2026-09论文中文摘要
医疗图像生成中的分词器-生成器耦合
Tokenizer-Generator Coupling in Medical Image Generation
arXiv 2608.07713 · 2026-09-29MaskFlow：精准、一致且无缝的区域图像编辑
MaskFlow: Precise, Consistent and Seamless Regional Image Editing
arXiv 2608.06929 · 2026-09-29前沿语言模型在引导压力下的发散响应模式
From Behavior to Mechanism: Tracing Divergent Response Modes in Frontier Language Models
arXiv 2608.06578 · 2026-09-29TransSLR：用于手语识别的轻量级Transformer
TransSLR: A Lightweight Transformer for Sign Language Recognition
arXiv 2608.06407 · 2026-09-29BaKron：基于克罗内克因式分解海森矩阵的高效量化方法
FastKron: Efficient Quantization with Kronecker-Factored Hessians
arXiv 2608.06291 · 2026-09-29无需展开预测任务难度
Predicting Task Difficulty Without Rollouts
arXiv 2608.05797 · 2026-09-29一次响应，始终是响应：通过潜在提示恢复检测大语言模型生成的文本
Once a Response, Always a Response: Detecting LLM-generated Text via Latent Prompt Restoration
arXiv 2608.05741 · 2026-09-29PhyLatent：为JEPA世界模型学习与动力学相关的表征
PhyLatent: Learning Dynamics-Relevant Representations for JEPA World Models
arXiv 2608.05720 · 2026-09-29超越帧选择：用多模态大语言模型重新思考长视频理解
Beyond Frame Selection: Rethinking Long-Video Understanding with MLLMs
arXiv 2608.05592 · 2026-09-29欧氏球的Steklov刚性
Geometric rigidity from the Dirichlet-to-Neumann operator
arXiv 2608.05577 · 2026-09-29随机性并非难点：先决条件DAG上教学排序的归约与复杂度
Stochasticity Is Not the Hard Part: Reduction and Complexity in Instructional Sequencing over Prerequisite DAGs
arXiv 2608.05455 · 2026-09-29SSTQ：基于子采样随机 TurboQuant 的隐私保护向量量化
SSTQ:Privacy-Preserving Vector Quantization via Subsampled Stochastic TurboQuant
arXiv 2608.05127 · 2026-09-29语音函数调用：面向大型音频语言模型的语音理解新视角
Spoken Function Calling: A New Perspective on Spoken Language Understanding for Large Audio Language Models
arXiv 2608.05126 · 2026-09-29EASy：迈向高效的基于大语言模型的智能体系统
E$^3$-Orch: Towards Effective, Efficient, and Extensible Agentic Orchestration with Reinforcement Learning
arXiv 2608.04588 · 2026-09-29相互作用玻色子模型中部分动力学对称性的量子信息指纹
Quantum-information fingerprints of partial dynamical symmetry in the interacting boson model
arXiv 2608.04486 · 2026-09-29并非所有偏差都应被抑制：在线策略蒸馏中的反事实可恢复性
Not Every Divergence Should Be Suppressed: Counterfactual Recoverability in On-Policy Distillation
arXiv 2608.04408 · 2026-09-29镜像审视：微调引发的副作用对齐偏差内省
Introspecting Alignment Shifts Beyond Behaviors Implanted Through Fine-Tuning
arXiv 2608.04347 · 2026-09-29OmniVR：用于恢复退化历史影片的联合视频-音频条件生成模型
OmniVR: Audio-Video Conditional Generation for Archival Footage Restoration
arXiv 2608.04224 · 2026-09-29SpecDrop：无参数类别条件路由的模块化专业化方法
SpecDrop: Parameter-Free Category-Conditioned Routing for Modular Specialization
arXiv 2608.04084 · 2026-09-29MissClick：利用序列化数字坐标攻击GUI定位模型
MissClick: Execution-Aware Adversarial Attacks on Coordinate Generation in GUI Grounding Models
arXiv 2608.03740 · 2026-09-29波浪水池中浮体运动的开源低成本单相机六自由度跟踪的实验验证
Experimental validation of an open-source low-cost single-camera 6-DOF tracking of floating-body motion in wave tanks
arXiv 2608.03241 · 2026-09-29时间反演对称晶体的四元数-凯勒几何
Quaternion-Kähler geometry of time reversal symmetric crystals
arXiv 2608.03178 · 2026-09-29积分希尔伯特空间与圈量子宇宙的动力学
Integral Hilbert spaces and the dynamics of loop quantum cosmos
arXiv 2608.02798 · 2026-09-29dots.tts.edit：基于连续自回归模型的精确可控语音编辑
dots.tts.edit: Precisely Controlled Speech Editing with a Continuous Autoregressive Model
arXiv 2608.02673 · 2026-09-29偏好而非安全：成对偏好并非临床安全的可靠替代指标
Preferred, Not Safer: Pairwise Preference Is a Poor Proxy for Clinical Safety
arXiv 2608.02617 · 2026-09-29树与星森林的二分Turán数
Bipartite Turán Numbers of Trees and Star Forests
arXiv 2608.01873 · 2026-09-29访问时间统计
Visiting time statistics
arXiv 2608.01453 · 2026-09-29DynActiveGS：面向动态场景重建的主动高斯溅射方法
DynActiveGS: Active Gaussian Splatting for Dynamic Scene Reconstruction
arXiv 2608.01178 · 2026-09-29q元汉明空间中作为二次偏差精确极小化元的完备码
Perfect codes as exact minimizers of quadratic discrepancy in q-ary Hamming spaces
arXiv 2608.01134 · 2026-09-29假设你已知：使用智能体AI修复流策略的认知语义
Assuming You Knew: Fixing an Epistemic Semantics for Flow Policies Using Agentic AI
arXiv 2608.00882 · 2026-09-29吸积盘中MRI产生的螺旋密度波
The generation of spiral density waves by MRI in accretion discs
arXiv 2608.00343 · 2026-09-29基于孔隙力学的多孔岩石中完全耦合的反应运移与岩土力学模拟框架
A Poromechanics-Based Framework for Fully Coupled Reactive Transport and Geomechanics in Porous Rocks
arXiv 2608.00310 · 2026-09-29AgentStream：自进化大语言模型智能体在流式任务下的表现如何？
AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?
arXiv 2608.00155 · 2026-09-29DASH-OPD：面向在线策略蒸馏的带滞后性的差异感知切换算法
DASH-OPD: Discrepancy-Aware Switching with Hysteresis for On-Policy Distillation
arXiv 2607.29078 · 2026-09-29基于低秩防御与电路引导代理的高效大语言模型对抗训练
Efficient LLM Adversarial Training via Low-Rank Defense and Circuit-Guided Surrogates
arXiv 2607.28959 · 2026-09-29NeSyFS：部分可观测场景下LLM智能体的神经符号快慢思考框架
NeSyFS: A Neuro-symbolic Fast-Slow Thinking Framework for LLM Agent under Partial Observability
arXiv 2607.28942 · 2026-09-29TELLER：用于表格实体链接的双路径迭代偏好优化
TELLER: Dual-Path Iterative Preference Optimization for Table Entity Linking
arXiv 2607.28680 · 2026-09-29相交族中的度幂问题
Degree Power Sums in Extremal Set Systems
arXiv 2607.28616 · 2026-09-29用Tayler-Spruit发电机对中子星并合遗迹进行自旋制动：全球模拟揭示大质量吸积盘与富中子抛射物的形成
Spinning down neutron-star merger remnants with the Tayler-Spruit dynamo: Global simulations reveal the formation of massive disks and neutron-rich ejecta
arXiv 2607.28556 · 2026-09-29你会步行去洗车吗？揭示大型语言模型在常识推理中的显著性偏差
Would You Walk to the Car Wash? Salience Bias in LLM Commonsense Reasoning
arXiv 2607.28478 · 2026-09-29对数莱夫谢茨不动点公式与共振边界指标
Logarithmic Lefschetz fixed point formulae and resonant boundary indices
arXiv 2607.28288 · 2026-09-29基于人工智能的评分系统低估了语言薄弱学生在物理解释中的概念理解能力
AI-based scoring systematically underestimates conceptual understanding of linguistically weak students' explanations in physics
arXiv 2607.28210 · 2026-09-29RedFlow：将失败重定向为流匹配VLA策略的动作级修正
RedFlow: Redirect Failure into Action-level Corrections for Flow-matching VLA Policy
arXiv 2607.27782 · 2026-09-29分数阶抛物型偏微分方程解的神经网络逼近
Fractional Parabolic Partial Differential Equations in Anisotropic Spectral Barron Spaces: Regularity and Neural Approximation
arXiv 2607.27781 · 2026-09-29关于Gowers与Littlewood猜想相关的一个问题
On a question of Gowers related to Littlewood's conjecture
arXiv 2607.27780 · 2026-09-29通过自监督语义扩散将技能像参数一样训练
Skill Training with Corruption and Reconstruction Loop
arXiv 2607.27557 · 2026-09-29THGFM：双分支时序异质图融合模型
THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model
arXiv 2607.27303 · 2026-09-29TurboVLA：在RTX 4090上以32 Hz运行、显存占用<1 GB的实时视觉-语言-动作模型
TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM
arXiv 2607.27205 · 2026-09-29InferScale：面向个性化大语言模型服务的原生GPU KV注入方案
InferScale: GPU-Native KV Injection for Personalized LLM Serving
arXiv 2607.27090 · 2026-09-29SciFigQual-Bench：面向全文本上下文的科学图片质量评估基准
SciFigQual-Bench: A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript Context
arXiv 2607.27084 · 2026-09-29
