2026-09论文中文摘要
单令牌稀疏自动编码器特征在因果关系上是必要的吗？层深度和SAE家族效应
Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
arXiv 2607.20596 · 2026-09-01DynamicMCPBench：用于实时MCP服务器上的大语言模型智能体的基于轨迹的效果评分基准测试
DynamicMCPBench: A Trace-Grounded, Effect-Scored Benchmark for LLM Agents over Live MCP Servers
arXiv 2607.20531 · 2026-09-01EvoSQL：用于文本到SQL的内存增强型评论家-生成器协同进化
EvoSQL: Memory-Augmented Critic-Generator Co-Evolution for Text-to-SQL
arXiv 2607.20489 · 2026-09-01给自己的笔记：大语言模型能从经验抽象中受益吗？
Notes to Self: Can LLMs Benefit from Experiential Abstractions?
arXiv 2607.20372 · 2026-09-01基于对映体构造的39维和43维球体填充
Optimal Extensions of Cross-Sections: Sphere Packings in Dimensions 38 to 43
arXiv 2607.20359 · 2026-09-01SimulS2ST-Omni：通过显式轨迹监督实现数据高效的流式语音到语音翻译
SimulS2ST-Omni: Data-Efficient Streaming Speech-to-Speech Translation via Explicit Trajectory Supervision
arXiv 2607.19810 · 2026-09-01脱离上下文的广义信赖域策略优化算法：利用特权信息学习解决难题
Off-Context GRPO: Learning to Reason on Hard Problems using Privileged Information
arXiv 2607.19313 · 2026-09-01意大利面架构师：一种抗污染、自带标签、多语言代码数据集生成器
Spaghetti Architect: A Contamination-Resistant, By-Construction-Labelled, Multi-Language Code Dataset Generator
arXiv 2607.18642 · 2026-09-01算法时代的良好治理：数字治国方略的基础
Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft
arXiv 2607.18483 · 2026-09-01模堆：同调方面
Heaps of modules: homological aspects
arXiv 2607.18328 · 2026-09-01世界杯竞技场：语言模型和深度研究代理在足球预测上的细粒度评估
WorldCupArena: Fine-Grained Evaluation of Language Models and Deep-Research Agents on Football Forecasting
arXiv 2607.18084 · 2026-09-01二维空间形式中某些凸域上第二诺伊曼特征函数的临界点
Quantitative characterization of deviations from the hot spots conjecture on convex domains in two-dimensional space forms
arXiv 2607.17882 · 2026-09-01SALT：用于长上下文压缩的显著性感知词汇字典树
SALT: Salience-Aware Lexical Trie for Long-Context Compression
arXiv 2607.17486 · 2026-09-01有限域中的连续非平方非本原元组
Consecutive non-square non-primitive tuples in finite fields
arXiv 2607.17267 · 2026-09-01InLiER：通过中间混合基数结构关键点令牌化实现无学习的异构激光雷达地点识别
InLiER: Learning-Free Heterogeneous LiDAR Place Recognition via Intermediate Mixed-Radix Structural Keypoint Tokenization
arXiv 2607.16862 · 2026-09-01稀疏图上消息传递中深度的价值：凯斯滕 - 斯蒂古姆二分法
The Value of Depth in Message Passing on Sparse Graphs: A Kesten-Stigum Dichotomy
arXiv 2607.16676 · 2026-09-01多模态和图增强的检索增强生成（RAG）何时有用？文档问答的对照评估
Controlled Evaluation of Graph and Multimodal Augmentation in RAG for Document Question Answering
arXiv 2607.16604 · 2026-09-01生成模型能把握时间吗？对合成序列表格数据的时间感知评估
Seq2Synth: Benchmarking Temporal Fidelity in Synthetic Sequential Tabular Data
arXiv 2607.15606 · 2026-09-01改进的 VBGS：实时连续变分贝叶斯高斯点云融合
ImprovedVBGS: Real-time Continual Variational Bayes Gaussian Splatting
arXiv 2607.15542 · 2026-09-01强混合下的宇宙学对撞机信号
Cosmological Collider Signals at Strong Mixing
arXiv 2607.14891 · 2026-09-01三的幂次二进制展开中的非周期性与子词复杂性
Aperiodicity and subword complexity in the binary expansion of powers of three
arXiv 2607.14774 · 2026-09-01MEMORA：基于自我中心视频的具身动作记忆用于推理与规划
MEMORA: Embodied Action Memory from Egocentric Videos for Reasoning and Planning
arXiv 2607.14252 · 2026-09-01VideoRAE：通过表示自动编码器驯服用于生成建模的视频基础模型
VideoRAE: Taming Video Foundation Models for Generative Modeling via Representation Autoencoders
arXiv 2607.14088 · 2026-09-01复对称矩阵的半代数维数与截断Toeplitz模型
Semialgebraic Dimension and Truncated Toeplitz Models for Complex Symmetric Matrices
arXiv 2607.14019 · 2026-09-01原初密度 - 势交叉相关系数与自旋因子分布之间的普遍关系
A Universal Relation Between Primordial Density-Potential Cross-correlation Coefficient and Spin Factor Distribution
arXiv 2607.13971 · 2026-09-01混乱的研究、认证与科学的货币化
The Value of Peer Review and the Reward to Reputation
arXiv 2607.13844 · 2026-09-01从弗雷德霍姆行列式到AdS/CFT可观测量：一个通用的强耦合框架
From Fredholm Determinants to AdS/CFT Observables: A Universal Strong-Coupling Framework
arXiv 2607.13140 · 2026-09-01阿尔特兰 - 齐恩鲍尔对称类中多重能带简并的拓扑表征
Topological characterization of multifold band degeneracies in Altland-Zirnbauer symmetry classes
arXiv 2607.13016 · 2026-09-01基于能量的聚类张拉整体结构物理信息找形
Energy-Based Physics-Informed Form Finding for Clustered Tensegrity Structures
arXiv 2607.12888 · 2026-09-01基于张量的降阶建模用于基于优化的反问题
Tensor-Based Reduced-Order Modeling for Optimization-Based Inverse Problems
arXiv 2607.12613 · 2026-09-01WikiSTAR：一个揭示科学维基百科文章隐藏历史的系统
WikiSTAR: A System for Shedding Light on the Hidden History of Scientific Wikipedia Articles
arXiv 2607.12441 · 2026-09-013M 魔力扣如何工作？一种断裂力学方法
How do 3M Command strips work? A fracture mechanics approach
arXiv 2607.11650 · 2026-09-01通过交替约束优化实现结构特征对齐的图学习
Decoupled Structure-Feature Alignment via Alternating Optimization for Graph Learning
arXiv 2607.11577 · 2026-09-01PRISM Edit：适用于所有时间答案的单一向量
PRISM Edit: One Vector for All Temporal Answers
arXiv 2607.11327 · 2026-09-01学习聚焦：用于医学图像分类的解剖学引导注意力正则化
Learning To Focus: Anatomy-Guided Attention Regularization for Medical Image Classification
arXiv 2607.10851 · 2026-09-01关于《HLL型MHD黎曼求解器中间状态的能量一致性》中HLLD型格式的数值扩散问题
An improved estimate of the intermediate internal energy in the energy-consistent HLLD scheme
arXiv 2607.10293 · 2026-09-01论强CP问题的起源
On the Origins of the Strong CP Problem
arXiv 2607.10272 · 2026-09-01全球导航卫星系统（GNSS）退化情况下学习到的小型无人机分离策略的运行时安全过滤
Runtime Safety Filtering for Learned Small UAS Separation Policies under GNSS Degradation
arXiv 2607.10014 · 2026-09-01关于丘成桐猜想的统计：方差渐近性
Statistics on Yau's conjecture: Variance asymptotics
arXiv 2607.09946 · 2026-09-01Agora：通过基于拍卖的任务分配增强大语言模型智能体推理
Agora: Enhancing LLM Agent Reasoning Via Auction-Based Task Allocation
arXiv 2607.09600 · 2026-09-01通过QCD求和规则对\(J/ψΩ\)质谱中隐藏魅五夸克候选态的分析
Analysis of the hidden-charm pentaquark candidates in the $J/ψΩ$ mass spectrum via the QCD sum rules
arXiv 2607.09430 · 2026-09-01极端质量比旋近（EMRIs）中偏心率调制的相位简并以及暗物质与吸积盘环境效应之间的可区分性
Eccentricity-Modulated Phase Degeneracy and Distinguishability between Dark Matter and Accretion Disk Environmental Effects in EMRIs
arXiv 2607.09214 · 2026-09-01存在但重新缩放：加法激活引导的聊天到智能体的转移
Activation Steering Transfer to Agents: One Gain Ratio Does Not Identify Potency and Efficacy
arXiv 2607.09156 · 2026-09-01关于目标跟踪中多径数据关联的置信传播收敛性
On the Convergence of Belief Propagation for Multipath Data Association in Target Tracking
arXiv 2607.08521 · 2026-09-01TFP：用于视觉运动学习的时间条件记忆融合策略
TFP: Temporally Conditioned Memory-Fusion Policies for Visuomotor Learning
arXiv 2607.08283 · 2026-09-01加权估计量的稳健推断
Robust Inference for Weighted Estimands
arXiv 2607.07524 · 2026-09-01SynthAVE：通过大语言模型领域验证实现电子商务的可扩展合成标注
SynthAVE: Scalable Synthetic Labeling for E-Commerce with LLM-Arena Validation
arXiv 2607.07469 · 2026-09-01关于非常一般的阿贝尔簇的周环及皮罗拉的一个问题
On the Chow ring of very general abelian varieties and a question of Pirola
arXiv 2607.07053 · 2026-09-01关于使轻量级博弈代理强大的因素的金标准研究
A Gold-Standard Study of What Makes a Lightweight Game-Playing Agent Strong
arXiv 2607.06854 · 2026-09-01MemDefrag：大语言模型的潜在内存碎片整理
MemDefrag: Latent Memory Defragmentation for Large Language Models
arXiv 2607.05969 · 2026-09-01
