High Energy Physics (Theory)
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 27 August
Can a finite quantum episode populate a classical future that the same prepared state never reaches deterministically? We show that it can, using inflation as a cosmological laboratory. In our supergravity realization, two classical continuations already exist, but the prepared state reaches only one without noise. A stable entropy direction temporarily becomes light without a tachyonic instability, allowing mode-derived phase-space fluctuations to access the competing continuation. Finite-horizon first passage assigns it a \(0.33\)–\(0.40\) conditional weight across three nearby CMB-normalized realizations, while suppressing the stochastic source restores its exact deterministic zero. The entropy sector then becomes heavy again, and the selected histories continue classically. A finite quantum episode can thus leave a lasting redistribution among classical futures.
Quantum Physics
Stochastic transport of a Goldstone mode in a self-organized atomic crystal
Spontaneous breaking of a continuous symmetry produces a massless Goldstone mode that can evolve across a degenerate manifold at zero energy cost. Goldstone modes have been identified primarily through excitation spectra, mode softening or collective oscillations. However, their time-domain transport under intrinsic fluctuations and dissipation has remained largely unexplored. Here we directly track the stochastic transport of a Goldstone mode in a self-organized atomic crystal inside an optical ring cavity. The ring cavity maps the order-parameter phase onto the real-space position of the emergent crystal. Without any external perturbation, fundamental photon-scattering recoil drives the collective transport, while cavity dissipation generates friction. We monitor individual trajectories of the atoms and their self-generated optical lattice by measuring the cavity output phase. We find that the diffusion constant decreases as $1/N$, indicating that all atoms move collectively as a rigid object rather than independently. By tuning the Langevin driving force and cavity-mediated damping, we show that the normalized diffusion constant collapses onto a single universal curve. This work extends the study of continuous symmetry breaking from excitation-frequency measurements to real-time tracking of transport, and opens routes for studying non-equilibrium collective transport, phonon dynamics, and defect formation in driven-dissipative quantum matter.
Read the full abstract
Discrete Mathematics
The Project Scheduling Interdiction Problem with Delay Groups
Large-scale projects are frequently delayed by correlated disruptions: when a shared input such as a common supplier, a specialized team, or a supporting platform degrades, all dependent activities are slowed down simultaneously. This paper introduces delay groups to capture such disruptions: a delay group is a set of activities whose delays stem from a common cause, described jointly by an uncertainty set. Our model takes the perspective of an interdictor that, subject to a budget of $k$ groups, selects which groups to disrupt so as to maximize the project makespan. The interdictor can extend activity durations within each disrupted group by delays from a group-specific uncertainty set, while non-disrupted activities keep their nominal duration. We study the complexity of the resulting Project Scheduling Interdiction Problem with Delay Groups (PSIP-DG), which provides a worst-case stress test of the schedule. The problem is computationally intractable ($N\!P$-hard) for general polyhedral uncertainty sets, even for a single delay group with a continuous knapsack constraint. For budgeted uncertainty sets, we prove $N\!P$-hardness both when all activities of a disrupted group are delayed and when only one activity per group may be delayed, and derive an inapproximability bound of $1-1/e+\epsilon$ for the former case. We further develop a greedy heuristic with approximation guarantee $k$ and two structure-based heuristics with initializations and neighborhoods from tractable special cases. Experiments on $5{,}000$-activity networks show that the heuristics match the solution quality of an exact solver at substantially lower running times, in some cases finding strictly better solutions.
Read the full abstract
Chemical Physics
Real-Time Emergence of Charge-Transfer-to-Solvent States from Core Excitation
Charge-transfer-to-solvent (CTTS) excitations provide a chemically central route to generating hydrated electrons and initiating redox chemistry in solution, yet the earliest stage of CTTS—the formation of the excited state itself—is usually treated as instantaneous. Here we present a time-domain perspective of how CTTS character builds up during core-level photoexcitation of an aqueous metal ion. Using time-dependent configuration interaction, we simulate the coherent evolution of a dense manifold of core-excited states and track the ultrafast flow of electronic charge from the initially localized site into solvent-supported final states. We find that the dynamics evolves from a few-state, oscillatory behavior to effectively irreversible delocalization, as the charge disperses among many coupled configurations, providing a microscopic mechanism for the early-time emergence of CTTS character. Our results offer a transparent real-time interpretation of what core-level spectroscopies and core-hole-clock-type measurements can probe in solutions, outlining experimental signatures for probing the build-up of CTTS states on the core-hole-lifetime timescale. Looking ahead, attosecond and sub-femtosecond X-ray pump–probe approaches at X-ray free-electron lasers provide a realistic route to directly time-resolve the core-excited CTTS wave packets in solution.
Read the full abstract
Robotics
FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference
Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution. This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. We introduce FlashVLA, a streaming action decoding framework that addresses both challenges in a unified formulation. FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. This design allows FlashVLA to produce one executable action chunk per inference step. Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. It can achieve $\geq$30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment.
Read the full abstract
Mesoscale and Nanoscale Physics
Scalable, Simple, and Versatile Encapsulation of 2D Materials and Devices
Air-sensitive 2D materials present a fundamental challenge for device integration. Encapsulation is often required to preserve intrinsic properties, yet conventional protection strategies often fail for thicker layers and complicate fabrication. Here, we demonstrate that electron-beam (e-beam) evaporated aluminum oxide ($\mathrm{AlO}_x$) serves as both an effective encapsulation layer and a platform for direct device fabrication. Unlike transfer-based approaches, this scalable method is compatible with thicker flakes and full device or wafer coverage. It requires no stacking procedures and enables contacts without post-encapsulation etching. Using rare-earth tritellurides ($\mathrm{RTe}_3$, R = La, Er), semimetallic $\mathrm{WTe}_2$, and superconducting $\mathrm{FeTe}_x\mathrm{Se}_{1-x}$, we show that $\mathrm{AlO}_x$ suppresses oxidation and preserves intrinsic optical and electronic properties. We establish substrate-dependent optimization of encapsulation across a range of flake thicknesses, demonstrate that ultrathin $\mathrm{AlO}_x$ preserves $\mathrm{WTe}_2$'s plasmonic response and maintains superconducting performance in $\mathrm{FeTe}_x\mathrm{Se}_{1-x}$. Thus we overcome the longstanding tradeoff between encapsulation and straightforward device fabrication in fragile quantum materials.
Read the full abstract
Game Theory
Token-Level Advertising
Generative AI is transforming how people access information, challenging traditional advertising mechanisms built around predefined slots. Towards generation-native advertising, we propose the Latent Advertiser Mixture Auction (LAMA), a token-level advertising mechanism that embeds advertiser influence directly into the generation process. Advertisers report local continuation values that induce advertiser-specific next-token policies, from which the platform decodes through a latent mixture while updating an allocation posterior. We show that LAMA satisfies Markov DSIC and IR, and achieves near-optimal KL-regularized welfare. We further develop a learning-based implementation that reconstructs the required reports online from learned local advantages and root values. Proof-of-concept experiments on real-world commercial-search query splits show that LAMA improves platform welfare and revenue while maintaining user-facing response quality, providing initial evidence for the feasibility of generation-native advertising.
Read the full abstract
Combinatorics
Boxicity and Threshold Dimension of Zero Divisor Graphs
The zero divisor graph $\Gamma(R)$ of a finite commutative ring $R$ has as vertices the non-zero zero divisors of $R$, with an edge between two elements exactly when their product is zero. We determine the boxicity and threshold dimension of $\Gamma(R)$ for two classes of finite commutative rings: reduced rings and quotients of principal ideal domains. Our proofs use a new combinatorial gadget, the integral covering graph, that captures the structure shared by both ring families and generalizes the disjointness graph on subsets of $[n]$, where two subsets are adjacent if and only if they are disjoint. In doing so, we answer two questions recently posed by L.~Sunil Chandran and Suraj Kumar Sahoo in Boxicity of Zero Divisor Graphs, Discrete Applied Mathematics 391 (2026).
Read the full abstract
Computation and Language
D2C-Routing: Dimension-to-Composition Evidence Routing for Mixed-Origin AI-Generated Text Detection
AI-generated text detection is commonly framed as a binary document-level judgment about whether a text is human-written or machine-generated. This framing breaks down for mixed-origin writing, where content origin and expression origin may differ. We cast mixed-origin detection as dimension-to-composition source attribution, inferring content origin and expression origin before composing them into four collaboration types. We propose Dimension-to-Composition Routing (D2C-Routing), which routes content-side and expression-side evidence to supervised dimension heads before a learned gated composition layer predicts the final label. On MixD2C, a reconstructed split derived from the HART mixed-origin benchmark, our disclosed D2C-Routing-based detector system reaches 0.8603 four-way Avg TPR@1%FPR, 6.5 points above the same-split RACE-local rerun. Core ablations support the routing design, while error analysis shows that distinguishing AI-content/human-expression from fully AI-generated text remains the hardest boundary. Code is available at https://github.com/bystander563/d2c-routing-artifact.
Read the full abstract
Instrumentation and Methods for Astrophysics
The MICADO first light imager for the ELT: first on-bench performance tests of the SCAO hybrid LQG+Integrator controller
The Extremely Large Telecope (ELT) is expected to begin scientific operations in 2030, with MICADO as its first-light imager. The Single Conjugate Adaptive Optics (SCAO) system for MICADO is currently being assembled and tested at the Observatoire de Paris. MICADO's baseline AO control strategy uses a hybrid approach: a Linear Quadratic Gaussian (LQG) regulator for low-order modes, optimised to adapt to the disturbance caused by vibrations and ELT windshake, and an integral action controller for higher-oder modes. Despite the non linearity of the Pyramid Wavefront Sensor (PyWFS), the so-called "optical gains", LQG control has proven to be highly effective in simulation. It has been shown that the LQG regulator is resilient to optical gain variations in the loop or to inaccuracies in their estimation. Increasing the number of modes controlled by the LQG regulator would therefore make the controller less sensitive to the optical gains, which are notoriously difficult to estimate during operation. A first hybrid controller, with only tip-tilt controlled with LQG control, has been validated using COMPASS, the end-to-end, GPU-accelerated, AO simulation platform. We will present the first experimental validation of this hybrid controller on the MICADO-SCAO bench. We will also present results obtained by increasing the number of modes controlled with LQG control and the work carried out to go towards an autonomous LQG regulator.
Read the full abstract
Fluid Dynamics
How well can Diffusion Models learn Lagrangian-Tracer Statistics in Non-reciprocal Turbulence?
Recent advances in generative artificial intelligence have led to significant potential applications in conventional fluid flows, including those that are turbulent. Can these methods be carried over to studies of novel types of turbulence, such as turbulence induced by non-reciprocity in binary-fluid systems? To answer this question, we analyze the statistics of Lagrangian-tracer particles in non-reciprocal binary-fluid turbulence, which has been studied recently in the non-reciprocal Cahn-Hilliard-Navier-Stokes (NRCHNS). We obtain our ground-truth data via extensive pseudospectral direct numerical simulations (DNSs) of the two-dimensionsl (2D) NRCHNS model. Our study yields a variety of intriguing results for probability distribution functions (PDFs) for particle accelerations and velocity-component PDFs; the latter turn out to be bimodal, completely unlike their 2D-fluid-turbulence counterparts. We relate this bimodality to lane-type structures in Eulerian-velocity components. Furthermore, we characterize Lagrangian multiscaling via Lagrangian velocity increments, their structure functions and flatnesses, and multiscaling exponent ratios, for the first time in non-reciprocal hydrodynamics. Finally, we use generative diffusion models to obtain synthetic Lagrangian trajectories for the NRCHNS system, assess how effectively they can emulate the Lagrangian statistics that we obtain from our DNSs, and highlight open challenges in the application of generative artificial intelligence in non-reciprocal systems.
Read the full abstract
Mesoscale and Nanoscale Physics
Mobility Enhancement in Si/SiGe Quantum Well Enabled by a Buried Si Layer Trapping Oxygen Impurities
Reducing disorder in undoped Si/SiGe field-effect heterostructures remains an important materials challenge for scalable quantum devices, particularly electron spin qubits. Background impurities such as oxygen have been identified as mobility-limiting, yet practical heterostructure-design strategies for suppressing their incorporation remain underexplored, and their influence across different transport regimes is not fully established. Here, we demonstrate a simple route to oxygen reduction and mobility enhancement in Si/SiGe quantum-well heterostructures grown by reduced-pressure chemical vapor deposition (RP-CVD) on 200 mm Si(100) substrates through the introduction of a thin, electrically passive buried Si layer within the lower SiGe barrier. Secondary-ion mass spectrometry shows that the buried Si layer reproducibly reduces the oxygen background in the subsequently grown SiGe by approximately a factor of five, without modifying the active quantum-well region. Density- and temperature-dependent magnetotransport measurements further show that this reduction increases the electron mobility, while leaving the percolation density and density-dependent mobility scaling largely unchanged. Upon cooling to 0.3 K, both high- and low-oxygen devices exhibit similar density-dependent fractional mobility enhancements, indicating that the reduced oxygen background improves momentum relaxation without substantially altering the dominant low-density disorder landscape. These results establish the buried Si layer as a straightforward and process-compatible heterostructure-design element for reducing oxygen incorporation and improving transport in 200 mm CVD-grown Si/SiGe quantum-device materials.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Near universality of nonlinear transverse and radial velocity responses in spherical collapse with arbitrary radial profiles
The nonlinear relation between density and expansion is usually formulated for a homogeneous spherical top hat, which has a single local Hubble rate. A smooth spherical profile instead expands differently along and across the radial direction-a direct signature of radial inhomogeneity. We show that, for growing-mode pressureless matter with a cosmological constant, the complete shellwise response is nevertheless fixed by one top-hat function. Given the local and enclosed density contrasts, a transverse response and its derivative determine the transverse and radial expansion rates. The linear and second-order limits are algebraic maps whose only dynamical input is the usual growth rate. An exact equal-age construction supplies the nonlinear response without integrating an evolution equation. We check that it reconstructs full Lambda-Lemaitre-Tolman-Bondi profiles to numerical precision. We also provide a derivative-aware, cosmology-independent three-term symbolic fit that requires only the growth rate and the local and enclosed density contrasts. Across a representative set of matter-curvature-redshift combinations and for a shell located in the compensated transition, the maximum relative errors are 0.3% and 0.7% in the transverse and radial responses, respectively. This compact formulation separates the production of a density profile from its expansion response and makes the effect of radial gradients explicit.
Read the full abstract
Number Theory
Delta theory of Anderson Modules II: Hodge-Pink structure
In this article, using the theory of $\delta$-geometry, we construct a canonical $z$-isocrystal $(\mathbf{H}_\delta(E), \mathfrak{f}^*)$ admitting a Hodge-Pink structure for any abelian Anderson module $E$. The Hodge-Pink structure on $\mathbf{H}_\delta(E)$ induces a natural filtration $(\mathbf{H}_\delta(E) \supset \mathbf{X}_{\mathrm{prim}}(E) \supset \{0\})$. The elements of $\mathbf{X}_{\mathrm{prim}}(E)$ are represented by primitive delta characters associated to $E$. We establish a natural morphism from $\mathbf{H}_\delta(E)$ to the associated de Rham cohomology module $\mathbf{H}^*_{\mathrm{dR}}(E)$, which is strictly compatible with the aforementioned filtration and the classical Hodge filtration $(\mathbf{H}^{*}_{\mathrm{dR}}(E)\supset {\mathrm{Lie}(E)^{*}}\supset \{0\})$ on $\mathbf{H}^*_{\mathrm{dR}}(E)$. Moreover, we show that the map induces an isomorphism between $\mathbf{X}_{\mathrm{prim}}(E)$ and $\mathrm{Lie}(E)^*$. Hence our isomorphism provides an interesting interpretation of the invariant differentials of $E$ as primitive delta characters of $E$. Furthermore, when $E$ is a Drinfeld module, we show that the constructed $z$-isocrystal $\mathbf{H}_\delta(E)$ is weakly admissible. Consequently, the positive equal characteristic analogue of the Fontaine functor associates a crystalline $z$-adic Galois representation to the $\delta$-geometric object $\mathbf{H}_\delta(E)$. In the case, when $E$ is the Carlitz module, we show that the Galois representation associated to $\mathbf{H}_\delta(E)$ is indeed the usual one coming from the Tate module.
Read the full abstract
Probability
Distribution-constrained optimal multiple stopping: the Root-type solution
We consider the distribution-constrained optimal stopping problem introduced by Bayraktar and Miller (Mathematical Finance, 2019) and Beiglbock et al. (PTRF, 2018). Motivated by the multi-marginal Skorokhod embedding problem (SEP) and applications in financial mathematics, we generalize the Root-type solution to the multi-marginal case. The key difficulty is that the associated stopping barriers need not be ordered, so the problem in general cannot be reduced to a sequence of one-marginal problems. First, we give a probabilistic characterization in terms of sequential optimal stopping for an auxiliary time-reversed process, in the same spirit as Cox et al. (PTRF, 2019). Then, we prove the optimality of this construction by martingale arguments for a general class of reward functions, including multi-marginal versions of all examples in Beiglbock et al. (PTRF, 2018) as special cases.
Read the full abstract
Astrophysics of Galaxies
Non-isothermal vertical distribution functions for the Milky Way
The vertical structure of our Galaxy has commonly been assumed to follow a pseudo-isothermal distribution. However, there is no a priori reason to expect this form to arise from scattering by giant molecular clouds (GMCs), since GMCs are confined to a narrow layer around the Galactic midplane and therefore do not randomise stars uniformly in vertical phase space. We first present a simple statistical argument to derive a new vertical distribution function (DF), in the limiting case of a razor-thin distribution of GMCs, which admits an extra factor of ${J_z}^{-\alpha}$, where $\alpha>0$ and varies for different potentials. In light of this, we revisit the diffusion coefficients for vertical heating from the mechanics of two-body scattering, in a more realistic, extended distribution of GMCs. The resulting drift and diffusion coefficients are strongly non-linear in $v_z$, particularly at low relative velocities, where the effective range of weak, local encounters can vanish. We determine stationary solutions, and also numerically solve the orbit-averaged Fokker–Planck equation with time-dependent coefficients. We find, in both cases, a vertical action distribution that deviates from that of the pseudo-isothermal case, and propose a new, family of vertical DFs that generalise the pseudo-isothermal form. We find the two primary agents of non-isothermality are: (i) a GMC distribution confined near the midplane, and (ii) the fact that the Coulomb logarithm, $\Lambda$ is strongly velocity-dependent at small velocities, and cannot be treated as a constant, contrary to the commonly adopted assumption.
Read the full abstract
Probability
Universality and sharp thresholds for ellipsoid fitting
We establish a sharp phase transition for fitting random vectors by an ellipsoid. The random vectors have independent subgaussian coordinates with mean zero, variance one, and a common fourth moment, and the number of vectors is proportional to the square of the dimension. We identify an explicit satisfiability threshold such that, with high probability, a positive definite ellipsoid passes through every data point below the threshold, whereas no positive semidefinite fit exists above it. We also determine the optimal squared fitting error throughout the unsatisfiable regime. In particular, the threshold depends on the coordinate distributions only through their common fourth moment, revealing a fourth moment universality phenomenon. For standard Gaussian data the threshold is $1/4$, resolving the ellipsoid fitting conjecture.
Read the full abstract
Robotics
Embodied Scene Rearrangement Planning
This paper introduces Embodied Scene Rearrangement Planning (ESRP), a novel task requiring embodied agents to rearrange furniture in 3D scenes to match a target configuration using only egocentric observations and a top-down target layout. Unlike prior rearrangement tasks, ESRP precludes global state access and introduces mutual object occlusions, reflecting the practical constraints of real-world robotic deployment. These factors make aligning partial egocentric observations with the global target layout particularly challenging for long-horizon planning. To facilitate research, we present ESRP-Bench, a comprehensive benchmark built on OmniGibson featuring over 5,400 scene pairs and 8,200 objects. We define three multi-level metrics to evaluate rearrangement quality and provide four baselines: a hierarchical task-and-motion planning method, a vision-language-model-based method, and two learning-based approaches (IL and RL). Experimental results demonstrate that current methods struggle to complete the task efficiently, highlighting ESRP as a challenging frontier for embodied agents in scene understanding and long-horizon task planning. This work serves as a stepping stone toward deploying intelligent agents in real-world scenarios. Project page: https://pie-lab.cn/ESRP/.
Read the full abstract
Computation and Language
Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090
Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities. Although strong open-source efforts already exist, including open-weight models and open-source training recipes, a cost-efficient, hardware-accessible, and open-source pretraining recipe has long been missing. Even at a small scale, training Llama-3.2-3B costs over \$1.5M, and reproducing SmolLM3-3B needs over \$700K. In this report, we present an open pretraining recipe designed to lower this barrier. Using this recipe, we train a collection of Puro-2B models from scratch on up to 1.4 trillion tokens with FP8 precision on consumer-grade RTX 5090 GPUs. The models in the collection differ in token budgets and selected recipe variants. Our best model is trained at a compute cost of less than \$6.9K and approaches Qwen2.5-1.5B performance under our evaluation protocol. This cost efficiency is enabled by a combination of approaches, including hardware selection, low-precision training, hyperball optimization, curriculum model averaging, and the data recipe. Beyond the recipe itself, we provide two additional results. First, across the Puro-2B collection, we derive a Puro Cost Scaling Law that relates training cost to average model performance; the fitted law suggests that about \$4.4K, less than \$5,090, is sufficient to reach the performance of Qwen2-1.5B. Second, as an end-to-end case study, we examine how pretraining data curricula shape downstream performance after post-training. Such controlled studies are enabled by having access to the full pretraining pipeline rather than model weights alone. We release the full training recipe for Puro-2B, including data, code, and model weights under Apache 2.0 at https://huggingface.co/collections/thu-pacman/puro-2b.
Read the full abstract
Statistical Computation
Deep-Control BSDE: Layerwise Brownian-Weighted Regression for High-Dimensional Semilinear PDEs
High-dimensional semilinear parabolic partial differential equations arise in stochastic control, financial engineering, and uncertainty quantification, but classical spatial discretizations suffer from the curse of dimensionality. Motivated by Gaussian perturbation and conditional regression in denoising score matching, we propose Deep-Control BSDE (\DCBSDE{}), a layerwise control-regression method for Markovian backward stochastic differential equations. From an implicit Euler scheme, we identify the discrete ideal control $z_n^\pi=\mathcal M_nu_{n+1}^\pi$ as the conditional projection coefficient of the successor value response onto the one-step Brownian increment. At each time level, the method freezes the successor value function, approximates the resulting Brownian conditional-moment target using finitely many branches, regresses the control, and then fits the value through the implicit BSDE relation. A frozen linear-response baseline and antithetic pairing reduce finite-branch fluctuations while preserving the conditional target, and a within-layer correction coordinates the value and control approximations. We establish backward stability through a contraction property of the Brownian projection and derive conditional consistency when discretization, local learning, numerical, and finite-branch errors vanish jointly. Experiments across six benchmarks demonstrate that \DCBSDE{} achieves a favorable overall balance among value accuracy, control accuracy, and dynamic consistency while exhibiting generally stable performance across random seeds. These results support the use of \DCBSDE{} in applications requiring reliable joint approximation of the value and control processes.
Read the full abstract
Optimization and Control
UGM: A Unified Framework and New Perspectives for Accelerated Gradient Methods in Smooth and Strongly Convex Optimization
In this paper, we propose a unified framework for accelerated gradient methods, dubbed UGM, which subsumes a wide range of accelerated and conventional gradient-type methods designed for minimizing $L$-smooth and $\mu$-strongly convex functions. We demonstrate that the iteration update of the proposed framework can be intrinsically interpreted as a hybrid combination of the heavy-ball method and vanilla gradient descent. This interpretation reveals that classical accelerated gradient methods essentially integrate a conservative gradient descent step into the fast yet unstable heavy-ball dynamics, which achieves a favorable trade-off between acceleration and stability. We further establish a unified convergence analysis using Lyapunov functions. Guided by our analysis, we develop a family of enhanced accelerated gradient algorithms that leverage the inner product relationship between gradient information and iterative variables to optimize iterative updates. Extensive numerical experiments on unconstrained quadratic optimization and logistic regression validate that the proposed algorithms achieve superior performance compared with existing baseline methods under typical structural conditions.
Read the full abstract
Computer Vision
Successive Capacity Growth: Task-Complexity-Driven Width and Depth Expansion for Vision Transformer Encoders in JEPA World Models
Joint-Embedding Predictive Architectures (JEPAs) for world modeling typically employ fixed-size Vision Transformer encoders that are over-provisioned for simple tasks and under-provisioned for complex ones, with significant redundancy across attention heads. We propose Successive Capacity Growth (SCG), a method that starts from a minimal encoder (1 head, 2 layers, 283K parameters) and grows incrementally in width (adding attention heads for low-level semantic capacity) or depth (adding transformer blocks for higher-order semantic abstraction), driven by a task-agnostic test-and-verify mechanism that exploits function-preserving expansion to safely trial architectural changes and roll back if they do not improve prediction loss. The Sketched Isotropic Gaussian Regularizer (SIGReg) ensures that all learned semantic dimensions remain statistically independent and aligned with the predictive objective, preventing collapse even as the architecture grows. On a 60-dimensional multi-object dynamics task, SCG naturally triggers depth expansion, improving prediction loss by 20.3% over the fixed small baseline with 56 times greater parameter efficiency than scaling to the fixed large model; on a 2D navigation task, a single width expansion yields even an 23% improvement over the fixed large model. Across all three tested environments of increasing complexity, the adaptive encoder matches or exceeds the fixed small baseline, with zero false-positive expansions and bit-exact function preservation (ratio = 1.0, absolute difference = 0.0). The take-away is that JEPA world model encoders need not be pre-allocated at maximum capacity - they can grow successively as the task demands, achieving significant compute and data efficiency while maintaining representation quality.
Read the full abstract
Information Retrieval
Stageboost: Recommending Signals Based on Counterfactual Estimation
Signals are short textual or visual snippets displayed on the eBay View-Item (VI) page, providing additional, contextual information for users about the viewed item. The aim of displaying these signals is to facilitate intelligent purchase and to incentivize engagement. In this paper, we present a 2 stage xgboost based model that optimally populates the VI page with signals. This approach has shown a 0.08% lift in overall GMB (Gross Merchandise Bought) and 0.58% increase in Parts and Accessories GMB, primarily due to increase in conversion of high average price items in online experimentation.
Read the full abstract
Computer Vision
KnockGS:interaction-Grounded Calibrationof Physical Gaussian Representations
Physics-integrated 3D Gaussian representations now allow reconstructed deformable objects to be simulated and rendered under explicit material models. Existing pipelines, however, assume that material parameters are known or manually specified, limiting their applicability when these parameters must be inferred from observed object dynamics. We propose KnockGS, an interaction-response PhysicalGS framework that estimates the elasticity and density scales of a 3D Gaussian object from its dynamics under a known applied force. Rather than treating physical simulation only as a forward process, we turn the force-induced response into a calibration signal: temporal response features are xtracted from the observed dynamics, the two material scales are estimated from those features, and the estimate is then frozen and written back into the same simulator so that it can be tested on an interaction it was never fitted to.We evaluate the framework on both parameter recovery and response-level fidelity. The estimated scales are compared against hidden ground truth, and the re-simulated object is measured against the target using 3D particle trajectories, response-curve statistics, and rendered-frame quality. Across five held-out material targets, our method recovers the scales substantially more accurately than response retrieval, global regression, or a fixed default material, and the frozen estimate remains predictive under interactions that differ in direction and in magnitude. Interaction response therefore carries enough information to calibrate material scales in physically grounded 3D Gaussian representations.Our study is a first step toward interactive PhysicalGS systems that calibrate a Gaussian asset whose rendered appearance and simulated response are consistent.
Read the full abstract
Artificial Intelligence
Sophistication in GenAI Use: Field Evidence from a Large Firm
We study how sophistication in generative AI (genAI) use varies among the back-office workforce of a large firm. Using proprietary data, we observe 713,564 employee prompts and their corresponding large language model responses from nearly 4,000 back-office employees across 15 functional areas over eight months in 2025. We document three main findings. First, senior employees exhibit more sophisticated genAI use, consistent with domain expertise complementing genAI capabilities. Second, sophistication varies considerably across functions and is highest in Strategy, Digital Innovation, and Project Management, three groups that share a focus on firmwide strategic initiatives and organizational change. Third, we observe neither improvements in sophistication over time nor lasting improvements following formal AI training, suggesting that sophisticated use can be difficult to change. Together, our study provides measures of and insights into sophisticated genAI use that managers can use to improve outcomes and that researchers can use in future research.
Read the full abstract
Quantum Physics
On supporting affine functionals for Entanglement of Formation
In several articles, the authors assume that the convex roof structure of the EoF and finite-dimensionality of subsystems $A$ and $B$ guarantee the existence of the (global) supporting affine functional for the EoF at any state of the system $AB$. This means that for any state $\rho$ of $AB$ there is a Hermitian operator $\Lambda_\rho$ on $\mathcal{H}_{AB}=\mathcal{H}_A\otimes\mathcal{H}_B$ such that $E_F(\rho)=\mathrm{Tr}\Lambda_\rho\rho$ and $E_F(\sigma)\geq\mathrm{Tr}\Lambda_\rho\sigma$ for any state $\sigma$ of $AB$. We present an explicit example showing that, when $\rho$ is degenerate, this is not true even in the simplest case when $A$ and $B$ are qubit systems. The construction is based on the fact that the existence of a supporting affine functional for the EoF at a state $\rho$ is equivalent to the Lipschitz lower semicontinuity of the EoF at this state $\rho$. We use Wootters' formula and the help of Claude Fable 5 to find a state $\rho$ of the system $AB$ for which the latter property does not hold. We also describe conditions for the existence the local and global supporting affine functionals for the EoF at a given state of both finite and infinite-dimensional bipartite quantum systems. These conditions allow us to find Lipschitz lower semicontinuity bounds for the EoF at a given finite rank state $\rho$ (i.e. inequalities of the form $\,E_F(\rho)-E_F(\sigma)\leq C_\rho\|\rho-\sigma\|_1$) with and without restrictions on the support of the state $\sigma$.
Read the full abstract
Statistics Applications
How exceptional was the Big Three era? Extremes and persistence in men's professional tennis
Three players won 66 of the 81 Grand Slam titles contested between 2003 and 2023, and their era is widely held to be the most dominant in the history of tennis. Assessing it means comparing players who never met, so that every comparison passes through the opponents each did face. We measure dominance by how far a player stands above the field of his own day, how many stand there at once, and how long they remain, the last through averages over fixed windows rather than through unbroken runs, which are unstable when the strengths beneath them are themselves estimated. Strengths come from a Bayesian dynamic Bradley-Terry state space model, and dominance is characterized through exceedances of a high threshold. Applied to 197,926 matches played between 1968 and 2025, only Djokovic separates from the field on peak strength, Federer and Nadal being indistinguishable from Borg, McEnroe and Lendl. What distinguishes the recent period is not how high its best players stood but how long three of them stood there together. At its height three players held a top-three position in at least four fifths of all weeks, where in the strongest comparable decade of the past those positions were shared among four. The sharper anomaly lies between: for twelve years from 1990 the upper tail was not merely unoccupied but compressed.
Read the full abstract
Materials Science
Temperature dependence of the charge density from first principles: application to the (222) forbidden reflection in silicon
Forbidden reflections (FRs) in X-ray diffraction have inherently weak intensity and have long been studied, in particular in semiconductors like silicon. They serve as sensitive probes of symmetry breaking, local strain, impurities, and weak charge redistribution. Despite extensive experimental work, the theory of their temperature dependence has typically relied on simplified models. While atomic Debye-Waller factors work well on allowed reflections, their applicability to the valence charge between atoms, which determines the intensity of FRs such as Si (222), is questionable, and previous agreement between theory and experiment relied on ad-hoc Debye-Waller corrections. We compute the temperature-dependent valence charge density $\rho(\mathbf{r},T)$ of silicon from first principles, using two methods: (i) perturbation theory, and (ii) averaging over thermally distorted supercells in a non-perturbative approach. The perturbative expression for the charge density is far more demanding than that for electronic energies, since it depends on the wavefunctions themselves and requires an explicit sum over unoccupied bands. We use an acoustic sum rule to express the second derivatives of the potential in terms of first derivatives, making the expression tractable within existing frameworks. The (222) FR then follows directly from the Fourier transform of $\rho(\mathbf{r},T)$, with no ad-hoc factors. Both methods give similar results, in reasonable agreement with experiment, with thermal expansion noticeably affecting the temperature dependence. The charge density answers a question the measured intensities could not settle: how the valence charge actually redistributes with temperature. Relative to the rigid model, we find more charge in the bonds and less in the core regions, a redistribution that shows up in the intensity as a somewhat weaker temperature dependence of the (222) FR.
Read the full abstract
Computation and Language
Your Voice Cloning System is Secretly a Voice Anonymizer
Speaker anonymization suppresses speaker-identifying attributes from speech while preserving linguistic content and quality. We propose repurposing XTTSv2, a multilingual voice cloning model trained on 27k hours of speech, for speaker anonymization without retraining. Our key insight is that XTTSv2's voice cloning capabilities preserve prosodic structure independently of speaker identity, enabling voice conversion by conditioning on a pseudo-speaker. We introduce an iterative refinement strategy that balances privacy and utility by maximizing a harmonic mean of speaker dissimilarity and intelligibility. Evaluated on seven European languages across CommonVoice and Multilingual LibriSpeech, our system achieves near-optimal privacy (EER $\approx$ 0.49), competitive intelligibility, and substantially better speech quality than dedicated anonymization baselines, while requiring no language-specific training. We release the code here: https://github.com/rm00cr/coqui-tts.
Read the full abstract
Astrophysics of Galaxies
$N$-body simulations of the open cluster population in the Milky Way and the impact of GMC encounters
We investigate the population of open clusters in the Solar Neighbourhood and model the effect of encounters with giant molecular clouds (GMCs) over the last 1 Gyr. We combine a Galactic model with $N$-body simulations of 20692 unique clusters in the mass range $[50-24000]$ $\mathrm{M}_{\odot}$. Each cluster is simulated twice: with and without tidal forces from the GMCs. We find that an initial cluster mass function truncated at $\sim 9000$ $\mathrm{M}_{\odot}$ best reproduces the observed mass function evolution. For the age function, the observations show a decline in clusters for ages older than $\sim 1$ Myr, whereas our simulated clusters show a decline after $\sim 50$ Myr. The observed early disruption suggests that some clusters form supervirial, whereas our simulated clusters are created in virial equilibrium. Low-mass ($<600 \, \mathrm{M}_{\odot}$) clusters are most sensitive to GMC encounters, which accelerate their disruption in the first $\sim 200$ Myr. After $\sim 500$ Myr, the impact of GMCs becomes irrelevant since the clusters will have been destroyed regardless of whether they experience GMC encounters or not. The survival of intermediate-mass ($600-6000\, \mathrm{M}_{\odot}$) clusters is significantly reduced by GMCs at ages up to 1 Gyr. High-mass ($>6000 \, \mathrm{M}_{\odot}$) clusters survive to 1 Gyr with minimal disruption with or without GMCs. We find that clusters that have had strong GMC encounters within the last $20$ Myr should have tidal tails that are randomly orientated with respect to the Galactic centre.