机器学习学术速递[6.5] na

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232篇

大模型相关(40篇)

【1】Self-Augmenting Retrieval for Diffusion Language Models
标题：扩散语言模型的自扩充检索
链接：https://arxiv.org/abs/2606.06474

【2】PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training
标题：PC层：用于改善LLM预训练的多项权重预处理
链接：https://arxiv.org/abs/2606.06470

【3】Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance
标题：学习忘记什么：通过习得代币级重要性来改善LLM忘记学习
链接：https://arxiv.org/abs/2606.06320

【4】Tangram: Unlocking Non-Uniform KV Cache for Efficient Multi-turn LLM Serving
标题：七巧板：解锁非均匀KV缓存以实现高效的多回合LLM服务
链接：https://arxiv.org/abs/2606.06302

【5】Generative Criticality in Large Language Model Temperature Scaling
标题：大型语言模型温度缩放中的生成临界性
链接：https://arxiv.org/abs/2606.06238

【6】Design a Reliable LLM-Integrated Interface for Mortality Forecasting
标题：设计可靠的LLM集成界面用于死亡率预测
链接：https://arxiv.org/abs/2606.06235

【7】Learning to Route LLMs from Implicit Cost-Performance Preferences via Meta-Learning
标题：学习通过元学习从隐性成本-绩效偏好中寻找LLM
链接：https://arxiv.org/abs/2606.06178

【8】IR3DE: A Linear Router for Large Language Models
标题：IR 3DE：大型语言模型的线性路由器
链接：https://arxiv.org/abs/2606.06098

【9】RedditPersona: A Modular Framework for Community-Conditioned LLM Adaptation from Reddit
标题：RedditPersona：Reddit的社区条件LLM改编模块化框架
链接：https://arxiv.org/abs/2606.06027

【10】LLM Explainability with Counterfactual Chains and Causal Graphs
标题：用反事实链和因果图解释LLM
链接：https://arxiv.org/abs/2606.05972

【11】Measuring the sensitivity of LLM-based structured extraction to prompt, model, and schema choices in clinical discharge summaries
标题：衡量基于LLM的结构化提取对临床出院摘要中提示、模型和模式选择的敏感性
链接：https://arxiv.org/abs/2606.05970

【12】Retrospective Harness Optimization: Improving LLM Agents via Self-Preference over Trajectory Rollouts
标题：回顾性收件箱优化：通过自我偏好而不是轨迹卷展来改进LLM代理
链接：https://arxiv.org/abs/2606.05922

【13】When Denser Credit Is Not Enough: Evidence-Calibrated Policy Optimization for Long-Horizon LLM Agent Training
标题：当更高的信用还不够时：长期LLM代理训练的循证校准政策优化
链接：https://arxiv.org/abs/2606.05885

【14】CaliDist: Calibrating Large Language Models via Behavioral Robustness to Distraction
标题：CalibDist：通过行为稳健性来校准大型语言模型
链接：https://arxiv.org/abs/2606.05799

【15】CollabBench: Benchmarking and Unleashing Collaborative Ability of LLMs with Diverse Players via Proactive Engagement
标题：CollabBench：通过积极主动的参与来基准测试和释放LLM与多元化参与者的协作能力
链接：https://arxiv.org/abs/2606.05793

【16】Can LLMs Write Correct TLA+ Specifications? Evaluating Natural-Language-to-TLA+ Generation
标题：LLM可以编写正确的TLA+规范吗？评估自然语言到TLA+生成
链接：https://arxiv.org/abs/2606.05792

【17】Domain-Adapted Small Language Models with Hybrid Post-Processing: Achieving Cost-Efficient, Low-Latency Multi-Label Structured Prediction via LoRA Fine-Tuning on Scarce Data
标题：具有混合后处理的域自适应小型语言模型：通过对稀缺数据进行LoRA微调实现成本效益、低延迟的多标签结构化预测
链接：https://arxiv.org/abs/2606.05781

【18】DRIFT: A Residual Flow Adapter for Decoding Continuous Outputs in Vision-Language Models
标题：DRFT：用于解码视觉语言模型中连续输出的剩余流量适配器
链接：https://arxiv.org/abs/2606.05758

【19】Let It Be Simple: One-Step Action Generation for Vision-Language-Action Models
标题：让它变得简单：视觉-语言-动作模型的一步动作生成
链接：https://arxiv.org/abs/2606.05737

【20】Automated Proving of Shannon-Type Entropy Inequalities via Fine-Tuned Language Models and Guided Tree Search
标题：通过精调语言模型和引导树搜索自动证明香农型熵不等式
链接：https://arxiv.org/abs/2606.05729

【21】Beyond Output Matching: Preserving Internal Geometry in NVFP4 LLM Distillatio
标题：超越输出匹配：在NVFP 4 LLM蒸馏中保留内部几何形状
链接：https://arxiv.org/abs/2606.05682

【22】CASS-RTL: Correctness-Aware Subspace Steering for RTL Generation with LLMs
标题：CASS-RTL：使用LLM生成RTL的正确性感知子空间引导
链接：https://arxiv.org/abs/2606.05680

【23】SlotGCG: Exploiting the Positional Vulnerability in LLMs for Jailbreak Attacks
标题：SlotGCG：利用LLM中的位置漏洞进行越狱攻击
链接：https://arxiv.org/abs/2606.05609

【24】Autoregressive Diffusion World Models for Off-Policy Evaluation of LLM Agents
标题：LLM代理非政策评估的自回归扩散世界模型
链接：https://arxiv.org/abs/2606.05558

【25】Less is MoE: Trimming Experts in Domain-Specialist Language Models
标题：MoE更少：精简领域专家语言模型中的专家
链接：https://arxiv.org/abs/2606.05538

【26】Almieyar-Oryx-BloomBench: A Bilingual Multimodal Benchmark for Cognitively Informed Evaluation of Vision-Language Models
标题：Almieyar-Oryx-BloomBench：视觉语言模型认知知情评估的双语多模式基准
链接：https://arxiv.org/abs/2606.05531

【27】Dominant-Layer ZO: A Single Layer Dominates Zeroth-Order Fine-Tuning of LLMs
标题：主导层Zero：单层主导LLM的零阶微调
链接：https://arxiv.org/abs/2606.05516

【28】Localizing Prompt Ambiguity in Large Language Models with Probe-Targeted Attribution
标题：具有针对探针的归因的大型语言模型中的提示歧义本地化
链接：https://arxiv.org/abs/2606.05486

【29】Can We Predict The Human Preference For Text-to-Image Content Prior To Generation And Is It Even Useful To Do So?
标题：我们可以预测人类在生成之前对文本到图像内容的偏好吗？这样做有用吗？
链接：https://arxiv.org/abs/2606.05478

【30】Selective-Advantage Entropy-Adaptive Horizon GRPO: Asymmetric Token-Level Discounting for Efficient Reinforcement Learning of Language Models
标题：选择性优势、自适应视野GRPO：非对称令牌级折扣，实现语言模型的高效强化学习
链接：https://arxiv.org/abs/2606.05434

【31】When Evidence is Sparse: Weakly Supervised Early Failure Alerting in Dialogs and LLM-Agent Trajectories
标题：当证据稀疏时：对话和LLM代理轨迹中弱监督的早期失败警报
链接：https://arxiv.org/abs/2606.05414

【32】Trust, but Don't Verify: Epistemic Blind Spots in LLM Source Evaluation
标题：信任，但不要验证：LLM来源评估中的认识盲点
链接：https://arxiv.org/abs/2606.05403

【33】Pattern Selectivity is Not Task-Causal Structure: A Cross-Architecture Mechanistic Study of Composed-Task Circuits in 1B-Class Language Models
标题：模式选择性不是任务因果结构：1B类语言模型中组合任务回路的跨体系结构机制研究
链接：https://arxiv.org/abs/2606.05378

【34】SHALA-LLM: Smartly Handling Ambiguous Labels in Aligning LLMs
标题：SHLA-LLM：智能处理调整LLM中的模糊标签
链接：https://arxiv.org/abs/2606.05376

【35】Statistically Reliable LLM-Based Ranking Evaluation via Prediction-Powered Inference
标题：通过预测动力推理进行统计可靠的基于LLM的排名评估
链接：https://arxiv.org/abs/2606.05308

【36】Aggregating LLM-Based Weak Verifiers for Spatial Layout Generation
标题：聚合基于LLM的弱验证器以生成空间布局
链接：https://arxiv.org/abs/2606.05268

【37】State commitment learning: training language models to distinguish computation from memory
标题：国家承诺学习：训练语言模型以区分计算与记忆
链接：https://arxiv.org/abs/2606.05201

【38】Temporal Preference Concepts and their Functions in a Large Language Model
标题：大语言模型中的时间偏好概念及其功能
链接：https://arxiv.org/abs/2606.05194

【39】ERRORQUAKE: Heavy-Tailed Error Severity Distributions in Open-Weight Large Language Models
标题：CLARQUAKE：开放权重大型语言模型中的重尾错误严重性分布
链接：https://arxiv.org/abs/2606.05170

【40】The Evaluation Blind Spot: A Stereological Theory of Benchmark Coverage for Large Language Models
标题：评估盲点：大型语言模型基准覆盖率的立体学理论
链接：https://arxiv.org/abs/2606.05169

Graph相关(图学习|图神经网络|图优化等)(10篇)

【1】Maximising the Set-Piece Return: Optimising Football Corner Tactics with Graph Reinforcement Learning
标题：最大化定位球回传：利用图强化学习优化足球角球战术
链接：https://arxiv.org/abs/2606.06353

【2】PAC-Bayesian Adversarially Robust Generalization for Message Passing Graph Neural Networks: A Sensitivity Analysis
标题：消息传递图神经网络的PAC-Bayesian对抗鲁棒推广：敏感性分析
链接：https://arxiv.org/abs/2606.06293

【3】Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation
标题：弥合语义协作差距：冷启动项目推荐的非对称图架构
链接：https://arxiv.org/abs/2606.06225

【4】A Machine Learning-Based Framework for Discovering Huntington's Disease Stages: Integrating Graph Representation Learning and clustering to Uncover Progression Dynamics in Longitudinal Enroll-HD Dataset
标题：用于发现亨廷顿氏症分期的基于机器学习的框架：集成图表示学习和集群以揭示纵向登记HD数据集中的进展动态
链接：https://arxiv.org/abs/2606.06196

【5】HoT-SSM:Higher-order Temporal Knowledge Graph Reasoning with State Space Models for Health Care
标题：HoT-RSM：使用状态空间模型的医疗保健高级时态知识图推理
链接：https://arxiv.org/abs/2606.05994

【6】Zero-Copy Semantic Contagion: An In-Memory Streaming Architecture for Evolving Attention Graphs
标题：零复制语义传染：一种用于进化注意力图的内存流媒体架构
链接：https://arxiv.org/abs/2606.05733

【7】Q-GNN: Query-Conditioned Graph Neural Networks with Type Awareness for Knowledge Graph Completion
标题：Q-GNN：具有知识图完成类型意识的查询条件图神经网络
链接：https://arxiv.org/abs/2606.05639

【8】StableRCA: Robust Graph-Agnostic Mechanism-Level Root Cause Analysis
标题：StableRCA：稳健的图形不可知机制级根本原因分析
链接：https://arxiv.org/abs/2606.05636

【9】HDST-GNN: Heterogeneous Dynamic Spatiotemporal Graph Neural Networks for Multi-Object Tracking in UAV Aerial Imagery
标题：HDST-GNN：用于无人机航空图像多目标跟踪的异类动态时空图神经网络
链接：https://arxiv.org/abs/2606.05587

【10】CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting
标题：CASEARCH ONE：基于时空图的因果建模，用于冷启动兴趣点登记预测
链接：https://arxiv.org/abs/2606.05413

Transformer(2篇)

【1】Consistency Training Along the Transformer Stack
标题：Transformer堆栈中的一致性训练
链接：https://arxiv.org/abs/2606.05817

【2】Transformer-Enhanced Reinforcement Learning: Fundamentals and Applications in Communication Networks
标题：转换器增强强化学习：通信网络的基础知识和应用
链接：https://arxiv.org/abs/2606.05208

GAN|对抗|攻击|生成相关(13篇)

【1】Attack Detection using Time Series Foundation Models
标题：基于时间序列基础模型的攻击检测
链接：https://arxiv.org/abs/2606.06347

【2】Quantifying the Privacy of Counterfactuals by Leveraging Membership Inference Attacks Against Synthetic Data
标题：通过利用针对合成数据的成员推断攻击来量化反事实的隐私
链接：https://arxiv.org/abs/2606.06334

【3】Diffusion Models for Adaptive Sequential Data Generation
标题：自适应序列数据生成的扩散模型
链接：https://arxiv.org/abs/2606.06007

【4】Steering Vectors are an Adversarial Attack Surface
标题：转向载体是对抗攻击面
链接：https://arxiv.org/abs/2606.05958

【5】GenAutoML: An Agentic Framework for Dynamic Architecture Generation and Optimization in Time-Series Analysis
标题：GenAutoML：时间序列分析中动态架构生成和优化的抽象框架
链接：https://arxiv.org/abs/2606.05860

【6】Next-Generation Parallel Decoder for LPDR: Architectural Optimization and Class-Balanced GAN-Augmentation
标题：下一代LPDR并行解码器：架构优化和类平衡GAN增强
链接：https://arxiv.org/abs/2606.05785

【7】Hybrid CNN-LSTM Framework for Intelligent Cyber Attack Detection and Prevention in U.S. Critical Digital Infrastructure: A Comparative Machine Learning Evaluation on CSE-CIC-IDS2018
标题：用于美国关键数字基础设施中智能网络攻击检测和预防的混合CNN-LSTM框架：CSE-CIC-IDS 2018上的比较机器学习评估
链接：https://arxiv.org/abs/2606.05714

【8】MolE-RAG: Molecular Structure-Enhanced Retrieval-Augmented Generation for Chemistry
标题：MolE-RAG：化学分子结构增强检索增强生成
链接：https://arxiv.org/abs/2606.05693

【9】Balancing Image Compression and Generation with Bootstrapped Tokenization
标题：通过引导令牌化平衡图像压缩和生成
链接：https://arxiv.org/abs/2606.05552

【10】REGEN: Reference-Guided Synthetic Multivariate Time Series Generation for Forecasting
标题：REGEN：用于预测的参考引导合成多元时间序列生成
链接：https://arxiv.org/abs/2606.05264

【11】NIV: Neural Axis Variations for Variable Font Generation
标题：NIV：可变字体生成的神经轴变体
链接：https://arxiv.org/abs/2606.05261

【12】Diff2SP: Diffusion Models for Correlated Scenario Generation in Stochastic Programming
标题：迪夫2SP：随机规划中相关场景生成的扩散模型
链接：https://arxiv.org/abs/2606.05649

【13】AlloGen: Conformation-Selective Binder Generation with Differential State Scoring
标题：AlloGen：具有差异状态评分的形态选择性粘合剂生成
链接：https://arxiv.org/abs/2606.05474

半/弱/无/有监督|不确定性|主动学习(5篇)

【1】Unsupervised Skill Discovery for Agentic Data Analysis
标题：用于统计数据分析的无监督技能发现
链接：https://arxiv.org/abs/2606.06416

【2】Unsupervised Pattern Analysis in Japanese Veterinary Toxicology: A Regulatory-Compliant Framework for Cross-Species Risk Assessment
标题：日本兽医毒理学中的无监督模式分析：跨物种风险评估的法规遵从性框架
链接：https://arxiv.org/abs/2606.06207

【3】Robust and sparse support vector machine via hybrid truncated loss for supervised classification
标题：通过混合截断损失进行监督分类的鲁棒稀疏支持载体机
链接：https://arxiv.org/abs/2606.05814

【4】T-SAR-JEPA: Self-Supervised Temporal Anomaly Detection in SAR Amplitude Stacks via Latent Prediction
标题：T-SAR-JEPA：通过潜在预测在SAR幅度堆栈中进行自监督时间异常检测
链接：https://arxiv.org/abs/2606.05700

【5】Evidence-Guided Neural Architecture Selection under Uncertainty for Subject-Specific Blood Glucose Forecasting
标题：不确定条件下基于证据的神经网络结构选择用于个体血糖预测
链接：https://arxiv.org/abs/2606.05373

迁移|Zero/Few/One-Shot|自适应(15篇)

【1】Regret Minimization with Adaptive Opponents in Repeated Games
标题：重复游戏中适应性对手的遗憾最小化
链接：https://arxiv.org/abs/2606.06486

【2】DAS-PINNs for high-dimensional partial differential equations: extending deep adaptive sampling to spacetime domains
标题：用于多维偏微方程的DAS-PINN：将深度自适应采样扩展到时空域
链接：https://arxiv.org/abs/2606.06314

【3】Reactive Flux Matching: Mechanism Discovery and Adaptive Sampling of Rare Events
标题：反应性通量匹配：机制发现与罕见事件的自适应采样
链接：https://arxiv.org/abs/2606.06295

【4】TLA-Prover: Verifiable TLA+ Specification Synthesis via Preference-Optimized Low-Rank Adaptation
标题：TLA-Prover：通过偏好优化的低等级自适应进行可验证的TLA+规范合成
链接：https://arxiv.org/abs/2606.06133

【5】Adaptive state-action abstractions via rate-distortion
标题：通过速率失真的自适应状态动作抽象
链接：https://arxiv.org/abs/2606.06123

【6】Step-adaptive multimodal fusion network with multi-scale cloud feature learning for ultra-short-term solar irradiance forecasting
标题：具有多尺度云特征学习的分步自适应多模式融合网络用于超短期太阳辐射预测
链接：https://arxiv.org/abs/2606.06102

【7】Adaptive Oscillatory-State Alignment for Time Series Forecasting
标题：时间序列预测的自适应振荡状态对齐
链接：https://arxiv.org/abs/2606.06010

【8】To Be Multimodal or Not to Be: Query-Adaptive Audio-Visual Person Retrieval via Active Modality Detection
标题：多模式还是非模式：通过主动模式检测的查询自适应视听人检索
链接：https://arxiv.org/abs/2606.05931

【9】Revisiting Prototype Rehearsal for Exemplar-Free Continual Learning: Manifold-Aware Boundary Sampling with Adaptive Class-Balanced Loss
标题：重新审视无示例连续学习的原型排练：具有自适应类平衡损失的Manifold Aware边界采样
链接：https://arxiv.org/abs/2606.05695

【10】Cross-Epoch Adaptive Rollout Optimization for RL Post-Training
标题：RL训练后的跨时代自适应推出优化
链接：https://arxiv.org/abs/2606.05606

【11】DP-MacAdam: Differentially Private Mechanism with Adaptive Clipping and Adaptive Momentum
标题：DP-MacAdam：具有自适应剪辑和自适应动量的差异私有机制
链接：https://arxiv.org/abs/2606.05435

【12】OLIVE: Online Low-Rank Incremental Learning for Efficient Adaptive Exoskeletons
标题：OLIVE：在线低等级增量学习，实现高效的自适应外骨骼
链接：https://arxiv.org/abs/2606.05234

【13】Effective Dimensionality as an Operator Invariant for Physics-Preserving Constraint Adaptation in Physics-Informed Neural Networks
标题：有效维度作为物理信息神经网络中物理保持约束自适应的操作不变量
链接：https://arxiv.org/abs/2606.06171

【14】Adaptive Learning Rates with Surrogate Probability for Follow-the-Perturbed-Leader
标题：跟随受干扰领导者的适应性学习率和替代概率
链接 ：https://arxiv.org/abs/2606.06043

【15】Harnessing Source Heterogeneity for Cluster-Structured Transfer Learning
标题：利用源异源进行机器人结构化迁移学习
链接：https://arxiv.org/abs/2606.05258

强化学习(7篇)

【1】Online KL-Regularized Reinforcement Learning with Function Approximation under Misspecification
标题：错误规范下具有函数逼近的在线KL正规强化学习
链接：https://arxiv.org/abs/2606.06053

【2】Merging model-based control with multi-agent reinforcement learning for multi-agent cooperative teaming strategies
标题：将基于模型的控制与多智能体强化学习相结合以实现多智能体合作团队策略
链接：https://arxiv.org/abs/2606.06011

【3】Representation Learning Enables Scalable Multitask Deep Reinforcement Learning
标题：表示学习实现可扩展多任务深度强化学习
链接：https://arxiv.org/abs/2606.05555

【4】Agentic Monte Carlo: Simulating Reinforcement Learning for Black-Box Agents
标题：抽象蒙特卡洛：模拟黑匣子代理的强化学习
链接：https://arxiv.org/abs/2606.05296

【5】Policy-Conditioned Counterfactual Credit for Verifiable Reinforcement Learning of Long-Horizon Language Agents
标题：长期语言代理的可验证强化学习的政策条件反事实信用
链接：https://arxiv.org/abs/2606.05263

【6】A New Quaternion-Joint Cable-Driven Redundant Manipulator Configuration and its Control Through FABRIK and Residual Reinforcement Learning
标题：基于FABRIK和残差强化学习的四元数索驱动冗余度机器人构型及其控制
链接：https://arxiv.org/abs/2606.05236

【7】Drag reduction or reward hacking? Recurrent multi-agent reinforcement learning that earns its reward
标题：减少阻力还是奖励黑客？获得回报的循环性多智能体强化学习
链接：https://arxiv.org/abs/2606.06227

符号|符号学习(2篇)

【1】Symb-xMIL: Symbolic Explanations for Multiple Instance Learning in Digital Pathology
标题：Symb-xMIL：数字病理学中多实例学习的符号解释
链接：https://arxiv.org/abs/2606.06224

【2】EML-CD: Causal Mechanism Recovery via EML Symbolic Trees in Structure Learning
标题：EML-CD：结构学习中通过EML符号树恢复因果机制
链接：https://arxiv.org/abs/2606.05942

医学相关(3篇)

【1】Learning to model pediatric asthma exacerbation from multiple risk factors: a case study in coastal Virginia
标题：学习根据多种风险因素对儿童哮喘急性发作进行建模：弗吉尼亚州沿海地区的一项案例研究
链接：https://arxiv.org/abs/2606.06174

【2】Quantifying the biophysical properties of stomatocytes in health and disease
标题：量化口腔细胞在健康和疾病中的生物物理特性
链接：https://arxiv.org/abs/2606.05227

【3】A differentiable machine learning small-angle X-ray scattering analysis framework for structure elucidation of lipid nanoparticles
标题：用于脂质纳米颗粒结构解析的可微机器学习小角度X射线散射分析框架
链接：https://arxiv.org/abs/2606.05200

蒸馏|知识提取(4篇)

【1】OPRD: On-Policy Representation Distillation
标题：OPRD：政策上代表蒸馏
链接：https://arxiv.org/abs/2606.06021

【2】Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation
标题：压缩-提取：推理痕迹压缩以实现高效知识提取
链接：https://arxiv.org/abs/2606.05988

【3】ViCuR: Visual Cues as Recoverable Privilege for Multimodal On-Policy Distillation
标题：ViCuR：视觉线索是多模式政策上蒸馏的可恢复特权
链接：https://arxiv.org/abs/2606.05718

【4】Flash-WAM: Modality-Aware Distillation for World Action Models
标题：Flash WAM：世界动作模型的模式感知蒸馏
链接：https://arxiv.org/abs/2606.05254

聚类(1篇)

【1】Central Description Length (CDL) Clustering Validation Index
标题：中心描述长度（CDL）集群验证指数
链接：https://arxiv.org/abs/2606.05230

自动驾驶|车辆|车道检测等(1篇)

【1】3D Underwater Path Planning via Generative Flow Field Surrogates
标题：基于生成流场代理的3D水下路径规划
链接：https://arxiv.org/abs/2606.06077

点云|SLAM|雷达|激光|深度RGBD相关(1篇)

【1】Data-efficient flood depth prediction through domain-aware coreset selection and tabular foundation models
标题：通过领域感知核心集选择和表格式基础模型进行数据高效的洪水深度预测
链接：https://arxiv.org/abs/2606.05265

推理|分析|理解|解释(9篇)

【1】RREDCoT: Segment-Level Reward Redistribution for Reasoning Models
标题：RRECDCoT：推理模型的分段级奖励重新分配
链接：https://arxiv.org/abs/2606.06475

【2】Latent Reasoning with Normalizing Flows
标题：具有规范化流程的潜在推理
链接：https://arxiv.org/abs/2606.06447

【3】Causal Atlases from Entropic Inference: Bayesian Networks beyond Optimal DAGs
标题：从博弈论推断因果关系：超越最佳DAB的Bayesian网络
链接：https://arxiv.org/abs/2606.06440

【4】A Vision-language Framework for Comparative Reasoning in Radiology
标题：放射学比较推理的视觉语言框架
链接：https://arxiv.org/abs/2606.06407

【5】Video-Rate Streaming Stylization on a Vision-Aware MLLM-Conditioned Edit Diffusion: Asymmetric Batched Inference on a Distilled UNet + MLLM Text Encoder
标题：视觉感知MLLM条件编辑扩散：蒸馏UNet + MLLM文本编码器上的不对称批量推理
链接：https://arxiv.org/abs/2606.05981

【6】Knowledge Manifold: A Riemannian Geometric Framework for Semantic Mapping and Geodesic Analysis of Scientific Literature
标题：知识库：科学文献语义映射和测地分析的Riemann几何框架
链接：https://arxiv.org/abs/2606.05907

【7】Critic-Guided Heterogeneous Multi-Agent Reasoning for Reliable Mathematical Problem Solving
标题：批判引导的异类多智能体推理用于可靠的数学问题解决
链接：https://arxiv.org/abs/2606.05704

【8】What Objects Enable, Not What They Are: Functional Latent Spaces for Affordance Reasoning
标题：对象启用什么，而不是它们是什么：负担能力推理的功能潜在空间
链接：https://arxiv.org/abs/2606.05533

【9】Symmetric Divergence and Normalized Similarity: A Unified Topological Framework for Representation Analysis
标题：对称发散与规范相似：表示分析的统一拓扑框架
链接：https://arxiv.org/abs/2606.06342

检测相关(6篇)

【1】Operation-Guided Progressive Human-to-AI Text Transformation Benchmark for Multi-Granularity AI-Text Detection
标题：用于多粒度AI文本检测的操作引导渐进式人机到人工智能文本转换基准
链接：https://arxiv.org/abs/2606.06481

【2】Event Detection for Parameter-to-KPI Dependency Learning for AI-RAN
标题：AI-RAN参数与KPI依赖性学习的事件检测
链接：https://arxiv.org/abs/2606.06459

【3】End-to-End Subgraph Detection with GraphDETR
标题：使用GraphDETR进行端到端子图检测
链接：https://arxiv.org/abs/2606.06364

【4】An Improved CNN-LSTM Based Intrusion Detection System for IoT Networks
标题：一种改进的基于CNN-LSTM的物联网入侵检测系统
链接：https://arxiv.org/abs/2606.05776

【5】Anomaly Detection for Electro-Hydrostatic Actuators using LSTM Autoencoder
标题：使用LSTM自动编码器检测电液致动器的异常
链接：https://arxiv.org/abs/2606.05274

【6】TabSODA: Tabular Diffusion based Imputation with Skip Pattern Detection and Ordinal Awareness
标题：TabSODA：具有跳过模式检测和有序意识的基于表格扩散的插补
链接：https://arxiv.org/abs/2606.05361

表征(2篇)

【1】Discrete Causal Representations from Heterogeneous Domains: A Bayesian Approach with Social Survey Applications
标题：来自异类领域的离散因果表示：具有社会调查应用的Bayesian方法
链接：https://arxiv.org/abs/2606.06288

【2】Environment-Robust Representation Learning with Empirical Bayes
标题：使用经验Bayes的环境稳健表示学习
链接：https://arxiv.org/abs/2606.05365

3D|3D重建等相关(1篇)

【1】Tracing the Oracle: Improving Diffusion Timestep Scheduling for 3D CT Reconstruction
标题：追踪Oracle：改进3D CT重建的扩散时步调度
链接：https://arxiv.org/abs/2606.06236

优化|敛散性(10篇)

【1】Double Preconditioning (DoPr): Optimization for Test-Time Performance, not Validation Loss
标题：双重预处理（DoPr）：优化测试时间性能，而不是验证损失
链接：https://arxiv.org/abs/2606.06418

【2】Your GFlowNet Secretly Learns an Optimal Transport Plan
标题：您的GFlowNet秘密学习最佳运输计划
链接：https://arxiv.org/abs/2606.06272

【3】MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following
标题：MDP-GRPO：多约束指令遵循的稳定群体相对政策优化
链接：https://arxiv.org/abs/2606.06058

【4】When Good Enough Is Optimal: Multiplication-Only Matrix Inversion Approximation for Quantized Gated DeltaNet
标题：当足够好时是最佳的：量化门控DelaNet的纯乘矩阵逆逼近
链接：https://arxiv.org/abs/2606.06034

【5】SALT: When More Rollouts Don't Help in Group-Based Policy Optimization and How to Make Them Matter
标题：SALT：当更多的卷出对基于组的策略优化没有帮助时，以及如何使它们发挥作用
链接：https://arxiv.org/abs/2606.05800

【6】Mitigating the Curse of Dimensionality in Uniform Convergence of Deep Neural Networks via Smooth Activations
标题：通过平滑激活减轻深度神经网络一致收敛中的收敛性诅咒
链接：https://arxiv.org/abs/2606.05599

【7】Sharp First-Order Lower Bounds for Higher-Order Smooth Nonconvex Optimization
标题：高级光滑非凸优化的尖锐一阶下界
链接：https://arxiv.org/abs/2606.05438

【8】Multimarginal flow matching with optimal transport potentials
标题：最优输运势的多边际流匹配
链接：https://arxiv.org/abs/2606.05327

【9】Alpha-RTL: Test-Time Training for RTL Hardware Optimization
标题：Alpha-RTL：RTL硬件优化的测试时训练
链接：https://arxiv.org/abs/2606.05253

【10】Fast and Robust Convergence Rate for TD(0) with Linear Function Approximation, Universal Learning Steps and I.I.D. Samples
标题：具有线性函数逼近、通用学习步骤和I.I.D.的TD（0）快速稳健的收敛率样品
链接：https://arxiv.org/abs/2606.05967

预测|估计(13篇)

【1】Learned Response-Field Inertia Operator for HEC-RAS 2D Water-Surface Elevation Prediction
标题：用于HEC-ras 2D水面海拔预测的学习响应场惯性操作器
链接：https://arxiv.org/abs/2606.06385

【2】Performance Evaluation of GraphCast for Medium-Range Weather Forecasting over Brazil
标题：巴西中期天气预报的GraphCast性能评估
链接：https://arxiv.org/abs/2606.06348

【3】Bridging Domain Expertise and Generalization for Performance Estimation
标题：将领域专业知识和通用化结合起来进行性能评估
链接：https://arxiv.org/abs/2606.06335

【4】Trust-Aware Predictive Emissions Monitoring for Gas Turbine Fleets with Limited Labelled Data
标题：具有有限标签数据的燃气涡轮机机队的可信预测排放监控
链接：https://arxiv.org/abs/2606.06156

【5】OrderGrad: Optimizing Beyond the Mean with Order-Statistic Policy Gradient Estimation
标题：Order Grad：通过顺序统计政策梯度估计进行超越平均值的优化
链接：https://arxiv.org/abs/2606.06096

【6】On Advantage Estimates for Max@K Policy Gradients
标题：关于Max@K政策支持者的优势估计
链接：https://arxiv.org/abs/2606.06080

【7】Causal Longitudinal Prior-Fitted Networks for Counterfactual Outcome Prediction
标题：用于反事实结果预测的因果纵向先验匹配网络
链接：https://arxiv.org/abs/2606.05797

【8】Benchmarking Counterfactual Prediction in Epidemic Time Series with Time-Varying Interventions
标题：采用时变干预措施在流行病时间序列中进行反事实预测基准
链接：https://arxiv.org/abs/2606.05692

【9】From Prediction to Self: Developmental Conditions for Agency in Minimal Neural Systems
标题：从预测到自我：最小神经系统中机构的发展条件
链接：https://arxiv.org/abs/2606.05605

【10】Field Validation of a Multi-Resolution ConvLSTM Framework for Retaining Wall Deformation Prediction
标题：挡土墙变形预测的多分辨率ConvLSTM框架的现场验证
链接：https://arxiv.org/abs/2606.05556

【11】Function-Space Priors for Bayesian Neural ODEs with Application to Vessel Trajectory Prediction
标题：Bayesian神经ODE的功能空间先验及其在船舶轨迹预测中的应用
链接：https://arxiv.org/abs/2606.06351

【12】Uncovering Extreme Event Mechanisms for Prediction and Control with Sensitivity-Balanced Projections
标题：利用敏感性平衡投影揭示极端事件机制进行预测和控制
链接：https://arxiv.org/abs/2606.05618

【13】The Language of Elution: Autoregressive Prediction of the Next Feature in Untargeted LC-HRMS Lipidomics
标题：洗出的语言：非靶向LC-HRT脂质组学中下一个特征的自回归预测
链接：https://arxiv.org/abs/2606.05225

其他神经网络|深度学习|模型|建模(41篇)

【1】TailLoR: Protecting Principal Components in Parameter-Efficient Continual Learning
标题：TailLoR：在参数高效的持续学习中保护主要成分
链接：https://arxiv.org/abs/2606.06494

【2】DNQ: Deep Nash Q-Network for Partially Observable n-Player Games
标题：DNQ：用于部分可观察n-玩家游戏的深度纳什Q网络
链接：https://arxiv.org/abs/2606.06480

【3】Pretraining Recurrent Networks without Recurrence
标题：预训练无回归的回归网络
链接：https://arxiv.org/abs/2606.06479

【4】In-Context Multiple Instance Learning
标题：上下文多实例学习
链接：https://arxiv.org/abs/2606.06458

【5】The Post-GCN Decade Revisited: Curvature-Stratified Evaluation of Relational Learning
标题：后GCN十年重温：关系学习的曲线分层评估
链接：https://arxiv.org/abs/2606.06397

【6】Plug-and-Play Guidance for Discrete Diffusion Models via Gradient-Informed Logit Correction
标题：通过受影响者知情的Logit纠正为离散扩散模型提供即插即用指导
链接：https://arxiv.org/abs/2606.06303

【7】Integrating Mechanistic and Data-Driven Models for Neurological Disorders through Differentiable Programming
标题：通过差异编程集成神经系统疾病的机械和数据驱动模型
链接：https://arxiv.org/abs/2606.06094

【8】Metamorphic Testing with the Rashomon Set: Explanation Faithfulness in Machine Learning
标题：使用罗生门集进行变形测试：机器学习中的忠实性解释
链接：https://arxiv.org/abs/2606.06056

【9】Learning solution operators of PDEs with sparse approximation methods
标题：用稀疏逼近方法学习偏出方程的解运算符
链接：https://arxiv.org/abs/2606.06046

【10】Catastrophic Forgetting as Accessibility Collapse: A Three-Level Framework for Knowledge Persistence in Continual Learning
标题：灾难性遗忘作为可及性崩溃：持续学习中知识持久性的三层次框架
链接：https://arxiv.org/abs/2606.06032

【11】Dead Directions: Geometric Singular Learning
标题：死方向：几何奇异学习
链接：https://arxiv.org/abs/2606.05957

【12】Short paper: Models in the dark -- Rectification and erasure under GDPR in ML supply chains
标题：短文：黑暗中的模型--ML供应链中GDPR下的纠正和删除
链接：https://arxiv.org/abs/2606.05946

【13】DBHN-Net: Dual-Branch Hybrid Neural Network For Low-Complexity Monaural Speech Enhancement
标题：DBHN-Net：用于低复杂度单耳语音增强的双分支混合神经网络
链接：https://arxiv.org/abs/2606.05911

【14】High-Dimensional Theory of LoRA Fine-Tuning in a Solvable Attention Model
标题：可解注意力模型中LoRA微调的多维理论
链接：https://arxiv.org/abs/2606.05899

【15】TS-ICL: A Flexible Time-Indexed Foundation Model for Time Series via In-Context Learning
标题：TS-ICL：通过上下文学习的时间序列灵活的时间索引基础模型
链接：https://arxiv.org/abs/2606.05878

【16】LadderMan: Learning Humanoid Perceptive Ladder Climbing
标题：LadderMan：学习类人感知爬梯
链接：https://arxiv.org/abs/2606.05873

【17】Deciphering Two Training Clocks in Grokking via Deep Linear Network Theory with Conditional ReLU Reduction
标题：通过深度线性网络理论和条件ReLU约简破解Grokking中的两个训练时钟
链接：https://arxiv.org/abs/2606.05863

【18】Intercomparison of Machine Learning Algorithms for Remote Sensing-based In-season Crop Mapping
标题：基于遥感的季节作物制图机器学习算法的相互比较
链接：https://arxiv.org/abs/2606.05731

【19】Causal Modeling of Selection in Evolution
标题：进化中选择的因果模型
链接：https://arxiv.org/abs/2606.05689

【20】Two-Way Is Better Than One: Bidirectional Alignment with Cycle Consistency for Exemplar-Free Class-Incremental Learning
标题：双向总比单向好：双向调整与循环一致性，实现无示例课堂增量学习
链接：https://arxiv.org/abs/2606.05675

【21】CLaaS: Continual learning as a service for sample efficient online learning
标题：Claas：将持续学习作为示例高效在线学习的服务
链接：https://arxiv.org/abs/2606.05559

【22】Towards Unified and Data-Efficient Prognostics and Health Management with Tabular Foundation Models
标题：利用表格基础模型实现统一且数据高效的预测和健康管理
链接：https://arxiv.org/abs/2606.05481

【23】GOTabPFN: From Feature Ordering to Compact Tokenization for Tabular Foundation Models on High-Dimensional Data
标题：GOTabPFN：从特征排序到多维数据上表格基础模型的紧凑代币化
链接：https://arxiv.org/abs/2606.05441

【24】Learning-Augmented Online Minimization with Dual Predictions
标题：具有双重预测的学习增强在线最小化
链接：https://arxiv.org/abs/2606.05380

【25】Mamba-Assisted Non-Markovian Closure for Reduced-Order Modeling
标题：用于降阶建模的Mamba辅助非马尔科夫闭合
链接：https://arxiv.org/abs/2606.05371

【26】Should Demand Models Incorporate Competitor Prices? Oblivious Learning and Algorithmic Collusion
标题：需求模型是否应该纳入竞争对手的价格？无意识学习与妄想共谋
链接：https://arxiv.org/abs/2606.05363

【27】A prism hierarchy of learning regimes in large linear autoencoders
标题：大型线性自动编码器中学习机制的棱镜层次结构
链接：https://arxiv.org/abs/2606.05335

【28】The Invisible Hand of Physics: When Video Diffusion Models Know More Than They Show
标题：物理学看不见的手：当视频扩散模型知道的比它们所展示的更多时
链接：https://arxiv.org/abs/2606.05328

【29】Learning Manifold and Itô Dynamics with Branched Neural Rough Differential Equations
标题：使用分支神经粗方程的学习Manifics和Itó动力学
链接：https://arxiv.org/abs/2606.05272

【30】Scaling Laws for Behavioral Foundation Models over User Event Sequences
标题：用户事件序列上行为基础模型的缩放定律
链接：https://arxiv.org/abs/2606.05257

【31】DiffSlack: Learning under Nonlinear Inequality Constraints via Learnable Slack Variables
标题：DiffSlack：通过可学习的松弛变量在非线性不等式约束下学习
链接：https://arxiv.org/abs/2606.05247

【32】Gradient Descent with Large Step Size Restores Symmetry in Deep Linear Networks with Multi-Pathway
标题：大步进的梯度下降恢复多路径深度线性网络的对称性
链接：https://arxiv.org/abs/2606.05219

【33】Epidemiology of Model Collapse: Modeling Synthetic Data Contamination via Bilayer SIR Dynamics
标题：模型崩溃的流行病学：通过双层Sir动力学对合成数据污染进行建模
链接：https://arxiv.org/abs/2606.05168

【34】Wall Shear Stress Reconstruction from Concentration: Differentiable Physics and Physics-Informed Neural Networks
标题：从浓度重建壁面剪应力：可微物理和物理信息神经网络
链接：https://arxiv.org/abs/2606.06313

【35】Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors
标题：扩散模型只观察从属：分数匹配错误的几何视角
链接：https://arxiv.org/abs/2606.06179

【36】$p$-adic Bi-Filtrations for Topological Machine Learning on Genomic Sequences
标题：$p$-adic双过滤用于基因组序列上的拓学机器学习
链接 ：https://arxiv.org/abs/2606.06117

【37】Cross-scale spatially-aware generative modeling of transcriptomic programs underlying neurodegenerative brain organization
标题：神经退行性大脑组织背后的转录组程序的跨规模空间感知生成建模
链接：https://arxiv.org/abs/2606.05870

【38】Gradient descent at the Edge of Stability: free energy model and kinetic description of the two-layer network
标题：稳定边缘的梯度下降：两层网络的自由能模型和动力学描述
链接：https://arxiv.org/abs/2606.05326

【39】The Score Hamiltonian: Mapping Diffusion Models to Adiabatic Transport
标题：分数哈密顿量：将扩散模型映射到绝热输运
链接：https://arxiv.org/abs/2606.05217

【40】Multi-Fidelity Learning with Shallow Recurrent Decoders for Reactor Physics
标题：反应堆物理中使用浅层循环解码器的多保真学习
链接：https://arxiv.org/abs/2606.05202

【41】An accurate nucleic acid-small molecule docking framework via geometric deep learning with large-scale pretraining
标题：通过大规模预训练的几何深度学习构建准确的核酸-小分子对接框架
链接：https://arxiv.org/abs/2606.05198

其他(46篇)

【1】HANDOFF: Humanoid Agentic Task-Space Whole-Body Control via Distilled Complementary Teachers
标题：交接：通过提炼补充教师的类人抽象任务空间全身控制
链接：https://arxiv.org/abs/2606.06493

【2】You Only Index Once: Cross-Layer Sparse Attention with Shared Routing
标题：您只需索引一次：通过共享路由实现跨层稀疏注意力
链接：https://arxiv.org/abs/2606.06467

【3】Proper Scoring Rules for Right-Censored Survival Data
标题：右审查生存数据的适当评分规则
链接：https://arxiv.org/abs/2606.06393

【4】Boosting Brain-to-Image Decoding with TRIBE v2 Data Augmentation
标题：利用TRIBE v2数据增强增强大脑到图像解码
链接：https://arxiv.org/abs/2606.06345

【5】Equivariant Neural Belief Propagation
标题：等变神经信念传播
链接：https://arxiv.org/abs/2606.06344

【6】Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability
标题：子空间感知的稀疏自编码器用于有效的机制解释性
链接：https://arxiv.org/abs/2606.06333

【7】Efficient Mean Curvature Computation on High-Dimensional Data Manifolds
标题：多维数据库上的高效平均弯曲计算
链接：https://arxiv.org/abs/2606.06329

【8】PAMF: Prior-Aware Multimodal Fusion for Incomplete Time Series Data
标题：PAMF：不完整时间序列数据的先验感知多模态融合
链接：https://arxiv.org/abs/2606.06328

【9】GRAMformer: Any-Order Modality Interactions via Volumetric Multimodal Cross-Attention
标题：GRAMformer：通过体积多模式交叉注意力的任何阶模式交互
链接：https://arxiv.org/abs/2606.06249

【10】Non-Negative Matrix Factorization for Event Data
标题：事件数据的非负矩阵分解
链接：https://arxiv.org/abs/2606.06205

【11】On the training of physics-informed neural operators for solving parametric partial differential equations
标题：关于训练物理信息神经运算符以求解参数偏微方程
链接：https://arxiv.org/abs/2606.06164

【12】Tight list replicability bounds via a novel sphere covering theorem
标题：通过新颖的球覆盖定理建立紧列表可复制性界限
链接：https://arxiv.org/abs/2606.06148

【13】A Sliced-Wasserstein Framework on Correlation Matrices for EEG Decoding
标题：用于脑电解码的相关矩阵切片-沃瑟斯坦框架
链接：https://arxiv.org/abs/2606.06104

【14】A Pre-Registered Causal Partition of Self-Consistency Elicitation and Reward Design in RLVR
标题：RLVR中自我和谐诱发与奖励设计的预记录因果划分
链接：https://arxiv.org/abs/2606.05932

【15】Addressing Imbalance in Multi-Label Data via Label-Specific Distance-based Oversampling
标题：通过特定标签距离的过采样解决多标签数据中的不平衡问题
链接：https://arxiv.org/abs/2606.05927

【16】Representing Research Attention as Contextually Structured Flows
标题：将研究注意力表示为上下文结构流
链接：https://arxiv.org/abs/2606.05895

【17】Beyond Soft Masks: Hard-Perturbation Mixup Explainer for Robust GNN Explainability
标题：超越软面具：强GNN解释性的硬扰动Mixup解释器
链接：https://arxiv.org/abs/2606.05756

【18】When Surface Form Changes Moderation Decisions: A Paired Study of Code-Mixed Workflow Instability
标题：当表面形式发生变化时调整决策：代码混合工作流程不稳定性的配对研究
链接：https://arxiv.org/abs/2606.05654

【19】When New Generators Arrive: Lifelong Machine-Generated Text Attribution via Ridge Feature Transfer
标题：当新生成器到来时：通过山脊特征转移实现终身机器生成文本属性
链接：https://arxiv.org/abs/2606.05626

【20】Self-Commitment Latency: A Reward-Free Probe for Prompted Implicit Hacking
标题：自我承诺潜伏期：对隐性黑客的免奖励调查
链接：https://arxiv.org/abs/2606.05625

【21】Fix the Mind, Not the Move: Interpretable AI Assistance via Knowledge-Gap Localization
标题：修复思想，而不是行动：通过知识差距本地化提供可解释的人工智能协助
链接：https://arxiv.org/abs/2606.05602

【22】AsyncWebRL: Efficient Multi-Step RL for Visual Web Agents
标题：AsyncWebRL：用于视觉Web代理的高效多步骤RL
链接：https://arxiv.org/abs/2606.05597

【23】Auditing Demonstration Curation Metrics: Action-Only Scorers Fail on the Structural Defects That Degrade Imitation Policies
标题：审计演示教程：仅限员工的评分者未能在导致模仿政策降级的结构性缺陷上取得成功
链接：https://arxiv.org/abs/2606.05588

【24】Monte Carlo Steklov Operators for Large-Scale Geometry Processing in the Wild
标题：用于野外大规模几何处理的Monte Carlo Steklov运算符
链接：https://arxiv.org/abs/2606.05581

【25】LEVANTE-bench: Multi-Scale Comparison of VLMs to Children Using Cognitive Tasks (or, "Is Your VLM Smarter Than a 5th Grader?")
链接：https://arxiv.org/abs/2606.05497

【26】Learned Subspace Compression for Communication-Efficient Pipeline Parallelism
标题：学习子空间压缩以实现通信高效的管道并行主义
链接：https://arxiv.org/abs/2606.05484

【27】Multilingual Coreference Resolution via Cycle-Consistent Machine Translation
标题：通过周期一致的机器翻译实现多语言共指解析
链接：https://arxiv.org/abs/2606.05444

【28】Executable Schema Contracts: From Automatic Ingestion to Multi-Source Retrieval
标题：可执行模式合同：从自动摄入到多源检索
链接：https://arxiv.org/abs/2606.05415

【29】Agents' Last Exam
标题：特工的最后一次考试
链接：https://arxiv.org/abs/2606.05405

【30】Harnessing Generalist Agents for Contextualized Time Series
标题：利用通才代理进行情境化时间序列
链接：https://arxiv.org/abs/2606.05404

【31】LeanMarathon: Toward Reliable AI Co-Mathematicians through Long-Horizon Lean Autoformalization
标题：LeanMarathon：通过长期精益自动化发展可靠的人工智能联合数学家
链接：https://arxiv.org/abs/2606.05400

【32】Generalized TV--$\ell_p$ Structured Priors for Bayesian $T_1$ Mapping
链接：https://arxiv.org/abs/2606.05381

【33】PJ-RoPE: A Fourier-Jet-Affine Position Space for Relative Attention
标题：PJ-RoPE：傅里叶-喷射-仿射位置空间，以获得相对注意力
链接：https://arxiv.org/abs/2606.05345

【34】Sharp Low-Degree Thresholds for Planted-vs-Planted Testing
标题：用于种植与种植测试的尖锐低度交叉点
链接：https://arxiv.org/abs/2606.05266

【35】Differentiable Efficient Operator Search
标题：差异化高效的运营商搜索
链接：https://arxiv.org/abs/2606.05232

【36】PyCC.id: A package for hypothesis-driven equation discovery with structural identifiability
标题：PyCC.id：具有结构可识别性的假设驱动方程发现包
链接：https://arxiv.org/abs/2606.05191

【37】Staged Factorial Screening for Budget-Constrained Micro-Pretraining
标题：预算限制微预训练的分阶段析因筛选
链接：https://arxiv.org/abs/2606.05186

【38】Drishti AI-Event Guardian: An Intelligent Real-Time Crowd Monitoring and Emergency Response System for Mass Gathering Events
标题：Drishti人工智能活动守护者：针对大规模集会活动的智能实时人群监控和应急响应系统
链接：https://arxiv.org/abs/2606.05185

【39】How abundant are good interpolators?
标题：好的插值器有多少？
链接：https://arxiv.org/abs/2606.06469

【40】Conformal Risk Sharing: Certified Cost Allocation with Participation Guarantees
标题：共形风险分担：认证的成本分配和参与保证
链接：https://arxiv.org/abs/2606.06391

【41】Anchor PCA
标题：锚PCA
链接：https://arxiv.org/abs/2606.06233

【42】Finding Most Influential Sets
标题：寻找最有影响力的集合
链接：https://arxiv.org/abs/2606.05919

【43】Conformal Risk-Averse Decision Making with Action Conditional Guarantee
标题：具有行动条件保证的共形风险规避决策
链接：https://arxiv.org/abs/2606.05551

【44】Sparse Functional Singular Value Decomposition for Biclustering and Triclustering Longitudinal Data
标题：双集群和三集群纵向数据的稀疏函数奇异值分解
链接：https://arxiv.org/abs/2606.05488

【45】Deterministic Envelopes for Tamed SGLD: Decoupling Stochastic-Gradient Noise and Localizing Taming
标题：驯服SGLD的确定性信封：随机梯度噪音去耦合和局部驯服
链接：https://arxiv.org/abs/2606.05242

【46】HyFAD: Hybrid Time-Frequency Diffusion with Frequency-Aware Embedding for Time Series Imputation
标题：HyFAD：混合时频扩散与频率感知嵌入的时间序列插补算法
链接：https://arxiv.org/abs/2606.05239

机器翻译由腾讯交互翻译提供，仅供参考

点击“阅读原文”获取带摘要的学术速递