机器学习学术速递[9.29]

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297篇

大模型相关(37篇)

【1】Language Models Can Learn from Verbal Feedback Without Scalar Rewards
标题：语言模型可以从口头反馈中学习，而无需获得量化奖励
链接：https://arxiv.org/abs/2509.22638

【2】Variational Reasoning for Language Models
标题：语言模型的变分推理
链接：https://arxiv.org/abs/2509.22637

【3】Benefits and Pitfalls of Reinforcement Learning for Language Model Planning: A Theoretical Perspective
标题：语言模型规划强化学习的好处和陷阱：理论视角
链接：https://arxiv.org/abs/2509.22613

【4】EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning
标题：EPO：LLM代理强化学习的熵正则化策略优化
链接：https://arxiv.org/abs/2509.22576

【5】Dynamic Experts Search: Enhancing Reasoning in Mixture-of-Experts LLMs at Test Time
标题：动态专家搜索：在测试时增强混合专家LL中的推理
链接：https://arxiv.org/abs/2509.22572

【6】REMA: A Unified Reasoning Manifold Framework for Interpreting Large Language Model
标题：REMA：用于解释大型语言模型的统一推理Manifold框架
链接：https://arxiv.org/abs/2509.22518

【7】Representing LLMs in Prompt Semantic Task Space
标题：在提示语义任务空间中表示LLM
链接：https://arxiv.org/abs/2509.22506

【8】Estimating the Empowerment of Language Model Agents
标题：评估语言模型代理的赋权
链接：https://arxiv.org/abs/2509.22504

【9】Evaluating the Limits of Large Language Models in Multilingual Legal Reasoning
标题 ：评估多语言法律推理中大型语言模型的局限性
链接：https://arxiv.org/abs/2509.22472

【10】MoveFM-R: Advancing Mobility Foundation Models via Language-driven Semantic Reasoning
标题：MoveFM-R：通过数据驱动的语义推理推进移动基础模型
链接：https://arxiv.org/abs/2509.22403

【11】SpinGPT: A Large-Language-Model Approach to Playing Poker Correctly
标题：SpinGPT：正确玩扑克的大型语言模型方法
链接：https://arxiv.org/abs/2509.22387

【12】Investigating Faithfulness in Large Audio Language Models
标题：研究大型音频语言模型中的忠实性
链接：https://arxiv.org/abs/2509.22363

【13】Fine-Grained Uncertainty Decomposition in Large Language Models: A Spectral Approach
标题：大型语言模型中的细粒度不确定性分解：谱方法
链接：https://arxiv.org/abs/2509.22272

【14】Lightweight error mitigation strategies for post-training N:M activation sparsity in LLMs
标题：LLM中训练后N：M激活稀疏性的轻量级错误缓解策略
链接：https://arxiv.org/abs/2509.22166

【15】Multi-Agent Path Finding via Offline RL and LLM Collaboration
标题：通过离线RL和LLM协作进行多智能体路径查找
链接：https://arxiv.org/abs/2509.22130

【16】Mind the Missing: Variable-Aware Representation Learning for Irregular EHR Time Series using Large Language Models
标题：注意缺失：使用大型语言模型的不规则EHR时间序列的变量感知表示学习
链接：https://arxiv.org/abs/2509.22121

【17】The Rogue Scalpel: Activation Steering Compromises LLM Safety
标题：流氓手术刀：激活操纵损害LLM安全性
链接：https://arxiv.org/abs/2509.22067

【18】Goal-Guided Efficient Exploration via Large Language Model in Reinforcement Learning
标题 ：强化学习中通过大语言模型实现目标引导的高效探索
链接：https://arxiv.org/abs/2509.22008

【19】ERGO: Efficient High-Resolution Visual Understanding for Vision-Language Models
标题：ERGO：视觉语言模型的高效高分辨率视觉理解
链接：https://arxiv.org/abs/2509.21991

【20】Think Smart, Not Hard: Difficulty Adaptive Reasoning for Large Audio Language Models
标题：聪明思考，不难思考：大型音频语言模型的自适应推理困难
链接：https://arxiv.org/abs/2509.21960

【21】Active Attacks: Red-teaming LLMs via Adaptive Environments
标题：主动攻击：通过自适应环境进行红色团队LLM
链接：https://arxiv.org/abs/2509.21947

【22】Extracting Actionable Insights from Building Energy Data using Vision LLMs on Wavelet and 3D Recurrence Representations
标题：使用基于子波和3D回归表示的视觉LLM从建筑能源数据中提取可操作见解
链接：https://arxiv.org/abs/2509.21934

【23】No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shaping
标题：不留任何提示：通过熵引导的优势塑造在LLM强化学习中利用零方差预算
链接：https://arxiv.org/abs/2509.21880

【24】Abductive Logical Rule Induction by Bridging Inductive Logic Programming and Multimodal Large Language Models
标题：通过连接归纳逻辑编程和多模式大型语言模型来归纳归纳逻辑规则
链接：https://arxiv.org/abs/2509.21874

【25】SBFA: Single Sneaky Bit Flip Attack to Break Large Language Models
标题：SBFA：单次Sneaky Bit Flip攻击破坏大型语言模型
链接：https://arxiv.org/abs/2509.21843

【26】Navigating the Impact of Structured Output Format on Large Language Models through the Compass of Causal Inference
标题：通过因果推理指南针导航结构化输出格式对大型语言模型的影响
链接：https://arxiv.org/abs/2509.21791

【27】Blockwise Hadamard high-Rank Adaptation for Parameter-Efficient LLM Fine-Tuning
标题：用于参数有效LLM微调的BHADAMARD高阶自适应
链接：https://arxiv.org/abs/2509.21637

【28】InvBench: Can LLMs Accelerate Program Verification with Invariant Synthesis?
标题：InvBench：LLM可以通过不变合成加速程序验证吗？
链接：https://arxiv.org/abs/2509.21629

【29】Guiding Audio Editing with Audio Language Model
标题：用音频语言模型指导音频编辑
链接：https://arxiv.org/abs/2509.21625

【30】Multi-Objective Reinforcement Learning for Large Language Model Optimization: Visionary Perspective
标题：面向大型语言模型优化的多目标强化学习：前瞻性观点
链接：https://arxiv.org/abs/2509.21613

【31】Evidence for Limited Metacognition in LLMs
标题：LLM元认知有限的证据
链接：https://arxiv.org/abs/2509.21545

【32】A circuit for predicting hierarchical structure in-context in Large Language Models
标题：用于预测大型语言模型中上下文中分层结构的电路
链接：https://arxiv.org/abs/2509.21534

【33】Preemptive Detection and Steering of LLM Misalignment via Latent Reachability
标题：通过潜在可达性先发制人地检测和引导LLM失准
链接：https://arxiv.org/abs/2509.21528

【34】Chasing the Tail: Effective Rubric-based Reward Modeling for Large Language Model Post-Training
标题：追尾：大型语言模型训练后的有效基于条目的奖励建模
链接：https://arxiv.org/abs/2509.21500

【35】d2: Improved Techniques for Training Reasoning Diffusion Language Models
标题：d2：用于训练推理扩散语言模型的改进技术
链接：https://arxiv.org/abs/2509.21474

【36】LLMs for Bayesian Optimization in Scientific Domains: Are We There Yet?
标题 ：科学领域的Bayesian优化法学硕士：我们已经到了吗？
链接：https://arxiv.org/abs/2509.21403

【37】Automated Machine Learning Pipeline for Training and Analysis Using Large Language Models
标题：使用大型语言模型进行训练和分析的自动机器学习管道
链接：https://arxiv.org/abs/2509.21647

Graph相关(图学习|图神经网络|图优化等)(14篇)

【1】Learning the Neighborhood: Contrast-Free Multimodal Self-Supervised Molecular Graph Pretraining
标题：学习邻居：无对比多峰自我监督分子图预训练
链接：https://arxiv.org/abs/2509.22468

【2】One Prompt Fits All: Universal Graph Adaptation for Pretrained Models
标题：一个提示适合所有：预训练模型的通用图适应
链接：https://arxiv.org/abs/2509.22416

【3】Transformers Can Learn Connectivity in Some Graphs but Not Others
标题：Transformer可以在某些图形中学习连通性，但不能在其他图形中学习
链接：https://arxiv.org/abs/2509.22343

【4】Wavelet-Induced Rotary Encodings: RoPE Meets Graphs
标题：微波诱导的旋转编码：RoPE与图形相遇
链接：https://arxiv.org/abs/2509.22259

【5】Kernel Regression of Multi-Way Data via Tensor Trains with Hadamard Overparametrization: The Dynamic Graph Flow Case
标题：通过Hadamard过度参数化的张量串进行多路数据的核回归：动态图流案例
链接：https://arxiv.org/abs/2509.22197

【6】SHAKE-GNN: Scalable Hierarchical Kirchhoff-Forest Graph Neural Network
标题：SHAKE-GNN：可扩展分层Kirchhoff-Forest图神经网络
链接：https://arxiv.org/abs/2509.22100

【7】AEGIS: Authentic Edge Growth In Sparsity for Link Prediction in Edge-Sparse Bipartite Knowledge Graphs
标题：AEGIS：边稀疏二分知识图中链接预测的稀疏性真实边增长
链接：https://arxiv.org/abs/2509.22017

【8】Graph of Agents: Principled Long Context Modeling by Emergent Multi-Agent Collaboration
标题：代理图：紧急多代理协作的原则性长上下文建模
链接：https://arxiv.org/abs/2509.21848

【9】Uncovering Alzheimer's Disease Progression via SDE-based Spatio-Temporal Graph Deep Learning on Longitudinal Brain Networks
标题：通过纵向脑网络上基于SDP的时空图深度学习揭示阿尔茨海默病进展
链接：https://arxiv.org/abs/2509.21735

【10】Exact Subgraph Isomorphism Network for Predictive Graph Mining
标题：用于预测图挖掘的精确子图同质网络
链接：https://arxiv.org/abs/2509.21699

【11】EEG-Based Consumer Behaviour Prediction: An Exploration from Classical Machine Learning to Graph Neural Networks
标题：基于EEG的消费者行为预测：从经典机器学习到图神经网络的探索
链接：https://arxiv.org/abs/2509.21567

【12】AutoClimDS: Climate Data Science Agentic AI -- A Knowledge Graph is All You Need
标题：AutoClimDS：气候数据科学统计人工智能--知识图谱即可满足您的需求
链接：https://arxiv.org/abs/2509.21553

【13】GraphPFN: A Prior-Data Fitted Graph Foundation Model
标题：GraphPFN：一个符合先验数据的图基础模型
链接：https://arxiv.org/abs/2509.21489

【14】SGNNBench: A Holistic Evaluation of Spiking Graph Neural Network on Large-scale Graph
标题：SGNNBench：一种基于大规模图的尖峰图神经网络整体评估方法
链接：https://arxiv.org/abs/2509.21342

Transformer(7篇)

【1】IIET: Efficient Numerical Transformer via Implicit Iterative Euler Method
标题：IET：通过隐式迭代欧拉方法的高效数值Transformer
链接：https://arxiv.org/abs/2509.22463

【2】Bridging Kolmogorov Complexity and Deep Learning: Asymptotically Optimal Description Length Objectives for Transformers
标题：弥合Kolmogorov复杂性和深度学习：Transformer的渐进最优描述长度目标
链接：https://arxiv.org/abs/2509.22445

【3】SoDaDE: Solvent Data-Driven Embeddings with Small Transformer Models
标题：SoDaDE：具有小型Transformer模型的溶剂数据驱动嵌入
链接：https://arxiv.org/abs/2509.22302

【4】Teaching Transformers to Solve Combinatorial Problems through Efficient Trial & Error
标题：教Transformer通过有效的试错解决组合问题
链接：https://arxiv.org/abs/2509.22023

【5】PreLoRA: Hybrid Pre-training of Vision Transformers with Full Training and Low-Rank Adapters
标题：PreLoRA：Vision Transformers与完整训练和低级别适配器的混合预训练
链接：https://arxiv.org/abs/2509.21619

【6】From Embeddings to Equations: Genetic-Programming Surrogates for Interpretable Transformer Classification
标题：从嵌入到方程：可解释Transformer分类的遗传编程替代品
链接：https://arxiv.org/abs/2509.21341

【7】Accurate typhoon intensity forecasts using a non-iterative spatiotemporal transformer model
标题：使用非迭代时空Transformer模型准确预测台风强度
链接：https://arxiv.org/abs/2509.21349

GAN|对抗|攻击|生成相关(17篇)

【1】Training-Free Synthetic Data Generation with Dual IP-Adapter Guidance
标题：具有双IP适配器引导的免训练合成数据生成
链接：https://arxiv.org/abs/2509.22635

【2】JointDiff: Bridging Continuous and Discrete in Multi-Agent Trajectory Generation
标题：JointDiff：在多智能体轨迹生成中弥合连续和离散
链接：https://arxiv.org/abs/2509.22522

【3】SurvDiff: A Diffusion Model for Generating Synthetic Data in Survival Analysis
标题：SurvDiff：生存分析中生成合成数据的扩散模型
链接：https://arxiv.org/abs/2509.22352

【4】NIFTY: a Non-Local Image Flow Matching for Texture Synthesis
标题：NIFTY：一种用于纹理合成的非局部图像流匹配
链接 ：https://arxiv.org/abs/2509.22318

【5】DragGANSpace: Latent Space Exploration and Control for GANs
标题：DragGAN Space：GAN的潜在空间探索和控制
链接：https://arxiv.org/abs/2509.22169

【6】Countering adversarial evasion in regression analysis
标题：回归分析中对抗对抗规避
链接：https://arxiv.org/abs/2509.22113

【7】Non-Linear Trajectory Modeling for Multi-Step Gradient Inversion Attacks in Federated Learning
标题：联邦学习中多步梯度逆攻击的非线性轨迹建模
链接：https://arxiv.org/abs/2509.22082

【8】Generation Properties of Stochastic Interpolation under Finite Training Set
标题：有限训练集下随机插值的生成性质
链接：https://arxiv.org/abs/2509.21925

【9】MolSpectLLM: A Molecular Foundation Model Bridging Spectroscopy, Molecule Elucidation, and 3D Structure Generation
标题：MolSpectLLM：连接光谱学、分子解析和3D结构生成的分子基础模型
链接：https://arxiv.org/abs/2509.21861

【10】Noise-to-Notes: Diffusion-based Generation and Refinement for Automatic Drum Transcription
标题：噪音笔记：基于扩散的自动鼓转录生成和细化
链接：https://arxiv.org/abs/2509.21739

【11】SpecMER: Fast Protein Generation with K-mer Guided Speculative Decoding
标题：SpecMER：利用K-mer引导的推测解码快速蛋白质生成
链接：https://arxiv.org/abs/2509.21689

【12】Generating Stable Placements via Physics-guided Diffusion Models
标题：通过物理引导的扩散模型生成稳定布局
链接：https://arxiv.org/abs/2509.21664

【13】What Happens Next? Anticipating Future Motion by Generating Point Trajectories
标题：接下来会发生什么？通过生成点轨迹预测未来运动
链接：https://arxiv.org/abs/2509.21592

【14】No Alignment Needed for Generation: Learning Linearly Separable Representations in Diffusion Models
标题：生成时不需要对齐：学习扩散模型中的线性可分离表示
链接：https://arxiv.org/abs/2509.21565

【15】DistillKac: Few-Step Image Generation via Damped Wave Equations
标题：DistillKac：通过衰减波方程的少步图像生成
链接：https://arxiv.org/abs/2509.21513

【16】ReGeS: Reciprocal Retrieval-Generation Synergy for Conversational Recommender Systems
标题：ReGeS：对话式推荐系统的交互检索生成协同作用
链接：https://arxiv.org/abs/2509.21371

【17】Comparative Analysis of GAN and Diffusion for MRI-to-CT translation
标题：核磁共振转CT转换的GAN和扩散比较分析
链接：https://arxiv.org/abs/2509.22049

半/弱/无/有监督|不确定性|主动学习(11篇)

【1】IA2: Alignment with ICL Activations Improves Supervised Fine-Tuning
标题：IA 2：与ICL激活保持一致改进了监督微调
链接：https://arxiv.org/abs/2509.22621

【2】Learn the Ropes, Then Trust the Wins: Self-imitation with Progressive Exploration for Agentic Reinforcement Learning
标题：学习绳索，然后相信胜利：自我模仿，并进行渐进式探索，以实现抽象强化学习
链接：https://arxiv.org/abs/2509.22601

【3】From Parameters to Behavior: Unsupervised Compression of the Policy Space
标题：从参数到行为：政策空间的无监督压缩
链接：https://arxiv.org/abs/2509.22566

【4】Clinical Uncertainty Impacts Machine Learning Evaluations
标题：临床不确定性影响机器学习评估
链接：https://arxiv.org/abs/2509.22242

【5】GenUQ: Predictive Uncertainty Estimates via Generative Hyper-Networks
标题：GenUQ：通过生成超网络进行预测不确定性估计
链接 ：https://arxiv.org/abs/2509.21605

【6】Machine Learning. The Science of Selection under Uncertainty
标题：机器学习不确定性下的选择科学
链接：https://arxiv.org/abs/2509.21547

【7】TRiCo: Triadic Game-Theoretic Co-Training for Robust Semi-Supervised Learning
标题：TRiCo：用于稳健半监督学习的三重游戏理论联合训练
链接：https://arxiv.org/abs/2509.21526

【8】Uncertainty-Aware Knowledge Tracing Models
标题：不确定性感知知识追踪模型
链接：https://arxiv.org/abs/2509.21514

【9】Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)
标题：具有真实数据监督的匹配模型的通用逆蒸馏（无GAN）
链接：https://arxiv.org/abs/2509.22459

【10】Multidimensional Uncertainty Quantification via Optimal Transport
标题：通过最优运输进行多维不确定性量化
链接：https://arxiv.org/abs/2509.22380

【11】Causal-EPIG: A Prediction-Oriented Active Learning Framework for CATE Estimation
标题：CASEARCH-EPIG：用于CATE估计的面向预测的主动学习框架
链接：https://arxiv.org/abs/2509.21866

迁移|Zero/Few/One-Shot|自适应(10篇)

【1】Adaptive Dual-Mode Distillation with Incentive Schemes for Scalable, Heterogeneous Federated Learning on Non-IID Data
标题：具有激励计划的自适应双模式蒸馏，用于非IID数据上的可扩展、异类联邦学习
链接：https://arxiv.org/abs/2509.22507

【2】Adaptive Policy Backbone via Shared Network
标题：通过共享网络自适应政策主干
链接：https://arxiv.org/abs/2509.22310

【3】Towards Understanding Feature Learning in Parameter Transfer
标题：了解参数传输中的特征学习
链接：https://arxiv.org/abs/2509.22056

【4】Task-Adaptive Parameter-Efficient Fine-Tuning for Weather Foundation Models
标题：天气基础模型的任务自适应参数高效微调
链接：https://arxiv.org/abs/2509.22020

【5】GRAM-TDI: adaptive multimodal representation learning for drug target interaction prediction
标题：GRAM-LTD：用于药物靶点相互作用预测的自适应多模式表示学习
链接：https://arxiv.org/abs/2509.21971

【6】Zubov-Net: Adaptive Stability for Neural ODEs Reconciling Accuracy with Robustness
标题：Zubov-Net：神经ODE的自适应稳定性，兼顾准确性和鲁棒性
链接：https://arxiv.org/abs/2509.21879

【7】MMPlanner: Zero-Shot Multimodal Procedural Planning with Chain-of-Thought Object State Reasoning
标题：MMPlanner：具有思想链对象状态推理的零拍摄多模态程序规划
链接：https://arxiv.org/abs/2509.21662

【8】DyME: Dynamic Multi-Concept Erasure in Diffusion Models with Bi-Level Orthogonal LoRA Adaptation
标题：DyME：具有双层垂直LoRA自适应的扩散模型中的动态多概念擦除
链接：https://arxiv.org/abs/2509.21433

【9】SADA: Safe and Adaptive Inference with Multiple Black-Box Predictions
标题：SADA：具有多个黑匣子预测的安全和适应性推理
链接：https://arxiv.org/abs/2509.21707

【10】Effective continuous equations for adaptive SGD: a stochastic analysis view
标题：自适应新元的有效连续方程：随机分析观点
链接：https://arxiv.org/abs/2509.21614

强化学习(8篇)

【1】Fairness-Aware Reinforcement Learning (FAReL): A Framework for Transparent and Balanced Sequential Decision-Making
标题：公平意识强化学习（FAReL）：透明、平衡的顺序决策框架
链接：https://arxiv.org/abs/2509.22232

【2】Reinforcement Learning for Durable Algorithmic Recourse
标题：强化学习实现持久的脑震荡求助
链接：https://arxiv.org/abs/2509.22102

【3】Structural Information-based Hierarchical Diffusion for Offline Reinforcement Learning
标题：基于结构信息的离线强化学习分层扩散
链接：https://arxiv.org/abs/2509.21942

【4】Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards
标题：位置：具有可验证奖励的强化学习的隐性成本和测量差距
链接：https://arxiv.org/abs/2509.21882

【5】Preference-Guided Learning for Sparse-Reward Multi-Agent Reinforcement Learning
标题：稀疏回报多智能体强化学习的偏好引导学习
链接：https://arxiv.org/abs/2509.21828

【6】Reinforcement Learning Based Traffic Signal Design to Minimize Queue Lengths
标题：基于强化学习的排队时间最小化交通信号灯设计
链接：https://arxiv.org/abs/2509.21745

【7】POLO: Preference-Guided Multi-Turn Reinforcement Learning for Lead Optimization
标题：POLO：用于潜在客户优化的偏好引导多回合强化学习
链接：https://arxiv.org/abs/2509.21737

【8】Towards Efficient Online Exploration for Reinforcement Learning with Human Feedback
标题：通过人类反馈实现强化学习的高效在线探索
链接：https://arxiv.org/abs/2509.22633

元学习(1篇)

【1】Enhancing Credit Risk Prediction: A Meta-Learning Framework Integrating Baseline Models, LASSO, and ECOC for Superior Accuracy
标题：增强信用风险预测：集成基线模型、LASO和ECOC以实现卓越准确性的元学习框架
链接：https://arxiv.org/abs/2509.22381

符号|符号学习(1篇)

【1】Beyond Formula Complexity: Effective Information Criterion Improves Performance and Interpretability for Symbolic Regression
标题：超越公式复杂性：有效的信息标准提高符号回归的性能和可解释性
链接：https://arxiv.org/abs/2509.21780

医学相关(4篇)

【1】Integrating Background Knowledge in Medical Semantic Segmentation with Logic Tensor Networks
标题：与逻辑张量网络集成医学语义分割中的背景知识
链接：https://arxiv.org/abs/2509.22399

【2】Expert-guided Clinical Text Augmentation via Query-Based Model Collaboration
标题：通过基于查询的模型协作进行专家指导的临床文本增强
链接：https://arxiv.org/abs/2509.21530

【3】The LongiMam model for improved breast cancer risk prediction using longitudinal mammograms
标题：LongiMam模型使用纵向乳房X光检查改善乳腺癌风险预测
链接：https://arxiv.org/abs/2509.21383

【4】COMPASS: Robust Feature Conformal Prediction for Medical Segmentation Metrics
标题：康普斯：医疗分割子索的稳健特征保形预测
链接：https://arxiv.org/abs/2509.22240

蒸馏|知识提取(2篇)

【1】Enriching Knowledge Distillation with Intra-Class Contrastive Learning
标题：通过课内对比学习丰富知识提炼
链接：https://arxiv.org/abs/2509.22053

【2】Score-based Idempotent Distillation of Diffusion Models
标题：基于分数的扩散势等蒸馏模型
链接：https://arxiv.org/abs/2509.21470

聚类(1篇)

【1】Scalable Second-order Riemannian Optimization for $K$-means Clustering
标题：$K$-均值集群的可扩展二阶Riemann优化
链接：https://arxiv.org/abs/2509.21675

超分辨率|去噪|去模糊|去雾(3篇)

【1】Conditional Denoising Diffusion Autoencoders for Wireless Semantic Communications
标题：无线语义通信的条件去噪扩散自动编码器
链接：https://arxiv.org/abs/2509.22282

【2】RED-DiffEq: Regularization by denoising diffusion models for solving inverse PDE problems with application to full waveform inversion
标题：RD-DiffEq：通过去噪扩散模型进行正规化，用于解决逆偏东方程问题，并应用于全波逆
链接：https://arxiv.org/abs/2509.21659

【3】Downscaling climate projections to 1 km with single-image super resolution
标题：利用单图像超分辨率将气候预测缩减至1公里
链接：https://arxiv.org/abs/2509.21399

自动驾驶|车辆|车道检测等(2篇)

【1】Lifelong Learning with Behavior Consolidation for Vehicle Routing
标题：车辆路线行为整合的终身学习
链接：https://arxiv.org/abs/2509.21765

【2】Interpretable Spectral Features Predict Conductivity in Self-Driving Doped Conjugated Polymer Labs
标题：可解释的光谱特征预测自驱动掺合聚合物实验室的导电性
链接：https://arxiv.org/abs/2509.21330

点云|SLAM|雷达|激光|深度RGBD相关(1篇)

【1】Convexity-Driven Projection for Point Cloud Dimensionality Reduction
标题：凸度驱动投影用于点云凸度降低
链接：https://arxiv.org/abs/2509.22043

联邦学习|隐私保护|加密(3篇)

【1】Role-Aware Multi-modal federated learning system for detecting phishing webpages
标题：角色感知用于检测网络钓鱼网页的多模式联邦学习系统
链接：https://arxiv.org/abs/2509.22369

【2】PQFed: A Privacy-Preserving Quality-Controlled Federated Learning Framework
标题：PQFed：一个保护隐私的质量控制联邦学习框架
链接：https://arxiv.org/abs/2509.21704

【3】Task-Agnostic Federated Continual Learning via Replay-Free Gradient Projection
标题：基于无重放梯度投影的任务无关联邦连续学习
链接：https://arxiv.org/abs/2509.21606

推理|分析|理解|解释(20篇)

【1】A Theoretical Analysis of Discrete Flow Matching Generative Models
标题：离散流匹配生成模型的理论分析
链接：https://arxiv.org/abs/2509.22623

【2】Quantile Advantage Estimation for Entropy-Safe Reasoning
标题：用于信息安全推理的分位数优势估计
链接：https://arxiv.org/abs/2509.22611

【3】TrueGradeAI: Retrieval-Augmented and Bias-Resistant AI for Transparent and Explainable Digital Assessments
标题：TrueGradeAI：用于透明且可解释的数字评估的检索增强和抗偏见人工智能
链接：https://arxiv.org/abs/2509.22516

【4】A Machine Learning Pipeline for Multiple Sclerosis Biomarker Discovery: Comparing explainable AI and Traditional Statistical Approaches
标题：用于多发性硬化生物标志物发现的机器学习管道：比较可解释的人工智能和传统统计方法
链接：https://arxiv.org/abs/2509.22484

【5】Progressive Weight Loading: Accelerating Initial Inference and Gradually Boosting Performance on Resource-Constrained Environments
标题：渐进权重加载：加速初始推理并逐步提高资源受限环境中的绩效
链接：https://arxiv.org/abs/2509.22319

【6】GSM-Agent: Understanding Agentic Reasoning Using Controllable Environments
标题：GSM-Agent：使用可控环境理解统计推理
链接：https://arxiv.org/abs/2509.21998

【7】Outlier Detection in Plantar Pressure: Human-Centered Comparison of Statistical Parametric Mapping and Explainable Machine Learning
标题：脚底压力中的异常值检测：以人为本的统计参数映射和可解释机器学习的比较
链接：https://arxiv.org/abs/2509.21943

【8】Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts
标题：弹性MoE：解锁混合专家的推理时可扩展性
链接：https://arxiv.org/abs/2509.21892

【9】Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts
标题：思想检索：通过重用思想进行高效推理
链接：https://arxiv.org/abs/2509.21743

【10】A Systematic Review of Conformal Inference Procedures for Treatment Effect Estimation: Methods and Challenges
标题：治疗效果估计的共形推理方法的系统评价：方法与挑战
链接：https://arxiv.org/abs/2509.21660

【11】DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Models
标题：DriftLite：用于扩散模型推断时间缩放的轻量级漂移控制
链接：https://arxiv.org/abs/2509.21655

【12】Understanding and Enhancing Mask-Based Pretraining towards Universal Representations
标题：理解和增强基于面具的泛表示预训练
链接：https://arxiv.org/abs/2509.21650

【13】Causal Abstraction Inference under Lossy Representations
标题：有损表示下的因果抽象推理
链接：https://arxiv.org/abs/2509.21607

【14】Automated and Interpretable Survival Analysis from Multimodal Data
标题：来自多模式数据的自动化和可解释生存分析
链接：https://arxiv.org/abs/2509.21600

【15】Interpretable time series analysis with Gumbel dynamics
标题：使用Gumbel动力学的可解释时间序列分析
链接：https://arxiv.org/abs/2509.21578

【16】High-Probability Analysis of Online and Federated Zero-Order Optimisation
标题：在线和联邦零阶优化的高概率分析
链接：https://arxiv.org/abs/2509.21484

【17】Talking Trees: Reasoning-Assisted Induction of Decision Trees for Tabular Data
标题：会说话的树：表格数据决策树的推理辅助归纳
链接：https://arxiv.org/abs/2509.21465

【18】A State-of-the-Art SQL Reasoning Model using RLVR
标题：使用WLVR的最先进SQL推理模型
链接：https://arxiv.org/abs/2509.21459

【19】Improving Autism Detection with Multimodal Behavioral Analysis
标题：利用多模式行为分析改进自闭症检测
链接：https://arxiv.org/abs/2509.21352

【20】Error Analysis of Discrete Flow with Generator Matching
标题：发电机匹配离散流的误差分析
链接：https://arxiv.org/abs/2509.21906

检测相关(2篇)

【1】Exploring the Relationships Between Physiological Signals During Automated Fatigue Detection
标题：探索自动疲劳检测过程中生理信号之间的关系
链接：https://arxiv.org/abs/2509.21794

【2】Leveraging Big Data Frameworks for Spam Detection in Amazon Reviews
标题：在亚马逊评论中利用大数据框架进行垃圾邮件检测
链接：https://arxiv.org/abs/2509.21579

分类|识别(5篇)

【1】Machine learning approaches to seismic event classification in the Ostrava region
标题：俄斯特拉发地区地震事件分类的机器学习方法
链接：https://arxiv.org/abs/2509.22574

【2】(Sometimes) Less is More: Mitigating the Complexity of Rule-based Representation for Interpretable Classification
标题：（有时）少即是多：缓解基于规则的可解释分类表示的复杂性
链接：https://arxiv.org/abs/2509.22384

【3】Stage-wise Dynamics of Classifier-Free Guidance in Diffusion Models
标题：扩散模型中无分类器引导的阶段动力学
链接：https://arxiv.org/abs/2509.22007

【4】Object Identification Under Known Dynamics: A PIRNN Approach for UAV Classification
标题：已知动态下的目标识别：无人机分类的PIRNN方法
链接：https://arxiv.org/abs/2509.21405

【5】Spiking Neural Networks for Mental Workload Classification with a Multimodal Approach
标题：尖峰神经网络用于多模式方法的心理状态分类
链接：https://arxiv.org/abs/2509.21346

表征(6篇)

【1】Mechanistic Independence: A Principle for Identifiable Disentangled Representations
标题：机械独立性：可识别的非纠缠表示的一个原则
链接：https://arxiv.org/abs/2509.22196

【2】BrainPro: Towards Large-scale Brain State-aware EEG Representation Learning
标题：BrainPro：迈向大规模大脑状态感知的脑电表示学习
链接：https://arxiv.org/abs/2509.22050

【3】ChaosNexus: A Foundation Model for Universal Chaotic System Forecasting with Multi-scale Representations
标题：ChaosNexus：具有多尺度表示的普适混乱系统预测的基础模型
链接：https://arxiv.org/abs/2509.21802

【4】Contrastive Mutual Information Learning: Toward Robust Representations without Positive-Pair Augmentations
标题：对比互信息学习：迈向没有正对增强的鲁棒表示
链接：https://arxiv.org/abs/2509.21511

【5】Linear Causal Representation Learning by Topological Ordering, Pruning, and Disentanglement
标题：通过布局排序、修剪和解纠缠进行线性因果表示学习
链接：https://arxiv.org/abs/2509.22553

【6】IndiSeek learns information-guided disentangled representations
标题：IndiSeek学习信息引导的解开表示
链接：https://arxiv.org/abs/2509.21584

优化|敛散性(17篇)

【1】OFMU: Optimization-Driven Framework for Machine Unlearning
标题：OFMU：机器取消学习的优化驱动框架
链接：https://arxiv.org/abs/2509.22483

【2】Nonlinear Optimization with GPU-Accelerated Neural Network Constraints
标题：具有GOP加速神经网络约束的非线性优化
链接：https://arxiv.org/abs/2509.22462

【3】Global Convergence in Neural ODEs: Impact of Activation Functions
标题：神经ODE的全球收敛：激活函数的影响
链接：https://arxiv.org/abs/2509.22436

【4】Distributed Associative Memory via Online Convex Optimization
标题：通过在线凸优化的分布式联想记忆
链接：https://arxiv.org/abs/2509.22321

【5】Efficiency Boost in Decentralized Optimization: Reimagining Neighborhood Aggregation with Minimal Overhead
标题：去中心化优化的效率提升：以最小的成本重新构想邻里聚集
链接：https://arxiv.org/abs/2509.22174

【6】Slicing Wasserstein Over Wasserstein Via Functional Optimal Transport
标题：通过功能性最优运输将沃瑟斯坦切片到沃瑟斯坦
链接：https://arxiv.org/abs/2509.22138

【7】Learning More with Less: A Dynamic Dual-Level Down-Sampling Framework for Efficient Policy Optimization
标题：少花钱多学：用于高效政策优化的动态双层下采样框架
链接：https://arxiv.org/abs/2509.22115

【8】MO-GRPO: Mitigating Reward Hacking of Group Relative Policy Optimization on Multi-Objective Problems
标题：MO-GRPO：减轻多目标问题群体相对政策优化的奖励黑客
链接：https://arxiv.org/abs/2509.22047

【9】FastGRPO: Accelerating Policy Optimization via Concurrency-aware Speculative Decoding and Online Draft Learning
标题：FastGRPO：通过并发感知的推测解码和在线草稿学习加速政策优化
链接：https://arxiv.org/abs/2509.21792

【10】Information-Theoretic Bayesian Optimization for Bilevel Optimization Problems
标题：二层优化问题的信息论Bayesian优化
链接：https://arxiv.org/abs/2509.21725

【11】Align2Speak: Improving TTS for Low Resource Languages via ASR-Guided Online Preference Optimization
标题：Alignn 2Speak：通过ASB引导的在线偏好优化改进低资源语言的TTC
链接：https://arxiv.org/abs/2509.21718

【12】New Algorithmic Directions in Optimal Transport and Applications for Product Spaces
标题：产品空间最优传输和应用的新数学方向
链接：https://arxiv.org/abs/2509.21502

【13】Preventing Model Collapse Under Overparametrization: Optimal Mixing Ratios for Interpolation Learning and Ridge Regression
标题：防止过度参数化下的模型崩溃：内插学习和岭回归的最佳混合比
链接：https://arxiv.org/abs/2509.22341

【14】A Random Matrix Perspective of Echo State Networks: From Precise Bias--Variance Characterization to Optimal Regularization
标题：回声状态网络的随机矩阵视角：从精确偏差--方差特征到最佳正规化
链接：https://arxiv.org/abs/2509.22011

【15】Sequential 1-bit Mean Estimation with Near-Optimal Sample Complexity
标题：具有近最优样本复杂度的序列1位均值估计
链接：https://arxiv.org/abs/2509.21940

【16】Automating Sensor Characterization with Bayesian Optimization
标题：使用Bayesian优化自动化传感器特征
链接：https://arxiv.org/abs/2509.21661

【17】Near-Optimal Experiment Design in Linear non-Gaussian Cyclic Models
标题：线性非高斯循环模型中的近优试验设计
链接：https://arxiv.org/abs/2509.21423

预测|估计(13篇)

【1】Learning from Delayed Feedback in Games via Extra Prediction
标题：通过额外预测从游戏中的延迟反馈中学习
链接：https://arxiv.org/abs/2509.22426

【2】NeuroScalar: A Deep Learning Framework for Fast, Accurate, and In-the-Wild Cycle-Level Performance Prediction
标题：NeuroScalar：一个深度学习框架，用于快速、准确和野外周期级性能预测
链接：https://arxiv.org/abs/2509.22410

【3】Improving accuracy in short mortality rate series: Exploring Multi-step Forecasting Approaches in Hybrid Systems
标题：提高短死亡率系列的准确性：探索混合系统中的多步预测方法
链接：https://arxiv.org/abs/2509.22395

【4】Aurora: Towards Universal Generative Multimodal Time Series Forecasting
标题：Aurora：走向通用生成多峰时间序列预测
链接：https://arxiv.org/abs/2509.22295

【5】Learnable Conformal Prediction with Context-Aware Nonconformity Functions for Robotic Planning and Perception
标题：用于机器人规划和感知的具有上下文感知非一致性函数的可学习共形预测
链接：https://arxiv.org/abs/2509.21955

【6】Wav2Arrest 2.0: Long-Horizon Cardiac Arrest Prediction with Time-to-Event Modeling, Identity-Invariance, and Pseudo-Lab Alignment
标题：Wav2Arrest 2.0：具有事件时间建模、身份不变性和伪实验室对齐的长视野心脏骤停预测
链接：https://arxiv.org/abs/2509.21695

【7】Filtering with Confidence: When Data Augmentation Meets Conformal Prediction
标题：自信地过滤：当数据增强满足保形预测时
链接：https://arxiv.org/abs/2509.21479

【8】Forecasting Seismic Waveforms: A Deep Learning Approach for Einstein Telescope
标题：预测地震波：爱因斯坦望远镜的深度学习方法
链接：https://arxiv.org/abs/2509.21446

【9】Smoothing-Based Conformal Prediction for Balancing Efficiency and Interpretability
标题：基于平滑的保形预测平衡效率和可解释性
链接：https://arxiv.org/abs/2509.22529

【10】CausalKANs: interpretable treatment effect estimation with Kolmogorov-Arnold networks
标题：Cairmogorov-Arnold网络的可解释治疗效果估计
链接：https://arxiv.org/abs/2509.22467

【11】Direct Bias-Correction Term Estimation for Propensity Scores and Average Treatment Effect Estimation
标题：倾向评分的直接偏差修正项估计和平均治疗效果估计
链接：https://arxiv.org/abs/2509.22122

【12】Multi-modal Bayesian Neural Network Surrogates with Conjugate Last-Layer Estimation
标题：多峰Bayesian神经网络的耦合最后层估计
链接：https://arxiv.org/abs/2509.21711

【13】Assessment of deep learning models integrated with weather and environmental variables for wildfire spread prediction and a case study of the 2023 Maui fires
标题：评估与天气和环境变量相结合的深度学习模型以预测野火蔓延以及2023年毛伊岛火灾案例研究
链接：https://arxiv.org/abs/2509.21327

其他神经网络|深度学习|模型|建模(54篇)

【1】See, Point, Fly: A Learning-Free VLM Framework for Universal Unmanned Aerial Navigation
标题：看、点、飞：通用无人空中导航的免学习VLM框架
链接：https://arxiv.org/abs/2509.22653

【2】Learning Admissible Heuristics for A*: Theory and Practice
标题：学习A* 的可接受启发式方法：理论与实践
链接：https://arxiv.org/abs/2509.22626

【3】From Formal Language Theory to Statistical Learning: Finite Observability of Subregular Languages
标题：从形式语言理论到统计学习：次规则语言的有限可观察性
链接：https://arxiv.org/abs/2509.22598

【4】Effective Policy Learning for Multi-Agent Online Coordination Beyond Submodular Objectives
标题：超越子模块目标的多主体在线协调的有效政策学习
链接：https://arxiv.org/abs/2509.22596

【5】Transport Based Mean Flows for Generative Modeling
标题：生成式建模的基于运输的平均流量
链接：https://arxiv.org/abs/2509.22592

【6】The Lie of the Average: How Class Incremental Learning Evaluation Deceives You?
标题：平均值的谎言：班级增量学习评估如何欺骗你？
链接：https://arxiv.org/abs/2509.22580

【7】Activation Function Design Sustains Plasticity in Continual Learning
标题：激活功能设计维持持续学习的可塑性
链接：https://arxiv.org/abs/2509.22562

【8】Learning to Price Bundles: A GCN Approach for Mixed Bundling
标题：学习为捆绑包定价：混合捆绑的GCN方法
链接：https://arxiv.org/abs/2509.22557

【9】ECHO: Toward Contextual Seq2Seq Paradigms in Large EEG Models
标题：ECHO：大型EEG模型中的上下文Seq2Seq范式
链接：https://arxiv.org/abs/2509.22556

【10】Overclocking Electrostatic Generative Models
标题：静电生成模型
链接：https://arxiv.org/abs/2509.22454

【11】Learning to Ball: Composing Policies for Long-Horizon Basketball Moves
标题：学习投球：制定长期篮球动作政策
链接：https://arxiv.org/abs/2509.22442

【12】Fast-Forward Lattice Boltzmann: Learning Kinetic Behaviour with Physics-Informed Neural Operators
标题：快进格子Boltzmann：使用了解物理的神经运算符学习动力学行为
链接：https://arxiv.org/abs/2509.22411

【13】ReLAM: Learning Anticipation Model for Rewarding Visual Robotic Manipulation
标题：ReLAM：奖励视觉机器人操纵的学习预期模型
链接：https://arxiv.org/abs/2509.22402

【14】Context and Diversity Matter: The Emergence of In-Context Learning in World Models
标题：背景和多样性很重要：世界模型中上下文学习的出现
链接：https://arxiv.org/abs/2509.22353

【15】Spectral Collapse Drives Loss of Plasticity in Deep Continual Learning
标题：光谱崩溃导致深度连续学习中可塑性丧失
链接：https://arxiv.org/abs/2509.22335

【16】HiGS: History-Guided Sampling for Plug-and-Play Enhancement of Diffusion Models
标题：HiGS：历史引导采样，用于即插即用增强扩散模型
链接：https://arxiv.org/abs/2509.22300

【17】Structured Sparse Transition Matrices to Enable State Tracking in State-Space Models
标题：结构化稀疏转移矩阵可在状态空间模型中实现状态跟踪
链接：https://arxiv.org/abs/2509.22284

【18】Towards a more realistic evaluation of machine learning models for bearing fault diagnosis
标题：对用于轴承故障诊断的机器学习模型进行更现实的评估
链接：https://arxiv.org/abs/2509.22267

【19】Learning Equivariant Functions via Quadratic Forms
标题：通过二次形式学习等变函数
链接：https://arxiv.org/abs/2509.22184

【20】Modeling Psychological Profiles in Volleyball via Mixed-Type Bayesian Networks
标题：基于混合型Bayesian网络的排球心理特征建模
链接：https://arxiv.org/abs/2509.22111

【21】MCGM: Multi-stage Clustered Global Modeling for Long-range Interactions in Molecules
标题：MCGM：分子长程相互作用的多级离散全局模拟
链接：https://arxiv.org/abs/2509.22028

【22】Concept-SAE: Active Causal Probing of Visual Model Behavior
标题：概念-AE：视觉模型行为的主动因果探索
链接：https://arxiv.org/abs/2509.22015

【23】Bilinear relational structure fixes reversal curse and enables consistent model editing
标题：双线性关系结构修复了逆转诅咒并实现一致的模型编辑
链接：https://arxiv.org/abs/2509.21993

【24】Multiplicative-Additive Constrained Models:Toward Joint Visualization of Interactive and Independent Effects
标题：乘加约束模型：迈向交互和独立效应的联合可视化
链接：https://arxiv.org/abs/2509.21923

【25】Closing the Oracle Gap: Increment Vector Transformation for Class Incremental Learning
标题：缩小Oracle差距：课堂增量学习的增量向量转换
链接：https://arxiv.org/abs/2509.21898

【26】Why High-rank Neural Networks Generalize?: An Algebraic Framework with RKHSs
标题：为什么要推广高级神经网络？：具有RKHS的代数框架
链接：https://arxiv.org/abs/2509.21895

【27】Beyond RAG vs. Long-Context: Learning Distraction-Aware Retrieval for Efficient Knowledge Grounding
标题：超越RAG与长上下文：学习注意力分散检索以实现高效的知识基础
链接：https://arxiv.org/abs/2509.21865

【28】Scaling Laws for Neural Material Models
标题：神经材料模型的比例定律
链接：https://arxiv.org/abs/2509.21811

【29】Machine Learning and AI Applied to fNIRS Data Reveals Novel Brain Activity Biomarkers in Stable Subclinical Multiple Sclerosis
标题：应用于fNIRS数据的机器学习和人工智能揭示了稳定亚临床多发性硬化症中的新型大脑活动生物标志物
链接：https://arxiv.org/abs/2509.21770

【30】HyperCore: Coreset Selection under Noise via Hypersphere Models
标题：HyperCore：基于超球模型的噪声下核集选择
链接：https://arxiv.org/abs/2509.21746

【31】Brain PathoGraph Learning
标题：脑病理图学习
链接：https://arxiv.org/abs/2509.21742

【32】A Unifying Framework for Parallelizing Sequential Models with Linear Dynamical Systems
标题：线性动态系统序列模型并行化的统一框架
链接：https://arxiv.org/abs/2509.21716

【33】Downscaling human mobility data based on demographic socioeconomic and commuting characteristics using interpretable machine learning methods
标题：使用可解释的机器学习方法基于人口社会经济和通勤特征缩减人类流动性数据
链接：https://arxiv.org/abs/2509.21703

【34】SlotFM: A Motion Foundation Model with Slot Attention for Diverse Downstream Tasks
标题：SlotFM：一个具有插槽关注度的Motion Foundation模型，用于多样化的下游任务
链接：https://arxiv.org/abs/2509.21673

【35】MORPH: Shape-agnostic PDE Foundation Models
标题：MORPH：形状不可知的PED基础模型
链接：https://arxiv.org/abs/2509.21670

【36】DIM: Enforcing Domain-Informed Monotonicity in Deep Neural Networks
标题：DIM：在深度神经网络中强制域信息单调性
链接：https://arxiv.org/abs/2509.21666

【37】Differentiable Structure Learning for General Binary Data
标题：一般二进制数据的可区分结构学习
链接：https://arxiv.org/abs/2509.21658

【38】LANCE: Low Rank Activation Compression for Efficient On-Device Continual Learning
标题：LANCE：低秩激活压缩，用于有效的设备上持续学习
链接：https://arxiv.org/abs/2509.21617

【39】SlimDiff: Training-Free, Activation-Guided Hands-free Slimming of Diffusion Models
标题：SlimDiff：无需训练、激活引导的免提瘦身扩散模型
链接：https://arxiv.org/abs/2509.21498

【40】Functional Encryption in Secure Neural Network Training: Data Leakage and Practical Mitigations
标题：安全神经网络训练中的函数加密：数据泄漏和实际缓解措施
链接：https://arxiv.org/abs/2509.21497

【41】Context-Aware Hybrid Routing in Bluetooth Mesh Networks Using Multi-Model Machine Learning and AODV Fallback
标题：基于多模型机器学习和AODV回退的蓝牙Mesh网络上下文感知混合路由
链接：https://arxiv.org/abs/2509.21490

【42】Neural Operators for Mathematical Modeling of Transient Fluid Flow in Subsurface Reservoir Systems
标题：地下储层系统中非稳定流体流动数学建模的神经运算符
链接：https://arxiv.org/abs/2509.21485

【43】Learning to Reason with Mixture of Tokens
标题：学习使用混合代币推理
链接：https://arxiv.org/abs/2509.21482

【44】Null-Space Filtering for Data-Free Continual Model Merging: Preserving Transparency, Promoting Fidelity
标题：无数据连续模型合并的零空间过滤：保持透明度，促进保真度
链接：https://arxiv.org/abs/2509.21413

【45】Impact of Loss Weight and Model Complexity on Physics-Informed Neural Networks for Computational Fluid Dynamics
标题：失重和模型复杂性对计算流体力学物理信息神经网络的影响
链接：https://arxiv.org/abs/2509.21393

【46】Debugging Concept Bottleneck Models through Removal and Retraining
标题：通过移除和再训练的概念瓶颈模型
链接：https://arxiv.org/abs/2509.21385

【47】ConQuER: Modular Architectures for Control and Bias Mitigation in IQP Quantum Generative Models
标题：ConQuER：IQP量子生成模型中控制和偏差缓解的模块化架构
链接：https://arxiv.org/abs/2509.22551

【48】Metrics for Parametric Families of Networks
标题：网络参数族的收件箱
链接：https://arxiv.org/abs/2509.22549

【49】Incorporating priors in learning: a random matrix study under a teacher-student framework
标题：总结学习中的先验：师生框架下的随机矩阵研究
链接：https://arxiv.org/abs/2509.22124

【50】Exploring the Early Universe with Deep Learning
标题：利用深度学习探索早期宇宙
链接：https://arxiv.org/abs/2509.22018

【51】A Nonparametric Discrete Hawkes Model with a Collapsed Gaussian-Process Prior
标题：具有高斯过程先验的非参数离散Hawkes模型
链接：https://arxiv.org/abs/2509.21996

【52】HuLA: Prosody-Aware Anti-Spoofing with Multi-Task Learning for Expressive and Emotional Synthetic Speech
标题：HuLA：具有多任务学习的韵律感知反欺骗，用于表达性和情感合成语音
链接：https://arxiv.org/abs/2509.21676

【53】Foundation models for high-energy physics
标题：高能物理基础模型
链接：https://arxiv.org/abs/2509.21434

【54】Seismic Velocity Inversion from Multi-Source Shot Gathers Using Deep Segmentation Networks: Benchmarking U-Net Variants and SeismoLabV3+
标题：使用深度分割网络从多源炮点聚集进行地震速度倒置：U-Net变体和SeismoLabV 3+基准测试
链接：https://arxiv.org/abs/2509.21331

其他(58篇)

【1】Scale-Wise VAR is Secretly Discrete Diffusion
标题：规模VAR是秘密离散的扩散
链接：https://arxiv.org/abs/2509.22636

【2】StateX: Enhancing RNN Recall via Post-training State Expansion
标题：StateX：通过训练后状态扩展增强RNN召回
链接：https://arxiv.org/abs/2509.22630

【3】SPARK: Synergistic Policy And Reward Co-Evolving Framework
标题：SPARK：协同政策和奖励共同演变框架
链接：https://arxiv.org/abs/2509.22624

【4】Nearly Tight Regret Bounds for Profit Maximization in Bilateral Trade
标题：双边贸易利润最大化几乎是一丝遗憾
链接：https://arxiv.org/abs/2509.22563

【5】Dual Optimistic Ascent (PI Control) is the Augmented Lagrangian Method in Disguise
标题：双重乐观上升（PI控制）是伪装的增广拉格朗日方法
链接：https://arxiv.org/abs/2509.22500

【6】Bayesian Transfer Operators in Reproducing Kernel Hilbert Spaces
标题：再生核Hilbert空间中的Bayesian转移运算符
链接：https://arxiv.org/abs/2509.22482

【7】Physics-informed GNN for medium-high voltage AC power flow with edge-aware attention and line search correction operator
标题：基于物理知识的GNN，用于中高压交流潮流，具有边缘感知的注意力和线路搜索纠正操作员
链接：https://arxiv.org/abs/2509.22458

【8】The Flood Complex: Large-Scale Persistent Homology on Millions of Points
标题：洪水综合体：数百万个点上的大规模持续同质性
链接：https://arxiv.org/abs/2509.22432

【9】Partial Parameter Updates for Efficient Distributed Training
标题：部分参数更新以实现高效分布式训练
链接：https://arxiv.org/abs/2509.22418

【10】Neural Feature Geometry Evolves as Discrete Ricci Flow
标题：神经特征几何作为离散Ricci流进化
链接：https://arxiv.org/abs/2509.22362

【11】Stochastic activations
标题：随机激活
链接：https://arxiv.org/abs/2509.22358

【12】HEAPr: Hessian-based Efficient Atomic Expert Pruning in Output Space
标题：HEAPr：输出空间中基于黑森的高效原子专家修剪
链接：https://arxiv.org/abs/2509.22299

【13】A Multi-Level Framework for Multi-Objective Hypergraph Partitioning: Combining Minimum Spanning Tree and Proximal Gradient
标题：多目标超图划分的多层框架：结合最小生成树和近端梯度
链接：https://arxiv.org/abs/2509.22294

【14】Unlocking the Power of Mixture-of-Experts for Task-Aware Time Series Analytics
标题：释放混合专家的力量用于任务感知时间序列分析
链接：https://arxiv.org/abs/2509.22279

【15】Erase or Hide? Suppressing Spurious Unlearning Neurons for Robust Unlearning
标题：隐藏还是隐藏？抑制伪去学习神经元以实现稳健去学习
链接：https://arxiv.org/abs/2509.22263

【16】ASSESS: A Semantic and Structural Evaluation Framework for Statement Similarity
标题：ASSESS：陈述相似性的语义和结构评估框架
链接：https://arxiv.org/abs/2509.22246

【17】Automatic Discovery of One Parameter Subgroups of $SO(n)$
标题：$SO（n）$的单参数子集的自动发现
链接：https://arxiv.org/abs/2509.22219

【18】A Law of Data Reconstruction for Random Features (and Beyond)
标题：随机特征（及其他）的数据重建定律
链接：https://arxiv.org/abs/2509.22214

【19】Reversible GNS for Dissipative Fluids with Consistent Bidirectional Dynamics
标题：具有一致双向动力学的消散流体的可逆GNS
链接：https://arxiv.org/abs/2509.22207

【20】Pushing Toward the Simplex Vertices: A Simple Remedy for Code Collapse in Smoothed Vector Quantization
标题：推向单纯形点：平滑载体量化中代码崩溃的简单补救措施
链接：https://arxiv.org/abs/2509.22161

【21】Latent Diffusion : Multi-Dimension Stable Diffusion Latent Space Explorer
标题：潜在扩散：多维稳定扩散潜在太空探索者
链接：https://arxiv.org/abs/2509.22038

【22】OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features
标题：OrtSAE：正交稀疏自动编码器揭示原子特征
链接：https://arxiv.org/abs/2509.22033

【23】FlowDrive: moderated flow matching with data balancing for trajectory planning
标题：FlowDrive：通过数据平衡进行适度的流匹配，以实现轨迹规划
链接：https://arxiv.org/abs/2509.21961

【24】Statistical Advantage of Softmax Attention: Insights from Single-Location Regression
标题：Softmax Attention的统计优势：单位置回归的见解
链接：https://arxiv.org/abs/2509.21936

【25】Discrete Guidance Matching: Exact Guidance for Discrete Flow Matching
标题：离散引导匹配：离散流匹配的精确引导
链接：https://arxiv.org/abs/2509.21912

【26】Beyond Johnson-Lindenstrauss: Uniform Bounds for Sketched Bilinear Forms
标题：超越约翰逊-林登施特劳斯：草图双线性形式的均匀边界
链接：https://arxiv.org/abs/2509.21847

【27】On the Complexity Theory of Masked Discrete Diffusion: From $\mathrm{poly}(1/ε)$ to Nearly $ε$-Free
链接：https://arxiv.org/abs/2509.21835

【28】Sharpness-Aware Minimization Can Hallucinate Minimizers
标题：敏锐的最小化会让最小化者产生幻觉
链接：https://arxiv.org/abs/2509.21818

【29】CubistMerge: Spatial-Preserving Token Merging For Diverse ViT Backbones
标题：CubistMerge：空间保留代币合并，以实现多元化ViT主干
链接：https://arxiv.org/abs/2509.21764

【30】Reparameterizing 4DVAR with neural fields
标题：用神经场重新参数化四维变分
链接：https://arxiv.org/abs/2509.21751

【31】SubZeroCore: A Submodular Approach with Zero Training for Coreset Selection
标题：SubZeroCore：一种用于核心集选择的零训练子模块方法
链接：https://arxiv.org/abs/2509.21748

【32】Self-Speculative Biased Decoding for Faster Live Translation
标题：自推测偏置解码，实现更快的实时翻译
链接：https://arxiv.org/abs/2509.21740

【33】UISim: An Interactive Image-Based UI Simulator for Dynamic Mobile Environments
标题：UISim：用于动态移动环境的基于图像的交互式UI模拟器
链接：https://arxiv.org/abs/2509.21733

【34】Prophecy: Inferring Formal Properties from Neuron Activations
标题：预言：从神经元激活推断形式性质
链接：https://arxiv.org/abs/2509.21677

【35】Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli
标题：神经探针：评估颅内对自然刺激的反应
链接：https://arxiv.org/abs/2509.21671

【36】Logic of Hypotheses: from Zero to Full Knowledge in Neurosymbolic Integration
标题：假设的逻辑：神经符号整合中从零知识到全知识
链接：https://arxiv.org/abs/2509.21663

【37】Limitations on Safe, Trusted, Artificial General Intelligence
标题：安全、可信、人工通用智能的局限性
链接：https://arxiv.org/abs/2509.21654

【38】MobiLLM: An Agentic AI Framework for Closed-Loop Threat Mitigation in 6G Open RANs
标题：ARMLLM：一个用于6G开放RAN中闭环威胁缓解的实用AI框架
链接：https://arxiv.org/abs/2509.21634

【39】Shoot from the HIP: Hessian Interatomic Potentials without derivatives
标题：从HIPP拍摄：黑森原子间势不含衍生物
链接：https://arxiv.org/abs/2509.21624

【40】OjaKV: Context-Aware Online Low-Rank KV Cache Compression with Oja's Rule
标题：OjaKV：基于Oja规则的上下文感知在线低秩KV Cache压缩
链接：https://arxiv.org/abs/2509.21623

【41】VLCE: A Knowledge-Enhanced Framework for Image Description in Disaster Assessment
标题：VLCE：灾害评估中图像描述的知识增强框架
链接：https://arxiv.org/abs/2509.21609

【42】Domain-Aware Speaker Diarization On African-Accented English
标题：非裔英语中的领域意识发言者Dialogue
链接：https://arxiv.org/abs/2509.21554

【43】Agribot: agriculture-specific question answer system
标题：Agribot：农业特定问答系统
链接：https://arxiv.org/abs/2509.21535

【44】$\mathbf{Li_2}$: A Framework on Dynamics of Feature Emergence and Delayed Generalization
链接：https://arxiv.org/abs/2509.21519

【45】VISION: Prompting Ocean Vertical Velocity Reconstruction from Incomplete Observations
标题：愿景：从不完整的观测中重建海洋垂直速度
链接：https://arxiv.org/abs/2509.21477

【46】Are Hallucinations Bad Estimations?
标题：幻觉是不好的估计吗？
链接：https://arxiv.org/abs/2509.21473