relative Value Iteration (RVI) algorithm. Then, we prove that the sampling decision has a threshold structure in the transmitter VAoI. Further, under reliable gossiping links, we show that transmission and idling has an VAoI-difference threshold structure with minimum VAoI activation. Numerical results obtained using RVI show that the optimal policy consistently outperforms natural baselines. These results characterize how source sampling, direct transmission, and receiver gossip should be jointly coordinated for cost-aware semantic freshness in gossip networks.
Read the full abstract

Probability

Spectral gaps and mixing rates of primitive matrices

We obtain lower bounds on spectral gap/upper bounds for mixing rate for arbitrary $n$-state (not necessarily reversible) primitive Markov chains which have exponent $-O(n)$ rather than $-O(n^2)$, a significant improvement on the order obtained from classical arguments. Our improvements come from using the irreducibility distance and chainable index (which are at most linear in $n$) rather than the primitivity index (which can be near $n^2$). This also yields corresponding bounds for arbitrary nonnegative primitive matrices.
Read the full abstract

Machine Learning

Interpretable Cyber-Physical Attack Detection in Smart Grids: A Leakage-Controlled Multi-Testbed Evaluation of Fusion, Robustness, and Observability Limits

Cyber-physical smart-grid detectors are typically treated as black boxes and validated exclusively on clean, temporally shuffled datasets, which risk overly optimistic performance claims. This work presents a multi-testbed evaluation study built around an interpretable, time-windowed fusion Random Forest, adopting the pessimistic result whenever evaluation protocols disagree. (i) On a public paired cyber-physical dataset, shuffled cross-validation is optimistic: macro-F1 falls from 0.96 to 0.88 +/- 0.06 under rolling-origin temporal evaluation. A topology-free graph convolutional network (GCN) serving as a control confirms that this lightweight model is competitive on these splits. (ii) The adaptive and adaptive-joint feature-space attacks, which are infeasible as realistic grid manipulations, degrade macro-F1 to 0.67 and 0.34, respectively. Adversarial training and randomized smoothing partially restore performance, and a physics-constrained evolution-strategy evasion on the IEEE 14-Bus testbed shows the same vulnerability pattern. (iii) On real IEC-104 traffic, the cyber view reliably catches structural attacks (recall of 0.87 to 0.98) but largely fails on semantic value manipulation (recall of 0.11 to 0.35). An observability taxonomy on the IEEE 14-Bus system reveals that stealthy false data injection defeats all three observation views. (iv) On a full AC weighted least-squares state-estimation testbed, a forecasting-aided temporal test is evaded by slow ramps precisely where a rate-invariant physics-manifold test sustains detection at a common 5% false-alarm target. A full end-to-end evaluation on IEEE 118-Bus confirms the ordering but bounds its strength: the manifold test detects every trial down to 0.05 degree/step and 0.48 at 0.02 degree/step, against 0.14 for the temporal test. All code and benchmarks are released.
Read the full abstract

All authors Muhammad Ridwanul Hoque, Sarah Sharif, Reza Saeed Kandezy, Yaser Banad

Robotics

Explicitly capturing demonstrator multi-modality for distillation into diffusion policies

The diversity of a distilled diffusion policy's outputs matters beyond how it impacts the policy's success rate. In particular, downstream applications like test-time search and policy improvement require exploration. Thus, we want to distill multi-modal demonstrators into diffusion policies which reproduce their conditional action distributions. Standard data collection for behavior cloning applies a sliding window to demonstration episodes, which pairs unique observation histories with actions. These datasets can be fit by a policy that is either multi-modal, or unimodal with high Lipschitz constant to capture a diversity of actions seen at nearby observation samples. Prior work finds that diffusion policies trained on sparse data learn the latter, memorizing one action per observation. We adapt vine sampling to behavior cloning data collection to resolve this ambiguity: at some states the demonstrator visits, query it $k$ times from exactly the same observation. On a synthetic task, we study how the number of actions per observation, $k$, affects the learned policy, and find that larger $k$ yields a policy that better matches the demonstrator's conditional distribution, including its multi-modality and diversity. On a 2D truck unloading task, vine sampling likewise leads to a policy closer to the demonstrator and substantially improves both best-of-$N$ and expert iteration performance.
Read the full abstract

All authors Shao Yuan Chew Chia, Rebecca H. Jiang, Abhinav Agarwal, Andreas Hofmann, Leslie Pack Kaelbling, Russ Tedrake

Probability

An Introduction to Applied Probability and Stochastic Processes: Open Lecture Notes

These open lecture notes provide a broad introduction to measure-theoretic probability and stochastic processes, with a focus on applications. Topics include probability spaces, conditional expectation, generating and characteristic functions, Markov chains in discrete and continuous time, convergence of random variables, Brownian motion, stochastic calculus, stochastic differential equations and their connection to partial differential equations, jump diffusions, stochastic control, optimal stopping, and Monte Carlo methods. Most chapters include exercises, and solutions are provided. The notes were written for a two-quarter course for first-year PhD students in applied mathematics, and should also be useful to graduate students in statistics, engineering, economics and finance. The LaTeX source is freely available on GitHub under a CC BY-SA 4.0 license, and readers are encouraged to fork the notes and contribute corrections, improvements, and new material.
Read the full abstract

Cryptography and Security

Validating AI-Discovered Cloud Threats on Live Infrastructure: A Multi-Model, Multi-Cloud Study

Cloud infrastructure threat modeling lacks empirical validation of AI-discovered threats. We present a three-phase methodology of multi-agent threat discovery, automated exploit synthesis, and non-destructive live validation, and apply it to three commercial foundation models on two cloud architectures, validating each threat natively on the cloud (AWS, GCP, or Azure) it was discovered against. Of 233 AI-discovered threats, 103 were confirmed and 34 refuted (75.2% precision, 95% CI [67.3, 81.7]); 44 were conditional (the named mechanism refuted but an equivalent impact shown through another path) and 52 indeterminate. Multi-model discovery helped through breadth, not agreement: the three models together found 2.2$\times$ the confirmed threats of the best single model at comparable precision, while threats reported by several models were confirmed no more often than single-model ones. False positives were not architecture-specific: they asserted access or system state the deployment did not have. Threats coded in advance as observable configuration resolved cleanly more often than multi-step behavioral threats (72.5% vs.\ 51.6%), which were more often conditional, suggesting that models more often misidentify the exploitation mechanism than the underlying weakness. AI-assisted discovery is a practical input to security workflows when paired with empirical validation.
Read the full abstract

Mesoscale and Nanoscale Physics

Imaging spin entanglement with a scanning tunneling microscope

Artificial spin arrays assembled on surfaces enable atomic-scale studies of quantum correlations. Here we establish a hierarchy of connections between spin inelastic electron tunneling spectroscopy (IETS) and entanglement in spin-$1/2$ systems. We first show that the number of spectroscopically active transitions in a spin-$1/2$ dimer serves as an entanglement witness. Using the leading-order tunneling conductance for the tip placed on spin $i$, $dI(i)/dV$, we then relate spectral moments $\mathcal{M}^{(n)}(i)=\int V^n(d^2I(i)/dV^2)\,dV$ to quantum correlations. The zeroth moment probes the one-tangle in systems with a gapped, nondegenerate ground state. For rotationally invariant Heisenberg systems, the first moment yields thermal spin correlators $\langle\boldsymbol S(i)\cdot\boldsymbol S(j)\rangle$, enabling reconstruction of nearest-neighbor concurrence and spatial imaging of entanglement in spin chains. It also provides access to the internal energy and magnetic specific heat, both usable as entanglement witnesses. Applying this approach to experimental IETS data from nanographene spin chains, we find good agreement with theory. We further discuss how the second moment probes longer-range correlations and the inverse moment relates to local spin susceptibility. For spin-$1/2$ trimers, we connect inter-multiplet inelastic spectral weight to genuine tripartite entanglement quantified by the concurrence fill. Finally, we reconcile the local character of IETS with its sensitivity to nonlocal entanglement through an analogy with digital quantum circuits, where Hamiltonian evolution implements multiqubit operations. These results establish IETS as a tool for quantifying and imaging entanglement in engineered on-surface spin systems.
Read the full abstract

All authors J. C. G. Henriques, B. Viña-Bausá, I. Brihuega, J. Fernández-Rossier

Machine Learning

CoupledFlow: One-Step Neural Operators for Coupled Multi-Physics PDEs

This paper introduces COUPLED FLOW, a new neural operator learning framework for multi-physics PDE systems that captures the co-evolution of multiple physical processes by modeling their cross-process interactions. Unlike existing neural operator methods that primarily target single-physics settings, COUPLED FLOW employs a novel structure-constrained flow matching algorithm to learn process- specific mean velocity fields conditioned on the full multi-physics state. This enables fast and accurate one-step prediction of coupled PDE systems that would otherwise require expensive numerical simulations. A key challenge in training COUPLED FLOW however lies in sampling structure-constrained conditional flow interpolants that capture cross-process correlations and serve as training examples for flow matching. To address this challenge, we derive a multi-physics mean-flow consistency identity that characterizes cross-process interactions and defines the structure-constrained mean-flow matching loss. To provide informative training samples for flow matching with this loss, we develop a principled method for sampling correlated conditional interpolants based on posterior multi-output Gaus- sian processes, which capture richer cross-process correlations than the commonly used endpoint-conditioned linear interpolants. We also establish a theoretical re- sult showing that the Wasserstein-2 distance between the predicted and oracle PDE solution distributions is upper-bounded in terms of the structure-constrained flow matching loss and the smoothness of the learned flow. Experiments across diverse coupled PDE benchmarks demonstrate that COUPLED FLOW consistently outperforms existing state-of-the-art neural operators and generative PDE solvers.
Read the full abstract

All authors Trong Khiem Tran, Minh Long Bui, Phi Le Nguyen, Mrinal Sen, Sanjay Srinivasan, Trong Nghia Hoang, Jana Doppa

Chemical Physics

Molecular Photoionization Spectra from Signed Dyson Orbitals and Anharmonic Nuclear Motion

Molecular photoelectron bands couple changes in nuclear structure to the energy-dependent scattering of the departing electron. Band intensities require both nuclear transition moments and electronic continuum amplitudes. We present SPARKS (Scattering and Photoionization: Amplitudes, Response Kernels and Spectra), a molecular photoionization approach that connects signed Dyson orbitals to complex molecular dipole amplitudes and harmonic or anharmonic nuclear transition moments. Compact B-spline and finite-element discrete-variable representations support spectral and sparse outgoing-boundary solvers. Independent molecular checks reproduce a reference H$_2^+$ spectrum and CO continuum amplitudes, while analytic hydrogen photoionization tests establish the continuum normalization. For H$_2$O/H$_2$O$^+$, electronic derivatives activate a Franck–Condon-dark fundamental and enter the same nuclear operator used in harmonic, perturbative and internal-coordinate variational calculations. Furan extends the nuclear treatment to all 21 vibrational coordinates in a selected configuration-interaction window, including low-lying out-of-plane states. Four epichlorohydrin ionization channels show how orbital mixing reshapes cross sections and angular distributions at similar Dyson norms. Energy conservation is imposed separately for every vibronic transition at fixed photon energy. The resulting spectra retain vibrational assignments and electronic interference, connecting weak-band activation, anharmonic spacings and photoelectron angular distributions within one physical description.
Read the full abstract

Statistical Methodology

Forecasting Replication Study Results from Meta-Analyses

Multisite replication projects often rely on published meta-analyses to plan their sample sizes, but meta-analyses tend to overestimate population effects because of publication bias and differences in study characteristics. We developed methods to forecast site-level estimates of a not-yet-conducted replication from a meta-analysis. Extending calibrated estimates from meta-regression, we constructed nonparametric prediction intervals that condition on each site's moderator levels and correct for publication bias through inverse-probability-of-publication weighting under an assumed severity. These intervals achieve nominal coverage asymptotically when population effects are normally distributed, with a bounded coverage error otherwise. We compared four forecasting models (intercept-only meta-analysis or meta-regression, each with and without publication bias correction) in simulation study (81 scenarios and blinded retrospective forecasts of two multisite replications in infant cognition: ManyBabies 1 (104 sites) and ManyBabies 4 (33 sites). Correcting for publication bias improved interval calibration, with gains that increased with bias severity. Meta-regression improved forecasts only when measured moderators explained effect variation and replication sites differed meaningfully on them. In practice, publication bias correction should be the default when forecasting replications from meta-analyses, and replication projects that deliberately vary measured moderators across sites would enable more informative forecasts.
Read the full abstract

All authors Yi Li, Alvin Wei Ming Tan, Francesco Margoni, Michael C. Frank, Maya B. Mathur

General Relativity and Quantum Cosmology

Thermodynamic, Optical, and Wave Signatures of Charged Letelier AdS Black Holes with PFDM and Quintessence

We study a four-dimensional charged Letelier–anti-de Sitter (AdS) black-hole metric containing a cloud of strings, a phenomenological perfect-fluid-dark-matter (PFDM) source, and a Kiselev-type quintessence term. We derive the curvature invariants, horizon temperature, entropy, electric potential, and horizon-based mass function, and analyze local thermal response and the associated extended-phase-space identities. A rectangular benchmark cycle gives $W=0.095000$ and $\eta\simeq15.39\%$ for the explicitly stated geometric-unit parameters; these values are model- and cycle-specific and are not presented as a universal efficiency enhancement. For geometric optics, we derive the photon-sphere condition, critical impact parameter, and finite-radius shadow angle. For a massless, minimally coupled scalar field, we obtain a radial-coordinate Helmholtz equation and identify frequency-dependent barriers, evanescent intervals, and candidate trapping regions; the eikonal barrier-top scale approaches the photon-sphere capture scale. Two qualifications delimit the interpretation. First, with the stated PFDM stress-tensor convention, $\rho_{\rm PFDM}=-\beta/(8\pi r^3)$, so positive $\beta$ describes a negative-density effective source. Second, the mass function follows from the horizon equation, but identifying it with a conserved asymptotic AdS mass requires a separate charge and boundary-normalization analysis. The potential-well analysis likewise does not determine complex quasinormal frequencies. We therefore present the results as a consistent, parameter-specific analysis of the adopted phenomenological geometry, while distinguishing algebraic thermodynamic identities from conclusions requiring asymptotic-charge or spectral calculations.
Read the full abstract

All authors B. Eslam Panah, S. Gurtas Dogan, A. Guvendi, O. Mustafa

Materials Science

Deintercalation Equilibria Control the Composition, Structure, and Electronic Quality in the Synthesis of 2$M$-WS$_2$

Alkali-metal intercalation-deintercalation (AMID) enables access to metastable layered materials such as 2$M$-WS$_2$. Mechanistic studies of AMID have focused on intercalation; how deintercalation proceeds and influences crystal properties is comparatively poorly understood. Here, we assess 34 deintercalation treatments of 2$M$-K$_x$WS$_2$ and find that K$^+$ removal proceeds through coupled redox and ion-exchange equilibria, whose positions set the composition, structure, and electronic quality of the products. Across a series of neat solvents, K$^+$ deintercalation amount increased with solvent polarity and was driven to near-completion upon addition of nonoxidizing crown ether chelators, regardless of solvent polarity. Spontaneous reintercalation occurred when 2$M$-WS$_2$ was placed in dilute aqueous solutions of the strong base KOH but not the neutral salt KCl. These results suggest that the extent of deintercalation is governed by the stability of K$^+$ in solution, rather than oxidant strength alone, consistent with a reaction under thermodynamic control. Treatments that bias the deintercalation equilibrium most strongly toward the products while avoiding deleterious side reactions produce the highest quality 2$M$-WS$_2$. A six-hour treatment with acidified K$_2$Cr$_2$O$_7$ gave the highest average residual resistivity ratio ($\overline{RRR}$). Residual K$^+$, quantified by ICP-MS, was found in all treated samples and appeared to impose an upper bound on $\overline{RRR}$. Treatments yielding crystals with $\overline{RRR}$ below this bound did not always afford 2$M$-WS$_2$; we observed two new structures in bleach-treated samples and signatures of alternative phases under other conditions. This work demonstrates that deintercalation is not merely a processing step in AMID but a mechanistically rich reaction that might be rationally tuned to improve and control metastable materials.
Read the full abstract

All authors Brenna C. Bierman, Taeran Yang, Jon Ramos, Sara Yarworth, James Lazarcik, Daniel A. Rhodes

Computer Vision

Semantic Synthesis over Brute-Force Search: LLM-Guided Transformation Design for Efficient Learning

Data transformation pipelines shape the training distribution in computer vision, but designing them is a computationally expensive problem. We investigate whether large language models can enable efficient discovery by synthesizing executable transformation programs outside predefined augmentation spaces. We compare nine strategies under a unified training protocol, including standard augmentation policies, brute-force search, direct code generation with two LLMs, and performance-guided generation using LLMs fine-tuned on accuracy-ordered transformation pairs. When the best candidates are retrained for 20 epochs, LLM-generated transformation programs lead on every dataset, reaching 0.8877 accuracy on CIFAR-10, 0.5825 on CIFAR-100, and 0.9588 on SVHN, surpassing brute-force search over 6,000 candidates. The best transform on CIFAR-10 and SVHN comes from a fine-tuned generator using 270 candidates, while the best on CIFAR-100 comes from an off-the-shelf generator using only 10. This needs 22x fewer candidate evaluations in the generation step and 600x fewer for off-the-shelf generations, relative to brute-force search. Beyond transformation discovery, we analyze the reliability of the reduced-budget evaluation commonly used for augmentation search. Correlating one-epoch selection scores with 20-epoch outcomes reveals strong, weak, and absent predictive relationships depending on the dataset and architecture, with correlations ranging from 0.756 to near zero. A positive control indicates that range restriction alone is unlikely to explain the low correlations. Our results indicate that semantic synthesis with code LLMs provides an efficient alternative to search-based transformation discovery and highlight the importance of validating evaluation fidelity when optimizing data transformations.
Read the full abstract

Materials Science

In-Band Boundary Resonances from Coalescing Zero-Group-Velocity Points

Boundary states generally reside in band gaps, whereas in-band confinement typically requires exact decoupling from propagating modes. We demonstrate an alternative mechanism in reciprocal, topologically trivial Hermitian lattices arising from the coalescence of two zero-group-velocity (ZGV) points. At coalescence, the group velocity and inverse effective mass vanish simultaneously, and the corresponding Bloch root acquires higher multiplicity. The coalesced root remains propagating, whereas nearby root splitting yields weakly evanescent channels that, together with a bulk-matched termination, generate a boundary-weighted resonance within the propagating spectrum. The mechanism is realized analytically in a two-leg ladder model through the coalescence of time-reversal-related ZGVs and extends to distinct ZGVs produced by longer-range interactions. Finite-size scaling distinguishes the resulting resonance from extended and defect-bound states. Scanning laser Doppler vibrometry on a 3D-printed elastic ladder reveals boundary enhancement in a nearly flat band region, consistent with operation slightly below criticality. ZGV coalescence thus provides a bulk framework for in-band boundary resonances in reciprocal Hermitian lattices.
Read the full abstract

Robotics

BRACE: Belief-guided Residual Adaptive Control for Safe Navigation under Unknown Disturbances

Disturbances encountered during autonomous navigation generally vary across space, often discontinuously, as obstacles redirect wind and changing terrain alters traction. Tracking controllers typically assume sufficient control authority to mitigate disturbances and maintain safety; the regime where this assumption fails remains underexplored. We present Belief-guided Residual Adaptive Control (BRACE), a hierarchical reinforcement learning (RL) framework for reach-avoid navigation under unknown, bounded, spatially varying disturbances. BRACE infers a disturbance belief from model prediction errors and uses that to condition a high-level policy that adapts the reference trajectory. The policy combines a nominal and a belief-guided residual term, trained jointly with a likelihood-ratio advantage reweighting heuristic. We evaluate BRACE in three simulated quadcopter environments with abruptly switching wind fields, and validate it on Crazyflie hardware with fan-generated disturbances. Across these settings, BRACE improves reach-avoid performance over model predictive control and RL baselines.
Read the full abstract

All authors Jaeyoun Choi, Niklas Schmid, Oswin So, Augustinos Saravanos, Chuchu Fan

Materials Science

Lifting mechanical zero modes without dispersion

Floppy frameworks carry flat bands of zero-frequency mechanisms with no finite-frequency resonance, and couplings that lift them usually spread them. We show that an on-site tensor lifts a mechanism band without first-order dispersion exactly when, projected onto the mechanisms, it acts as wavevector-independent scalars on orthogonal pieces. Three structures achieve this: local normal modes, invariant symmetry sectors, and a balance law. A mechanism moves both ends of every bond between two vertex classes equally along it, and where the inter-class bond tensors are class scalars this balance law fixes its norm shares and hence the flat value from geometry alone. For the fluorite net the band lies outside the tensor spectrum, at 14/9 of the contact stiffness to first order, and stays flat to all orders in a fixed cage. A placement inside a tiling dual that keeps the common symmetry supplies such tensors, because each contact star inherits its site symmetry. Ten catalogued dual pairs lift flat; half are eigenvalue cases a resolved count matches, and the balance law explains flat bands in random cages too. Duality makes the rigid cage common. In a printable beam-frame lattice, contacts that lock joint rotation remove first-order bending dispersion whenever bonded mechanisms are orthogonal local normals.
Read the full abstract

Computation and Language

Can small transformers learn the language of Coxeter groups?

The field of mechanistic interpretability is well-suited to collaborative projects between mathematicians and AI researchers. This work describes one such project. We trained a small transformer to learn the language of the Coxeter group affine A2, best known as the symmetries of the tiling of the Euclidean plane by equilateral triangles. We then formulated and proved a word-theoretic conjecture about affine A2 after observing the attention patterns of the trained model.
Read the full abstract

All authors J. Elisenda Grigsby, Yiyao Huang, Arushi Katyal, Ryan Lin, Pranav Ramakrishnan, Weiya Wang

Metric Geometry

Compact metric spaces as exponentiable metric compact Hausdorff spaces

The class of compact metric spaces (metric spaces which induce a compact topology), despite being an important class of metric spaces, is not closed under many natural constructions. For instance, infinite metric products of compact metric spaces will often fail to induce a compact topology, while on the other hand, the topological product of any family of compact metric spaces is always a compact topological space, evidencing a discrepancy between topological and metric products. The notion of metric compact Hausdorff space – compact Hausdorff spaces equipped with a lower semi-continuous metric – provides a class of spaces which keep track of metric data and topological data separately, thereby forming a complete and cocomplete category, while including all compact metric spaces. Our main contribution is that this inclusion can be characterised via universal property: the compact metric spaces are the exponentiable metric compact Hausdorff spaces with respect to a natural monoidal structure on the respective category.
Read the full abstract

Solar and Stellar Astrophysics

Stellar Disruption and Black Hole Retention in Massive Collisions: The Role of Thermonuclear Feedback

A stellar-mass black hole (BH) that collides with a massive star can be captured by drag and settle at its center as a black hole star. Hydrodynamic simulations that neglect nuclear reactions have found that the star survives and becomes a black hole star (BH*) for BH-to-star mass ratios $q\lesssim0.2$. We perform hydrodynamic simulations of head-on collisions with the \athenak\ code, including CNO burning coupled to the gas, for $q=0.1$ and $0.2$ BHs, stellar masses of $100$ and $500\,M_{\odot}$, solar, metal-deficient, and Helium-rich compositions. In all models (with $q\le0.2$), the energy dissipated by drag is comparable to or less than the binding energy that the captured BH adds to the gas so that it alone does not disrupt the star. But, under compression around the captured BH, gas is heated to $\gtrsim2\times10^8$\,K and undergoes thermonuclear burning at the $\beta$-limited hot-CNO rate. Within about one dynamical time after the BH settles at the center, the nuclear energy released during the encounter becomes a significant fraction of the star's binding energy $\tfrac32\bar{\beta}\int P\,dV$. The star loses most of its mass once the cumulative release reaches this binding energy. Below this threshold, the release rate declines. Merger between a solar-composition star and a BH with a mass ratio $q=0.2$ leads to the disruption of the star after releasing most of its total mass. But a metal-poor star, with nearly the same binding energy and total yield, survives with minor release after merger with an identical BH. After coalescence with a less massive ($q=0.1$) BH, stars with both solar and metal-poor composition survive as black hole stars, burning at a rate $\gtrsim 10^{44}\,{\rm erg\,s^{-1}}$, far from thermal equilibrium. Very massive and helium-rich stars, which are weakly bound, are the most easily disrupted.
Read the full abstract