机器学习学术速递[5.5]

点击阅读原文访问arxivdaily.com，涵盖CS|物理|数学|经济|统计|金融|生物|电气领域，更有搜索、收藏等功能！

cs.LG 方向，今日共计359篇

大模型相关(25篇)

【1】U-Define: Designing User Workflows for Hard and Soft Constraints in LLM-Based Planning
标题：U-define：在基于LLM的规划中设计针对硬约束和软约束的用户工作流
链接：https://arxiv.org/abs/2605.02765

【2】Bolek: A Multimodal Language Model for Molecular Reasoning
标题：Bolek：一种用于分子推理的多模式语言模型
链接：https://arxiv.org/abs/2605.02745

【3】Gradient-Gated DPO: Stabilizing Preference Optimization in Language Models
标题：学生门控DPO：稳定语言模型中的偏好优化
链接：https://arxiv.org/abs/2605.02626

【4】On Training Large Language Models for Long-Horizon Tasks: An Empirical Study of Horizon Length
标题：关于为长期任务训练大型语言模型：地平线长度的实证研究
链接：https://arxiv.org/abs/2605.02572

【5】Statistically-Lossless Quantization of Large Language Models
标题：大型语言模型的统计无损量化
链接：https://arxiv.org/abs/2605.02404

【6】When Correct Isn't Usable: Improving Structured Output Reliability in Small Language Models
标题：当无法使用正确时：提高小型语言模型中的结构化输出可靠性
链接：https://arxiv.org/abs/2605.02363

【7】Break the Block: Dynamic-size Reasoning Blocks for Diffusion Large Language Models via Monotonic Entropy Descent with Reinforcement Learning
标题：打破障碍：通过单调性熵下降和强化学习扩散大型语言模型的动态大小推理模块
链接：https://arxiv.org/abs/2605.02263

【8】Perturbation Dose Responses in Recursive LLM Loops: Raw Switching, Stochastic Floors, and Persistent Escape under Append, Replace, and Dialog Updates
标题：循环LLM循环中的扰动剂量响应：附加、替换和对话框更新下的原始切换、随机下限和持续逃逸
链接：https://arxiv.org/abs/2605.02236

【9】Pair2Score: Pairwise-to-Absolute Transfer for LLM-Based Essay Scoring
标题：Pair 2 Score：基于LLM的论文评分的成对到绝对转移
链接：https://arxiv.org/abs/2605.02069

【10】Real-Time Text Transmission via LLM-Based Entropy Coding over Fixed-Rate Channels
标题：固定速率通道上通过基于LLM的熵编码进行实时文本传输
链接：https://arxiv.org/abs/2605.01991

【11】Learn-to-learn on Arbitrary Textual Conditioning: A Hypernetwork-Driven Meta-Gated LLM
标题：任意文本条件下的学着学：超网络驱动的元门控LLM
链接：https://arxiv.org/abs/2605.01973

【12】RefusalGuard: Geometry-Preserving Fine-Tuning for Safety in LLMs
标题：Refminster Guard：为LLM的安全而进行几何保留微调
链接：https://arxiv.org/abs/2605.01913

【13】Molecular Representations for Large Language Models
标题：大型语言模型的分子表示
链接：https://arxiv.org/abs/2605.01822

【14】Mitigating Multimodal LLMs Hallucinations via Relevance Propagation at Inference Time
标题：通过推理时的相关传播减轻多模态LLM幻觉
链接：https://arxiv.org/abs/2605.01766

【15】SplitZip: Ultra Fast Lossless KV Compression for Disaggregated LLM Serving
标题：SplitZip：用于分解LLM服务的超快速无损KV压缩
链接：https://arxiv.org/abs/2605.01708

【16】Importance-Guided Basis Selection for Low-Rank Decomposition of Large Language Models
标题：大型语言模型低等级分解的重要性引导基选择
链接：https://arxiv.org/abs/2605.01627

【17】Evaluating LLMs on Large-Scale Graph Property Estimation via Random Walks
标题：通过随机游动评估大规模图属性估计的LLM
链接：https://arxiv.org/abs/2605.01484

【18】Active Reasoning Vision-Language Models via Sequential Experimental Design
标题：通过顺序实验设计的主动推理视觉语言模型
链接：https://arxiv.org/abs/2605.01345

【19】The Partial Testimony of Logs: Evaluation of Language Model Generation under Confounded Model Choice
标题：NPS的部分证词：混淆模型选择下语言模型生成的评估
链接：https://arxiv.org/abs/2605.01311

【20】Activation Compression in LLMs: Theoretical Analysis and Efficient Algorithm
标题：LLM中的激活压缩：理论分析和高效算法
链接：https://arxiv.org/abs/2605.01255

【21】When Embedding-Based Defenses Fail: Rethinking Safety in LLM-Based Multi-Agent Systems
标题：当基于嵌入的防御失败时：重新思考基于LLM的多智能体系统中的安全性
链接：https://arxiv.org/abs/2605.01133

【22】LLM Ghostbusters: Surgical Hallucination Suppression via Adaptive Unlearning
标题：LLM捉鬼敢死队：通过适应性遗忘抑制手术幻觉
链接：https://arxiv.org/abs/2605.01047

【23】CLEAR: Revealing How Noise and Ambiguity Degrade Reliability in LLMs for Medicine
标题：Clearar：揭示噪音和模糊性如何降低医学LLM的可靠性
链接：https://arxiv.org/abs/2605.01011

【24】Generalized Category Discovery under Domain Shifts: From Vision to Vision-Language Models
标题：领域转移下的广义类别发现：从视觉到视觉语言模型
链接：https://arxiv.org/abs/2605.00906

【25】H-Probes: Extracting Hierarchical Structures From Latent Representations of Language Models
标题：H-Probes：从语言模型的潜在表示中提取分层结构
链接：https://arxiv.org/abs/2605.00847

Graph相关(图学习|图神经网络|图优化等)(13篇)

【1】A Closed-Form Persistence-Landmark Pipeline for Certified Point-Cloud and Graph Classification
标题：用于认证点云和图形分类的封闭式持久性地标管道
链接：https://arxiv.org/abs/2605.02836

【2】Fine-Grained Graph Generation through Latent Mixture Scheduling
标题：通过潜在混合调度生成细粒度图
链接：https://arxiv.org/abs/2605.02780

【3】Graph Federated Unlearning for Privacy Preservation
标题：用于隐私保护的图联邦遗忘算法
链接：https://arxiv.org/abs/2605.02297

【4】Large margin classifier with graph-based adaptive regularization
标题：具有基于图的自适应正规化的大余量分类器
链接：https://arxiv.org/abs/2605.02027

【5】Misclassification Rate and Privacy-Utility Trade-offs in Graph Convolutional Networks via Subsampling Stability
标题：基于子采样稳定性的图卷积网络误分类率和隐私-效用权衡
链接：https://arxiv.org/abs/2605.01987

【6】Federated Semi-Supervised Graph Neural Networks with Prototype-Guided Pseudo-Labeling for Privacy-Preserving Gestational Diabetes Mellitus Prediction
标题：具有原型引导伪标记的联邦半监督图神经网络用于隐私保护妊娠糖尿病预测
链接：https://arxiv.org/abs/2605.01810

【7】GraphSculptor: Sculpting Pre-training Coreset for Graph Self-supervised Learning
标题：GraphSculptor：为图形自我监督学习雕刻预训练核心集
链接：https://arxiv.org/abs/2605.01310

【8】Spectral Graph Sparsification Preserves Representation Geometry in Graph Neural Networks
标题：谱图稀疏保留图神经网络中的表示几何
链接：https://arxiv.org/abs/2605.01136

【9】Graph Rewiring in GNNs to Mitigate Over-Squashing and Over-Smoothing: A Survey
标题：GNN中的图形重新布线以缓解过度挤压和过度平滑：一项调查
链接：https://arxiv.org/abs/2605.00951

【10】PhaseNet++: Phase-Aware Frequency-Domain Anomaly Detection for Industrial Control Systems via Phase Coherence Graphs
标题：PhaseNet++：通过相一致图对工业控制系统进行相感知频域异常检测
链接：https://arxiv.org/abs/2605.00929

【11】Trees and Graphs with Non Log-concave Dominating Set Sequence via AI Tools
标题：通过人工智能工具具有非log-凹支配集序列的树和图
链接：https://arxiv.org/abs/2605.02193

【12】An ALE-Consistent Graph Neural Operator-Transformer Framework for Fluid-Structure Interaction
标题：流固相互作用的ALE-consistent图神经操作器-变换器框架
链接：https://arxiv.org/abs/2605.00937

【13】NAKUL-Med: Spectral-Graph State Space Models with Dynamics Kernels for Medical Signals
标题：NAKUL-Med：具有医学信号动力学核的谱图状态空间模型
链接：https://arxiv.org/abs/2605.00871

Transformer(11篇)

【1】Trust, but Verify: Peeling Low-Bit Transformer Networks for Training Monitoring
标题：信任，但要验证：剥离低位Transformer网络用于训练监控
链接：https://arxiv.org/abs/2605.02853

【2】InfiltrNet: Dual-Branch CNN-Transformer Architecture for Brain Tumor Infiltration Risk Prediction
标题：InfiltrNet：用于脑肿瘤渗透风险预测的双分支CNN-Transformer架构
链接：https://arxiv.org/abs/2605.02230

【3】Projection-Free Transformers via Gaussian Kernel Attention
标题：通过高斯核注意力的无投影Transformer
链接：https://arxiv.org/abs/2605.02144

【4】QHyer: Q-conditioned Hybrid Attention-mamba Transformer for Offline Goal-conditioned RL
标题：QHyer：Q调节混合注意力曼巴Transformer，用于离线目标调节RL
链接：https://arxiv.org/abs/2605.01862

【5】Concepts Whisper While Syntax Shouts: Spectral Anti-Concentration and the Dual Geometry of Transformer Representations
标题：概念低语而尖叫：光谱反集中和Transformer表示的双重几何
链接：https://arxiv.org/abs/2605.01609

【6】CGFformer: Cluster-Guidance Frequency Transformer for Pansharpening
标题：CGFformer：用于Panshiring的直升机-制导频率Transformer
链接：https://arxiv.org/abs/2605.01490

【7】Multi-Perspective Transformers in ARC-AGI-2 Challenge
标题：AR-AGI-2挑战赛中的多视角Transformer
链接：https://arxiv.org/abs/2605.01154

【8】LEAP: Layer-wise Exit-Aware Pretraining for Efficient Transformer Inference
标题：LEAP：逐层退出感知预训练，以实现高效的Transformer推理
链接：https://arxiv.org/abs/2605.01058

【9】Reconstructing conformal field theoretical compositions with Transformers
标题：用Transformer重建保形场论成分
链接：https://arxiv.org/abs/2605.01072

【10】Robust Cross-Domain WiFi Fall Detection via Physics-Driven Attention-Enhanced Transformers
标题：通过物理驱动的注意力增强变形器进行稳健的跨域WiFi跌落检测
链接：https://arxiv.org/abs/2605.00869

【11】1BT: One-Block Transformer for EEG-Based Cognitive Workload Assessment
标题：1BT：一种基于EEG的认知功能评估的单块Transformer
链接：https://arxiv.org/abs/2605.00856

GAN|对抗|攻击|生成相关(23篇)

【1】A decoupled diffusion planner that adapts to changing cost limits by using cost-conditioned generation for safety and reward gradients for performance
标题：一个脱钩的扩散规划器，通过使用成本条件发电来实现安全性和回报梯度来实现性能，来适应不断变化的成本限制
链接：https://arxiv.org/abs/2605.02777

【2】Efficient Preference Poisoning Attack on Offline RLHF
标题：离线WLHF的高效偏好中毒攻击
链接：https://arxiv.org/abs/2605.02495

【3】Anomaly-Preference Image Generation
标题：异常偏好图像生成
链接：https://arxiv.org/abs/2605.02439

【4】The Compliance Trap: How Structural Constraints Degrade Frontier AI Metacognition Under Adversarial Pressure
标题：合规陷阱：结构性约束如何在对抗压力下降级前沿人工智能元认知
链接：https://arxiv.org/abs/2605.02398

【5】Decoding-Time Debiasing via Process Reward Models: From Controlled Fill-in to Open-Ended Generation
标题：通过流程奖励模型进行解码时去偏置：从受控填充到开放式生成
链接 ：https://arxiv.org/abs/2605.02348

【6】Can Causal Discovery Algorithms Help in Generating Legal Arguments?
标题：因果发现算法可以帮助生成法律论据吗？
链接：https://arxiv.org/abs/2605.02318

【7】DurableUn: Quantization-Induced Recovery Attacks in Machine Unlearning
标题：DurableUn：机器学习中的量化诱导恢复攻击
链接：https://arxiv.org/abs/2605.02196

【8】Manifold-Constrained Adversarial Training for Long-Tailed Robustness via Geometric Alignment
标题：通过几何对齐实现长尾鲁棒性的Manifold约束对抗训练
链接：https://arxiv.org/abs/2605.02183

【9】Adversarial Update-Based Federated Unlearning for Poisoned Model Recovery
标题：基于对抗更新的联邦撤销学习用于中毒模型恢复
链接：https://arxiv.org/abs/2605.02110

【10】Detecting Adversarial Data via Provable Adversarial Noise Amplification
标题：通过可证明的对抗性噪音放大检测对抗性数据
链接：https://arxiv.org/abs/2605.02109

【11】TRAP: Tail-aware Ranking Attack for World-Model Planning
标题：TRAP：针对世界模型规划的尾部感知排名攻击
链接：https://arxiv.org/abs/2605.01950

【12】Remote Action Generation: Remote Control with Minimal Communication
标题：远程动作生成：通过最少通信进行远程控制
链接：https://arxiv.org/abs/2605.01833

【13】Skipping the Zeros in Diffusion Models for Sparse Data Generation
标题：跳过扩散模型中的零来生成稀疏数据
链接：https://arxiv.org/abs/2605.01817

【14】Adversarial Imitation Learning with General Function Approximation: Theoretical Analysis and Practical Algorithms
标题：通用函数逼近的对抗性模仿学习：理论分析和实用算法
链接：https://arxiv.org/abs/2605.01778

【15】Anticipation-VLA: Solving Long-Horizon Embodied Tasks via Anticipation-based Subgoal Generation
标题：预期-VLA：通过基于预期的子目标生成来解决长期预定任务
链接：https://arxiv.org/abs/2605.01772

【16】Robust Linear Dueling Bandits with Post-serving Context under Unknown Delays and Adversarial Corruptions
标题：未知延迟和对抗性腐蚀下服务后环境下的鲁棒线性决斗强盗
链接：https://arxiv.org/abs/2605.01752

【17】Model-Based Proactive Cost Generation for Learning Safe Policies Offline with Limited Violation Data
标题：基于模型的主动成本生成，用于学习具有有限违规数据的安全离线策略
链接：https://arxiv.org/abs/2605.01356

【18】GA-VisAgent: A Multi-Agent application for code generation and visualization in interactive learning
标题：GA-VisAgent：用于交互式学习中代码生成和可视化的多Agent应用程序
链接：https://arxiv.org/abs/2605.01299

【19】CADFit: Precise Mesh-to-CAD Program Generation with Hybrid Optimization
标题：CADFit：采用混合优化的精确网格到CAD程序生成
链接：https://arxiv.org/abs/2605.01171

【20】Almost for Free: Crafting Adversarial Examples with Convolutional Image Filters
标题：几乎免费：使用卷积图像过滤器制作对抗示例
链接：https://arxiv.org/abs/2605.01098

【21】PPO guided Agentic Pipeline for Adaptive Prompt Selection and Test Case Generation
标题：PPO引导的自适应快速选择和测试用例生成的并行流水线
链接：https://arxiv.org/abs/2605.00942

【22】Fusing Urban Structure and Semantics: A Conditional Diffusion Model for Cross-City OD Matrix Generation
标题：融合城市结构和语义：跨城市OD矩阵生成的条件扩散模型
链接：https://arxiv.org/abs/2605.00938

【23】Autonomous QA Agent: A Retrieval-Augmented Framework for Reliable Selenium Script Generation
标题：自主QA代理：用于可靠生成Selenium脚本的检索增强框架
链接：https://arxiv.org/abs/2601.06034

半/弱/无/有监督|不确定性|主动学习(12篇)

【1】Unsupervised Machine Learning for Detecting Structural Anomalies in European Regional Statistics
标题：无监督机器学习用于检测欧洲区域统计中的结构异常
链接：https://arxiv.org/abs/2605.02884

【2】Gradient-Discrepancy Acquisition for Pool-Based Active Learning
标题：基于池的主动学习的学生差异获取
链接：https://arxiv.org/abs/2605.02609

【3】Mixture Prototype Flow Matching for Open-Set Supervised Anomaly Detection
标题：开集监督异常检测的混合原型流匹配
链接：https://arxiv.org/abs/2605.02438

【4】Benchmarking Single-Pose Docking, Consensus Rescoring, and Supervised ML on the LIT-PCBA Library: A Critical Evaluation of DiffDock, AutoDock-GPU, GNINA, and DiffDock-NMDN
标题：在LIT-PCBA库上对单姿态对接、共识重新评分和监督ML进行基准测试：对迪夫Dock、AutoDock-图形处理器、GNINA和迪夫Dock-NMDN的批判性评估
链接：https://arxiv.org/abs/2605.01681

【5】PACE: Parameter Change for Unsupervised Environment Design
标题：PACE：无监督环境设计的参数更改
链接：https://arxiv.org/abs/2605.01358

【6】Rhamba: Region-Aware Hybrid Attention-Mamba Framework for Self-Supervised Learning in Resting-State fMRI
标题：Rhamba：用于静息状态fMRI中自我监督学习的区域感知混合注意力-曼巴框架
链接：https://arxiv.org/abs/2605.01240

【7】Divergence is Uncertainty: A Closed-Form Posterior Covariance for Flow Matching
标题：分歧是不确定性：流量匹配的封闭后验协方差
链接：https://arxiv.org/abs/2605.00941

【8】CGM-JEPA: Learning Consistent Continuous Glucose Monitor Representations via Predictive Self-Supervised Pretraining
标题：CGM-JEPA：通过预测性自我监督预训练学习一致的连续葡萄糖监测仪表示
链接：https://arxiv.org/abs/2605.00933

【9】Dimensionality-Aware Anomaly Detection in Learned Representations of Self-Supervised Speech Models
标题：自监督语音模型学习表示中的感知异常检测
链接：https://arxiv.org/abs/2605.02715

【10】A Semi-Supervised Kernel Two-Sample Test
标题：半监督核双样本检验
链接 ：https://arxiv.org/abs/2605.01775

【11】MU-SHOT-Fi: Self-Supervised Multi-User Wi-Fi Sensing with Source-free Unsupervised Domain Adaptation
标题：MU-SHOT-Fi：具有无源无监督域自适应的自监督多用户Wi-Fi感知
链接：https://arxiv.org/abs/2605.01369

【12】Autonomous Reliability Qualification of Ga$_2$O$_3$-based Hydrogen and Temperature Sensors via Safe Active Learning
标题：通过安全主动学习对Ga$2$O$3 $基氢和温度传感器进行自主可靠性鉴定
链接：https://arxiv.org/abs/2605.00868

迁移|Zero/Few/One-Shot|自适应(23篇)

【1】SpecKV: Adaptive Speculative Decoding with Compression-Aware Gamma Selection
标题：SpecKV：具有压缩感知伽玛选择的自适应推测解码
链接：https://arxiv.org/abs/2605.02888

【2】Adaptive Interpolation-Synthesis for Motion In-Betweening on Keyframe-Based Animation
标题：基于关键帧动画的自适应插补合成
链接：https://arxiv.org/abs/2605.02742

【3】ProPACT: A Proactive AI-Driven Adaptive Collaborative Tutor for Pair Programming
标题：ProPACT：一个主动的人工智能驱动的配对编程自适应协作导师
链接：https://arxiv.org/abs/2605.02703

【4】Reference-Sampled Boltzmann Projection for KL-Regularized RLVR: Target-Matched Weighted SFT, Finite One-Shot Gaps, and Policy Mirror Descent
标题：KL正规化WLVR的参考抽样Boltzmann投影：目标匹配加权SFT、有限单次间隙和政策镜像下降
链接：https://arxiv.org/abs/2605.02469

【5】Dueling DDQN-Based Adaptive Multi-Objective Handover Optimization for LEO Satellite Networks
标题：基于决斗DDQN的LEO卫星网络自适应多目标切换优化
链接：https://arxiv.org/abs/2605.02416

【6】Closed-Loop CO2 Storage Control With History-Based Reinforcement Learning and Latent Model-Based Adaptation
标题：基于历史强化学习和潜在模型自适应的闭环CO2存储控制
链接：https://arxiv.org/abs/2605.02405

【7】Demographic-Aware Transfer Learning for Sleep Stage Classification in Clinical Polysomnography
标题：临床多导睡眠图中睡眠阶段分类的人口感知迁移学习
链接：https://arxiv.org/abs/2605.02245

【8】Heterogeneous Model Fusion for Privacy-Aware Multi-Camera Surveillance via Synthetic Domain Adaptation
标题：通过合成域自适应实现隐私感知多摄像机监控的异类模型融合
链接：https://arxiv.org/abs/2605.02169

【9】Flexi-LoRA with Input-Adaptive Ranks: Efficient Finetuning for Speech and Reasoning Tasks
标题：具有输入自适应排名的CLAR-LoRA：语音和推理任务的高效微调
链接：https://arxiv.org/abs/2605.01959

【10】Selector-Guided Autonomous Curriculum for One-Shot Reinforcement Learning from Verifiable Rewards
标题：选择者引导的自主课程，用于从可验证奖励中进行一次性强化学习
链接：https://arxiv.org/abs/2605.01823

【11】Zero-Shot, Safe and Time-Efficient UAV Navigation via Potential-Based Reward Shaping, Control Lyapunov and Barrier Functions
标题：通过基于潜力的回报整形、控制李雅普诺夫和障碍函数实现Zero-Shot、安全且省时的无人机导航
链接：https://arxiv.org/abs/2605.01787

【12】Class-Aware Adaptive Differential Privacy in Deep Learning for Sensor-Based Fall Detection
标题：基于传感器的跌倒检测的深度学习中的类感知自适应差异隐私
链接：https://arxiv.org/abs/2605.01679

【13】Adaptive Pluralistic Alignment: A pipeline for dynamic artificial democracy
标题：自适应多元化联盟：动态人工民主的管道
链接：https://arxiv.org/abs/2605.01642

【14】The Banach-Butterfly Invariant: Influence-Adaptive Walsh Geometry for Ternary Polynomial Threshold Functions
标题：Banach-Butterfly不变量：三元多项阈值函数的影响自适应Walsh几何
链接：https://arxiv.org/abs/2605.01637

【15】Chebyshev-Augmented One-Shot Transfer Learning for PINNs on Nonlinear Differential Equations
标题：非线性方程PINN的Chebyshev增强单次传输学习
链接：https://arxiv.org/abs/2605.01634

【16】GEODE: Angle-Adaptive OOD Detection with Universal Scorer Compatibility
标题：GEODE：具有通用评分器兼容性的角度自适应OOD检测
链接 ：https://arxiv.org/abs/2605.01063

【17】GAZE: Grounded Agentic Zero-shot Evaluation with Viewer-Level Tools and Literature Retrieval on Rare Brain MRI
标题：GAZE：使用观察者级工具和文献检索对罕见脑MRI进行的接地磁共振零激发评价
链接：https://arxiv.org/abs/2605.00876

【18】Adaptive Alarm Threshold Prediction in 4G Mobile Networks: A Percentile-Guided Deep Learning Framework with Interpretable Outputs
标题：4G移动网络中的自适应警报预测阈值：具有可解释输出的百分位引导深度学习框架
链接：https://arxiv.org/abs/2605.00838

【19】Active multiple matrix completion with adaptive confidence sets
标题：具有自适应置信集的主动多重矩阵完成
链接：https://arxiv.org/abs/2605.02458

【20】Adaptive Estimation and Inference in Semi-parametric Heterogeneous Clustered Multitask Learning via Neyman Orthogonality
标题：半参数异质离散多任务学习中的自适应估计与推理
链接：https://arxiv.org/abs/2605.01907

【21】Transfer Learning for Tonal Noise Prediction in VRF Units Using Thermodynamic and Vibration Signals
标题：使用热力学和振动信号进行VRF单元音调噪音预测的传递学习
链接：https://arxiv.org/abs/2605.00895

【22】An Adaptive Spatiotemporal Clustering Framework for 3D Ocean Subsurface Temperature Reconstruction
标题：用于3D海底温度重建的自适应时空聚集框架
链接：https://arxiv.org/abs/2605.00860

【23】Foundation Model Guided Dual-Branch Co-Adaptation for Source-Free EEG Decoding
标题：基础模型引导的双分支协同适应无源脑电解码
链接：https://arxiv.org/abs/2605.00857

强化学习(15篇)

【1】Federated Reinforcement Learning for Efficient Mobile Crowdsensing under Incomplete Information
标题：不完全信息下高效移动人群感知的联邦强化学习
链接：https://arxiv.org/abs/2605.02705

【2】Recurrent Deep Reinforcement Learning for Chemotherapy Control under Partial Observability
标题：部分可观察性下化疗控制的循环深度强化学习
链接：https://arxiv.org/abs/2605.02552

【3】Beyond Specialization: Robust Reinforcement Learning Navigation via Procedural Map Generators
标题：超越专业化：通过程序地图生成器进行稳健的强化学习导航
链接：https://arxiv.org/abs/2605.02528

【4】Binary Rewards and Reinforcement Learning: Fundamental Challenges
标题：二元奖励和强化学习：根本挑战
链接：https://arxiv.org/abs/2605.02375

【5】A Meta Reinforcement Learning Approach to Goals-Based Wealth Management
标题：基于目标的财富管理的Meta强化学习方法
链接：https://arxiv.org/abs/2605.02300

【6】Experience Constrained Hierarchical Federated Reinforcement Learning for Large-scale UAV Teams in Hazardous Environments
标题：危险环境中大型无人机团队的经验约束分层联邦强化学习
链接：https://arxiv.org/abs/2605.02165

【7】Combining Trained Models in Reinforcement Learning
标题：在强化学习中结合训练模型
链接：https://arxiv.org/abs/2605.02159

【8】Coopetition-Gym v1: A Formally Grounded Platform for Mixed-Motive Multi-Agent Reinforcement Learning under Strategic Coopetition
标题：Cooperation-Gym v1：战略合作竞争下混合动机多智能体强化学习的正式平台
链接：https://arxiv.org/abs/2605.02063

【9】MAGIC: Multi-Step Advantage-Gated Causal Influence for Multi-agent Reinforcement Learning
标题：MAGIC：多智能体强化学习的多步骤阈值门控因果影响
链接：https://arxiv.org/abs/2605.01805

【10】Hybrid Quantum Reinforcement Learning with QAOA for Improved Vehicle Routing Optimization
标题：基于QAOA的混合量子强化学习车辆路径优化
链接：https://arxiv.org/abs/2605.01574

【11】Feedback-Normalized Developer Memory for Reinforcement-Learning Coding Agents: A Safety-Gated MCP Architecture
标题：用于增强学习编码代理的反馈规范化开发人员记忆：安全门控LCP架构
链接：https://arxiv.org/abs/2605.01567

【12】Protein-Conditioned Multi-Objective Reinforcement Learning for Full-Length mRNA Design
标题：用于全长mRNA设计的蛋白质条件多目标强化学习
链接：https://arxiv.org/abs/2605.01513

【13】Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression
标题：通过深度不平衡回归的强化学习将分布意识注入MLLM
链接：https://arxiv.org/abs/2605.01402

【14】Separation Assurance between Heterogeneous Fleets of Small Unmanned Aerial Systems via Multi-Agent Reinforcement Learning
标题：通过多智能体强化学习实现小型无人机系统异类机队之间的隔离保证
链接：https://arxiv.org/abs/2605.01041

【15】Middle-mile logistics through the lens of goal-conditioned reinforcement learning
标题：通过目标条件强化学习的视角实现中英里物流
链接：https://arxiv.org/abs/2605.02461

符号|符号学习(1篇)

【1】Deep Variational Inference Symbolic Regression
标题：深度变分推理符号回归
链接：https://arxiv.org/abs/2605.01067

分层学习(1篇)

【1】Hierarchical Federated Learning for Networked AI: From Communication Saving to Architecture-Aware Design
标题：网络人工智能的分层联邦学习：从通信节省到架构感知设计
链接：https://arxiv.org/abs/2605.00931

医学相关(5篇)

【1】M\textsuperscript{4}Fuse: Lightweight State-Space MoE with a Cross-Scale Gating Bridge for Brain Tumor Segmentation
标题：M extsuperScript{4}：具有跨规模门控桥的轻量级状态空间MoE用于脑肿瘤分割
链接：https://arxiv.org/abs/2605.02444

【2】GeoSAE: Geometric Prior-Guided Layer-Wise Sparse Autoencoder Annotation of Brain MRI Foundation Models
标题：GeoAE：脑MRI基础模型的几何优先引导分层稀疏自动编码器注释
链接：https://arxiv.org/abs/2605.01829

【3】ECG-biometrics-bench: A Unified Framework for Reproducible Benchmarking of ECG Biometrics
标题：心电生物识别台：心电图生物识别可重复基准的统一框架
链接：https://arxiv.org/abs/2605.01548

【4】Linking spatial biology and clinical histology via Haiku
标题：通过Haiku将空间生物学和临床组织学联系起来
链接：https://arxiv.org/abs/2605.00925

【5】A Hybrid Windkessel-Neural Approach for Improved Noninvasive Blood Pressure Monitoring
标题：改进无创血压监测的Windkessel-神经混合方法
链接：https://arxiv.org/abs/2605.00858

蒸馏|知识提取(6篇)

【1】Standing on the Shoulders of Giants: Stabilized Knowledge Distillation for Cross--Language Code Clone Detection
标题：站在巨人的肩膀上：跨语言代码克隆检测的稳定知识提炼
链接：https://arxiv.org/abs/2605.02860

【2】TIJERE: A Novel Threat Intelligence Joint Extraction Model Based on Analyst Expert Knowledge
标题：TIJERE：一种基于分析师专家知识的新型威胁情报联合提取模型
链接：https://arxiv.org/abs/2605.02041

【3】LIE: LiDAR-only HD Map Construction with Intensity Enhancement via Online Knowledge Distillation
标题：LIE：通过在线知识提炼增强强度的纯LiDART高清地图构建
链接：https://arxiv.org/abs/2605.01478

【4】MAD-OPD: Breaking the Ceiling in On-Policy Distillation via Multi-Agent Debate
标题：MAD-OPD：通过多代理辩论打破政策蒸馏的上限
链接：https://arxiv.org/abs/2605.01347

【5】Benchmarking local Hebbian learning rules for memory storage and prototype extraction
标题：对内存存储和原型提取的本地Hebbian学习规则进行基准测试
链接：https://arxiv.org/abs/2605.01074

【6】Selective Correlation Based Knowledge Distillation for Ground Reaction Force Estimation
标题：基于选择相关的地面反作用力估计知识提取
链接：https://arxiv.org/abs/2605.00888

超分辨率|去噪|去模糊|去雾(1篇)

【1】Denoising data using convex relaxations
标题：使用凸松弛去噪数据
链接：https://arxiv.org/abs/2605.02327

自动驾驶|车辆|车道检测等(1篇)

【1】From Packets to Patterns: Interpreting Encrypted Network Traffic as Longitudinal Behavioral Signals
标题：从数据包到模式：将加密网络流量解释为纵向行为信号
链接：https://arxiv.org/abs/2605.01616

点云|SLAM|雷达|激光|深度RGBD相关(2篇)

【1】SplAttN: Bridging 2D and 3D with Gaussian Soft Splatting and Attention for Point Cloud Completion
标题：SplAttN：利用高斯软飞溅连接2D和3D并关注点云完成
链接：https://arxiv.org/abs/2605.01466

【2】Machine Learning Enhanced Laser Spectroscopy for Multi-Species Gas Detection in Complex and Harsh Environments
标题：机器学习增强激光光谱在复杂和恶劣环境中用于多物种气体检测
链接：https://arxiv.org/abs/2605.01306

联邦学习|隐私保护|加密(4篇)

【1】FedPLT: Scalable, Resource-Efficient, and Heterogeneity-Aware Federated Learning via Partial Layer Training
标题：FedPLT：通过部分层训练可扩展、资源高效且具有异类意识的联邦学习
链接：https://arxiv.org/abs/2605.02337

【2】Personalized Federated Learning for Gradient Alignment
标题：用于梯度对齐的个性化联邦学习
链接：https://arxiv.org/abs/2605.02143

【3】FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
标题：FedQueue：跨设施高性能计算训练的客户感知联合学习
链接：https://arxiv.org/abs/2605.02125

【4】Toward Resilient 5G Networks: Comparative Analysis of Federated and Centralized Learning for RF Jamming Detection
标题：迈向弹性5G网络：用于RF干扰检测的联邦学习和集中式学习的比较分析
链接：https://arxiv.org/abs/2605.01705

推理|分析|理解|解释(19篇)

【1】Enhancing RL Generalizability in Robotics through SHAP Analysis of Algorithms and Hyperparameters
标题：通过算法和超参数的SHAP分析增强机器人中RL的泛化能力
链接：https://arxiv.org/abs/2605.02867

【2】Visual Latents Know More Than They Say: Unsilencing Latent Reasoning in MLLMs
标题：视觉潜意识知道的比他们说的更多：在MLLM中消除潜在推理
链接：https://arxiv.org/abs/2605.02735

【3】When Attention Collapses: Residual Evidence Modeling for Compositional Inference
标题：当注意力崩溃时：成分推理的剩余证据建模
链接：https://arxiv.org/abs/2605.02323

【4】Metric Unreliability in Multimodal Machine Unlearning: A Systematic Analysis and Principled Unified Score
标题：多模式机器去学习中的指标不可靠性：系统分析和原则统一评分
链接：https://arxiv.org/abs/2605.02206

【5】Weight Clipping for Robust Conformal Inference under Unbounded Covariate Shifts
标题：无界协变量位移下鲁棒共形推理的权重剪裁
链接：https://arxiv.org/abs/2605.02072

【6】Robust and Explainable Divide-and-Conquer Learning for Intrusion Detection
标题：用于入侵检测的稳健且可解释的分治学习
链接：https://arxiv.org/abs/2605.02015

【7】ViM-Q: Scalable Algorithm-Hardware Co-Design for Vision Mamba Model Inference on FPGA
标题：ViM-Q：基于现场可扩展的视觉Mamba模型推理框架-硬件协同设计
链接：https://arxiv.org/abs/2605.01935

【8】Stochastic Sparse Attention for Memory-Bound Inference
标题：记忆限制推理的随机稀疏注意力
链接：https://arxiv.org/abs/2605.01910

【9】How Label Imbalance Shapes Geometry: A General Spectral Analysis of Multi-Label Neural Collapse
标题：标签失衡如何影响几何：多标签神经崩溃的一般谱分析
链接：https://arxiv.org/abs/2605.01897

【10】Chart-FR1: Visual Focus-Driven Fine-Grained Reasoning on Dense Charts
标题：Chart-FR 1：密集图表上的视觉焦点驱动细粒度推理
链接：https://arxiv.org/abs/2605.01882

【11】Segment-Aligned Policy Optimization for Multi-Modal Reasoning
标题：多模式推理的分段一致策略优化
链接：https://arxiv.org/abs/2605.01327

【12】CombinationTS: A Modular Framework for Understanding Time-Series Forecasting Models
标题：CombinationTS：理解时间序列预测模型的模块化框架
链接：https://arxiv.org/abs/2605.01231

【13】Attention Sinks in Massively Multilingual Neural Machine Translation:Discovery, Analysis, and Mitigation
标题：大规模多语言神经机器翻译中的注意力下沉：发现、分析和缓解
链接：https://arxiv.org/abs/2605.01229

【14】When Less is Enough: Efficient Inference via Collaborative Reasoning
标题：当更少就足够：通过协作推理进行高效推理
链接：https://arxiv.org/abs/2605.01111

【15】Finite-Sample Analysis of Elimination in Active Hypothesis Testing
标题：主动假设检验中消除的样本分析
链接：https://arxiv.org/abs/2605.01039

【16】Understanding Emergent Misalignment via Feature Superposition Geometry
标题：通过特征叠加几何理解紧急失准
链接：https://arxiv.org/abs/2605.00842

【17】Agentopic: A Generative AI Agent Workflow for Explainable Topic Modeling
标题：可解释主题建模的生成式AI Agent工作流
链接：https://arxiv.org/abs/2605.00833

【18】Understanding the Performance Plateau in Text-to-Video Retrieval: A Comprehensive Empirical and Linguistic Analysis
标题：了解文本到视频检索的性能平台：全面的实证和语言分析
链接：https://arxiv.org/abs/2605.00826

【19】Minimum Specification Perturbation: Robustness as Distance-to-Falsification in Causal Inference
标题：最低规格扰动：因果推理中与证伪距离的稳健性
链接：https://arxiv.org/abs/2605.01579

检测相关(10篇)

【1】Beating the Style Detector: Three Hours of Agentic Research on the AI-Text Arms Race
标题：击败风格检测器：对人工智能文本军备竞赛的三小时详细研究
链接：https://arxiv.org/abs/2605.02620

【2】Evaluating Tabular Representation Learning for Network Intrusion Detection
标题：网络入侵检测的表格表示学习评估
链接：https://arxiv.org/abs/2605.02519

【3】Open-access model for detecting openly dumped dispersed municipal solid waste from crowdsourced UAV imagery in Sub-Saharan Africa
标题：从撒哈拉以南非洲众包无人机图像中检测公开倾倒的分散城市固体废物的开放获取模型
链接：https://arxiv.org/abs/2605.02316

【4】CLaC at SemEval-2026 Task 6: Response Clarity Detection in Political Discourse
标题：CLaC参加SemEval-2026任务6：政治话语中的反应清晰度检测
链接：https://arxiv.org/abs/2605.02170

【5】Retrieval with Multiple Query Vectors through Anomalous Pattern Detection
标题：通过异常模式检测的多查询载体检索
链接：https://arxiv.org/abs/2605.01965

【6】From Flat Facts to Sharp Hallucinations: Detecting Stubborn Errors via Gradient Sensitivity
标题：从平淡的事实到尖锐的幻觉：通过梯度灵敏度检测顽固错误
链接：https://arxiv.org/abs/2605.00939

【7】EventADL: Open-Box Anomaly Detection and Localization Framework for Events in Cloud-Based Service Systems
标题：EventADL：基于云的服务系统中事件的开箱异常检测和定位框架
链接：https://arxiv.org/abs/2605.00936

【8】Latent Space Probing for Adult Content Detection in Video Generative Models
标题：视频生成模型中成人内容检测的潜在空间探测
链接：https://arxiv.org/abs/2605.00874

【9】Pi-Change: A Prior-Informed Multiple Change Point Detection Algorithm
标题：Pi-Change：一种事先知情的多变点检测算法
链接：https://arxiv.org/abs/2605.01003

【10】An Algorithm for On-Sensor Agnostic Detection of Changes in Human Activity for Ultra-Low-Power Applications
标题：超低功耗应用中人类活动变化的传感器不可知检测算法
链接：https://arxiv.org/abs/2605.00870

分类|识别(7篇)

【1】VideoNet: A Large-Scale Dataset for Domain-Specific Action Recognition
标题：VideoNet：用于特定领域动作识别的大规模数据集
链接：https://arxiv.org/abs/2605.02834

【2】HARMES: A Multi-Modal Dataset for Wearable Human Activity Recognition with Motion, Environmental Sensing and Sound
标题：HARMES：用于可穿戴人类活动识别的多模式数据集，具有运动、环境传感和声音
链接：https://arxiv.org/abs/2605.02596

【3】Physics-Informed Neural Learning for State Reconstruction and Parameter Identification in Coupled Greenhouse Climate Dynamics
标题：基于物理信息的神经网络在温室气候动力学中的状态重构和参数识别
链接：https://arxiv.org/abs/2605.02524

【4】Rethinking Multi-Label Node Classification: Do Tuned Classic GNNs Suffice?
标题：重新思考多标签节点分类：调整后的经典GNN足够吗？
链接：https://arxiv.org/abs/2605.01403

【5】Networked Information Aggregation for Binary Classification
标题：二进制分类的网络信息聚合
链接：https://arxiv.org/abs/2605.01082

【6】Skeleton-Based Posture Classification to Promote Safer Walker-Assisted Gait in Older Adults
标题：基于普林斯顿的姿势分类以促进老年人更安全的步行辅助步态
链接：https://arxiv.org/abs/2605.00890

【7】Deep Learning for Multi-Antenna Modulation Recognition of Radio Signals
标题：用于无线电信号多天线调制识别的深度学习
链接：https://arxiv.org/abs/2605.00849

表征(10篇)

【1】Representation learning from OCT images
标题：来自光学断层扫描图像的表示学习
链接：https://arxiv.org/abs/2605.02589

【2】Statistical Consistency and Generalization of Contrastive Representation Learning
标题：对比表示学习的统计一致性和推广
链接：https://arxiv.org/abs/2605.02116

【3】Complex Diffusion Maps with $ω$-Parameterized Kernels Revealing Inherent Harmonic Representations
标题：具有$ω$参数核的复扩散映射的内调和表示
链接：https://arxiv.org/abs/2605.01691

【4】PRIME: Protein Representation via Physics-Informed Multiscale Equivariant Hierarchies
标题：PRIME：通过物理信息多尺度等变分层结构的蛋白质表示
链接：https://arxiv.org/abs/2605.01625

【5】A framework for analyzing concept representations in neural models
标题：分析神经模型中概念表示的框架
链接：https://arxiv.org/abs/2605.01381

【6】Continuous Temporal Representations of Event-Based Signals via Interference-Based Wave Modeling
标题：通过基于干扰的波建模实现基于事件的信号的连续时间表示
链接：https://arxiv.org/abs/2605.01270

【7】Diffusion Operator Geometry of Feedforward Representations
标题：前向表示的扩散运算符几何
链接：https://arxiv.org/abs/2605.01107

【8】Differentiable Multiphysics Co-Optimization via Implicit Neural Representations: A Transient Hamburger-Cooking Benchmark
标题：通过隐式神经表示的可区分多物理场协同优化：瞬时汉堡烹饪基准
链接：https://arxiv.org/abs/2605.01040

【9】Physiology-Aware Masked Cross-Modal Reconstruction for Biosignal Representation Learning
标题：用于生物信号表示学习的生理感知掩蔽跨模式重建
链接：https://arxiv.org/abs/2605.00973

【10】Benchmarking Wireless Representations: High-Dimensional vs. Compressed Embeddings for Efficiency and Robustness
标题：无线表示基准测试：多维与压缩嵌入以提高效率和鲁棒性
链接：https://arxiv.org/abs/2605.02009

3D|3D重建等相关(1篇)

【1】GETA-3DGS: Automatic Joint Structured Pruning and Quantization for 3D Gaussian Splatting
标题：GETA-3DGS：用于3D高斯溅射的自动联合结构化修剪和量化
链接：https://arxiv.org/abs/2605.02086

编码器(4篇)

【1】The Model Knows, the Decoder Finds: Future Value Guided Particle Power Sampling
标题：模型知道，解码器发现：未来价值引导的粒子功率采样
链接：https://arxiv.org/abs/2605.02427

【2】HELIX: Hybrid Encoding with Learnable Identity and Cross-dimensional Synthesis for Time Series Imputation
标题：HEIX：具有可学习身份和跨维合成的混合编码用于时间序列插补
链接：https://arxiv.org/abs/2605.02278

【3】SURGE: SuperBatch Unified Resource-efficient GPU Encoding for Heterogeneous Partitioned Data
标题：SURGE：针对异类分区数据的SuperBlock统一资源高效的图形处理器编码
链接：https://arxiv.org/abs/2605.01060

【4】How Well Can We Decode Vowels from Auditory EEG -- A Rigorous Cross-Subject Benchmark with Honest Assessment
标题：从听觉脑电中解码元音的效果如何--具有诚实评估的严格跨学科基准
链接：https://arxiv.org/abs/2605.00865

优化|敛散性(17篇)

【1】Inducing Permutation Invariant Priors in Bayesian Optimization for Carbon Capture and Storage Applications
标题：碳捕获和储存应用的Bayesian优化中引入排列不变先验
链接：https://arxiv.org/abs/2605.02409

【2】A Near-optimal SQ Lower Bound for Smoothed Agnostic Learning of Boolean Halfspaces
标题：布尔半空间平滑不可知学习的接近最优SJ下界
链接：https://arxiv.org/abs/2605.02350

【3】ANO: A Principled Approach to Robust Policy Optimization
标题：ANO：稳健政策优化的原则方法
链接：https://arxiv.org/abs/2605.02320

【4】On the Optimal Sample Complexity of Offline Multi-Armed Bandits with KL Regularization
标题：KL正规化离线多臂盗贼的最佳样本复杂性
链接：https://arxiv.org/abs/2605.02141

【5】LUMINA: A Grid Foundation Model for Benchmarking AC Optimal Power Flow Surrogate Learning
标题：LUMINA：交流最优潮流替代学习基准的电网基础模型
链接：https://arxiv.org/abs/2605.02133

【6】Bringing Order to Asynchronous SGD: Towards Optimality under Data-Dependent Delays with Momentum
标题：为非同步新元带来秩序：在数据相关延迟下实现动量最优
链接：https://arxiv.org/abs/2605.02043

【7】Towards Systematic Generalization for Power Grid Optimization Problems
标题：电网优化问题的系统概括
链接：https://arxiv.org/abs/2605.02026

【8】Training Non-Differentiable Networks via Optimal Transport
标题：通过最佳传输训练不可区分的网络
链接：https://arxiv.org/abs/2605.01928

【9】Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise
标题：利用数据对称性选择标签噪音下的最佳训练数据子集
链接：https://arxiv.org/abs/2605.01874

【10】Value Functions for Temporal Logic: Optimal Policies and Safety Filters
标题：时态逻辑的值函数：最佳策略和安全过滤器
链接：https://arxiv.org/abs/2605.01051

【11】Fast Log-Domain Sinkhorn Optimal Transport with Warp-Level GPU Reductions
标题：快速日志域Sinkhorn优化传输，通过扭曲级图形处理器缩减
链接：https://arxiv.org/abs/2605.00837

【12】Polynomial-Time Optimal Group Selection via the Double-Commutator Eigenvalue Problem
标题：基于双交换器特征值问题的多项时间最优群选择
链接：https://arxiv.org/abs/2605.00834

【13】Black-box optimization of noisy functions with unknown smoothness
标题：平滑度未知的有噪函数的黑匣子优化
链接：https://arxiv.org/abs/2605.02462

【14】Foundations of Riemannian Geometry for Riemannian Optimization: A Monograph with Detailed Derivations
标题：Riemannian优化的Riemannian几何基础：一本具有详细推导的专著
链接：https://arxiv.org/abs/2605.02279

【15】A Parameter-Free First-Order Algorithm for Non-Convex Optimization with $\tilde{\mkern1mu O}(ε^{-5/3})$ Global Rate
链接：https://arxiv.org/abs/2605.02127

【16】Data-Driven, Geometry-Aware Optimal-Transport Calibration of Flavor Tagger
标题：基于数据驱动的几何感知最优传输的风味标签标定
链接：https://arxiv.org/abs/2605.01363

【17】An Efficient Spatial Branch-and-Bound Algorithm for Global Optimization of Gaussian Process Posterior Mean Functions
标题：高斯过程后验均值函数全局优化的高效空间分支定界算法
链接：https://arxiv.org/abs/2605.00855

预测|估计(16篇)

【1】MSMixer: Learned Multi-Scale Temporal Mixing with Complementary Linear Shortcut for Long-Term Time Series Forecasting
标题：MSMixer：用于长期时间序列预测的学习多尺度时间混合和互补线性插值
链接：https://arxiv.org/abs/2605.02689

【2】CARD: Coarse-to-fine Autoregressive Modeling with Radix-based Decomposition for Transferable Free Energy Estimation
标题：CARD：用于可传递自由能估计的粗到细自回归建模
链接：https://arxiv.org/abs/2605.02657

【3】Selective Prediction from Agreement: A Lipschitz-Consistent Version Space Approach
标题：根据一致性选择性预测：利普希茨一致的版本空间方法
链接：https://arxiv.org/abs/2605.02611

【4】A Novel Preprocessing-Driven Approach to Remaining Useful Life (RUL) Prediction Using Temporal Convolutional Networks (TCN)
标题：使用时间卷积网络（TCN）预测剩余使用寿命（RUL）的新型预处理驱动方法
链接：https://arxiv.org/abs/2605.02507

【5】Predicting Post Virality with Temporal Cross-Attention over Trend Signals
标题：通过对趋势信号的时间交叉注意预测病毒后
链接：https://arxiv.org/abs/2605.02358

【6】H3: A Healthcare Three-Hop Index for Physician Referral Network Prediction
标题：H3：医生推荐网络预测的医疗保健三跳指数
链接：https://arxiv.org/abs/2605.02150

【7】PepSpecBench: A Unified Evaluation Benchmark for Peptide Tandem Mass Spectrometry Prediction
标题：PepSpecBench：肽串联MS预测的统一评估基准
链接：https://arxiv.org/abs/2605.01945

【8】SwiftChannel: Algorithm-Hardware Co-Design for Deep Learning-Based 5G Channel Estimation
标题：SwiftChannel：基于深度学习的5G频道估计的框架-硬件协同设计
链接：https://arxiv.org/abs/2605.01931

【9】Robust Conditional Conformal Prediction via Branched Normalizing Flow
标题：基于分支正规化流的鲁棒条件保形预测
链接：https://arxiv.org/abs/2605.01868

【10】Geospatial foundation-model embeddings improve population estimation unevenly across space and scale
标题：地理空间基础模型嵌入改善了空间和规模上不均匀的人口估计
链接：https://arxiv.org/abs/2605.01650

【11】Decision-Focused Learning via Tangent-Space Projection of Prediction Error
标题：基于预测误差切空间投影的决策学习
链接：https://arxiv.org/abs/2605.01361

【12】Local Hessian Spectral Filtering for Robust Intrinsic Dimension Estimation
标题：用于鲁棒内维估计的局部黑森谱过滤
链接：https://arxiv.org/abs/2605.01221

【13】Online Generalised Predictive Coding
标题：在线广义预测编码
链接：https://arxiv.org/abs/2605.02675

【14】Stable Localized Conformal Prediction via Transduction
标题：通过转换进行稳定的局部保形预测
链接：https://arxiv.org/abs/2605.01452

【15】A Deep Learning Model for Battery State Prediction towards Intelligent Energy Management
标题：智能能源管理的电池状态预测深度学习模型
链接：https://arxiv.org/abs/2605.00898

【16】Earth System Foundation Model (ESFM): A unified framework for heterogeneous data integration and forecasting
标题：地球系统基础模型（ESFM）：异构数据集成和预测的统一框架
链接：https://arxiv.org/abs/2605.00850

其他神经网络|深度学习|模型|建模(59篇)

【1】Learning Equivariant Neural-Augmented Object Dynamics From Few Interactions
标题：从少量交互学习等变神经增强对象动力学
链接：https://arxiv.org/abs/2605.02699

【2】Spectral Model eXplainer: a chemically-grounded explainability framework for spectral-based machine learning models
标题：Spectral Model eXplayer：基于光谱的机器学习模型的基于化学的解释性框架
链接：https://arxiv.org/abs/2605.02684

【3】MPCS: Neuroplastic Continual Learning via Multi-Component Plasticity and Topology-Aware EWC
标题：MPCS：通过多成分可塑性和具有布局意识的EWC进行神经可塑性持续学习
链接：https://arxiv.org/abs/2605.02509

【4】Spatial-Temporal Learning-Based Distributed Routing for Dynamic LEO Satellite Networks
标题：动态LEO卫星网络基于时空学习的分布式路由
链接：https://arxiv.org/abs/2605.02413

【5】Differentiable Kernel Ridge Regression for Deep Learning Pipelines
标题：深度学习管道的可区分核岭回归
链接：https://arxiv.org/abs/2605.02313

【6】Variational Matrix-Learning Fourier Networks for Parametric Multiphysics Surrogates
标题：参数多物理场代理的变分矩阵学习傅里叶网络
链接：https://arxiv.org/abs/2605.02280

【7】MultiSense-Pneumo: A Multimodal Learning Framework for Pneumonia Screening in Resource-Constrained Settings
标题：MultiSense-Bio：资源受限环境下肺炎筛查的多模式学习框架
链接：https://arxiv.org/abs/2605.02207

【8】RAFNet: Region-Aware Fusion Network for Pansharpening
标题：RAFNet：用于Panshering的区域感知融合网络
链接：https://arxiv.org/abs/2605.02184

【9】Ultrasound Vision-Language Alignment via Contrastive Learning
标题：通过对比学习实现超声视觉与语言对齐
链接：https://arxiv.org/abs/2605.02126

【10】Geometric and Spectral Alignment for Deep Neural Network II
标题：深度神经网络的几何和谱对齐II
链接：https://arxiv.org/abs/2605.02111

【11】Geometric and Spectral Alignment for Deep Neural Network I
标题：深度神经网络的几何和谱对齐I
链接：https://arxiv.org/abs/2605.02108

【12】Bridging the Gap Between Average and Discounted TD Learning
标题：缩小平均和折扣TD学习之间的差距
链接：https://arxiv.org/abs/2605.02103

【13】Stochastic Modeling of Human-Machine Authentication Channels under Partial Information Leakage
标题：部分信息泄露下人机认证通道的随机建模
链接：https://arxiv.org/abs/2605.02102

【14】NeuroViz: Real-time Interactive Visualization of Forward and Backward Passes in Neural Network Training
标题：NeuViz：神经网络训练中正向和反向传递的实时交互可视化
链接：https://arxiv.org/abs/2605.02044

【15】How Can One Choose the Best CAM-Based Explainability Method for a CNN Model?
标题：如何为CNN模型选择最佳的基于CAM的解释方法？
链接：https://arxiv.org/abs/2605.02007

【16】RamanBench: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy
标题：RamanBench：拉曼光谱机器学习的大规模基准
链接：https://arxiv.org/abs/2605.02003

【17】Deep learning-based pavement performance modeling using multiple distress indicators and road work history
标题：使用多个遇险指标和道路施工历史进行基于深度学习的路面性能建模
链接：https://arxiv.org/abs/2605.01914

【18】Learning Koopman operators for coupled systems via information on governing equations of subsystems
标题：通过子系统控制方程的信息学习耦合系统的Koopman运算符
链接：https://arxiv.org/abs/2605.01835

【19】CoAction: Cross-task Correlation-aware Pareto Set Learning
标题：协同行动：跨任务相关性感知帕累托集学习
链接：https://arxiv.org/abs/2605.01712

【20】Floating-Point Networks with Automatic Differentiation Can Represent Almost All Floating-Point Functions and Their Gradients
标题：具有自动求导的浮点网络可以代表几乎所有浮点函数及其派生词
链接：https://arxiv.org/abs/2605.01702

【21】Towards Efficient and Expressive Offline RL via Flow-Anchored Noise-conditioned Q-Learning
标题：通过流锚定噪音条件Q学习实现高效且富有表现力的离线RL
链接：https://arxiv.org/abs/2605.01663

【22】The Case for ESM3 as a General-Purpose AI Model with Systemic Risk Under the EU AI Act
标题：根据欧盟人工智能法案，ESM 3作为具有系统性风险的通用人工智能模型的理由
链接：https://arxiv.org/abs/2605.01611

【23】Model Merging: Foundations and Algorithms
标题：模型合并：基础和算法
链接：https://arxiv.org/abs/2605.01580

【24】Barriers to Counterfactual Credit Attribution for Autoregressive Models
标题：自回归模型反事实信用归因的障碍
链接：https://arxiv.org/abs/2605.01425

【25】Quantifying Multimodal Capabilities: Formal Generalization Guarantees in Pairwise Metric Learning
标题：量化多模式能力：成对度量学习中的形式概括保证
链接：https://arxiv.org/abs/2605.01424

【26】Sequential Learning and Catastrophic Forgetting in Differentiable Resistor Networks
标题：可微电阻网络中的顺序学习和灾难性遗忘
链接：https://arxiv.org/abs/2605.01383

【27】Toward a foundational thermal model for residential buildings
标题：建立住宅建筑的基础热力模型
链接：https://arxiv.org/abs/2605.01364

【28】Robust Parameter Learning for Uncertain MDPs
标题：不确定MDP的鲁棒参数学习
链接：https://arxiv.org/abs/2605.01339

【29】Rethinking Model Selection in VLM Through the Lens of Gromov-Wasserstein Distance
标题：从Gromov-Wasserstein距离的角度重新思考VLM中的模型选择
链接：https://arxiv.org/abs/2605.01325

【30】Autonomous Drift Learning in Data Streams: A Unified Perspective
标题：数据流中的自主漂移学习：统一视角
链接：https://arxiv.org/abs/2605.01295

【31】Congestion-Aware Dynamic Axonal Delay for Spiking Neural Networks
标题：尖峰神经网络的拥挤感知动态轴突延迟
链接：https://arxiv.org/abs/2605.01291

【32】A Theory of Saddle Escape in Deep Nonlinear Networks
标题：深度非线性网络中马鞍逃逸理论
链接 ：https://arxiv.org/abs/2605.01288

【33】S^3-R1: Learning to Retrieve and Answer Step-by-Step with Synthetic Data
标题：S#3-R1：学习使用合成数据逐步调试和回答
链接：https://arxiv.org/abs/2605.01248

【34】Focus and Dilution: The Multi-stage Learning Process of Attention
标题：专注与稀释：注意力的多阶段学习过程
链接：https://arxiv.org/abs/2605.01199

【35】A Theory of Generalization in Deep Learning
标题：深度学习中的概括理论
链接：https://arxiv.org/abs/2605.01172

【36】Forager: a lightweight testbed for continual learning with partial observability in RL
标题：Forager：RL中具有部分可观察性的连续学习轻量级测试平台
链接：https://arxiv.org/abs/2605.01131

【37】Machine Learning-Augmented Acceleration of Iterative Ptychographic Reconstruction
标题：迭代重叠重建的机器学习增强加速
链接：https://arxiv.org/abs/2605.01122

【38】Learning to Race in Minutes: Infoprop Dyna on the Mini Wheelbot
标题：几分钟内学习比赛：Mini Wheelbot上的Infoprop Dyna
链接：https://arxiv.org/abs/2605.01096

【39】Learning Discriminators for Resampling in the Ensemble Gaussian Mixture Filter through a Normalizing Flow Approach
标题：通过正规化流方法在积分高斯混合过滤器中重新分配的学习鉴别器
链接：https://arxiv.org/abs/2605.01089

【40】Learning in the Fisher Subspace: A Guided Initialization for LoRA Fine-Tuning
标题：Fisher子空间中的学习：LoRA微调的引导收件箱
链接：https://arxiv.org/abs/2605.01046

【41】Continual Learning of Feedback-based Molecular Communication
标题：基于反馈的分子通讯的持续学习
链接：https://arxiv.org/abs/2605.01020

【42】Interpretable experiential learning based on state history and global feedback
标题：基于州历史和全球反馈的可解释体验式学习
链接：https://arxiv.org/abs/2605.00940

【43】Watch Your Step: Information Injection in Diffusion Models via Shadow Timestep Embedding
标题：注意你的步骤：通过阴影时间步嵌入的扩散模型中的信息注入
链接：https://arxiv.org/abs/2605.00935

【44】TRIP-Evaluate: An Open Multimodal Benchmark for Evaluating Large Models in Transportation
标题：TRIP-Evaluate：用于评估交通运输大型模型的开放式多模式基准
链接：https://arxiv.org/abs/2605.00907

【45】On the explainability of max-plus neural networks
标题：最大加神经网络的可解释性
链接：https://arxiv.org/abs/2605.00889

【46】OceanPile: A Large-Scale Multimodal Ocean Corpus for Foundation Models
标题：OceanPill：用于基础模型的大规模多模式海洋数据库
链接：https://arxiv.org/abs/2605.00877

【47】2026 Roadmap on Artificial Intelligence and Machine Learning for Smart Manufacturing
标题：2026年智能制造人工智能和机器学习路线图
链接：https://arxiv.org/abs/2605.00839

【48】From Euler to Dormand-Prince: ODE Solvers for Flow Matching Generative Models
标题：从欧拉到多曼-普林斯：流匹配生成模型的ODE求解器
链接：https://arxiv.org/abs/2605.00836

【49】Synthetic Designed Experiments for Diagnosing Vision Model Failure
标题：诊断视觉模型故障的综合设计实验
链接：https://arxiv.org/abs/2605.00832

【50】Multi-fidelity surrogates for mechanics of composites: from co-kriging to multi-fidelity neural networks
标题：复合材料力学的多保真替代物：从协同克里金法到多保真神经网络
链接：https://arxiv.org/abs/2605.02871

【51】Universality in Deep Neural Networks: An approach via the Lindeberg exchange principle
标题：深度神经网络的普适性：基于Lindeberg交换原则的方法
链接：https://arxiv.org/abs/2605.02771

【52】ParaRNN: An Interpretable and Parallelizable Recurrent Neural Network for Time-Dependent Data
标题：ParaRNN：一个可解释且可并行化的时间相关数据的回归神经网络
链接：https://arxiv.org/abs/2605.02692

【53】MIRA: A Score for Conditional Distribution Accuracy and Model Comparison
标题：MIRA：条件分布准确性和模型比较的评分
链接：https://arxiv.org/abs/2605.02014

【54】Extrapolation in Statistical Learning with Extreme Value Theory
标题：用极小值理论进行统计学习的外推
链接：https://arxiv.org/abs/2605.01909

【55】Distributional Causal Mediation via Conditional Generative Modeling
标题：通过条件生成建模的分布因果调解
链接：https://arxiv.org/abs/2605.01765

【56】Missingness-aware Data Imputation via AI-powered Bayesian Generative Modeling
标题：通过人工智能支持的Bayesian生成建模实现缺失感知数据插补
链接：https://arxiv.org/abs/2605.01676