Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Tuesday 8 September

Quantum Physics

Quantum algorithm for PageRank computation through multistep quantum resonant transitions

We present a quantum algorithm for obtaining a quantum state that encodes the PageRank vector of the Google matrix through multistep quantum resonant transition (mQRT). In the algorithm, the PageRank vector is encoded in the ground state of a problem Hamiltonian associated with the Google matrix. By dividing the web graph corresponding to the Google matrix into a series of subgraphs with nested structure, we construct a sequence of Hamiltonians based on the subgraphs to form a Hamiltonian evolution path from a simple initial Hamiltonian to the problem Hamiltonian. The ground state of the problem Hamiltonian is obtained by going through ground states of the intermediate Hamiltonians via QRT step by step. This algorithm requires only one ancillary qubit, and the runtime of the algorithm is proportional to the number of steps. It provides a new way for efficiently obtaining the quantum state of the PageRank vector of large-scale networks.
Read the full abstract

Computation and Language

IGT @ FinMMEval 2026 Task 2: Question-Type Prompting with Targeted Extraction for Multilingual Financial QA

We present the IGT system for PolyFiQA Task 2 of the FinMMEval Lab at CLEF 2026, a multilingual financial question answering task over English SEC filings and multilingual news articles (English, Chinese, Japanese, Spanish, Greek) for four companies. Our central observation is that the 344 development questions divide into two families requiring fundamentally different approaches: structured numeric types (R&D ratio, cash flow, capital expenditure) are best answered by direct keyword extraction on filing text, while synthesis types (investment strategy, capital allocation, top-three revenue focuses) require rule-based multilingual news passage selection. A dataset analysis reveals that 17-18 of 19 ground-truth reference answers per synthesis type share an exact evidence label prefix, whose unigram tokens contribute directly to ROUGE-1 overlap. The final system achieves development ROUGE-1 approximately 0.395, a 60% relative improvement over a generic RAG baseline (approximately 0.247), and ranks 3rd of 12 teams on the official test set with ROUGE-1 = 0.3071, Precision = 0.2821, and Recall = 0.4044.
Read the full abstract

Mathematical Physics

Energy diffusion in a pinned harmonic chain with free boundary condition and interactions of increasing range

We consider a finite two-dimensional pinned harmonic chain with an interaction range that diverges with the system size and with conservative rotational noise. Under a joint scaling in which the interaction range diverges microscopically but vanishes macroscopically, we prove that the expected em- pirical energy measures converge to the unique measure-valued weak solution of the heat equation.
Read the full abstract

Number Theory

The Fibonacci numbers are not an additive uniqueness set for multiplicative functions

Let $(F_n)_{n\geq 0}$ be the Fibonacci sequence. We show that a multiplicative function $f$ satisfying \[ f(F_n+F_m)=f(F_n)+f(F_m)\qquad(n,m\geq 1) \] need not be the identity function, even when $f$ takes positive integer values. This answers negatively a question posed by Spiro in 1992. The smallest example presented here is obtained from $F_{31}=557\cdot 2417$. We prove the required divisibility equivalence, formulate an abstract prime-signature construction, and give a practical criterion producing further examples. We also describe the AI-assisted search that led to the construction and provide a reproducible certificate checker.
Read the full abstract

Machine Learning

GPU-Enabled Large-Scale Optimization Using Randomized Linear Algebra

This paper introduces rlaopt, a PyTorch-based package for large-scale optimization and scientific computing using randomized numerical linear algebra (RandNLA). Despite substantial progress in RandNLA-based algorithms, few implementations combine GPU acceleration with a simple interface for specifying optimization problems. rlaopt addresses this gap by providing GPU-enabled solvers for positive-definite linear systems and convex empirical risk minimization with constraints and regularizers. These solvers use RandNLA to accelerate conjugate gradient (NystromPCG), operator splitting (NysADMM), and stochastic gradient methods (SAPPHIRE). Moreover, rlaopt includes a modeling language that lets users specify problems using natural mathematical syntax. rlaopt automatically checks compatibility with the selected solver and performs the required problem decomposition. The solvers also support differentiation through their iterations, enabling applications such as hyperparameter tuning. Experiments on ridge regression, bounded multinomial logistic regression, and bounded elastic net identify when randomized preconditioning improves performance and demonstrate substantial speedups from GPU execution. The package is open-source under an Apache license, with source code at https://github.com/udellgroup/rlaopt and version 0.1.0 available on PyPI.
Read the full abstract

All authors Pratik Rathore, Zachary Frangella, Parth Nobel, Xuning Hu, Madeleine Udell

Machine Learning

KBBQ: A Predictive Noise Law and the Limits of Spectrum Flattening in FP4 Quantization

We develop a second-order theory of quantization noise in matrix multiplication in which the quantization format is characterized by the variance it assigns to each element. The constant variance profile of integer quantization recovers existing integer-noise theory, while the multiplicative profile of floating-point rounding reduces the data dependence to a scalar, the participation factor $\kappa$, yielding a closed-form signal-to-noise-ratio law. The resulting functional also admits a closed-form upper bound $\kappa^{*}$ that no function-preserving linear transform can exceed and that is attained by a recent state-of-the-art method. Building on this analysis, we introduce KBBQ (Kappa-Braked Blockwise Quantization), which parameterizes the extent to which a transform approaches this ceiling. At W4A4, across four base models and two FP4 formats, KBBQ outperforms the prior state of the art without additional deployment-time computation.
Read the full abstract

Strongly Correlated Electrons

Observation of multipartite spin entanglement in a cuprate chain

Quantum materials are believed to host highly entangled states of matter, but probing and quantifying such entanglement has long remained experimentally elusive. Here, we use resonant inelastic X-ray scattering (RIXS) as a probe of multipartite spin entanglement. By projecting the RIXS cross section onto the appropriate spin excitation channel, we show that the Quantum Fisher Information can be extracted directly from the measured spin fluctuation spectrum. We detect at least 7-partite spin entanglement in the model one-dimensional cuprate Sr$_2$CuO$_3$, revealing one of the largest entanglement depths yet reported in a solid. The observed entanglement depth agrees with Hubbard-model expectations and persists to elevated temperatures. Our results establish RIXS as an experimental probe of many-body spin entanglement and provide a platform for extensions to charge and orbital sectors and nonequilibrium settings.
Read the full abstract

All authors S. F. R. TenHuisen, Z. Shen, V. Bhartiya, V. Menon, P. Sharma, H. Padma, Z. Guan, W. He, M. K. Lajer, J. Li, D. Banerjee, J. Pelliciari, I. A. Zaliznyak, G. D. Gu, M. D. Lukin, S. Johnston, M. P. M. Dean, V. Bisogni, Y. Wang, M. Mitrano

Machine Learning

Sparse Data Augmentation for Optimization with Provable Guarantees

In nonconvex optimization problems arising in geometric machine learning, data augmentation is commonly used to promote invariance by averaging empirical losses over transformations of the data. Computing the fully augmented objective, however, requires access to every element of the transformation group $G$, which may be prohibitively expensive when $G$ is large or accessible only through sampling. We study whether full augmentation can instead be approximated using a small, fixed sample of transformations acquired before optimization and reused thereafter. Under suitable regularity conditions, we show that, with probability at least $1-\delta$, gradient descent (GD) on the resulting sparsely augmented objective returns an $\varepsilon$-stationary point of the fully augmented objective using $\mathcal{O}\bigl((\log |G|+\log(1/\delta))/\varepsilon^2\bigr)$ group-transformation-oracle queries. By comparison, standard group stochastic gradient descent (group-SGD), which samples a fresh transformation at every iteration, uses $\mathcal{O}(1/\varepsilon^4)$ transformation queries. Therefore, gradient descent with fixed sparse augmentation requires fewer transformation queries than both GD applied to the fully augmented objective and group-SGD. Our proof techniques, which may be of independent interest, establish a uniform approximation of the full group-averaged gradient field by a random group average using spectral properties of group-induced operators and tools from representation theory.
Read the full abstract

Probability

Thick points of log-correlated Gaussian fields do not depend on the mollifier

The log-correlated Gaussian field (LGF) on $\mathbb{R}^d$ ($d \geq 2$) is a centered Gaussian random tempered distribution, defined modulo additive constants, whose covariance kernel is $\log(1/|x-y|)$. In two dimensions, the LGF coincides with the whole-plane Gaussian Free Field (GFF). Because the field is a distribution, studying its pointwise behavior requires regularization via convolution with a mollifier. A point is called $\alpha$-thick if the mollified field at that point grows like $\alpha \log(1/\epsilon)$ as the mollification scale $\epsilon \to 0$. A natural question is whether the set of $\alpha$-thick points depends on the choice of mollifier. We prove that for any two admissible mollifiers $\rho$ and $\sigma$ satisfying some mild conditions, the thick point sets coincide almost surely. The result holds in all dimensions $d \geq 2$, and in the special case $d = 2$ extends to the zero-boundary GFF on any open domain with harmonically non-trivial boundary via the Markov property.
Read the full abstract

Computation and Language

Jacap: Robust KV Cache Eviction via Jacobian-Based Nonlinear Information Capacity Preservation

Key-value (KV) cache eviction is essential for scaling long-context inference in Large Language Models. However, existing policies predominantly rely on empirical heuristics, lacking a rigorous characterization of token utility under the inherently nonlinear softmax attention mechanism. In this work, we rethink KV cache eviction through the lens of local information geometry, modeling the attention process as a nonlinear Gaussian communication channel. By performing a first-order Taylor expansion of the attention mapping, we derive the Jacobian Information Capacity, a novel objective that explicitly captures query relevance, softmax sensitivity, and structural diversity. Guided by this theory, we introduce Jacap, a capacity-aware eviction method that utilizes softmax-aware importance weighting and statistical leverage scores for subset selection. Extensive experiments across diverse architectures and benchmarks demonstrate that Jacap delivers superior performance in most scenarios, particularly in high-compression regimes.
Read the full abstract

All authors Jiaming Yang, Chenwei Tang, Liangli Zhen, Chenyang Zhang, Jiancheng Lv

Computer Vision

Observe Before You Alert: Adaptive Driver Alerting with Vision-Language Models

Driver alerting from dashcam video requires sequential decision-making under partial observability: a system must decide not only whether a scene is risky, but also when the evidence is sufficient to warn. Most existing accident anticipation models output a binary risk score, leaving ambiguous scenes to be handled by thresholding. We propose VLAlert, a vision-language alerting framework that casts warning generation as a tri-action policy over SILENT, OBSERVE, and ALERT. The OBSERVE action acts as an internal evidence-gathering decision that delays uncertain warnings and changes the next observation window, creating a lightweight perception-action loop for adaptive alerting. VLAlert uses Qwen3-VL-4B as a safety-evidence generator and pools hidden states from structured belief spans to form compact representations for danger estimation and policy prediction. We evaluate VLAlert on VLAlert-Bench, a unified per-tick benchmark from four real-world dashcam alert datasets, and further test transfer to held-out naturalistic ADAS takeover clips. On VLAlert-Bench validation, VLAlert achieves the highest deployment-oriented utility among tested baselines, with DAUS 0.4878 compared with 0.4752 for Open-BADAS, and improves AUROC, AP_tick, F1_t, and balanced accuracy from 0.610, 0.176, 0.276, and 0.581 to 0.689, 0.195, 0.297, and 0.648, respectively. On 221 held-out ADAS-TO-Critic clips, VLAlert improves R@5s from 74.2% to 88.7% and F1 from 0.585 to 0.686. These results indicate that adaptive observation and safety-focused VLM representations provide measurable gains for driver-facing alert decisions.
Read the full abstract

Algebraic Geometry

Intrinsic Restriction Traces and Toric Dynamics

We prove Morita invariance of the Campbell–Lind–Malkiewich–Ponto–Zakharevich restriction-system trace after passage to perfect modules. It therefore defines an intrinsic integral restriction trace for an exact endofunctor of a small idempotent-complete stable $\infty$-category. On $\pi_0$, the $m$-th ghost is the laced trace of the $m$-fold iterate, compatibly with Frobenius. For lattice-graded algebras, the ghost targets have twisted cocenters in degree zero; over the open parameter torus, this applies to the cyclic bimodule of Dinkins–Karpov–Krylov. For finite monomial endomorphisms of toric varieties, we construct motivic restriction classes whose ghosts are sums over cones fixed by the iterates. In one example, two classes have the same first ghost, while their second ghosts differ after rational Betti realization.
Read the full abstract

General Relativity and Quantum Cosmology

Scalarized Einstein-Euler-Heisenberg black holes at the approach to extremality

We investigate extremal black holes with scalar hair in the generic Einstein-Euler-Heisenberg (EEH)-scalar theory with two scalar couplings to the Maxwell term. One is an exponential coupling with coupling constant $\alpha$ and the other is its polynomial coupling. We first construct scalarized black holes on the cold (C)-horizon of EEH black holes described by mass $M$ and magnetic charge $P$ both at the linear level, through the existence curves $\alpha_n(q)$ with $q=P/M$, and as fully backreaction solutions. As extremality ($q=q_e$) is approached, all existence curves accumulate, following $\alpha_n-\alpha_c\sim 1/\ln^2(q_e-q)$, at the critical branch $\alpha_c$ fixed by the Breitenlohner-Freedman bound for the near-horizon (AdS$_2\times S^2$) throat. We obtain scalarized extremal black hole (SEBH) with constant secondary hair and it is recovered exactly from the entropy function approach working on the near-horizon throat. Its entropy is an attractor invariant, being independent of the asymptotic modulus, so the scalar hair remains secondary. Exploiting this constant scalar, we seek further SEBHs with scalar hair keeping its charge $Q_s$ for the zero asymptotic scalar ($\phi_\infty=0$). In the $(q,\alpha)$ plane, we observe that these extremal solutions occupy $q\ge q_e$. Hence, the existence curves $\alpha_n(q)$ existing for $q\le q_e$ and the extremal branches bound the scalarized domain from opposite sides and they meet only at $q=q_e$.
Read the full abstract

Dynamical Systems

A distributed-delay model for the El Niño Southern Oscillation with a minimum delay: a case study of the shifted linear chain trick

When modelling a system of interest with a delay differential equation, a distributed delay may be the appropriate modelling choice when the delayed response occurs over a significant range of times rather than with a single constant delay. In systems with a minimum physical transit or processing time, however, the delay kernel should respect a fixed and positive minimum delay. A shifted Erlang kernel is a convenient choice, because the shifted linear chain trick allows one to replace the convolution associated with this type of kernel by a finite auxiliary-variable representation with one constant delay. As we demonstrate with a case study of the Ghil-Zaliapin-Thompson (GZT) model of the El Niño Southern Oscillation with seasonal forcing and distributed delayed oceanic feedback, this reduction makes it possible to perform a bifurcation analysis with numerical continuation techniques for any width of the Erlang distribution. Specifically, we present the bifurcation and resonance structure of the distributed-delay GZT model in the plane of forcing strength and delay for different widths of the distribution. To ensure a like-for-like comparison and determine the influence of delay distribution, we present these results in terms of the effective delay and, moreover, rescale the feedback strength to account for its attenuation with increasing width of the distribution.
Read the full abstract

Artificial Intelligence

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how scheming arises from the interaction of key factors, such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. Prior work examines only a small number of scenarios, limiting the ability to isolate how these conditions shape an agent's propensity or capability to scheme. This limited scale and task diversity also restrict coverage of realistic deployment settings and the range of scheming strategies that can be observed. To this end, we introduce SCHEMEARENA, a 400-scenario benchmark for scalable scheming stress testing, constructed through a factorized scenario synthesis framework spanning diverse safety-relevant tool domains, instrumental goals, oversight conditions, and pressure mechanisms. To enable scalable and reliable monitoring, we further propose SCOUT, a scheming monitor that grounds multi-criteria judgments in evidence drawn from agents' reasoning and actions. Across controlled stress tests on five LLM agents, we find that explicit instrumental goals are the strongest driver of scheming propensity. Strategic hints play a distinct role by helping agents translate scheming reasoning into concrete covert behavior. Oversight has mixed effects: in several closed models, action-only monitoring increases scheming, suggesting that partial oversight can act as an optimization constraint rather than a deterrent. CoT is a useful but incomplete monitoring signal: it can reveal latent scheming before execution, yet action-only scheming shows that covert behavior may occur without explicit reasoning evidence. We release the benchmark, code, and monitor at: https://github.com/launchnlp/SchemeArena.
Read the full abstract

All authors Jie Ruan, Inderjeet Nair, Amy Liu, Muhammad Khalifa, Yusheng Zhou, Lu Wang

Algebraic Topology

Regular Bundles on Orbifolds: A Short Proof of Presentability

Let $\X$ be a Hausdorff second-countable smooth orbifold without boundary, possibly noncompact and ineffective. Suppose $\dim\X\leq n$ and $|G_x|\leq B$ for integers $n\geq0$ and $B\geq1$, where $G_x$ is the full stabilizer at $x$. We construct a smooth Hermitian bundle of explicit rank $R(n,B)$ whose fibre at $x$ is a positive multiple of $\C[G_x]$. Adams operations cancel the Bott obstructions on the boundary spheres of a locally finite good triangulation, and the connectivity of Stiefel manifolds gives actual bundles representing the resulting virtual classes. Smoothing preserves all stabilizer representations. Unitary frames give $\X\simeq[M/U(R(n,B))]$ for a smooth manifold $M$ with a proper locally free action; $M$ is compact exactly when $|\X|$ is compact. Thus every compact smooth orbifold is presentable.
Read the full abstract

Algebraic Geometry

Negative contacts in genus one: a comparison of punctured and root stack Gromov-Witten theories

Let $D$ be a smooth divisor in a smooth projective complex variety $X$. For connected curves of arithmetic genus one with prescribed signed contact orders, we prove that the refined punctured cycle of Battistella–Nabijou–Ranganathan and the negative-contact cycle of Fan–Wu–You agree after pushforward to the common moduli space of stable maps with divisor evaluations. Thus the pushed-forward refined punctured cycle is the constant coefficient of the pushed-forward root-stack virtual class, normalized by one power of the root order for each negative contact. The key step is a comparison for the universal target. After restricting to finite-type open substacks determined by the fixed pair $(X,D)$ and numerical data $\Gamma$, we prove that the positive BNR space maps finitely and with generic degree one onto Crumplin's main component. Using Crumplin's genus-one component description and degree formulas, we identify this component's fundamental cycle with the constant coefficient of the universal orbifold virtual class under comparison of root orders. Refined zero-section pullback recovers the negative contacts, and compatible virtual pullbacks and root-forgetting pushforwards transfer the resulting identity to $(X,D)$.
Read the full abstract

Robotics

DISEIL: Demonstration Distillation for Sample-Efficient Imitation Learning

A robot that can be taught a new task from a handful of demonstrations has to work out for itself what it still cannot do, and then ask for exactly that. Interactive imitation learning takes a step in that direction by letting a policy practice on its own and calling an expert when it goes wrong. Existing methods decide when to interrupt the learner. A further 2 decisions are left to whichever episode happened to trigger the interruption: which failure to correct, and where the demonstration should start. This paper is a first attempt at making both of them deliberately. DISEIL (Demonstration dIstillation for Sample-Efficient Imitation Learning) marks each failed episode at the step where the policy first becomes unreliable, represents that moment with a geometric descriptor, and groups the failures into recurring failure modes. A vision-language model and a language model read the selected mode and write a request for the next demonstration, and a store of task constraints checks that the request can be carried out before any expert time is spent. No model produces a robot action. Across 5 simulated tasks under state and image observations, changing only what the expert is asked for gives the highest mean held-out success rate in all 10 settings, with a tie in 1, and the margin is widest at the smallest budget we tested. The scope is narrow: a single round of practice at a time, in simulation, with experts that are mostly scripted. The longer-term aim is a learner that also tracks what its demonstration set already covers, and that asks a human teacher for the missing behavior in proportion to the effort each request costs them.
Read the full abstract

Probability

Target-adapted Green-Bessel SVGD: uniform-in-time propagation of chaos and last-iterate consistency

We prove uniform-in-time propagation of chaos and last-iterate consistency for a target-adapted Stein variational gradient descent (SVGD) flow on compact connected manifolds. The target has a smooth positive density, and the particles start independently from a fixed smooth nonnegative density ratio. The construction uses the Green–Bessel operator $Q_{r,\pi}=A_\pi^{-1}(\mathrm{Id}+A_\pi)^{-r}$ of the reversible target Langevin generator. Sufficient Bessel smoothing gives a scalar kernel with finite diagonal, and a positive matrix lift realizes its potential force as a Stein velocity. Population and empirical flows then dissipate the same finite target discrepancy. Population entropy and the target spectral gap give decay of this discrepancy; a finite-time particle comparison reaches a time after which common-energy monotonicity controls every later time. The resulting expected uniform discrepancy is $O((\log N)^{-1/2})$, with a corresponding logarithmic $W_1$ bound and consistency along every sequence $t_N\to\infty$. We also prove an exact finite-mode approximation theorem with an explicit spatial-resolution error and a feature-factorized particle implementation. For confining Euclidean targets, we establish static kernel and moment results and give a conditional dynamical extension under explicit population-regularity and transport hypotheses.
Read the full abstract

Computer Vision

Hyperspectral Anomaly Detection via Group Sparse Low-Rank Tensor Factorization With Automatic Anomaly Grouping

Low-rank tensor modeling has become an effective tool for hyperspectral anomaly detection. However, existing methods still suffer from high computational cost and limited flexibility in characterizing spatially structured anomalies. To address these issues, this paper proposes a hyperspectral anomaly detection method based on group sparse low-rank tensor factorization with automatic anomaly grouping (GSAA). Specifically, the low tubal rank background is characterized by imposing group sparsity on tensor factors, which provides an efficient alternative to direct tensor rank regularization. For anomaly modeling, a latent grouping map is introduced to build an automatic anomaly grouping penalty, allowing anomaly groups to be adaptively inferred from the data rather than predefined at the pixel level. To further exploit complementary spectral and spatial information, GSAA is applied in both domains, and the resulting detection maps are fused to form a spectral–spatial version of GSAA, termed GSAA-SS. An efficient linearized alternating direction method of multipliers algorithm with convergence guarantee is developed to solve the resulting model. Experimental results on five real hyperspectral datasets demonstrate that the proposed method achieves superior detection performance and competitive computational efficiency compared with several state-of-the-art methods.
Read the full abstract

Strongly Correlated Electrons

Textures as a phase-transition probe for quantum spin chains

The idea of quantum texture has been recently proposed and used as a tool for quantifying coherences and for quantum gate identification. In this work we offer a study on its usage to quantum phase transitions, demonstrating the rugosity metric as a simple tool for effective phase-transition probing. We establish the link between rugosity in the computational basis and the hierarchy of spin correlators, and analyze rugosities defined in the global ground-state and in ground-states belonging to different magnetization sectors (to which we refer to as global vs symmetry-resolved rugosities) to study the phase diagram of the Heisenberg XXZ model. We find distinct rugosity signatures at both transition points. In particular, a sharp feature appears at $\Delta=1$ already for small systems, revealing a pronounced sensitivity of the correlation hierarchy encoded by the texture to this point. Since the BKT transition coincides with the isotropic $SU(2)$ point of the XXZ model, this behavior may reflect a particular sensitivity of rugosity to the structure of the spin-correlation hierarchy at isotropy.
Read the full abstract

All authors Heitor P. Casagrande, Isaac M. Carvalho, William J. Munro, Krissia Zawadzki

Image and Video Processing

GC360IQ: Generic-to-Individualized Quality Assessment for Stitched 360-Degree Panoramas

Existing image quality assessment (IQA) methods typically predict mean opinion scores (MOSs) but struggle to capture variations in individual subject behaviors. This limitation is highly pronounced in immersive visual applications such as stitched 360-degree panoramas, where user opinions diverge drastically based on personal sensitivity to blending-induced luminance inconsistency, detail loss, and geometric misalignment. Here we propose GC360IQ, a novel Generic-to-Individualized IQA framework that establishes a learned explicit feature space to characterize individual subject behaviors. First, we construct a specialized 360-degree panorama database focusing on blending-induced luminance and detail degradation while minimizing geometric misalignment, providing multidimensional quality ratings alongside complete individual scores. Second, we develop a generic quality model that utilizes unblended views as a perceptual reference. Dual feature extraction branches capture gradient and structural information specifically around stitching regions to predict baseline quality. Third, we construct a compact preference embedding space that acts as an explicit feature domain to model each subject's deviation from generic quality perceptions. We also introduce a maximum a posteriori (MAP) adaptation mechanism. By leveraging a preference prior learned within our explicit feature space, a new subject's unique behavioral embedding is mapped progressively with an increasing number of anchor ratings. Experiments demonstrate that the generic model provides accurate stitching quality predictions and that subject adaptation further improves individual score predictions. Deeper analysis of the collected ratings and learned embeddings reveals that observer differences contain structured variation related to scoring tendencies and sensitivity to stitching artifacts, rather than merely random rating noise.
Read the full abstract

Computer Vision

DriveMotion: A Large-Scale Multi-Source Benchmark for Driver Motion Sequence Modeling and Forecasting

Driver motion can provide cues to ongoing behavior, attention, and near-term driving intent. However, most existing driver-centric datasets focus on recognizing predefined driver behaviors from short video clips, while human motion forecasting benchmarks largely target motion outside the vehicle. We introduce DriveMotion, a multi-source benchmark for continuous driver motion forecasting. DriveMotion contains 393 hours of 133-keypoint motion sequences at 10 Hz from 360 drivers, integrating naturalistic driving data, curated public in-cabin videos, and the AIDE dataset into a unified representation with per-joint validity masks and synchronized driving context. Naturalistic driving contains long periods of limited body movement, making uniformly sampled evaluation dominated by persistence and less sensitive to brief but behaviorally meaningful motion. To address this, we use dynamics-anchored evaluation, placing forecasting windows around vehicle maneuvers identified offline from CAN signals without providing CAN to the model at inference. Arm motion in pre-maneuver windows is 3.4x greater than in route-matched stable-driving controls. On these anchored windows, learned models reduce forecasting error over persistence by up to 15%, while maneuver-enriched training improves forecast-derived Part-State F1 by 44% over the zero-motion reference. Training on the full multi-source corpus further reduces forecasting error on held-out web drivers by 38% compared with BATON-only training. DriveMotion provides identity-disjoint splits, fixed evaluation subsets, and reference implementations for reproducible evaluation of continuous driver motion forecasting. The dataset and benchmark are available at https://huggingface.co/datasets/HenryYHW/DriveMotion
Read the full abstract

All authors Yuhang Wang, Chuheng Wei, Jingxin Yang, Xishun Liao, Hao Zhou

Superconductivity

Signatures of nodal superconductivity in stoichiometric FeTe

Superconductivity in stoichiometric FeTe opens access to the FeTe endpoint of the Fe(Se,Te) phase diagram, yet the nature of its superconducting pairing state remains unresolved. In this work, we combine scanning superconducting quantum interference device (SQUID) microscopy, electrical transport, scanning tunneling microscopy and spectroscopy (STM/S), and mean-field calculations to investigate the local superfluid response and pairing state of FeTe thin films with tunable stoichiometry. Even in stoichiometric FeTe, we observe micrometer-scale spatial variations in both superfluid stiffness and superconducting transition temperature $T_c$, while the London penetration depth remains non-saturating down to 0.02$T_c$ and follows a power-law temperature dependence with an exponent of approximately 1-1.5. Together with a V-shaped low-energy density of states and two-gap modeling, these results indicate a superconducting state with gap nodes or deep minima, consistent with either a $d$-wave or nodal $s$-wave superconducting state. Our findings establish stoichiometric FeTe as a distinct superconducting regime that departs from the trend toward more isotropic gaps at intermediate Se/Te compositions, providing a new benchmark for modern microscopic theories of iron-chalcogenide superconductivity. Our work also reveals a crossover from weak to rapid suppression of $T_c$ as superfluid stiffness decreases, connecting FeTe to the broader phenomenology observed in unconventional superconductors.
Read the full abstract

All authors Cequn Li, Zi-Jie Yan, Yang Ge, Zihao Wang, Bing Xia, Stephen Paolini, Pu Xiao, Lok-Kan Lai, Jiatao Song, Austin R. Kaczmarek, Lujin Min, Kenji Yasuda, Peter J. Hirschfeld, Jiabin Yu, Cui-Zu Chang, Katja C. Nowack

Artificial Intelligence

Router Prior Bias: Preserving Base Routing Structure in MoE Post-Training

Mixture-of-Experts (MoE) pretraining relies on an auxiliary load-balancing loss (LBL) to drive per-expert utilization toward uniformity. Post-training inherits a different situation: the base router already encodes non-uniform expert co-activation structure, which a re-imposed uniformity objective flattens away. We show that downstream performance depends instead on holding this inherited routing softly, a principle we term soft router anchoring, and instantiate it as Router Prior Bias (RPB), a training-time bias that pulls the router logits toward a prior read off the frozen base router while leaving the router itself trainable. On math post-training of Moonlight-16B-A3B, RPB attains 45.77 in-domain accuracy against 31.91 under re-applied LBL and 29.44 under unanchored fine-tuning, and retains more out-of-domain capability than either. The ordering against LBL reproduces on a second model family (Qwen3-30B-A3B-Base), and the advantage over LBL is resolvable on an independently sourced corpus. Anchors defined on the router weights, on its logits, or on its output distribution perform comparably with no consistent ordering, which places the effect in the softness of the constraint rather than in the particular prior RPB supplies. Retained community structure in the expert co-activation graph tracks these gains wherever the base router is non-uniform enough for communities to form, yet enforcing the same prior as a hard assignment preserves that structure while performance falls sharply. Community structure is therefore a footprint of soft anchoring rather than its source, and the practical lesson is that inherited routing should be held softly during post-training, since both flattening it toward uniformity and enforcing it absolutely carry a downstream cost. Our code will be released at https://github.com/naver-ai/rpb.
Read the full abstract

All authors Jaedeok Lee, Keonwoo Kim, Dongyoon Han, Sangdoo Yun, Yera Choi, Haanju Yoo

Materials Science

Atomistic modeling of molecular beam epitaxy growth of SrTiO3 and Sr2TiO4 thin films

Molecular beam epitaxy (MBE) is renowned for its potential for atomic layer control, but unexpected growth mechanisms can potentially compromise this level of precision. In this study, we employ first-principles calculations to investigate the atomistic processes governing the MBE growth of perovskite SrTiO3 and Ruddlesden-Popper Sr2TiO4 films on a SrTiO3 substrate. We systematically explore the potential molecular species in the gas phase and their reactions and diffusion dynamics on the film surfaces and layer edges. Our analyses uncover three mechanisms of importance for understanding this type of growth. First, oxygen vacancies can be dynamically induced during surface diffusion on defect-free substrate and noticeably accelerate the diffusion processes. Second, while the SrO layer is expected to grow in a single-layer growth mode, the presence of potential TiSr defects may promote the formation of SrO islands. Lastly, adsorbed Ti atoms and TiO2 molecules on the SrO bilayer can insert into SrO bilayers, resulting in an unexpected growth sequence. These findings may have broader implications for the MBE growth of metal oxide films and provide guidance for achieving improved control over their growth processes.
Read the full abstract

Differential Geometry

Non-Kähler Critical Hermitian Metrics of the Dinew–Popovici Functional

The Dinew–Popovici functional is an energy functional for Hermitian symplectic metrics in a fixed Aeppli cohomology class. Its vanishing characterizes the Kähler metrics in that class, providing a variational approach to Kähler geometry. Dinew and Popovici proved that every critical point is Kähler in complex dimension three. We give a negative answer to Erfan Soheil's question about higher dimensions by constructing non-Kähler critical metrics on products of two Kähler surfaces $(S_1,\eta_1)$ and $(S_2,\eta_2)$. These metrics are critical under every variation on the product 4-fold. We characterize the critical product metrics and prove that non-Kähler critical products exist in the Aeppli class $[\eta_1+\eta_2]_A$ precisely when the canonical bundles of the two surface factors are smoothly trivial. As a by-product, we develop a geometric flow driven by the torsion tensor and converging to the fixed Kähler background when this background metric has nonnegative holomorphic bisectional curvature.
Read the full abstract

Differential Geometry

Formation and structural stability of nondegenerate neckpinches

In this paper, we study the formation, precise asymptotics, and structural stability of neckpinch singularities in mean curvature flow. Motivated by the static rigidity of cylindrical self-shrinkers established by Colding, Ilmanen and Minicozzi, we first prove a dynamical rigidity result: mean curvature flow of hypersurfaces that are initially graphically close to a generalized cylinder on a sufficiently large scale, subject to a localized quadratic upward bending, inevitably develop a neckpinch singularity in finite time. We also establish sharp asymptotic expansions for the profile functions of these locally evolving graphs. We show that the rescaled graphical radius converges to a specific polynomial profile with quadratic bending, proving that the resulting singularities are nondegenerate. Finally, we establish an openness theorem showing that nondegenerate neckpinches are structurally stable under C^2 perturbations of the initial data. Combined with recent density theorems for the 3-dimensional mean curvature flow by Szekelyhidi, our results confirm that nondegenerate neckpinches constitute a generic and stable phenomenon in 3-dimensional mean curvature flow.
Read the full abstract

High Energy Physics (Theory)

Bogomol'nyi equations for Kruglov strings

We construct the Bogomol'nyi equations for Abelian gauge–Higgs vortices in which the Maxwell gauge sector is replaced by Kruglov nonlinear electrodynamics, a power-law family that interpolates between Maxwell theory, Born–Infeld electrodynamics, and exponential electrodynamics, characterized by a dimensionless exponent $\sigma$. Using the stressless method, we derive a pair of first-order equations directly from the vanishing of the spatial stress tensor, without assuming the Higgs potential a priori. For generic $\sigma$, the gauge and Higgs sectors are coupled through an implicit algebraic relation. We therefore introduce a constitutive map $\Phi(Y;\sigma)$ and analyze its monotonicity and range to determine the conditions for a smooth admissible Bogomol'nyi branch. For $\sigma>1/2$, the constitutive map is strictly monotonic and unbounded, whereas for $0<\sigma<1/2$ it possesses a finite maximum; the marginal case $\sigma=1/2$ is bounded. These properties yield explicit bounds on the nonlinear parameter $\beta$ for the latter cases. We further obtain closed-form constitutive relations, BPS potentials, and gauge-field equations for six representative values of $\sigma$, spanning linear, quadratic, and cubic algebraic structures. The corresponding vortex profiles are then computed numerically. The resulting BPS string tension is purely topological, $\mu_{\rm BPS}=2\pi n$, independent of both $\sigma$ and $\beta$.
Read the full abstract