2026-09论文中文摘要
ASSEMBLE：面向证据支撑的视频推理的原子技能
ASSEMBLE: Atomic Skills for Evidence-Grounded Video Reasoning
arXiv 2609.32466 · 2026-09-29PolyStepOR：无需最优决策的学习决策
PolyStepOR: Learning to Decide Without Optimal Decisions
arXiv 2609.32465 · 2026-09-29适应非平稳多输出高斯过程用于贝叶斯优化
Adapting Nonstationary Multi-output Gaussian Processes to Bayesian Optimization
arXiv 2609.32464 · 2026-09-29运动-语言模型能否理解结构？STRIDE：评估评估者
Can Motion-Language Models Ground Structure? STRIDE for Evaluating the Evaluators
arXiv 2609.32462 · 2026-09-29最优伪随机 $K_4$-自由图
Optimally pseudorandom $K_4$-free graphs
arXiv 2609.32461 · 2026-09-29REMEDY：视频生成距离医学教育世界模型还有多远？
REMEDY: How Far Is Video Generation from Medical Education World Models?
arXiv 2609.32460 · 2026-09-29超越模型：揭秘软件工程智能体中的框架效应
Beyond the Model: Demystifying Harness Effects in Software Engineering Agents
arXiv 2609.32459 · 2026-09-29流线型反思进化：面向任务自适应自我精炼流水线
Streamlined Reflective Evolution for Task-Adaptive Self-Refinement Pipelines
arXiv 2609.32458 · 2026-09-29写回 $\Delta$：以全新表示重新审视相同 Token
Write Back the $Δ$: Revisiting the Same Tokens with Fresh Representations
arXiv 2609.32457 · 2026-09-29看见部分，推理世界：部分观测下的视觉推理
Seeing Parts, Reasoning about Worlds: Visual Inference under Partial Observation
arXiv 2609.32456 · 2026-09-29QuacamFM：用于相机位姿估计的四元数约束流匹配
QuacamFM: Quaternion-Constrained Flow Matching for Camera Pose Estimation
arXiv 2609.32455 · 2026-09-29基于凸包自适应平移的骨架动作识别去偏方法
De-biasing Skeleton-based Action Recognition with Convex Hull Adaptive Shift
arXiv 2609.32454 · 2026-09-29反射器应看到什么？反射式提示优化中证据的实证研究
What Should the Reflector See? An Empirical Study of Evidence in Reflective Prompt Optimization
arXiv 2609.32452 · 2026-09-29AGN 喷流宿主星系 II：从半解析模型和 LoTSS 样本中获取基本信息
AGN hosting jets. II: Deriving fundamental information from semi-analytical models and the LoTSS sample
arXiv 2609.32451 · 2026-09-29关于某些替换相关的拉格朗日常数的序关系与单射性
On the Ordering and Injectivity of Lagrange Constants Associated with Certain Substitutions
arXiv 2609.32450 · 2026-09-29自我报告不能识别自我模型：反事实报告的可识别性测试
Self-Reports Do Not Identify Self-Models: An Identifiability Test for Counterfactual Reports
arXiv 2609.32449 · 2026-09-29ForkLeft：面向前缀对齐的自回归到扩散蒸馏的熵优先展开
ForkLeft: Entropy-First Rollouts for Prefix-Aligned Autoregressive-to-Diffusion Distillation
arXiv 2609.32448 · 2026-09-29并行扫描下的长度无关状态跟踪
Length-Independent State Tracking Under a Parallel Scan
arXiv 2609.32447 · 2026-09-29退火欠阻尼朗之万动力学
Annealed Underdamped Langevin Dynamics
arXiv 2609.32446 · 2026-09-29掩蔽高频词元锐化直接偏好优化
Masking Frequent Tokens Sharpens Direct Preference Optimization
arXiv 2609.32445 · 2026-09-29重新思考大语言模型强化学习中的训练-推理不匹配：其来源与纠正方法
Rethinking Training-Inference Mismatch in LLM Reinforcement Learning: Where It Arises and How to Correct It
arXiv 2609.32444 · 2026-09-29经典特征值不等式的改进 I：二维球面上的定量 Hersch 亏量与联合谱范围
Improvements of Classical Eigenvalue Inequalities I: Quantitative Hersch-deficit and the Joint Spectral Range on the Two-Sphere
arXiv 2609.32443 · 2026-09-29低维拓扑绝缘体的绝对与相对不变量
Absolute and relative invariants for low-dimensional topological insulators
arXiv 2609.32442 · 2026-09-29低维块鞍点问题无需精确内层求解
Saddle-Point Problems with a Low-Dimensional Block Do Not Need Accurate Inner Solves
arXiv 2609.32441 · 2026-09-29多面体的面距离与直径
Facial distance and diameter
arXiv 2609.32440 · 2026-09-29某些完美图和无圈图类别的支配染色
Dominated Coloring of Some Classes of Perfect and Unicyclic Graphs
arXiv 2609.32439 · 2026-09-29关于 $\ell$-正则共轭类的数量
On the number of $\ell$-regular conjugacy classes
arXiv 2609.32438 · 2026-09-29一种具有无回跳工作和低导通电压的20-kV光学注入增强型4H-SiC反向导通IGBT
A 20-kV Optical-Injection-Enhanced 4H-SiC Reverse-Conducting IGBT With Snapback-Free Operation and Low On-State Voltage
arXiv 2609.32437 · 2026-09-29通过多指标偏好选择实现可控的GNN解释
Controllable GNN Explanations via Multi-Metric Preference Selection
arXiv 2609.32436 · 2026-09-29逃离太阳俘获：用于LZ 248 keV事件的Majoron介导弹性暗物质
Escaping Solar Capture: Majoron mediated Elastic Dark Matter for the LZ 248~keV Event
arXiv 2609.32435 · 2026-09-29从潜在表示到连线：大型语言模型上的外科手术式后期编辑
From Latents to Wires: Surgical Post-Editing on Large Language Models
arXiv 2609.32434 · 2026-09-29基于反馈的量子优化：低深度与低测量开销
Feedback-based quantum optimization with low depth and measurement
arXiv 2609.32433 · 2026-09-29超越流形假设：用于流匹配的混合谱参数化
Beyond the Manifold Hypothesis: Hybrid Spectral Parameterizations for Flow Matching
arXiv 2609.32432 · 2026-09-29DualGuard：逻辑保持数据增强的双模式质量控制
DualGuard: Dual-Mode Quality Control for Logic-Preserving Data Augmentation
arXiv 2609.32431 · 2026-09-29基于主动子结构感知策略优化的多智能体系统搜索
Multi-Agent System Search via Active Substructure-aware Policy Optimization
arXiv 2609.32430 · 2026-09-29PrismQuant：分组量化器的最优零空间旋转
PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers
arXiv 2609.32429 · 2026-09-29授权闭包图：面向指令演化的LLM智能体的最小修复
Authorization Closure Graph: Minimal Repair for LLM Agents with Evolving User Instructions
arXiv 2609.32428 · 2026-09-29CityToolVQA：面向城市低空环境三维空间认知的工具增强视觉问答
CityToolVQA: Tool-Augmented Visual Question Answering for 3D Spatial Cognition in Urban Low-Altitude Environments
arXiv 2609.32427 · 2026-09-29HoTS：同质性感知的温度缩放用于图神经网络校准
HoTS: Homophily-Aware Temperature Scaling for Graph Neural Network Calibration
arXiv 2609.32426 · 2026-09-29对称分支通过Laurent现象代数
Symmetric Branching via Laurent Phenomenon Algebras
arXiv 2609.32425 · 2026-09-29CyberClear：面向APT攻击链溯源的大语言模型智能体系统基准
CyberClear: A Benchmark for LLM Agent Systems on APT Attack Chain Provenance
arXiv 2609.32424 · 2026-09-29PluginRSI：利用可复用插件递归改进智能体框架
PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins
arXiv 2609.32423 · 2026-09-29MergeHEIR：缓解模型合并带来的多模态幻觉税
MergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging
arXiv 2609.32422 · 2026-09-29Carnator：基于生成原生兼容性引导的跨请求复用的快速文本到视频生成
Carnator: Fast Text-to-Video Generation with Generation-Native Compatibility-Guided Cross-Request Reuse
arXiv 2609.32420 · 2026-09-29$\Omega^-\bar{\Omega}^+$ 对中四维纠缠的束流极化启用见证
Beam-polarization-enabled witness of four-dimensional entanglement in $Ω^-\barΩ^+$ pairs
arXiv 2609.32419 · 2026-09-29TESS振荡光谱自动化分析揭示$\delta$ Scuti星中的多重规则频率间隔
Multiple regular frequency spacings in $δ$ Scuti stars from automated analysis of TESS oscillation spectra
arXiv 2609.32418 · 2026-09-29UMD空间中带分数阶导数的非自治演化方程的随机极大$L^p$-正则性
Stochastic maximal $L^p$-regularity for non-autonomous evolution equations with fractional derivative in UMD spaces
arXiv 2609.32417 · 2026-09-29RE-0：通过局部在线蒸馏实现具身代码即策略智能体的验证递归改进
RE-0: Verified Recursive Improvement of Embodied Code-as-Policy Agents through Local On-Policy Distillation
arXiv 2609.32416 · 2026-09-29RefAdapt-DiT：用于参考条件扩散Transformer的自适应联合注意力
RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers
arXiv 2609.32415 · 2026-09-29AI如何改变DevOps性能：一种基于机制的模拟
How AI Changes DevOps Performance: A Mechanism-Based Simulation
arXiv 2609.32414 · 2026-09-29
