2026-09论文中文摘要
随机梯度方法中的算术简单性
Arithmetic Simplicity in Stochastic Gradient Methods
arXiv 2609.32240 · 2026-09-29联邦子空间引导的视觉-语言-动作策略蒸馏用于非独立同分布多机器人操作
Federated Subspace Guided Vision-Language-Action Policy Distillation for Non-IID Multi-Robot Manipulation
arXiv 2609.32239 · 2026-09-29FRTB-IMA资本的精确交易级归因
Exact Trade-Level Attribution of FRTB-IMA Capital
arXiv 2609.32238 · 2026-09-29带宽，而非浮点运算次数：Apple M6 上的 FFT 内核、矩阵单元与 SAR 成像
Bandwidth, Not FLOPS: FFT Kernels, Matrix Units and SAR Imaging on Apple M6
arXiv 2609.32237 · 2026-09-29RoboFFT：通过前向过程的在线强化学习微调生成式机器人策略
RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process
arXiv 2609.32236 · 2026-09-29解决每一步还不够：里程碑提示揭示大语言模型数学推理中的组合缺口
Solving Every Step Is Not Enough: Milestone Oracles Reveal a Composition Gap in LLM Math Reasoning
arXiv 2609.32235 · 2026-09-29KinyaMed：种子，而非行——以错误单位编写的语料需求未能约束什么
KinyaMed: Seeds, Not Rows -- What a Corpus Requirement Written in the Wrong Unit Fails to Constrain
arXiv 2609.32234 · 2026-09-29Navier-Stokes方程在涡-波数据下的全局时间无粘极限
Global-in-time inviscid limit of the Navier-Stokes equations for vortex-wave data
arXiv 2609.32233 · 2026-09-29超流体溃坝中的量子起伏涌潮、彩虹与事件视界
Quantum undular bores, rainbows, and event horizons in superfluid dam breaks
arXiv 2609.32232 · 2026-09-29流中的骨架：用于人体运动预测的图结构化流匹配
Skeletons in Flow: Graph Structured Flow Matching for Human Motion Prediction
arXiv 2609.32231 · 2026-09-29“我能在哪里信任你？”：代理模型保真度的边界感知评估
"Where Can I Trust You?": Boundary-Aware Evaluation of Surrogate Fidelity
arXiv 2609.32230 · 2026-09-29高维偏微分方程的有限表达式逼近：无维数灾难
Finite Expression Approximation of High-Dimensional PDEs Without the Curse of Dimensionality
arXiv 2609.32229 · 2026-09-29CompassPlay：对提议者引导求解器的方向给予奖励
CompassPlay: Rewarding the Proposer for Where It Moves the Solver
arXiv 2609.32228 · 2026-09-29OptiArena：在固定资源预算下，大型语言模型能否改进可执行算法？
OptiArena: Can LLMs Improve Executable Algorithms under Fixed Resource Budgets?
arXiv 2609.32227 · 2026-09-29迈向智能体光网络：LLM智能体驱动的自主生命周期管理愿景
Toward Agentic Optical Networks: A Vision of LLM Agent-Driven Autonomous Lifecycle Management
arXiv 2609.32226 · 2026-09-29LaMET-Agent：用于大动量有效理论分析的智能体框架
LaMET-Agent: An Agent Framework for Large-Momentum Effective Theory Analysis
arXiv 2609.32225 · 2026-09-29RAO-Nav：探索全语言模型用于零样本语义视听导航
RAO-Nav: Probing Omni-Language Models for Zero-shot Semantic Audio-Visual Navigation
arXiv 2609.32224 · 2026-09-29反馈线性化表示的物理信息学习
Physics-Informed Learning of Feedback-Linearizing Representations
arXiv 2609.32223 · 2026-09-29面向原始3D点云的保几何盲水印
Geometry-Preserving Blind Watermarking for Raw 3D Point Clouds
arXiv 2609.32222 · 2026-09-29用于电力系统稳定性分析的神经-解析能量函数
Neural-Analytica Energy Functions for Power System Stability Analysis
arXiv 2609.32221 · 2026-09-29一种通过基于评分准则的剧本归纳构建的双语AI听力学家，在模拟病例盲评中表现优于人类听力学家
A bilingual AI audiologist built through rubric-guided playbook induction outperforms human audiologists in a blinded evaluation of simulated cases
arXiv 2609.32220 · 2026-09-29用于电力系统稳定性分析的神经分散式条件
Neural Decentralized Conditions for Power System Stability Analysis
arXiv 2609.32219 · 2026-09-29扩展Hořava引力的Hamilton-Jacobi与辛分析
The Hamilton--Jacobi and Symplectic Analysis for Extended Hořava Gravity
arXiv 2609.32218 · 2026-09-29面向云原生HPC的自适应运行时系统中的无重启弹性机制
No-Restart Elasticity in an Adaptive Runtime System for Cloud-Native HPC
arXiv 2609.32217 · 2026-09-29理性对话者模型：理解与生成的统一
A model of rational interlocutors: Unification of comprehension and production
arXiv 2609.32216 · 2026-09-29DP-Rec：面向高效长序列推荐的自适应补丁化方法
DP-Rec: Towards Dynamic Patching for Efficient Long-Sequence Recommendation
arXiv 2609.32215 · 2026-09-29Weyl交错式的根估计
Rootwise estimates for Weyl alternants
arXiv 2609.32214 · 2026-09-29HM-ROUTER：面向智能体系统的联合模型与框架路由
HM-ROUTER: Joint Model and Harness Routing for Agentic Systems
arXiv 2609.32213 · 2026-09-29手性孤子晶格的高阶形式描述
Higher-Form Description of Chiral Soliton Lattice
arXiv 2609.32212 · 2026-09-29排名置信序列：随时有效的排行榜
Rank Confidence Sequences:Anytime-valid Leaderboards
arXiv 2609.32211 · 2026-09-290/1立方体中多面体的Chvátal秩的二次上界
A quadratic upper bound on the Chvátal rank of polytopes in the 0/1-cube
arXiv 2609.32210 · 2026-09-29Fracast-0：仅85K参数的时间序列基础模型的分形权重共享
Fracast-0: Fractal Weight Sharing for a Time Series Foundation Model with Only 85K Parameters
arXiv 2609.32209 · 2026-09-29Witness：交互式谜题环境中的发现、解读与顿悟
Witness: Discovery, Deciphering, and Epiphany in Interactive Puzzle Environments
arXiv 2609.32208 · 2026-09-29关于超划分上奇数最小缺失项之和
On the sum of odd minimal excludants over overpartitions
arXiv 2609.32207 · 2026-09-29多物种等离子体Vlasov-Ampère模型的保正半拉格朗日间断伽辽金方法
Positivity-preserving semi-Lagrangian discontinuous Galerkin methods for multi-species Vlasov-Ampère models of plasma
arXiv 2609.32206 · 2026-09-29逆向设计的纳米束腔用于硅中高协同性自旋-光子界面
Inverse-Designed Nanobeam Cavities for High-Cooperativity Spin-Photon Interfaces in Silicon
arXiv 2609.32205 · 2026-09-29DegreeSpar：面向高效安全Transformer推理的结构化度数稀疏化
DegreeSpar: Structured Degree Sparsity for Efficient Secure Transformer Inference
arXiv 2609.32204 · 2026-09-29基于核的CLIP引导：利用视觉-语言模型偏好
Kernel-Based Steering of CLIP with Vision-Language Model Preferences
arXiv 2609.32203 · 2026-09-29轻粒子尖锐引力产生的普适类
Universality classes of sharp gravitational production of light particles
arXiv 2609.32202 · 2026-09-29五维AdS-Chern-Simons引力中的非均匀尘埃坍缩
Inhomogeneous dust collapse in five-dimensional AdS-Chern-Simons gravity
arXiv 2609.32200 · 2026-09-29神经语言模型学习轨迹中的泛化与记忆：分类的几何解释
Generalization and Memorization along the Learning Trajectory of Neural Language Models: A Geometric Account of Categorization
arXiv 2609.32199 · 2026-09-29作为软件的智能体：面向智能体可靠性的编程语言议程
Agents as Software: A Programming Languages Agenda for Agent Reliability
arXiv 2609.32198 · 2026-09-29SparSP：利用通信稀疏性实现序列并行视频扩散Transformer
SparSP: Exploiting Communication Sparsity for Sequence-Parallel Video DiTs
arXiv 2609.32197 · 2026-09-29裁判并非其孪生：后训练使模型的写作更具可预测性，但作为裁判，其品味几乎未向可预测写作倾斜
The Judge Is Not Its Twin: Post-training makes a model's writing more predictable but barely moves its taste, as a judge, toward predictable writing
arXiv 2609.32196 · 2026-09-29$\eta$-重尾梯度噪声下SGD的高概率保证
High-Probability Guarantees for SGD under $β$-Heavy-Tailed Gradient Noise
arXiv 2609.32195 · 2026-09-29利用表面磁通输运模型成功重现太阳轨道器极区磁场观测
Successfully Reproducing Solar Orbiter Polar Field Observations with a Surface Flux Transport Model
arXiv 2609.32194 · 2026-09-29螺旋调和积
Spiral Harmonic Products
arXiv 2609.32191 · 2026-09-29评估基于单组学和多组学的可解释人工智能（MOXAI）用于成人型弥漫性胶质瘤分子亚类分类
Evaluating Single and Multi-Omics Based Explainable Artificial Intelligence (MOXAI) for Molecular Subclass Classification of Adult-Type Diffuse Gliomas
arXiv 2609.32190 · 2026-09-29ScopeIF：通过分级奖励建模提升大语言模型中范围感知的精确指令遵循能力
ScopeIF: Improving Scope-Aware Precise Instruction-Following in Large Language Models via Graded Reward Modeling
arXiv 2609.32189 · 2026-09-29存在并非忠实：文生图中的喻体侵入
Presence Is Not Faithfulness: Figurative Vehicle Intrusion in Text-to-Image Generation
arXiv 2609.32188 · 2026-09-29
