2026-10论文中文摘要
UniWAM技术报告：通过混合流世界-动作建模与操作锚点姿态监督实现统一移动操作
UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision
arXiv 2609.39388 · 2026-10-01超越仿真：用于真实世界自适应的保留-修复神经算子
Beyond Simulation: Retain-and-Repair Neural Operators for Real-World Adaptation
arXiv 2609.39387 · 2026-10-01何时而非多少：评估时间序列基础模型在稀疏事件上的表现
When, Not How Much: Evaluating Time-Series Foundation Models on Sparse Events
arXiv 2609.39386 · 2026-10-01TTLab at Daleel 2026: STAR-Ar，用于阿拉伯语论证识别的序列标注
TTLab at Daleel 2026: STAR-Ar, Sequence Tagging for Argument Recognition in Arabic
arXiv 2609.39385 · 2026-10-01RoboAssist：面向长时程手术辅助的交互式人形机器人规划
RoboAssist: Interactive Human-Humanoid Planning for Long-Horizon Surgical Assistance
arXiv 2609.39384 · 2026-10-01从搜索到信号：自动启发式设计中的在线后训练
From Search to Signal: Online Post-Training in Automatic Heuristic Design
arXiv 2609.39383 · 2026-10-01SkillFM：通过潜在流匹配为LLM智能体生成技能
SkillFM: Generating Skills for LLM Agents via Latent Flow Matching
arXiv 2609.39382 · 2026-10-01对撞机上共振与非共振新物理搜索的新观测量
New Observables for Resonant and Non-Resonant New Physics Searches at Colliders
arXiv 2609.39381 · 2026-10-01InfoAgent：基于证据的信息图可追溯生成与修复
InfoAgent: Traceable Generation and Repair of Evidence-Grounded Infographics
arXiv 2609.39380 · 2026-10-01全球AI天气模型中的蝴蝶效应确认：来自热带气旋预报的证据
Butterfly Effect Confirmed in Global AI Weather Models: Evidence from Tropical Cyclone Forecasting
arXiv 2609.39379 · 2026-10-01复杂网络上的恢复随机游走与极端事件
Recovery Random Walks and Extreme Events on Complex Networks
arXiv 2609.39377 · 2026-10-01近全能仿星器中α粒子的相空间不可积性
Phase-Space Non-Integrability of Alpha Particles in Near-Omnigeneous Stellarators
arXiv 2609.39376 · 2026-10-01超越当前场景：基于事件引用的主动视角选择抓取
Beyond the Current Scene: Event-Referential Grasping with Active View Selection
arXiv 2609.39375 · 2026-10-01自由费米子演化下费米子魔幻态的高效学习
Efficient Learning of Fermionic Magic States under Free-Fermion Evolution
arXiv 2609.39372 · 2026-10-01EHR-RobustGym：用于鲁棒临床推理的智能体基准测试与训练
EHR-RobustGym: Benchmarking and Training Agents for Robust Clinical Reasoning
arXiv 2609.39371 · 2026-10-01探索异构模型合并方法用于复杂知识迁移
Exploring Heterogeneous Model Merging Approach for Complex Knowledge Transfer
arXiv 2609.39369 · 2026-10-01使网格束搜索不那么贪婪
Making Grid Beam Search Less Greedy
arXiv 2609.39368 · 2026-10-01持续学习中LoRA的动力学理论
A Dynamical Theory of LoRA in Continual Learning
arXiv 2609.39367 · 2026-10-01COBICount：分离目标与背景响应以实现无需目标数据训练的遥感目标计数
COBICount: Separating Object and Background Responses for Remote Sensing Object Counting Without Training on Target Data
arXiv 2609.39366 · 2026-10-01层流三维钝体尾迹中的不完美分岔：俯仰和偏航的影响
Imperfect bifurcations in a laminar 3D bluff body wake: effect of pitch and yaw
arXiv 2609.39364 · 2026-10-01重新思考多图像理解中的多图像再表示
Rethinking Multi-Image Re-Representation in Multi-Image Understanding
arXiv 2609.39363 · 2026-10-01隐私保护知识图谱上的链接推断攻击
Link Inference Attack on Privacy-Preserving Knowledge Graphs
arXiv 2609.39362 · 2026-10-01LampAttention：面向专用加速器的前瞻混合精度FlashAttention
LampAttention: Look-Ahead Mixed-Precision FlashAttention for Dedicated Accelerators
arXiv 2609.39361 · 2026-10-01通信高效的 $(1+\varepsilon)\Delta$-边着色与 Lovász 局部引理
Communication-Efficient $(1+\varepsilon)Δ$-Edge Coloring and Lovász Local Lemma
arXiv 2609.39359 · 2026-10-01绕开计算上限：Galahad 中的字节精确内存使 LLM 阅读成为一次性成本
Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost
arXiv 2609.39358 · 2026-10-01通过软节流增强异步SGD的鲁棒性
Robustifying Asynchronous SGD via Soft Throttling
arXiv 2609.39357 · 2026-10-01广义Boussinesq方程在偶-奇扰动下孤立波附近的全局动力学
Global dynamics near solitary waves for the generalized Boussinesq equation under even-odd perturbations
arXiv 2609.39356 · 2026-10-01不变测度作为估计量：二阶随机展开与偏差缩减
Invariant Measures as Estimators: Second-Order Stochastic Expansions and Bias Reduction
arXiv 2609.39355 · 2026-10-01基于同步压缩曲波-拉东约束的共炮点道集中相干混叠信号迭代分离
Iterative separation of coherent blended signals in common shot gathers using synchrosqueezed curvelet-Radon constraints
arXiv 2609.39354 · 2026-10-01多维椭圆型问题的良态 Birkhoff 配点法
Well-Conditioned Birkhoff-Collocation Methods for Elliptic-type Problems in Multiple Dimensions
arXiv 2609.39353 · 2026-10-01隐藏于众目睽睽之下：在LLM智能体中解耦借口与执行以实现技能投毒
Hiding in Plain Sight: Decoupling Pretext from Actuation for Skill Poisoning in LLM Agents
arXiv 2609.39352 · 2026-10-01关于基于偏好的老虎机问题的复杂性
On the Complexity of Preference-Based Bandits
arXiv 2609.39351 · 2026-10-01HAPMoE：面向混合专家模型训练的异构感知自动并行规划
HAPMoE: Heterogeneity-Aware Automatic Parallelism Planning for Mixture-of-Experts Models Training
arXiv 2609.39350 · 2026-10-01小需求的多维资源调度
Multidimensional Resource Scheduling with Small Demands
arXiv 2609.39349 · 2026-10-01完全发展的主动湍流：通过非平衡相变定义
Fully developed active turbulence defined through a non-equilibrium phase transition
arXiv 2609.39348 · 2026-10-01加性-乘性图的直径和半径
Diameter and radius of an additive-multiplicative graph
arXiv 2609.39347 · 2026-10-01离线指导，在线推理：为小型语言模型重用LLM反馈
Offline Guidance, Online Reasoning: Reusing LLM Feedback for Small Language Models
arXiv 2609.39346 · 2026-10-01量子码中的常数开销注入
Constant-Overhead Injection into Quantum Codes
arXiv 2609.39345 · 2026-10-01谁说了什么，它会被记住吗？评估跨会议中的持久说话人归属
Who Said What, and Will It Be Remembered? Evaluating Persistent Speaker Attribution Across Meetings
arXiv 2609.39344 · 2026-10-01黄金路径假说：扩散缓存中的可复用调度
The Golden Path Hypothesis: Reusable Schedules in Diffusion Caching
arXiv 2609.39343 · 2026-10-01基于信念的最大占用原则与主动推断
Belief-Based Maximum Occupancy Principle and Active Inference
arXiv 2609.39342 · 2026-10-01理解即无套利：有界荷兰赌作为语言模型的定义与训练目标
Understanding as No-Arbitrage: Bounded Dutch Books as a Definition and Training Objective for Language Models
arXiv 2609.39341 · 2026-10-01ElectrolyteFM：通过跨属性知识学习统一电解质属性预测
ElectrolyteFM: Unifying Electrolyte Property Prediction through Cross-Property Knowledge Learning
arXiv 2609.39340 · 2026-10-01匹配问题的统一对偶方法
A Unified Dual Method for Matching Problems
arXiv 2609.39339 · 2026-10-01超越完全模仿的学习：任务保持的知识蒸馏
Learning Beyond Full Imitation: Task-Preserving Knowledge Distillation
arXiv 2609.39338 · 2026-10-01WinoTS：基于小波的自蒸馏时间序列模型
WinoTS: Wavelet-based Self-Distillation for Time Series Models
arXiv 2609.39337 · 2026-10-01跨域学习中需要多少样本才足够？
How Many Samples Are Enough for Learning Across Domains?
arXiv 2609.39336 · 2026-10-01TexTailor：通过自适应服装条件实现纹理保持的视频虚拟试穿
TexTailor: Texture-Preserving Video Virtual Try-On via Adaptive Garment Conditioning
arXiv 2609.39335 · 2026-10-01驯服投机搜索以实现LLM服务中的测试时扩展
Taming Speculative Search for Test-Time Scaling in LLM Serving
arXiv 2609.39334 · 2026-10-01NarrativeSteward：智能体辅助交互式叙事创作中的协调委托、指导与验证
NarrativeSteward: Coordinating Delegation, Guidance, and Verification in Agent-Assisted Interactive Narrative Authoring
arXiv 2609.39333 · 2026-10-01
