Numerical Analysis
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 11 September
Stochastic variance reduced gradient (SVRG) is a variant of stochastic gradient descent and is a promising iterative method for solving large-scale inverse problems. Nevertheless, the development of theoretically grounded a posteriori stopping rules for SVRG remains an open challenge. In this work, we provide a convergence analysis of SVRG equipped with the discrepancy principle, the most well-known a posteriori stopping rule, for solving a class of linear inverse problems in Hilbert spaces. We establish the regularizing property of SVRG, and moreover, under suitable source conditions, we derive convergence rates of SVRG iterates. To the best of our knowledge, these are the first convergence rate results of any stochastic iterative method for inverse problems under the a posteriori stopping rule. The theoretical findings are supported by numerical experiments.
Artificial Intelligence
Is Gaussian Splatting Becoming Neural Again? A Taxonomy and Controlled Study of Learned Parameterization
Three-dimensional Gaussian Splatting (3DGS) combines explicit primitives with efficient rasterization, yet recent systems increasingly use neural networks to generate or share Gaussian parameters. We characterize this trend along five axes: attribute decoding, spatial sharing, view-conditioned decoding, topology generation, and amortized inference. An analysis of 19 representative methods shows that these choices address different limitations and cannot be reduced to a binary neural label. We also isolate three forms of neural parameterization in a controlled mip-NeRF 360 study. Sharing appearance and opacity improves reconstruction quality, while decoding geometric structure offers no further gain. The evidence favors selective neuralization: shared functions help when they capture reusable correlations without sacrificing the local geometric freedom of explicit splats.
Read the full abstract
Artificial Intelligence
BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents
Mobile GUI agents are shifting from multi-module frameworks to native models trained end-to-end, yet industrial deployment faces three persistent gaps. Sandbox training produces a distribution mismatch with production environments; expensive real-device failures remain underutilized; and fixed benchmarks saturate, losing the power to guide iteration. We present BlueLM-GUI, a 35B-A3B mobile GUI agent built as a real-device-centric flywheel that closes these gaps through three principles. Every Sample Matters: a dual-track pipeline with Heterogeneous Triple-System Consensus evaluation and an Error Correction & Derivation Module salvages every trajectory into usable supervision. Every Rollout Is Real: a three-stage recipe—continual pre-training, supervised fine-tuning, and agentic reinforcement learning on hundreds of real phones—grounds every rollout in real production environments, so the capability the model learns transfers directly to deployment. Every Query Evolves: a quota-driven benchmark methodology with three orthogonal axes enables precise attribution and allows the benchmark to be systematically upgraded as the model improves. BlueLM-GUI achieves 87.4 on MobileGUI-VBench, surpassing the best closed-source model by 5.1 points, and 84.9 on AndroidWorld, the best result among open-source models and competitive with closed-source models. These results demonstrate that grounding model training and iterative improvement in both real devices and the three Every principles yields strong, robust, and transferable mobile GUI capability.
Read the full abstract
Nuclear Theory
Quantum convolutional neural network for predicting nuclear charge radii
Quantum machine learning has the potential to become a new tool for understanding complex nuclear structures. In this work, we apply a hybrid quantum convolutional neural network (QCNN) to nuclear charge-radius prediction for the first time, aiming to explore the feasibility of quantum machine learning in nuclear-physics data analysis. Based on a classical convolutional neural network (CNN) framework, a small variational quantum convolutional filter is introduced as a quantum feature map to extract local correlations on the nuclear chart. The QCNN shows promising predictive accuracy and training stability, and provides a reliable description of the charge-radius evolution along several representative isotopic chains. These results support further investigation of quantum convolutional architectures for nuclear-structure data analysis.
Read the full abstract
Signal Processing
Feedback-Efficient Beam-User Association for Near-Field mmWave Hybrid Beamforming Systems
Near-field multiuser hybrid beamforming (HBF) requires joint angle-distance codebooks whose size, and hence reporting overhead, grows with the array aperture. For the extremely large array considered here, reporting one quality metric per codeword already incurs more overhead than full channel state information (CSI) feedback. This letter develops a feedback-efficient beam–user equipment (UE) association framework. The focusing codebook is sampled at beam-depth spacing within an effective beamfocusing Rayleigh distance (EBRD)-aware focusing region, with one far-field codeword per angular direction beyond that region, so that its radial law and size follow from the array geometry. Each UE probes this codebook but reports only its M strongest candidates, and the base station associates UEs and beams with a proportional-fair metric that consumes the leakage terms carried by this report together with the codeword correlations it knows, thereby allowing co-angular UEs to be multiplexed by focal distance. Simulations show that M=3 suffices: the proposed scheme stays within 0.6% of an optimistic full-metric reporting reference while using 0.8% of its feedback, a 99.11% reduction with respect to full-CSI feedback, and the interference-aware metric contributes up to 18.7% of the sum spectral efficiency over its interference-blind counterpart.
Read the full abstract
Combinatorics
The maximum $t$-diversity of $t$-intersecting families of permutations
The study of $t$-intersecting families in symmetric groups received lots of attention in the last two decades. In this paper, we study two different kinds of stability results for $t$-intersecting families in symmetric groups. Let $\Sigma_n$ denote the symmetric group on $\{1,2,\ldots,n\}$. The $t$-diversity $\gamma_t(\mathcal{F})$ of $\mathcal{F}\subseteq\Sigma_n$ is defined as the minimum number of members of $\mathcal{F}$ whose deletion results in a family with transversal number $t$. The star $t$-diversity $\gamma_t^{\star}(\mathcal{F})$ is defined as the minimum number of members of $\mathcal{F}$ whose deletion results in a $t$-star. For $n$ relatively large with respect to $t$, we determine the best possible bounds for both $\gamma_t(\mathcal{F})$ and $\gamma_t^{\star}(\mathcal{F})$ over all $t$-intersecting families $\mathcal{F}\subseteq\Sigma_n$. The equality holding conditions are characterized. For $\gamma_t(\mathcal{F})$, the extremal case is generated by all $2t$-subsets of a $3t$-partial permutation or, when $t=2$, the family complements of the lines of the Fano plane. For $\gamma_t^{\star}(\mathcal{F})$, the extremal family is generated by all $(t+1)$-subsets of a $(t+2)$-partial permutation.
Read the full abstract
Astrophysics of Galaxies
A systematic study of the long-term mid-infrared color variations of Seyfert 1 galaxies
We present a comparative analysis of long-term mid-infrared (MIR) color variations (MCVs) for 1,977 narrow-line Seyfert 1 (NLSy1) and 4,113 broad-line Seyfert 1 (BLSy1) galaxies at $z < 0.3$. Using 14 years of WISE and NEOWISE data, we quantify the correlation between $W1-W2$ color and $W1$ magnitude. We categorize sources into bluer-when-brighter (BWB), redder-when-brighter (RWB), and weak/no MCV populations. Our results show that the MCV properties of BLSy1s are consistent with those of NLSy1s. This suggests that host-contaminated dust reprocessing is a universal mechanism across Seyfert 1s. Bolometric luminosity ($\Lbol$) is the primary driver of MCV behavior. However, at fixed $\Lbol$, the fraction of RWB sources decreases with increasing Eddington ratio ($\REdd$). This reveals a secondary dependence on the accretion state. Ensemble structure function analysis further shows that higher-$\REdd$ sources exhibit flatter structure functions at a given $\Lbol$, indicating smaller dust torus radii. These findings suggest that the central engine's radiation and accretion state shape the circumnuclear dust geometry. Radio-loud sources follow the same trends as radio-quiet counterparts, implying that jet emission does not significantly impact long-term MIR variability. Finally, modified blackbody modeling indicates that grains with sizes $a \gtrsim 0.1~\mu$m are responsible for the observed MCVs. Our results establish a unified framework for MCVs in Seyfert~1s, where the Eddington ratio modulates both variability patterns and the physical scale of the dust torus. Furthermore, given the similarity between the MCV distributions of CLAGNs and RWB sources, we estimate that $\lesssim 6\%$ of Seyfert~1s may be CLAGN candidates, offering a potential MIR-based pre-selection.
Read the full abstract
Information Theory
Error-Rate Reduction in LDPC Decoding via Bit-Aligned Temporal Reinforcement in Parallel Probabilistic-Bit Dynamics
Probabilistic bits (p-bits) provide a physical and algorithmic primitive for stochastic inference, but highly parallel updates can alter their collective dynamics. We study the decoding of random-regular (3, 6) low-density parity-check (LDPC) codes using Jacobi-type p-bit annealing with stochastic partial activation. An additive response-path rule stores each bit's saturated response and reuses it before stochastic readout. High-statistics simulations with independent parameter optimization for each method show pooled bit-error-rate reductions of 33.5%, 74.8%, and 81.8% relative to memoryless probabilistic simulated annealing (pSA) for representative codes of block lengths 96, 192, and 288, respectively. Static gain, normalized averaging, response shuffling, and same-bit binary-state feedback with only its coefficient tuned under the same nonmemory parameters do not reproduce the full saturated-response benefit. Trajectory analysis links the improvement to acquisition of the channel-consistent valid-codeword basin and enhanced post-acquisition stability; the acquisition advantage persists from random and channel-hard-decision starts under the tested conditions. Across all 30 independent code realizations, fixed additive parameter-and-readout packages achieve lower bit- and frame-error rates than separately optimized pSA-specific packages, with neither package retuned for individual codes. These results show that the computational effect of temporal state depends on the retained quantity and its reinjection into stochastic dynamics.
Read the full abstract
Signal Processing
RF-VoID: Towards Bandwidth-Efficient Exterior Tile Void Detection via Narrowband Radio-Frequency Representation Learning
Hidden debonding behind exterior ceramic tiles is a falling-tile hazard, and millimeter-wave radar offers a non-contact way to find it. Conventional interpretation first reconstructs a range profile, so its reliability is bounded by the available bandwidth, yet bandwidth is what sets the cost, the acquisition time, and the regulatory footprint of a deployed system. This work asks whether that bandwidth can be traded for computation. A 4-40 GHz stepped-frequency system scans twelve exterior-wall specimens containing 0.5-1.0 mm air voids at different depths and interfaces, and the bandwidth dependence of A-scan, B-scan, and C-scan interpretation is analyzed to establish the resolution bound. RF-VoID is then proposed, which decides directly on the narrowband complex response: the sub-band is kept in its measured frequency order with amplitude and phase alongside the in-phase and quadrature channels, a dual-branch encoder reads it along the physical frequency axis using relative position encoding and a distance-dependent locality bias, and an inspection-oriented objective handles the class imbalance and the asymmetric error cost of facade screening. Under a mixed-sample protocol the method attains 98.61% accuracy and a 95.84% F1-score with 0.5 GHz of bandwidth, a seventy-two-fold reduction relative to the full sweep, without range-profile reconstruction, deconvolution, or depth-slice selection; on specimens held out entirely from training it remains the strongest of the compared models, with a mean macro F1-score of 62.12% at 0.5 GHz that rises to 68.57% at 1 GHz.
Read the full abstract
General Economics
Do Sanctions Backfire? New Evidence on the Macroeconomic Effects of Supporting Ukraine
Using a balanced panel of 129 countries from 2000 to 2024, we estimate the macroeconomic costs of sanctions on Russia. To that end, we employ a novel two-wave common-correlated-effects imputation design that separates the 2014 sanctions from the 2022 escalation and accounts for heterogeneous exposure to global shocks. We find the main boomerang effect of these sanctions operates through prices rather than output. Specifically, relative to 2019-2021, we find a significant post-2022 increase in prices among sanctioning countries that remains robust under alternative counterfactuals, reference windows, and inference procedures. Conversely, we find no robust evidence of aggregate GDP-per-capita losses.
Read the full abstract
Machine Learning
Split Conformal Prediction with Label-Shift-Adjusted Bayesian Scores
Conformal prediction provides distribution-free uncertainty quantification under exchangeability. However, this assumption is violated by label shift, where the marginal distribution of labels changes while the conditional distribution of inputs given labels remains stable. Under such shifts, standard conformal procedures no longer maintain their intended coverage behavior. Existing approaches address this via importance weighting. They pair the reweighting with residual-based nonconformity scores that ignore predictive uncertainty. The resulting intervals have uniform width. Bayesian conformal methods produce adaptive intervals by leveraging predictive distributions. They evaluate conformity under the source predictive, which is misaligned with the target domain under label shift. We propose the Label-Shift-Adjusted Bayesian Score (LSA score), a nonconformity score derived from a posterior predictive tilting identity. This identity shows that the target predictive is an importance-weighted transformation of the source predictive. We use it to derive a direct correction to the Bayesian score. We evaluate the method on molecular property prediction under controlled label shift. The LSA score consistently yields shorter intervals than residual-based and source-based Bayesian scores. Coverage in the target domain remains comparable. Under stronger shift, all methods incur some coverage loss due to pseudo-label-based density-ratio estimation. The LSA score is defined for any source predictive with a tractable log-density. We instantiate it with Bayesian Ridge Regression, where the correction admits a closed form.
Read the full abstract
Representation Theory
Cocenter of Hecke algebras of Kac-Moody groups
Let $H$ be the generic Hecke algebra over $\mathbb{Z}[\mathbf{q}^{\pm 1}]$ associated to a split Kac–Moody group $G$, arising as the deformation of the group algebra of its Weyl group $W$. The cocenter $\overline{H} = H/[H, H]$ encodes the trace and character theory of $H$, playing a fundamental role in representation theory and harmonic analysis. Through deep combinatorial results on cyclic reductions in Coxeter groups, each conjugacy class $\mathcal{O}$ of $W$ determines a canonical element $T_{\mathcal{O}}$ in $\overline{H}$, and these elements are known to span the cocenter. However, establishing their linear independence has remained an open problem outside of finite and affine types. In this paper, we solve this problem: the canonical elements form a $\mathbb{Z}[\mathbf{q}^{\pm 1}]$-basis of the cocenter $\overline{H}$. Our approach differs from earlier representation-theoretic methods in finite and affine types. To construct explicit functionals that separate all conjugacy classes, we develop a new framework based on re-normalized orbital integrals. This framework synthesizes parabolic induction, traces of infinite-dimensional bimodules, and Kac–Moody harmonic analysis into an almost-dual basis for the cocenter. As a key local ingredient, we establish a generic duality theorem for finite groups of Lie type relating the cocenter to regular semisimple conjugacy classes, and determine precisely when this pairing is non-degenerate. Finally, we deduce the existence and uniqueness of generic class polynomials for $W$, and prove a uniform “dimension=degree” theorem for basic Deligne–Lusztig varieties of the split Kac–Moody group $G$.
Read the full abstract
Robotics
Understanding Whole-Body Robot Teleoperation Strategies Under Diverse Task Objectives and Constraints
This work investigates the control strategies of complex whole-body robot teleoperation that coordinate active perception, bimanual manipulation, and navigation. We developed a hybrid control framework, combining the free-form and constrained control, for the whole-body teleoperation of the TIAGo mobile manipulator. We conducted a user study to explore people's control strategies under different task constraints such as limited time and low tolerance of errors. Our results highlight the effective use of coordinated control in improving task efficiency and reducing the risk of reaching individual joint limits. We discuss our results and their implications for designing future whole-body robot teleoperation systems.
Read the full abstract
Analysis of PDEs
On energy equality of the 3D anisotropic Navier-Stokes equations
In the spirit of the recent work by Demmel and Wiedemann \cite{[DW]}, we investigate the validity of the energy balance law for weak solutions to the three-dimensional Navier-Stokes equations with only horizontal dissipation. We prove that famous Lions-Shinbrot type energy conservation criteria, originally established for the classical isotropic Navier-Stokes equations, continue to hold in this anisotropic setting. Furthermore, we present two distinct approaches that handle the torus case and the whole space case, respectively.
Read the full abstract
Computation and Language
Representation-based Masked Diffusion Model
Masked Diffusion Models (MDMs) have emerged as a compelling paradigm for language modeling, offering the capability for efficient parallel text generation. However, existing parallel sampling methods typically update multiple masked tokens independently and ignore the complex mutual dependencies among the masked tokens. This independent updating mechanism lacks global coordination and might lead to incoherent outputs. To address this limitation, we propose Representation-based Masked Diffusion Model (RMDM), a framework that leverages the text representation to explicitly encode global semantics and help to parallel update tokens more precisely. Specifically, we first encode text into a continuous semantic space using a pretrained encoder and learn an invertible transformation that normalizes the representation distribution to a Gaussian prior, facilitating efficient sampling during generation. Conditioned on this latent semantic representation, we train a masked diffusion model to learn the conditional text distribution, where the representation serves as global semantic guidance to coordinate parallel token updates and faithfully approximate the target distribution. Empirical results demonstrate that RMDM significantly improves generation quality, particularly in aggressive few-step sampling regimes.
Read the full abstract
Dynamical Systems
The Ergodicity of Geodesic Flows on Rank One Manifolds of Nonpositive Curvature
In this paper, we study the ergodicity of geodesic flows on closed rank-one manifolds of nonpositive sectional curvature. We introduce the infinite-order vanishing set of the Gaussian curvature in dimension two and of the fiberwise second moment of the reduced Jacobi determinant in higher dimensions. We bound the Liouville measure of the singular set in terms of the volume of this subset and, for surfaces, obtain a corresponding Hausdorff dimension bound. In particular, if this subset has zero volume, then the singular set has zero Liouville measure, and the geodesic flow is ergodic. We also give a geometric criterion for finitely many exceptional regions whose fundamental groups have virtually Abelian images of rank smaller than the dimension of the manifold. Under strict negativity of sectional curvature outside these regions, we obtain a Hausdorff dimension bound for the singular set and prove that the geodesic flow is ergodic with respect to Liouville measure. By extending this method to real-analytic metrics, we characterize the singular set as the zero set of a nontrivial real-analytic determinant constructed from curvature operators, thereby establishing ergodicity in this setting without further assumptions.
Read the full abstract
High Energy Astrophysical Phenomena
Evolution of first-interaction second-harmonic anisotropy in cosmic-ray air showers
The longitudinal evolution of second-harmonic anisotropy inherited from the first interaction in extensive air showers is investigated with CORSIKA simulations. Events with large initial anisotropy are selected. Second-harmonic observables are evaluated separately using the coordinate and momentum azimuths at several atmospheric depths and compared with their corresponding initial harmonic orientations. More than 84% of the initial second-harmonic signal is attenuated within the first 20 g cm$^{-2}$ of shower development. In coordinate space, the second-harmonic correlation with the initial orientation persists to ground level at approximately 0.09% of its initial magnitude. This residual is evaluated against a matched source-isotropized reference, constructed for each selected event by independently randomizing the azimuths of the first-interaction final-state particles and simulating the shower under identical atmospheric conditions. The positive coordinate-space ensemble-mean difference at ground level is statistically resolved, and the two samples show little event-level separation. In momentum space, a weak ensemble-level residual remains at 200 g cm$^{-2}$, and the original and source-isotropized signals become statistically indistinguishable by 300 g cm$^{-2}$ and remain so at ground level.
Read the full abstract
Distributed Computing
ForgeMegakernel: A General Framework for Efficient Auto-Regressive Model Decode Megakernels
Auto-regressive model decode is bandwidth-bound, since every weight and key/value-cache byte crosses high-bandwidth memory once per token. A megakernel is an ideal solution, but existing automatic megakernel generation approaches cannot achieve both generalization across models and correctness guarantees. We present ForgeMegakernel, which generates a per-model high-performance decode megakernel using coding agents. ForgeMegakernel pairs a universal knowledge base of ten progressive milestones with an independent mid-state test oracle. The milestones provide the megakernel's structural properties: a fine-grained instruction stream for each SM, dependency counters replacing the global synchronization, and a shared-memory buffer pool for workload balance across SMs and greater parallelism. The test oracle derives the mid-states of the megakernel and checks the performance, error and precision during the generation process, guaranteeing a correct and trustworthy forged megakernel. We evaluated ForgeMegakernel on 14 representative decoding operations across eight model families spanning 0.6B-13B parameters. The generated megakernels achieved 50.5-85.9% MBU and geometric mean speedups of 1.21x over SGLang 0.5.18 and 1.54x over a megakernel compiler under identical configurations. Inside SGLang, evaluated on GSM8K with ragged prompts, all 14 megakernels decoded faster than the SGLang engine at comparable answer accuracy.
Read the full abstract
Cryptography and Security
An Open-Source End-to-End FHE Implementation for Privacy-Preserving Llama 3 8B Inference
Cloud LLM services typically require users to send prompts to a model provider, creating a privacy risk. Fully homomorphic encryption (FHE) lets a server perform inference without decrypting the input, but representing data as ciphertexts adds storage and computational overhead. In CKKS-based LLM inference, the packing scheme maps logical tensors to ciphertexts and slots. It therefore determines the ciphertext count and the homomorphic cost of linear layers, and it constrains how data pass between linear layers, attention, and nonlinear computation. As models and sequences grow, inefficient layouts accumulate encoding, compute, and layout-conversion overhead. We present Odin, an FHE inference system that co-designs ciphertext packing and model execution for Llama. Starting from a THOR-style baseline whose bottleneck is weight encoding, Odin uses a feature-major cross-layer layout to unify residual connections and layer interfaces, and builds transient intra-operator layouts for linear projections and attention. This reduces redundant plaintext encoding of weights in wide projections. Within attention, QK^T produces scores that Softmax can consume directly, and PV consumes the resulting probabilities, avoiding intermediate repacking. For nonlinear ops, we use minimax polynomial approximation with input-range control and joint error allocation guided by model quality, reducing polynomial degree and multiplicative depth. To our knowledge, Odin is the first open-source end-to-end GPU CKKS implementation of Llama-3. With Llama-3-8B weights and a 128-token input, Odin evaluates all 32 Transformer layers on a single NVIDIA H100 80 GB GPU. Server-side end-to-end FHE evaluation takes 366.4 s and 58.9 GiB peak device memory. Under the same model, input, CKKS parameters, and hardware, THOR takes 1651.9 s, a 4.51x speedup.
Read the full abstract
High Energy Astrophysical Phenomena
A Localized Current-Sheet Magnetic-Diffusion and Heating Prescription for Ideal-GRMHD Simulations of M87*-like Accretion Flows
Ideal-GRMHD calculations provide the bulk accretion flows used in black-hole emission models. A localized dissipation prescription can add a resolved map of current-sheet activity while retaining that dynamical reference. We implement such a prescription in Athena++, using a smooth threshold on the dimensionless current proxy $\alpha_J$ to select the magnetic-diffusion coefficient. The constrained-transport update is coupled to a conservative energy-flux correction. Gas heating follows from primitive recovery; $q_{Ohm}=\eta J^2$ is recorded as a diagnostic, without a second energy source. Static Fourier and traveling Alfvén tests measure the evolved thermal gain, and an open-boundary Schwarzschild test closes the Killing-energy ledger. The relative total-energy residual is below $2\times10^{-15}$ in these tests. A two-resolution Harris matrix gives localized rate exponents $0.510\pm0.027$ and $0.488\pm0.028$, consistent with $1/2$ over the specified early interval. The $\eta=10^{-3}$ rate exceeds the same-resolution ideal control by a factor of 5.96. In matched axisymmetric M87*-like runs, localized diffusion changes the early mean magnetic flux, accretion rate, and normalized flux by less than $0.02\%$, with a radial-density distance of $0.005\%$. During the late active phase, a common-restart on/off comparison keeps these mean shifts below $1.6\%$, within the intrinsic temporal variability, and gives a density distance of $0.29\%$. Uniform diffusion at the tested coefficient reduces the raw flux and accretion rate by $85\%$ and $99.8\%$. These comparisons establish a spatially selective diffusion prescription with explicit energy accounting and a quantified, small bulk-flow response.
Read the full abstract
Statistical Methodology
Dynamic Amplification of Risk-Estimate Bias Through Differential Detection: A Markov Model for History-Based Covariates
History-based risk factors, such as known family history, recorded personal history, and documented regional history, are often treated as covariates. Because these variables depend on testing, diagnosis, and recording, their observed values are influenced by detection. However, monitoring data are often unavailable, making it difficult to distinguish true history effects from detection-driven associations. We study the feedback loop that arises when observed history affects future monitoring and future monitoring affects which histories become observed. Prostate cancer screening serves as a case in point: knowing a family history may raise awareness and increase testing compared with having no known history. We develop Markov models for true and observed history states and show that differential detection can distort both observed risk ratios and the distribution of the observed history variable, and can create an apparent history effect even when true event risk does not depend on history. With history-dependent true risk, the observed history process is generally not Markovian, although the joint true-observed process is. Uniform incomplete detection can attenuate true history effects by shifting individuals with missed events into less recent observed-history states. We also develop an inverse sensitivity-analysis framework that combines a published observed association, a baseline event probability, and plausible detection probabilities to obtain the implied true risk contrast. Numerical analyses illustrate the distortions, and a prostate cancer family-history example demonstrates the sensitivity calculation using external calibration values. The framework is intended for registry, electronic health record, and surveillance studies that use documented history as a risk factor.
Read the full abstract
Information Retrieval
ChronicleRec: Pre-training Temporally Anchored Tokens for Lifelong User Modeling
Modeling ultra-long user behavior sequences is crucial for industrial recommendation and online advertising, yet directly feeding thousands of historical actions into ranking models is computationally prohibitive, while truncation discards long-range signals. Existing lifelong-interest methods retrieve target-relevant behaviors for each candidate, coupling long-sequence modeling with candidate scoring and repeated online cost. Recent target-independent compression methods enable cached user summaries, but often append query tokens at the sequence end and use bidirectional encoding, producing unordered and redundant summaries that overlook temporal structure. We propose ChronicleRec, a pre-train-and-transfer framework that compresses an ultra-long behavior sequence once into a chronologically ordered set of Chronicle Tokens. ChronicleRec applies a recency-aware multi-granularity merge, preserving recent behaviors while coarsening distant history. It then interleaves query tokens with the merged sequence and uses a causal encoder, so each query summarizes only the history before its temporal anchor. A multi-horizon design masks different recent-history windows across parallel branches to learn complementary long-range interests. The compressor is pre-trained with a mask-and-predict objective that reconstructs held-out recent behaviors from compressed older history, aligning historical signals with near-present intent. Since Chronicle Tokens are target-independent, they can be cached per user, decoupling ultra-long sequence modeling from online candidate scoring. Experiments on KuaiRand and Tencent AdLive show that ChronicleRec outperforms recent-window and single-pass compression baselines while approaching full-attention performance. Token analyses reveal temporally organized and complementary representations, and a seven-day online A/B test confirms significant production gains.
Read the full abstract
Computer Vision
Context-Aware Causal Gaze Forecasting for Human-Vehicle Interaction During In-Cabin Tracking Dropouts
Dashboard-mounted gaze trackers often lose sight of the driver's eyes during large head rotations, including shoulder checks, mirror glances, and intersection scanning. These maneuvers occur when information about the driver's visual attention is most useful. Offline gap-filling methods may reconstruct a missing interval using observations from both sides, but an online driver-monitoring system cannot rely on measurements that have not yet occurred. We therefore formulate causal gaze recovery: forecasting unavailable gaze at time t without target-tracker gaze at t or later. We introduce the Causal Context-Gated Forecaster (CCGF), which encodes a 60-frame pre-dropout history of gaze and head pose and combines it with DINOv3 scene features. A learned reliability gate controls the contribution of the history and scene representations as the dropout progresses. We evaluate two scene conditions: Live, in which the scene representation continues to update during tracker loss, and Frozen, in which the final pre-dropout representation is used throughout the missing interval. We evaluate CCGF on 2,047 eligible, naturally occurring GazeSense head\_lost events drawn from 10.5 h of naturalistic driving by ten drivers. Across all recordings, head\_lost accounts for 8.5 percent of GazeSense recording time. Synchronized gaze coordinates from a head-mounted Neon tracker provide supervision and evaluation targets but are never used as model inputs. Under leave-one-driver-out evaluation, CCGF achieves a mean per-driver median error of 175.7 px (10.5 deg) with Live scene updates, a 33 percent reduction relative to history-only causal forecasting. With Frozen scene input, the error increases to 210.8 px (12.9 deg), indicating that scene observations acquired during the dropout provide useful predictive information. We will release the dataset, evaluation protocol, and causal baselines.
Read the full abstract
Artificial Intelligence
Toward Robust Personalized Alignment for LLMs: Mitigating Persona Drift in Multi-Turn Dialogue
Persona drift remains a central challenge for personalized language models, as user profiles evolve over long interactions rather than remain permanently fixed. Models must therefore revise persistent persona states when preferences genuinely change, while avoiding updates driven by transient, ambiguous, or unresolved observations. We propose CORE, which separates turn-local evidence from persistent persona-state revision and selectively updates grounded user preferences through uncertainty-aware belief revision. We also introduce PERSIST, a held-out post-anchor benchmark for persona-state robustness under sequential interaction stress, covering ambiguity, conflict, and controlled social influence. Across ALOE, PersonaChat, and PERSIST, CORE improves personalized alignment and robustness, with complementary gains in normalized closed-slot state fidelity. Human evaluation and mechanistic controls further support explicit update control beyond stronger generation or persistent memory alone.
Read the full abstract
Combinatorics
4-Arc-Pancyclicity of Regular Multipartite Tournaments
A multipartite tournament is an orientation of a complete multipartite graph. We prove that every $r$-regular $c$-partite tournament with common partite-set cardinality $\alpha$ is $4$-arc-pancyclic whenever $c\ge93$; that is, every arc belongs to a cycle of each length from $4$ to $c\alpha$. This confirms the conjecture of Zhou and Zhang for all sufficiently large $c$ and provides a multipartite analog of Alspach's arc-pancyclicity theorem. Moreover, we also give a construction to show that 4-arc-pancyclic is the best possible. Next, we prove that every arc belongs to at least $c\alpha-\alpha-1$ cycles of pairwise distinct lengths when $c\ge7$ and $\alpha\ge2$. For regular $3$-partite tournaments with common partite-set cardinality $\alpha\ge2$, we obtain the sharp lower bound $\alpha$, settling the remaining case of a conjecture of Xia, Cai, Guo, and Wang.
Read the full abstract
Robotics
READ: Learning Risk-Informed Fields for End-to-End Autonomous Driving
Autonomous driving requires more than recognizing what is present in a scene: a planner must determine how road structure, surrounding agents, and their motion states should influence a future maneuver. Existing learning-based planners can capture these influences through latent scene features and trajectory decoders, but the relationship between environmental factors and candidate actions often remains implicit. This limits the ability to inspect, diagnose, or refine how scene context affects the safety of a predicted trajectory. Classical safety fields provide an explicit spatial representation of this relationship, but their risk shapes and relative weights are prescribed in advance and do not adapt to each scene. We introduce READ, a framework that learns an explicit, planning-aligned risk representation from complementary geometric and behavioral constraints. READ instantiates this representation as a continuous spatiotemporal field, enabling differentiable queries along candidate trajectories. The learned field connects scene understanding with action selection by encouraging predicted trajectories to align with low-risk regions, while retaining a differentiable interface for trajectory evaluation and refinement. READ integrates with both end-to-end planners and Vision-Language-Action models. Experiments on NAVSIM show consistent gains across matched end-to-end backbones and strong performance in a VLA setting; READ also achieves competitive results on NAVSIM v2. These results establish learned spatial risk as an explicit, adaptable representation for safe planning.
Read the full abstract
Earth and Planetary Astrophysics
High water D/H ratio of the interstellar object 3I/ATLAS is consistent with a low-metallicity origin
Recent JWST observations have revealed unusually high $^{12}$C/$^{13}$C ratios in carbon-bearing molecules of the interstellar object 3I/ATLAS, consistent with formation in a lower-metallicity environment than the present-day local interstellar medium (ISM). 3I/ATLAS also exhibits an exceptionally high water D/H ratio, exceeding those in Solar System comets and nearby low-mass star-forming regions. Here we investigate whether this high water D/H ratio can be reproduced in a low-metallicity formation scenario, using gas-ice astrochemical models. Assuming that the water observed in 3I/ATLAS was inherited from the parent molecular cloud and core, we perform a grid of astrochemical models covering the cloud to core stages, varying the gas density, ultraviolet radiation field ($\chi$), cosmic-ray ionization rate ($\zeta$), and metallicity, while solving thermal balance for the gas temperature. We find that lower metallicity enhances H$_3^+$ deuteration and, more importantly, its transfer to water ice. In contrast, water D/H ratio depends non-monotonically on $\chi$ and $\zeta$, because of competing chemical and thermal effects. In our models, the observed water D/H ratio is most readily reproduced at subsolar metallicities, $\lesssim0.5Z_\odot$, and relatively high cloud densities of $\sim$10$^4$ cm$^{-3}$ without strong constraints on either $\chi$ or $\zeta$, as long as $\zeta<10^{-15}$ s$^{-1}$. The D/H ratio of methane normalized by that of water is not sensitive to the metallicity, being consistent with the similar values observed in 67P/Churyumov-Gerasimenko and 3I/ATLAS. These results suggest that water deuteration may provide a complementary probe of the metallicity and physical condition of the parent molecular cloud and dense core of interstellar objects.
Read the full abstract
Numerical Analysis
A spectral-element method for computing many eigenvalues and their asymptotics of the Schrödinger operator with Robin boundary condition
We propose spectral and spectral-element methods for accurately computing many eigenvalues of the Schrödinger operator with Robin boundary condition on simple and complex geometries, respectively. Due to their spectral-type accuracy in approximating those eigenfunctions corresponding to high-index eigenvalues, which are usually highly oscillatory, the proposed approaches have excellent resolution in computing thousands of eigenvalues accurately and efficiently with the resolution property that the number of eigenvalues with reasonable accuracy is proportional to the number of degrees of freedom. Using a standard HPC node with 128 GB of memory, we can obtain numerically more than 5,000 reliable eigenvalues with relative errors below $10^{-8}$ for different complex domains in two dimensions (2D). Based on these computed eigenvalues, we first confirm some theoretical results on the Robin-to-Neumann (RtN) gaps of the Laplacian operator in 2D, which were recently studied by Rudnick et al. [Comm. Math. Phys. 388 (2021)]. Then we systematically study the RtN gaps of the Schrödinger operator and their convergence rates. Based on our extensive numerical results, we formulate a unified conjecture on the cumulative averages of the RtN gaps of the Schrödinger operator.
Read the full abstract
Functional Analysis
A localization framework for $\mathbb{T}^m$-Invariant Toeplitz Algebras: Applications to Gelfand Theory
We study $\mathbb T^m$-invariant Toeplitz operators on weighted Bergman spaces over the unit ball. A natural unitary transformation represents each such operator as a direct sum of restrictions of Toeplitz operators acting on a polyball. Under an $L^\infty$-valued angular continuity assumption, the corresponding symbols extend from $\mathbb N_0^m$ to a norm-continuous family indexed by the maximal ideal space of the $C^*$-algebra generated by $k$-quasi-radial Toeplitz operators. This extension yields a localization framework for Toeplitz operator algebras. As an application, we consider commutative Banach algebras obtained by adjoining Toeplitz operators whose symbols are invariant under mixed unitary and circle actions. We identify their local quotient algebras on finite-coordinate strata and at infinity and use these identifications to describe their maximal ideal spaces and Gelfand transforms explicitly.
Read the full abstract
Machine Learning (Statistics)
Membership Inference via Pairwise Likelihood Ratios
Membership inference attacks (MIAs) are the standard tool for auditing the privacy risks of machine learning models. Given a query point, an MIA aims to determine whether that point was used to train the target model. In practice, such inference must rely on the statistical signals exposed by the model's outputs, such as confidence scores, logits, and intermediate feature representations. However, existing methods often fail to efficiently summarize and combine these statistical signals. To address this limitation, we propose Pairwise Likelihood MIA (PL-MIA), a unified method that combines a Gaussian likelihood-ratio (GLR) statistic with population calibration and the Cauchy combination test. We characterize theoretically how the GLR retains variance-contraction signals and establish conditions under which population calibration and Cauchy combination improve attack power. We obtain $p$-values from pairwise comparisons between the query point and reference points not used for training, and aggregate these continuous signals using the Cauchy combination test. This preserves the evidence strength that is discarded when each pairwise comparison is reduced to a binary vote. Extensive experiments demonstrate that PL-MIA outperforms strong baselines, improving the true positive rate (TPR) by over 25% in the critical low-false-positive regime, corroborating our theoretical findings. These results demonstrate how statistical principles can turn noisy model outputs into more powerful, calibrated, and reproducible evidence for membership privacy auditing.