2026-09论文中文摘要
聚焦条件：推理时自对比引导提升大语言模型中的条件文本嵌入
Focusing Condition: Inference-Time Self-Contrastive Steering Elicits Better Conditional Text Embeddings in LLMs
arXiv 2609.32684 · 2026-09-29可重构线性光学变换在单个集成多模波导中的实现
Reconfigurable Linear Optical Transformations in a Single Integrated Multimode Waveguide
arXiv 2609.32683 · 2026-09-29基于真实世界30公里30 Gbps实验演示的光子太赫兹通信的现实建模
Realistic modeling of Photonic Terahertz Communication based on a real-world 30 km 30 Gbps experimental demonstration
arXiv 2609.32682 · 2026-09-29RCVLA：用于自动驾驶的4D雷达接地语义推理与轨迹仲裁
RCVLA: 4D Radar-Grounded Semantic Reasoning and Trajectory Arbitration for Autonomous Driving
arXiv 2609.32681 · 2026-09-29GUI不是状态：诊断GUI世界模型中的状态别名问题
The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models
arXiv 2609.32679 · 2026-09-29E-过程的参考零校准阈值及其在共形鞅中的应用
Reference-Null Calibrated Thresholds for E-Processes with Applications to Conformal Martingales
arXiv 2609.32678 · 2026-09-29当更好变得更糟：自适应世界中自我改进智能体的改进保真度
When Better Gets Worse: Improvement Fidelity for Self-Improving Agents in Adaptive Worlds
arXiv 2609.32677 · 2026-09-29SIFT：通过语义不变性和结构保真微调增强时间序列基础模型
SIFT: Enhancing Time Series Foundation Models via Semantic Invariance and Structural Fidelity Fine-Tuning
arXiv 2609.32676 · 2026-09-29具有最大延迟的子可加服务成本在线覆盖
Online Covering with Maximum Delay under Subadditive Service Costs
arXiv 2609.32675 · 2026-09-29期望推理步回报统一了基于策略的奖励与教师学习
Expected Reasoning-Step Return Unifies On-Policy Learning from Rewards and Teachers
arXiv 2609.32674 · 2026-09-29自守L-函数矩的下界
Lower Bounds for Moments of Automorphic L-functions
arXiv 2609.32673 · 2026-09-29STAMP：无需目标数据预测分布外泛化
STAMP: Predicting Out-of-Distribution Generalization without Target Data
arXiv 2609.32672 · 2026-09-29概念补充稠密语义：学习用于文本-图像检索的紧凑稀疏空间
Concepts Complement Dense Semantics: Learning Compact Sparse Spaces for Text-Image Retrieval
arXiv 2609.32671 · 2026-09-29什么会证伪它？关于自解释机制性主张的变量特定证据标准
What Would Falsify It? A Variable Specific Evidence Standard for Mechanistic Claims About Self Explanation
arXiv 2609.32670 · 2026-09-29多模态 Transformer 中的细化对称性
Refinement Symmetry in Multimodal Transformers
arXiv 2609.32669 · 2026-09-29具有拟极大第三陈类的无挠层的模空间
The moduli space of torsion-free sheaves with quasi-maximal third Chern class
arXiv 2609.32668 · 2026-09-29向深思熟虑的教师学习：用于数学推理的自适应在线策略自蒸馏
Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning
arXiv 2609.32667 · 2026-09-29Rademacher 类型与 Enflo 类型再探
Rademacher Type and Enflo Type Revisited
arXiv 2609.32666 · 2026-09-29反向Kolmogorov输运：从轨迹数据采样可逆扩散的不变律
Backward Kolmogorov Transport: Sampling Invariant Laws of Reversible Diffusions from Trajectory Data
arXiv 2609.32664 · 2026-09-29Distance-KV：利用相对距离实现高效长上下文推理
Distance-KV: Exploiting Relative Distance for Efficient Long-Context Inference
arXiv 2609.32663 · 2026-09-29AsynCodeBench：软件工程中异步多智能体系统协作的基准测试
AsynCodeBench: Benchmarking Collaboration of Asynchronous Multi-Agent Systems in Software Engineering
arXiv 2609.32662 · 2026-09-29等变神经原始-对偶分配用于最大公共边子图
Equivariant Neural Primal-Dual Assignment for Maximum Common Edge Subgraphs
arXiv 2609.32661 · 2026-09-29InterTab：面向多模态表格推理的交错视觉-结构对齐
InterTab: Interleaved Visual-Structure Alignment for Multi-Modal Table Reasoning
arXiv 2609.32660 · 2026-09-29量化感知预训练与约束经验权重分布
Quantization-Aware Pre-Training with Constrained Empirical Weight Distribution
arXiv 2609.32659 · 2026-09-29契约记忆编译器：先解析，后遍历
Contract Memory Compiler: Resolve, Then Traverse
arXiv 2609.32658 · 2026-09-29具有可预测长时域边际的世界模型
World Models with Predictable Long-Horizon Marginals
arXiv 2609.32657 · 2026-09-29MixBench-TS：一个通道混合有效的多变量时间序列预测基准
MixBench-TS: A Multivariate Time Series Forecasting Benchmark Where Channel Mixing Pays Off
arXiv 2609.32656 · 2026-09-29盘型相变存储器性能增强与可扩展性的电热起源
Electro-Thermal Origins of Enhanced Performance and Scalability in Disc-Type Phase-Change Memory
arXiv 2609.32654 · 2026-09-29从知晓到弃权：弥合视觉语言模型中的表示-行动差距
From Knowing to Abstaining: Bridging the Representation-Action Gap in Vision-Language Models
arXiv 2609.32653 · 2026-09-29素数维半域与Kaplansky猜想的反例
Semifields in prime dimensions and counterexamples to Kaplansky's conjecture
arXiv 2609.32651 · 2026-09-29全局样条拟合：从GeV到最高能量的宇宙线能谱与质量组成的统一数据驱动视角
Global Spline Fit: A unified data-driven view of the cosmic-ray spectrum and mass composition from GeV to the highest energies
arXiv 2609.32649 · 2026-09-29曲线带深度：一种用于无参数平面曲线的基于带的深度
Curve Band Depth: A Band-Based Data Depth for Unparameterized Planar Curves
arXiv 2609.32648 · 2026-09-29二维不可压缩欧拉方程稳态$H^1_{\rm loc}$解的压力估计与分类
Pressure estimates and classifications for steady $H^1_{\rm loc}$-solutions to two-dimensional incompressible Euler equations
arXiv 2609.32647 · 2026-09-29通过抛物容量得到的 Morrey 测度对偶条件
Duality Conditions for Morrey Measures via Parabolic Capacity
arXiv 2609.32646 · 2026-09-29从场景图到答案：面向自动驾驶的选择性神经符号推理
From Scene Graphs to Answers: Selective Neuro-Symbolic Reasoning for Autonomous Driving
arXiv 2609.32645 · 2026-09-29二维磁体中自旋对称性跨越的有限温度相变
Finite-temperature phase transitions across spin symmetries in two-dimensional magnets
arXiv 2609.32644 · 2026-09-29商业账户入侵检测：基于智能体AI与LLM驱动的知识发现
Business Compromise Detection with Agentic AI and LLM-driven Knowledge Discovery
arXiv 2609.32643 · 2026-09-29只问不答：本地小语言模型在不泄露任务意图的前提下咨询云端大语言模型
Ask Without Telling: Local SLMs Consult Cloud LLMs Without Revealing Task Intent
arXiv 2609.32642 · 2026-09-29通过随机化算子草图实现极快且紧凑的二进制图表示
Extremely Fast and Compact Binary Graph Representations via Randomized Operator Sketching
arXiv 2609.32641 · 2026-09-29Schur-Neural KF：扩展卡尔曼滤波器的学习型Schur一致修正
Schur-Neural KF: Learned Schur-Consistent Corrections to the Extended Kalman Filter
arXiv 2609.32640 · 2026-09-29Connes-Moscovici算子的大特征值
Large eigenvalues of the Connes-Moscovici operator
arXiv 2609.32639 · 2026-09-29开放权重大语言模型（LLMs）能否模拟人类调查总体？一项跨工具校准研究
Can Open-Weight Large Language Models (LLMs) Simulate Human Survey Populations? A Cross-Instrument Calibration Study
arXiv 2609.32638 · 2026-09-29生成式人工智能与两种形式的脱钩
Generative AI & Two Forms of Decoupling
arXiv 2609.32637 · 2026-09-29一类具有弱相依重尾新息的非平稳线性过程的极限理论
Limit theory for a class of non-stationary linear processes with weakly dependent heavy-tailed innovations
arXiv 2609.32636 · 2026-09-29信任品牌，失去控制：身份如何劫持LLM智能体编排
Trust the Brand, Lose Control: How Identity Hijacks LLM Agent Orchestration
arXiv 2609.32635 · 2026-09-29PF-RL：基于目标条件价值几何的进度场强化学习用于视觉-语言-动作模型
PF-RL: Progress Field Reinforcement Learning via Goal-Conditioned Value Geometry for Vision-Language-Action Models
arXiv 2609.32634 · 2026-09-29学习评估什么：用于多目标贝叶斯优化的相关性感知解耦
Learning What to Evaluate: Correlation-Aware Decoupling for Multiobjective Bayesian Optimization
arXiv 2609.32632 · 2026-09-29SWE-MILE：面向长程软件工程智能体的异步势诱导里程碑信用分配
SWE-MILE: Asynchronous Potential-Induced Milestone Credit Assignment for Long-Horizon Software Engineering Agents
arXiv 2609.32631 · 2026-09-29ExpVoyager：动态智能体技能合成的直接经验导航
ExpVoyager: Direct Experience Navigation for Dynamic Agent Skill Synthesis
arXiv 2609.32630 · 2026-09-29各向异性拟线性椭圆系统，涉及具有负临界项的非线性项
Anisotropic quasilinear elliptic systems involving nonlinearities with negative critical terms
arXiv 2609.32629 · 2026-09-29
