|
创作新主题 |
社区所有版块导航
DATA
| docker Elasticsearch |
WEB开发
| linux MongoDB Redis DATABASE NGINX 其他Web框架 web工具 zookeeper tornado NoSql Bootstrap js peewee Git bottle IE MQ Jquery |
机器学习
| 机器学习算法 |
产品
| 短视频 |
印度
| 印度 |
一周十大热门主题
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445篇
大模型相关(74篇)
【1】Learning to Interpret Weight Differences in Language Models
标题：学习解释语言模型中的权重差异
链接：https://arxiv.org/abs/2510.05092
备注：The weight diffs and DIT adapters trained in the paper can be found at this https URL
【2】Test-Time Scaling in Diffusion LLMs via Hidden Semi-Autoregressive Experts
标题：通过隐藏半自回归专家在扩散LLM中进行测试时间缩放
链接：https://arxiv.org/abs/2510.05040
【3】Inoculation Prompting: Instructing LLMs to misbehave at train-time improves test-time alignment
标题：接种：指示LLM在训练时行为不端，改善测试时间一致性
链接：https://arxiv.org/abs/2510.05024
【4】Reinforce-Ada: An Adaptive Sampling Framework for Reinforce-Style LLM Training
标题：Reinforce-Ada：Reinforce式LLM训练的自适应抽样框架
链接：https://arxiv.org/abs/2510.04996
备注：16 pages, 6 figures
【5】Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (short paper)
标题：注意语气：调查迅速的礼貌如何影响法学硕士准确性（短文）
链接：https://arxiv.org/abs/2510.04950
备注：5 pages, 3 tables; includes Limitations and Ethical Considerations sections; short paper under submission to Findings of ACL 2025
【6】The Geometry of Truth: Layer-wise Semantic Dynamics for Hallucination Detection in Large Language Models
标题：真理的几何学：大型语言模型中幻觉检测的分层语义动力学
链接：https://arxiv.org/abs/2510.04933
备注：Comments: 14 pages, 14 figures, 5 tables. Code available at: this https URL
【7】HyperVLA: Efficient Inference in Vision-Language-Action Models via Hypernetworks
标题：HyperVLA：基于超网络的视觉-语言-动作模型的高效推理
链接：https://arxiv.org/abs/2510.04898
【8】SocialHarmBench: Revealing LLM Vulnerabilities to Socially Harmful Requests
标题：SocialHarmBench：揭示LLM对社会有害请求的漏洞
链接：https://arxiv.org/abs/2510.04891
【9】Revealing Interconnections between Diseases: from Statistical Methods to Large Language Models
标题：揭示疾病之间的相互联系：从统计方法到大型语言模型
链接：https://arxiv.org/abs/2510.04888
【10】RL Is a Hammer and LLMs Are Nails: A Simple Reinforcement Learning Recipe for Strong Prompt Injection
标题：RL是锤子，LLM是钉子：用于强即时注入的简单强化学习食谱
链接：https://arxiv.org/abs/2510.04885
【11】Alignment Tipping Process: How Self-Evolution Pushes LLM Agents Off the Rails
标题：对齐提示流程：自我进化如何推动LLM代理脱离正轨
链接：https://arxiv.org/abs/2510.04860
【12】LEGOMem: Modular Procedural Memory for Multi-agent LLM Systems for Workflow Automation
标题：LEGOMem：用于工作流程自动化的多代理LLM系统的模块化程序存储器
链接：https://arxiv.org/abs/2510.04851
【13】Distribution Preference Optimization: A Fine-grained Perspective for LLM Unlearning
标题：分布偏好优化：LLM遗忘的细粒度视角
链接：https://arxiv.org/abs/2510.04773
备注：20 pages
【14】ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMs
标题：ChelBench：了解扩散LLM中并行解码的权衡
链接：https://arxiv.org/abs/2510.04767
备注：Project Page: this https URL
【15】BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs
标题：BrokenMath：用LLM证明定理的谄媚基准
链接：https://arxiv.org/abs/2510.04721
【16】Evaluating LLMs for Demographic-Targeted Social Bias Detection: A Comprehensive Benchmark Study
标题：评估针对人口统计学目标的社会偏见检测的LLM：一项全面的基准研究
链接：https://arxiv.org/abs/2510.04641
备注：17 pages, 7 figures, 7 tables
【17】Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models
标题：抽象的上下文工程：为自我改进的语言模型不断进化的上下文
链接：https://arxiv.org/abs/2510.04618
【18】A Case for Declarative LLM-friendly Interfaces for Improved Efficiency of Computer-Use Agents
标题：声明性LLM友好界面以提高计算机使用代理效率的案例
链接：https://arxiv.org/abs/2510.04607
【19】Language Model Based Text-to-Audio Generation: Anti-Causally Aligned Collaborative Residual Transformers
标题：基于语言模型的文本到音频生成：反因果对齐的协作剩余变形器
链接：https://arxiv.org/abs/2510.04577
备注：Accepted to EMNLP 2025
【20】LaDiR: Latent Diffusion Enhances LLMs for Text Reasoning
标题：LaDiR：潜在扩散增强了文本推理的LLM
链接：https://arxiv.org/abs/2510.04573
【21】GILT: An LLM-Free, Tuning-Free Graph Foundational Model for In-Context Learning
标题：GILT：一个无LLM、无调谐的用于上下文学习的图基础模型
链接：https://arxiv.org/abs/2510.04567
【22】MedCLM: Learning to Localize and Reason via a CoT-Curriculum in Medical Vision-Language Models
标题：MedCLM：通过医学视觉语言模型中的CoT课程学习本地化和推理
链接：https://arxiv.org/abs/2510.04477
【23】SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations
标题：SECA：引发LLM幻觉的语义等效和一致攻击
链接：https://arxiv.org/abs/2510.04398
备注：Accepted at NeurIPS 2025. Code is available at this https URL
【24】Unmasking Backdoors: An Explainable Defense via Gradient-Attention Anomaly Scoring for Pre-trained Language Models
标题：揭开后门：通过预训练语言模型的学生注意力异常评分进行可解释的防御
链接：https://arxiv.org/abs/2510.04347
备注：15 pages total (9 pages main text + 4 pages appendix + references), 12 figures, preprint version. The final version may differ
【25】Read the Scene, Not the Script: Outcome-Aware Safety for LLMs
标题：阅读场景，而不是剧本：LLM的结果感知安全
链接：https://arxiv.org/abs/2510.04320
【26】FairAgent: Democratizing Fairness-Aware Machine Learning with LLM-Powered Agents
标题：FairAgent：通过LLM-Powered Agents使公平意识机器学习民主化
链接：https://arxiv.org/abs/2510.04317
备注：Accepted by ICDM 2025 Demo Workshop
【27】On the Importance of Task Complexity in Evaluating LLM-Based Multi-Agent Systems
标题：任务复杂性在基于LLM的多Agent系统评价中的重要性
链接：https://arxiv.org/abs/2510.04311
【28】Epistemic Diversity and Knowledge Collapse in Large Language Models
标题：大型语言模型中的认识多样性和知识崩溃
链接：https://arxiv.org/abs/2510.04226
备注：16 pages; 8 figures, 4 tables
【29】MLLMEraser: Achieving Test-Time Unlearning in Multimodal Large Language Models through Activation Steering
标题：MLLMEraser：通过激活转向实现多模态大型语言模型的测试时遗忘
链接：https://arxiv.org/abs/2510.04217
【30】Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models
标题：超越下一个代币预测：扩散语言模型与自回归语言模型的性能描述
链接：https://arxiv.org/abs/2510.04146
备注：11 pages, 5 figures
【31】Can Linear Probes Measure LLM Uncertainty?
标题：线性探针可以衡量LLM的不确定性吗？
链接：https://arxiv.org/abs/2510.04108
【32】Slow-Fast Policy Optimization: Reposition-Before-Update for LLM Reasoning
标题：慢-快策略优化：LLM推理的先重置后更新
链接：https://arxiv.org/abs/2510.04072
【33】Principled and Tractable RL for Reasoning with Diffusion Language Models
标题：用于使用扩散语言模型进行推理的有原则且可跟踪的RL
链接：https://arxiv.org/abs/2510.04019
【34】A Mathematical Explanation of Transformers for Large Language Models and GPTs
标题：大型语言模型和GPT转换器的数学解释
链接：https://arxiv.org/abs/2510.03989
【35】Distilling Reasoning into Student LLMs: Local Naturalness for Selecting Teacher Data
标题：将推理提炼到学生法学硕士中：选择教师数据的当地自然性
链接：https://arxiv.org/abs/2510.03988
备注：Preprint
【36】Quantifying Risks in Multi-turn Conversation with Large Language Models
标题：量化使用大型语言模型的多轮对话中的风险
链接：https://arxiv.org/abs/2510.03969
【37】LLM Chemistry Estimation for Multi-LLM Recommendation
标题：多LLM推荐的LLM化学评估
链接：https://arxiv.org/abs/2510.03930
备注：20 pages, 5 figures, 5 tables
【38】LLM as an Algorithmist: Enhancing Anomaly Detectors via Programmatic Synthesis
标题：LLM作为一名学者：通过程序合成增强异常检测器
链接：https://arxiv.org/abs/2510.03904
【39】Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
标题：通过强化学习探索解锁LLM中的推理能力
链接：https://arxiv.org/abs/2510.03865
【40】Adaptive and Explainable AI Agents for Anomaly Detection in Critical IoT Infrastructure using LLM-Enhanced Contextual Reasoning
标题：使用LLM增强上下文推理在关键物联网基础设施中进行异常检测的自适应和可解释的人工智能代理
链接：https://arxiv.org/abs/2510.03859
备注：22 pages
【41】Small Language Models for Agentic Systems: A Survey of Architectures, Capabilities, and Deployment Trade offs
标题：大型系统的小型语言模型：体系结构、能力和部署权衡的调查
链接：https://arxiv.org/abs/2510.03847
备注：9 Pages
【42】On Using Large Language Models to Enhance Clinically-Driven Missing Data Recovery Algorithms in Electronic Health Records
标题：使用大型语言模型增强电子健康记录中临床驱动的缺失数据恢复算法
链接：https://arxiv.org/abs/2510.03844
【43】TROLL: Trust Regions improve Reinforcement Learning for Large Language Models
标题：TROLL：信任区域改进大型语言模型的强化学习
链接：https://arxiv.org/abs/2510.03817
【44】A Trustworthy Industrial Fault Diagnosis Architecture Integrating Probabilistic Models and Large Language Models
标题：集成概率模型和大型语言模型的值得信赖的工业故障诊断架构
链接：https://arxiv.org/abs/2510.03815
备注：1tables,6 figs,11pages
【45】Diverse Text-to-Image Generation via Contrastive Noise Optimization
标题：通过对比噪音优化实现多样化的文本到图像生成
链接：https://arxiv.org/abs/2510.03813
【46】GuidedSampling: Steering LLMs Towards Diverse Candidate Solutions at Inference-Time
标题：GuidedSampling：在推理时引导LLM走向多样化的候选解决方案
链接：https://arxiv.org/abs/2510.03777
【47】EvoEngineer: Mastering Automated CUDA Kernel Code Evolution with Large Language Models
标题：EvoEngineer：掌握使用大型语言模型的自动化CUDA核心代码进化
链接：https://arxiv.org/abs/2510.03760
备注：Under Review of ICLR 2026
【48】Operationalizing Data Minimization for Privacy-Preserving LLM Prompting
标题：隐私保护LLM算法中的数据最小化
链接：https://arxiv.org/abs/2510.03662
【49】LLM-Guided Evolutionary Program Synthesis for Quasi-Monte Carlo Design
标题：准蒙特卡罗设计的LLM引导的进化程序综合
链接：https://arxiv.org/abs/2510.03650
【50】Predicting Stock Price Movement with LLM-Enhanced Tweet Emotion Analysis
标题：利用LLM增强推文情绪分析预测股价走势
链接：https://arxiv.org/abs/2510.03633
备注：17th International Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (KDIR 2025), Marbella, Spain, Oct. 22-24, 2025 (to appear) Best Student Paper Finalist
【51】Can an LLM Induce a Graph? Investigating Memory Drift and Context Length
标题：LLM可以推导出图形吗？调查内存漂移和上下文长度
链接：https://arxiv.org/abs/2510.03611
备注：2025 IEEE International Conference on Knowledge Graph (ICKG)
【52】Efficient Test-Time Scaling for Small Vision-Language Models
标题：小型视觉语言模型的高效测试时间扩展
链接：https://arxiv.org/abs/2510.03574
【53】Machine Unlearning Meets Adversarial Robustness via Constrained Interventions on LLMs
标题：机器放弃学习通过对LLM的约束干预来满足对抗鲁棒性
链接：https://arxiv.org/abs/2510.03567
【54】Reactive Transformer (RxT) -- Stateful Real-Time Processing for Event-Driven Reactive Language Models
标题：反应式Transformer（RxT）--事件驱动反应式语言模型的状态实时处理
链接：https://arxiv.org/abs/2510.03561
备注：25 pages, 13 figures
【55】Certifiable Safe RLHF: Fixed-Penalty Constraint Optimization for Safer Language Models
标题：可认证的安全WLHF：更安全语言模型的固定罚约束优化
链接：https://arxiv.org/abs/2510.03520
【56】RAPID: An Efficient Reinforcement Learning Algorithm for Small Language Models
标题：RAID：一种针对小语言模型的高效强化学习算法
链接：https://arxiv.org/abs/2510.03515
【57】ALHD: A Large-Scale and Multigenre Benchmark Dataset for Arabic LLM-Generated Text Detection
标题：ALHD：用于阿拉伯语LLM生成文本检测的大规模、多流派基准数据集
链接：https://arxiv.org/abs/2510.03502
备注：47 pages, 15 figures. Dataset available at Zenodo: this https URL Codebase available at GitHub: this https URL
【58】Memory-Efficient Backpropagation for Fine-Tuning LLMs on Resource-Constrained Mobile Devices
标题：资源受限移动设备上微调LLM的内存高效反向传播
链接：https://arxiv.org/abs/2510.03425
【59】KVComm: Enabling Efficient LLM Communication through Selective KV Sharing
标题：KVComm：通过选择性KN共享实现高效的LLM通信
链接：https://arxiv.org/abs/2510.03346
【60】Semantic-Aware Scheduling for GPU Clusters with Large Language Models
标题：具有大型语言模型的图形处理器集群的语义感知调度
链接：https://arxiv.org/abs/2510.03334
【61】Scaling Laws Revisited: Modeling the Role of Data Quality in Language Model Pretraining
标题：重新审视标度律：建模语言模型预训练中数据质量的作用
链接：https://arxiv.org/abs/2510.03313
备注：18 pages, 6 figures
【62】Predicting Effects, Missing Distributions: Evaluating LLMs as Human Behavior Simulators in Operations Management
标题：预测效果，缺失分布：评估LLM作为运营管理中的人类行为模拟器
链接：https://arxiv.org/abs/2510.03310
【63】CAFL-L: Constraint-Aware Federated Learning with Lagrangian Dual Optimization for On-Device Language Models
标题：CAFL-L：针对设备上语言模型的拉格朗日二元优化的约束感知联邦学习
链接：https://arxiv.org/abs/2510.03298
备注：Accepted by 39th NeurIPS - Constrained Optimization for Machine Learning
【64】UniPruning: Unifying Local Metric and Global Feedback for Scalable Sparse LLMs
标题：UniPruning：统一可扩展稀疏LLM的局部度量和全局反馈
链接：https://arxiv.org/abs/2510.03291
【65】Edge-FIT: Federated Instruction Tuning of Quantized LLMs for Privacy-Preserving Smart Home Environments
标题：Edge-FIT：针对隐私保护的智能家居环境对量化LLM进行联合指令调整
链接：https://arxiv.org/abs/2510.03284
备注：7 pages, 1 figure
【66】MACE: A Hybrid LLM Serving System with Colocated SLO-aware Continuous Retraining Alignment
标题：VCE：具有共址SLO感知持续再训练对齐的混合LLM服务系统
链接：https://arxiv.org/abs/2510.03283
备注：14 pages, 15 figures
【67】Training Optimal Large Diffusion Language Models
标题：训练最佳大扩散语言模型
链接：https://arxiv.org/abs/2510.03280
【68】SDQ-LLM: Sigma-Delta Quantization for 1-bit LLMs of any size
标题：SDQ-LLM：用于任何大小的1位LLM的Σ-Δ量化
链接：https://arxiv.org/abs/2510.03275
【69】Quant-dLLM: Post-Training Extreme Low-Bit Quantization for Diffusion Large Language Models
标题：Quant-dLLM：用于扩散大语言模型的训练后极低比特量化
链接：https://arxiv.org/abs/2510.03274
【70】Decision Potential Surface: A Theoretical and Practical Approximation of LLM's Decision Boundary
标题：决策势面：LLM决策边界的理论和实践逼近
链接：https://arxiv.org/abs/2510.03271
备注：Source code: this https URL
【71】PT$^2$-LLM: Post-Training Ternarization for Large Language Models
标题：PT $#2 $-LLM：大型语言模型的训练后模块化
链接：https://arxiv.org/abs/2510.03267
【72】SciTS: Scientific Time Series Understanding and Generation with LLMs
标题：SciTS：利用LLM理解和生成科学时间序列
链接：https://arxiv.org/abs/2510.03255
【73】Solving the Granularity Mismatch: Hierarchical Preference Learning for Long-Horizon LLM Agents
标题：解决粒度不匹配：长期LLM代理的分层偏好学习
链接：https://arxiv.org/abs/2510.03253
备注：Preprint
【74】PARS: Low-Latency LLM Serving via Pairwise Learning-to-Rank
标题：PARS：低延迟LLM通过成对学习排名提供服务
链接：https://arxiv.org/abs/2510.03243
Graph相关(图学习|图神经网络|图优化等)(12篇)
【1】TopInG: Topologically Interpretable Graph Learning via Persistent Rationale Filtration
标题：TopInG：通过持续理由过滤进行的可拓拓解图学习
链接
：https://arxiv.org/abs/2510.05102
备注：submitted to ICML 2025
【2】Graph-Aware Diffusion for Signal Generation
标题：用于信号生成的图形感知扩散
链接：https://arxiv.org/abs/2510.05036
【3】Graph-based Tabular Deep Learning Should Learn Feature Interactions, Not Just Make Predictions
标题：基于图形的表格深度学习应该学习特征交互作用，而不仅仅是做出预测
链接：https://arxiv.org/abs/2510.04543
备注：9 pages, 6 figures, submitted to position track NeurIPS 2025
【4】Toward a Unified Geometry Understanding: Riemannian Diffusion Framework for Graph Generation and Prediction
标题：走向统一的几何理解：图生成和预测的Riemann扩散框架
链接：https://arxiv.org/abs/2510.04522
备注：Accepted by NeuIPS 2025
【5】Causality-aware Graph Aggregation Weight Estimator for Popularity Debiasing in Top-K Recommendation
标题：Top-K推荐中受欢迎度去偏置的因果关系感知图聚集权重估计
链接：https://arxiv.org/abs/2510.04502
备注：Accepted by CIKM 2025
【6】Fractional Heat Kernel for Semi-Supervised Graph Learning with Small Training Sample Size
标题：小训练样本量半监督图学习的分数热核
链接：https://arxiv.org/abs/2510.04440
【7】Diffusion-Assisted Distillation for Self-Supervised Graph Representation Learning with MLPs
标题：基于扩散辅助蒸馏的MLP自监督图表示学习
链接：https://arxiv.org/abs/2510.04241
【8】ICEPool: Enhancing Graph Pooling Networks with Inter-cluster Connectivity
标题：ICEPool：通过集群间连接增强图池网络
链接：https://arxiv.org/abs/2510.03987
【9】On the Convergence and Size Transferability of Continuous-depth Graph Neural Networks
标题：连续深度图神经网络的收敛性和大小可移植性
链接：https://arxiv.org/abs/2510.03923
【10】Generalization of Graph Neural Network Models for Distribution Grid Fault Detection
标题：配电网故障检测图神经网络模型的推广
链接：https://arxiv.org/abs/2510.03571
备注 ：This paper has been submitted and accepted for IEEE SmartGridComm 2025
【11】LHGEL: Large Heterogeneous Graph Ensemble Learning using Batch View Aggregation
标题：LHGEL：使用批视图聚合的大型异类图集合学习
链接：https://arxiv.org/abs/2510.03432
备注：Accepted by ICDM 2025
【12】Interpretable Neuropsychiatric Diagnosis via Concept-Guided Graph Neural Networks
标题：通过概念引导图神经网络进行可解释的神经精神诊断
链接：https://arxiv.org/abs/2510.03351
Transformer(19篇)
【1】AWARE, Beyond Sentence Boundaries: A Contextual Transformer Framework for Identifying Cultural Capital in STEM Narratives
标题：意识到，超越句子界限：识别STEM叙事中文化资本的上下文Transformer框架
链接：https://arxiv.org/abs/2510.04983
【2】From News to Returns: A Granger-Causal Hypergraph Transformer on the Sphere
标题：从新闻到收益：球面上的Granger因果超图Transformer
链接：https://arxiv.org/abs/2510.04357
备注：6th ACM International Conference on AI in Finance
【3】FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow Fields
标题：FoilDiff：一个用于二维翼型流场扩散建模的混合Transformer主干
链接：https://arxiv.org/abs/2510.04325
【4】SliceMoE: Routing Embedding Slices Instead of Tokens for Fine-Grained and Balanced Transformer Scaling
标题：SliceMoE：路由嵌入切片而不是令牌，以实现细粒度和平衡的Transformer扩展
链接：https://arxiv.org/abs/2510.04286
备注：EMNLP 2025 Main, 8 pages, 9 figures
【5】Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention
标题：低精度Transformer训练为何失败：闪光注意力分析
链接：https://arxiv.org/abs/2510.04212
备注：19 pages, 10 figures
【6】Allocation of Parameters in Transformers
标题：Transformer中的参数分配
链接：https://arxiv.org/abs/2510.03784
【7】Unsupervised Transformer Pre-Training for Images: Self-Distillation, Mean Teachers, and Random Crops
标题：图像的无监督Transformer预训练：自蒸馏，平均教师和随机裁剪
链接：https://arxiv.org/abs/2510.03606
【8】FieldFormer: Physics-Informed Transformers for Spatio-Temporal Field Reconstruction from Sparse Sensors
标题：FieldFormer：用于从稀疏传感器重建时空场的物理信息转换器
链接：https://arxiv.org/abs/2510.03589
【9】CrossLag: Predicting Major Dengue Outbreaks with a Domain Knowledge Informed Transformer
标题：CrossLag：利用领域知识丰富的Transformer预测重大登革热疫情
链接：https://arxiv.org/abs/2510.03566
备注：(C) 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
【10】Platonic Transformers: A Solid Choice For Equivariance
标题：柏拉图式Transformer：等变的可靠选择
链接：https://arxiv.org/abs/2510.03511
【11】Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis
标题：通过分层注意力和激活分析解开Transformer模型中的回忆和推理
链接：https://arxiv.org/abs/2510.03366
【12】Understanding Transformers for Time Series: Rank Structure, Flow-of-ranks, and Compressibility
标题：了解时间序列的Transformer：等级结构、等级流和可压缩性
链接：https://arxiv.org/abs/2510.03358
备注：42 pages
【13】Pool Me Wisely: On the Effect of Pooling in Transformer-Based Models
标题：明智地汇集我：关于基于变形者的模型中汇集的影响
链接：https://arxiv.org/abs/2510.03339
【14】Convolutional Neural Nets vs Vision Transformers: A SpaceNet Case Study with Balanced vs Imbalanced Regimes
标题：卷积神经网络与视觉变形者：平衡与不平衡机制的SpaceNet案例研究
链接：https://arxiv.org/abs/2510.03297
备注：5 pages, 1 figure, 9 tables. Code and artifacts: this https URL (release v1.0.1)
【15】Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework
标题：通过混合属性和修剪框架发现Transformer电路
链接：https://arxiv.org/abs/2510.03282
备注：Accepted to the NeurIPS 2025 Workshop on Mechanistic Interpretability (Mechinterp) and the NeurIPS 2025 Workshop on New Perspectives in Graph Machine Learning
【16】PDE-Transformer: A Continuous Dynamical Systems Approach to Sequence Modeling
标题：PDE-Transformer：序列建模的连续动态系统方法
链接：https://arxiv.org/abs/2510.03272
【17】Share Your Attention: Transformer Weight Sharing via Matrix-based Dictionary Learning
标题：分享您的注意力：通过基于矩阵的词典学习共享Transformer权重
链接：https://arxiv.org/abs/2508.04581
【18】ReplaceMe: Network Simplification via Depth Pruning and Transformer Block Linearization
标题：ReplaceMe：通过深度修剪和Transformer块线性化简化网络
链接：https://arxiv.org/abs/2505.02819
【19】Atlas-free Brain Network Transformer
标题：无脑脑网络Transformer
链接：https://arxiv.org/abs/2510.03306
GAN|对抗|攻击|生成相关(14篇)
【1】Flow-Matching Based Refiner for Molecular Conformer Generation
标题：基于流匹配的分子适形物生成精炼机
链接：https://arxiv.org/abs/2510.04878
【2】Parameter-free Algorithms for the Stochastically Extended Adversarial Model
标题：随机扩展对抗模型的无参数算法
链接：https://arxiv.org/abs/2510.04685
备注：Accepted to NeurIPS 2025
【3】Predictive Feature Caching for Training-free Acceleration of Molecular Geometry Generation
标题：预测特征缓存，免训练加速分子几何生成
链接：https://arxiv.org/abs/2510.04646
备注：Accepted at the AI for Science Workshop @ NeurIPS 2025
【4】MASC: Boosting Autoregressive Image Generation with a Manifold-Aligned Semantic Clustering
标题：MASC：使用流形对齐的语义聚类来增强自回归图像生成
链接：https://arxiv.org/abs/2510.04220
【5】ObCLIP: Oblivious CLoud-Device Hybrid Image Generation with Privacy Preservation
标题：ObCLIP：具有隐私保护的不经意CLoud-设备混合图像生成
链接：https://arxiv.org/abs/2510.04153
备注：Accepted by NeurIPS 2025
【6】Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning
标题：把它拴在绳子上：可控的伪标签生成走向现实的长尾半监督学习
链接：https://arxiv.org/abs/2510.03993
备注：The paper is accepted by NeurIPS 2025
【7】Pilot Contamination Attacks Detection with Machine Learning for Multi-User Massive MIMO
标题：基于机器学习的多用户MIMO导频污染攻击检测
链接：https://arxiv.org/abs/2510.03831
备注：This version of the article has been accepted for publication, after peer review and is subject to Springer Nature's AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1007/s11235-024-01163-0
【8】Curriculum-Augmented GFlowNets For mRNA Sequence Generation
标题：用于mRNA序列生成的课程扩充GFlowNet
链接：https://arxiv.org/abs/2510.03811
【9】Merge and Guide: Unifying Model Merging and Guided Decoding for Controllable Multi-Objective Generation
标题：合并与引导：统一模型合并和引导解码以实现可控多目标生成
链接：https://arxiv.org/abs/2510.03782
备注：Work in progress
【10】From Theory to Practice: Evaluating Data Poisoning Attacks and Defenses in In-Context Learning on Social Media Health Discourse
标题：从理论到实践：评估社交媒体健康话语的上下文学习中的数据中毒攻击和防御
链接：https://arxiv.org/abs/2510.03636
【11】Neon: Negative Extrapolation From Self-Training Improves Image Generation
标题：霓虹灯：自我训练的负外推改善图像生成
链接：https://arxiv.org/abs/2510.03597
【12】Attack logics, not outputs: Towards efficient robustification of deep neural networks by falsifying concept-based properties
标题：攻击逻辑，而不是输出：通过伪造基于概念的属性实现深度神经网络的高效鲁棒化
链接：https://arxiv.org/abs/2510.03320
备注：13 pages, 2 figures, accepted by "7th OVERLAY" workshop
【13】SVDefense: Effective Defense against Gradient Inversion Attacks via Singular Value Decomposition
标题：SV Defense：通过奇异值分解有效防御梯度倒置攻击
链接：https://arxiv.org/abs/2510.03319
【14】Adversarial training with restricted data manipulation
标题：具有限制数据操作的对抗性训练
链接：https://arxiv.org/abs/2510.03254
备注：21 page, 5 figures
半/弱/无/有监督|不确定性|主动学习(14篇)
【1】HybridFlow: Quantification of Aleatoric and Epistemic Uncertainty with a Single Hybrid Model
标题：HybridFlow：使用单混合模型量化感性和认识性不确定性
链接：https://arxiv.org/abs/2510.05054
备注：Reviewed and published in TMLR at https://openreview.net/forum?id=xRiEdSyVjY
【2】Latent Uncertainty Representations for Video-based Driver Action and Intention Recognition
标题：基于视频的驾驶员行为和意图识别的潜在不确定性表示
链接：https://arxiv.org/abs/2510.05006
备注：16 pages, 8 figures, 7 tables, under submission
【3】Unsupervised Active Learning via Natural Feature Progressive Framework
标题：通过自然特征渐进框架的无监督主动学习
链接：https://arxiv.org/abs/2510.04939
备注：Under review at IEEE TPAMI
【4】Federated Self-Supervised Learning for Automatic Modulation Classification under Non-IID and Class-Imbalanced Data
标题：非IID和类别不平衡数据下自动调制分类的联邦自监督学习
链接：https://arxiv.org/abs/2510.04927
【5】How Different from the Past? Spatio-Temporal Time Series Forecasting with Self-Supervised Deviation Learning
标题：与过去有何不同？自监督偏差学习的时空时间序列预测
链接：https://arxiv.org/abs/2510.04908
备注：Accepted at NeurIPS 2025
【6】Utility-Learning Tension in Self-Modifying Agents
标题：自我修改代理人中的实用学习张力
链接：https://arxiv.org/abs/2510.04399
【7】Rethinking Consistent Multi-Label Classification under Inexact Supervision
标题：在不精确监督下重新思考一致的多标签分类
链接：https://arxiv.org/abs/2510.04091
【8】Using predefined vector systems as latent space configuration for neural network supervised training on data with arbitrarily large number of classes
标题：使用预定义的载体系统作为潜在空间配置，对任意大量类的数据进行神经网络监督训练
链接：https://arxiv.org/abs/2510.04090
备注：28 pages, 12 figures, 10 tables, 12 equations, 1 algorithm
【9】BONSAI: Structure-exploiting robust Bayesian optimization for networked black-box systems under uncertainty
标题：BONSAI：不确定性下网络黑匣子系统的结构利用鲁棒Bayesian优化
链接：https://arxiv.org/abs/2510.03893
备注：Published in Computers and Chemical Engineering, 2025
【10】HydroFusion-LMF: Semi-Supervised Multi-Network Fusion with Large-Model Adaptation for Long-Term Daily Runoff Forecasting
标题：HydroFusion-LMF：具有大模型自适应的半监督多网络融合，用于长期日径流预测
链接：https://arxiv.org/abs/2510.03744
备注：V1
【11】Multi-task neural diffusion processes for uncertainty-quantified wind power prediction
标题：不确定性量化风电预测的多任务神经扩散过程
链接：https://arxiv.org/abs/2510.03419
备注：36 pages, 13 figures, 2 tables,
【12】Conditional Pseudo-Supervised Contrast for Data-Free Knowledge Distillation
标题：无数据知识提炼的条件伪监督对比
链接：https://arxiv.org/abs/2510.03375
备注：13 pages
【13】Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning
标题：元意识增强推理模型：自对齐强化学习
链接：https://arxiv.org/abs/2510.03259
备注：preprint
【14】Towards Multimodal Active Learning: Efficient Learning with Limited Paired Data
标题：迈向多模式主动学习：利用有限的配对数据进行高效学习
链接：https://arxiv.org/abs/2510.03247
迁移|Zero/Few/One-Shot|自适应(25篇)
【1】Boomerang Distillation Enables Zero-Shot Model Size Interpolation
标题：回旋镖蒸馏实现Zero-Shot模型尺寸插值
链接：https://arxiv.org/abs/2510.05064
备注：10 pages, 7 figures in main text
【2】Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization
标题：通过基于模型的深度学习优化框架实现自适应记忆动量
链接：https://arxiv.org/abs/2510.04988
【3】Speak, Edit, Repeat: High-Fidelity Voice Editing and Zero-Shot TTS with Cross-Attentive Mamba
标题：说话、编辑、重复：高保真语音编辑和Zero-ShotTTC，具有交叉注意力的曼巴
链接：https://arxiv.org/abs/2510.04738
【4】Noise or Signal? Deconstructing Contradictions and An Adaptive Remedy for Reversible Normalization in Time Series Forecasting
标题：噪音还是信号？时间序列预测中的矛盾解构和可逆规范化的自适应补救措施
链接：https://arxiv.org/abs/2510.04667
备注：9pages, 6 figures
【5】Data-Driven Adaptive PID Control Based on Physics-Informed Neural Networks
标题：基于物理信息神经网络的数据驱动自适应PI控制
链接：https://arxiv.org/abs/2510.04591
备注：This work has been submitted to the IEEE Transactions on Control Systems Technology for possible publication
【6】Demystifying MaskGIT Sampler and Beyond: Adaptive Order Selection in Masked Diffusion
标题：揭开MaskGIT采样器及超越的神秘面纱：掩蔽扩散中的自适应顺序选择
链接：https://arxiv.org/abs/2510.04525
备注：23 pages
【7】Adaptive Weighted Loss for Sequential Recommendations on Sparse Domains
标题：稀疏域上顺序推荐的自适应加权损失
链接：https://arxiv.org/abs/2510.04375
【8】DoRAN: Stabilizing Weight-Decomposed Low-Rank Adaptation via Noise Injection and Auxiliary Networks
标题：DoRAN：通过噪音注入和辅助网络稳定权重分解的低等级自适应
链接：https://arxiv.org/abs/2510.04331
备注：Nghiem T. Diep, Hien Dang, and Tuan Truong contributed equally to this work
【9】HoRA: Cross-Head Low-Rank Adaptation with Joint Hypernetworks
标题：HoRA：使用联合超网络的交叉头低等级适应
链接：https://arxiv.org/abs/2510.04295
备注：Nghiem T. Diep, Dung Le, and Tuan Truong contributed equally to this work
【10】A KL-regularization framework for learning to plan with adaptive priors
标题：用于学习使用自适应先验进行计划的KL正规化框架
链接：https://arxiv.org/abs/2510.04280
备注：Preprint
【11】Adaptive Federated Learning via Dynamical System Model
标题：通过动态系统模型的自适应联邦学习
链接：https://arxiv.org/abs/2510.04203
【12】Adaptive kernel-density approach for imbalanced binary classification
标题：不平衡二元分类的自适应核密度方法
链接：https://arxiv.org/abs/2510.04046
【13】Technical note on Sequential Test-Time Adaptation via Martingale-Driven Fisher Prompting
标题：关于通过Martingale驱动Fisher预算进行顺序测试时间自适应的技术说明
链接：https://arxiv.org/abs/2510.03839
【14】Optimizing Fine-Tuning through Advanced Initialization Strategies for Low-Rank Adaptation
标题：通过高级调整策略优化微调以实现低等级适应
链接：https://arxiv.org/abs/2510.03731
【15】Balancing Interpretability and Performance in Reinforcement Learning: An Adaptive Spectral Based Linear Approach
标题：强化学习中的可解释性和性能平衡：基于自适应谱的线性方法
链接：https://arxiv.org/abs/2510.03722
【16】EmbodiSwap for Zero-Shot Robot Imitation Learning
标题：Zero-Shot机器人模仿学习的交换
链接：https://arxiv.org/abs/2510.03706
备注：Video link: this https URL
【17】SAFA-SNN: Sparsity-Aware On-Device Few-Shot Class-Incremental Learning with Fast-Adaptive Structure of Spiking Neural Network
标题：SAFA-SNN：稀疏感知的设备上Few-Shot类增量学习，具有尖峰神经网络的快速自适应结构
链接：https://arxiv.org/abs/2510.03648
【18】Deep Domain Adaptation for Turbofan Engine Remaining Useful Life Prediction: Methodologies, Evaluation and Future Trends
标题：Turbofan发动机剩余使用寿命预测的深域适应：方法论、评估和未来趋势
链接：https://arxiv.org/abs/2510.03604
【19】Task-Level Contrastiveness for Cross-Domain Few-Shot Learning
标题：跨领域Few-Shot学习的任务级对比
链接：https://arxiv.org/abs/2510.03509
备注：None
【20】Dynamic Meta-Learning for Adaptive XGBoost-Neural Ensembles
标题：自适应XGBoost神经集成的动态元学习
链接：https://arxiv.org/abs/2510.03301
【21】CoDA: Coding LM via Diffusion Adaptation
标题：CoDA：通过扩散适应编码LM
链接：https://arxiv.org/abs/2510.03270
【22】Semantic-Inductive Attribute Selection for Zero-Shot Learning
标题：Zero-Shot学习的语义归纳属性选择
链接：https://arxiv.org/abs/2510.03260
备注：26 pages, 9 figures, code available at this https URL
【23】POEM: Explore Unexplored Reliable Samples to Enhance Test-Time Adaptation
标题：POEM：探索未经探索的可靠样本以增强测试时间适应性
链接：https://arxiv.org/abs/2510.03258
备注：11pages,6 figures
【24】Modular and Adaptive Conformal Prediction for Sequential Models via Residual Decomposition
标题：通过剩余分解进行序列模型的模块化和自适应共形预测
链接：https://arxiv.org/abs/2510.04406
备注：11 pages, (37 with appendix), 15 figures
【25】Adaptive Coverage Policies in Conformal Prediction
标题：共形预测中的自适应覆盖策略
链接：https://arxiv.org/abs/2510.04318
备注：Code at: this https URL
强化学习(15篇)
【1】MARS: Optimizing Dual-System Deep Research via Multi-Agent Reinforcement Learning
标题：MARS：通过多智能体强化学习优化双系统深度研究
链接：https://arxiv.org/abs/2510.04935
备注：Ongoing Work
【2】Video Game Level Design as a Multi-Agent Reinforcement Learning Problem
标题：作为多智能体强化学习问题的视频游戏级设计
链接：https://arxiv.org/abs/2510.04862
备注：11 pages, 7 tables, 5 figures, published as full technical paper at the AAAI conference on Artificial Intelligence and Interactive Digital Entertainment 2025
【3】Learning on the Job: Test-Time Curricula for Targeted Reinforcement Learning
标题：在职学习：针对性强化学习的测试时间课程
链接：https://arxiv.org/abs/2510.04786
【4】Tail-Safe Hedging: Explainable Risk-Sensitive Reinforcement Learning with a White-Box CBF--QP Safety Layer in Arbitrage-Free Markets
标题：尾部安全对冲：使用白盒CBF的可解释风险敏感强化学习--无套利市场中的QP安全层
链接：https://arxiv.org/abs/2510.04555
备注：32 pages including appendices; 5 figures. Primary subject class: q-fin.TR. Cross-lists: cs.LG; q-fin.RM
【5】Wavelet Predictive Representations for Non-Stationary Reinforcement Learning
标题：非平稳强化学习的子波预测表示
链接：https://arxiv.org/abs/2510.04507
【6】Achieve Performatively Optimal Policy for Performative Reinforcement Learning
标题：实现执行强化学习的执行最佳策略
链接：https://arxiv.org/abs/2510.04430
【7】Closing the Loop: Coordinating Inventory and Recommendation via Deep Reinforcement Learning on Multiple Timescales
标题：闭环：通过多个时间尺度上的深度强化学习协调库存和推荐
链接：https://arxiv.org/abs/2510.04272
【8】RLRF: Competitive Search Agent Design via Reinforcement Learning from Ranker Feedback
标题：RL RF：通过来自排名反馈的强化学习进行竞争性搜索代理设计
链接：https://arxiv.org/abs/2510.04096
【9】Offline Reinforcement Learning in Large State Spaces: Algorithms and Guarantees
标题：大状态空间中的离线强化学习：算法和保证
链接：https://arxiv.org/abs/2510.04088
备注：To appear in Statistical Science
【10】Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models
标题：作为规划的时空预测：一种具有生成世界模型的基于模型的强化学习方法
链接：https://arxiv.org/abs/2510.04020
【11】Distributed Area Coverage with High Altitude Balloons Using Multi-Agent Reinforcement Learning
标题：使用多智能体强化学习的高空气球分布式区域覆盖
链接：https://arxiv.org/abs/2510.03823
【12】Token Hidden Reward: Steering Exploration-Exploitation in Group Relative Deep Reinforcement Learning
标题：代币隐藏奖励：引导群体相对深度强化学习中的探索利用
链接：https://arxiv.org/abs/2510.03669
【13】Deep Reinforcement Learning for Multi-Agent Coordination
标题：用于多智能体协调的深度强化学习
链接：https://arxiv.org/abs/2510.03592
备注：11 pages, 8 figures, 1 table, presented at SWARM 2022, to be published in Journal of Artificial Life and Robotics
【14】Long-Term Mapping of the Douro River Plume with Multi-Agent Reinforcement Learning
标题：利用多智能体强化学习进行杜鲁河羽流的长期映射
链接：https://arxiv.org/abs/2510.03534
【15】Dissecting Larval Zebrafish Hunting using Deep Reinforcement Learning Trained RNN Agents
标题：使用深度强化学习训练的RNN代理剖析幼虫斑马狩猎
链接：https://arxiv.org/abs/2510.03699
医学相关(10篇)
【1】KEEP: Integrating Medical Ontologies with Clinical Data for Robust Code Embeddings
标题：KEEP：将医疗实体与临床数据集成以实现稳健的代码嵌入
链接：https://arxiv.org/abs/2510.05049
备注：None
【2】A Clinical-grade Universal Foundation Model for Intraoperative Pathology
标题：临床级的手术中病理学通用基础模型
链接：https://arxiv.org/abs/2510.04861
【3】Forecasting-Based Biomedical Time-series Data Synthesis for Open Data and Robust AI
标题：基于预测的生物医学时间序列数据合成，用于开放数据和稳健的人工智能
链接：https://arxiv.org/abs/2510.04622
备注：Under Review
【4】Fast Witness Persistence for MRI Volumes via Hybrid Landmarking
标题：通过混合地标实现MRI收件箱的快速见证持久性
链接
：https://arxiv.org/abs/2510.04553
【5】SSM-CGM: Interpretable State-Space Forecasting Model of Continuous Glucose Monitoring for Personalized Diabetes Management
标题：SSM-CGM：用于个性化糖尿病管理的持续血糖监测的可解释状态空间预测模型
链接：https://arxiv.org/abs/2510.04386
备注：Shakson Isaac and Yentl Collin contributed equally
【6】Attending on Multilevel Structure of Proteins enables Accurate Prediction of Cold-Start Drug-Target Interactions
标题：关注蛋白质的多层次结构可以准确预测冷启动药物-靶标相互作用
链接：https://arxiv.org/abs/2510.04126
【7】Optimizing Resources for On-the-Fly Label Estimation with Multiple Unknown Medical Experts
标题：与多名未知医学专家一起优化实时标签估计资源
链接：https://arxiv.org/abs/2510.03954
备注：7 pages, 3 figures, 3 tables, Accepted at IEEE BHI 2025
【8】Linguistic and Audio Embedding-Based Machine Learning for Alzheimer's Dementia and Mild Cognitive Impairment Detection: Insights from the PROCESS Challenge
标题：阿尔茨海默氏痴呆症和轻度认知障碍检测的基于语言和音频嵌入的机器学习：来自Process挑战的见解
链接：https://arxiv.org/abs/2510.03336
【9】Thin Bridges for Drug Text Alignment: Lightweight Contrastive Learning for Target Specific Drug Retrieval
标题：药物文本对齐的瘦桥：针对特定目标药物检索的轻量级对比学习
链接：https://arxiv.org/abs/2510.03309
【10】A Benchmark Study of Deep Learning Methods for Multi-Label Pediatric Electrocardiogram-Based Cardiovascular Disease Classification
标题：基于多标签儿科心电图的心血管疾病分类的深度学习方法基准研究
链接：https://arxiv.org/abs/2510.03780
备注：8 pages, 5 figures
蒸馏|知识提取(6篇)
【1】ERDE: Entropy-Regularized Distillation for Early-exit
标题：ERDE：提前退出的熵正规蒸馏
链接：https://arxiv.org/abs/2510.04856
【2】Beyond Random: Automatic Inner-loop Optimization in Dataset Distillation
标题：超越随机：数据集蒸馏中的自动内循环优化
链接：https://arxiv.org/abs/2510.04838
【3】Learning from All: Concept Alignment for Autonomous Distillation from Multiple Drifting MLLMs
标题：向所有人学习：多个漂移MLLM自主蒸馏的概念一致
链接：https://arxiv.org/abs/2510.04142
【4】Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation
标题：通过对比微调和蒸馏实现轻量级且可推广的声学场景表示
链接：https://arxiv.org/abs/2510.03728
【5】MECKD: Deep Learning-Based Fall Detection in Multilayer Mobile Edge Computing With Knowledge Distillation
标题：MECKD：具有知识提炼的多层移动边缘计算中基于深度学习的跌倒检测
链接：https://arxiv.org/abs/2510.03601
备注：15 pages, 7 figures, and published in IEEE Sensors Journal
【6】Learning without Global Backpropagation via Synergistic Information Distillation
标题：基于协同信息蒸馏的无全局反向传播学习
链接：https://arxiv.org/abs/2510.03273
聚类(1篇)
【1】DECOR: Deep Embedding Clustering with Orientation Robustness
标题：DECOR：具有方向稳健性的深度嵌入集群
链接：https://arxiv.org/abs/2510.03328
超分辨率|去噪|去模糊|去雾(1篇)
【1】Diffusion-Based, Data-Assimilation-Enabled Super-Resolution of Hub-height Winds
标题：基于扩散、数据同化使能的中心高度风的超分辨率
链接：https://arxiv.org/abs/2510.03364
自动驾驶|车辆|车道检测等(2篇)
【1】Comparative Analysis of YOLOv5, Faster R-CNN, SSD, and RetinaNet for Motorbike Detection in Kigali Autonomous Driving Context
标题：基加利自动驾驶环境下用于摩托车检测的YOLOv 5、Faster R-CNN、SSD和RetinaNet的比较分析
链接：https://arxiv.org/abs/2510.04912
备注：3 figures, 2 tables
【2】Estimating link level traffic emissions: enhancing MOVES with open-source data
标题：估计链路级流量排放：利用开源数据增强MOVES
链接：https://arxiv.org/abs/2510.03362
点云|SLAM|雷达|激光|深度RGBD相关(3篇)
【1】On residual network depth
标题：关于剩余网络深度
链接：https://arxiv.org/abs/2510.03470
【2】Matching the Optimal Denoiser in Point Cloud Diffusion with (Improved) Rotational Alignment
标题：用（改进的）旋转对齐匹配点云扩散中的最佳降噪器
链接：https://arxiv.org/abs/2510.03335
备注：under review
【3】Fast frequency reconstruction using Deep Learning for event recognition in ring laser data
标题：使用深度学习快速频率重建用于环形激光数据中的事件识别
链接：https://arxiv.org/abs/2510.03325
联邦学习|隐私保护|加密(6篇)
【1】Federated Learning for Surgical Vision in Appendicitis Classification: Results of the FedSurg EndoVis 2024 Challenge
标题：阑尾炎分类中手术视觉的联邦学习：FedSurg EndoVis 2024挑战赛的结果
链接：https://arxiv.org/abs/2510.04772
备注：A challenge report pre-print (31 pages), including 7 tables and 8 figures
【2】Trade-off in Estimating the Number of Byzantine Clients in Federated Learning
标题：估计联邦学习中拜占庭客户数量的权衡
链接：https://arxiv.org/abs/2510.04432
【3】On Provable Benefits of Muon in Federated Learning
标题：联邦学习中μ on的可证明好处
链接：https://arxiv.org/abs/2510.03866
【4】Personalized federated prototype learning in mixed heterogeneous data scenarios
标题：混合异类数据场景中的个性化联邦原型学习
链接：https://arxiv.org/abs/2510.03726
【5】A Lightweight Federated Learning Approach for Privacy-Preserving Botnet Detection in IoT
标题：用于物联网中保护隐私的僵尸网络检测的轻量级联邦学习方法
链接：https://arxiv.org/abs/2510.03513
备注：This work has been published in the Proceedings of the 2025 IEEE International Conference on Applied Cloud and Data Science and Applications (ACDSA). The final published version is available via IEEE Xplore at https://doi.org/10.1109/ACDSA65407.2025.11165820
【6】A Robust Clustered Federated Learning Approach for Non-IID Data with Quantity Skew
标题：具有数量偏差的非IID数据的鲁棒交叉联邦学习方法
链接：https://arxiv.org/abs/2510.03380
推理|分析|理解|解释(28篇)
【1】From Noisy Traces to Stable Gradients: Bias-Variance Optimized Preference Optimization for Aligning Large Reasoning Models
标题：从有噪音的痕迹到稳定的结果：用于对齐大型推理模型的偏差方差优化偏好优化
链接：https://arxiv.org/abs/2510.05095
【2】Less is More: Recursive Reasoning with Tiny Networks
标题：少即是多：微型网络的回归推理
链接：https://arxiv.org/abs/2510.04871
【3】Synthesising Counterfactual Explanations via Label-Conditional Gaussian Mixture Variational Autoencoders
标题：通过标签条件高斯混合变分自动编码器合成反事实解释
链接：https://arxiv.org/abs/2510.04855
【4】On Predicting Post-Click Conversion Rate via Counterfactual Inference
标题：关于通过反事实推理预测点击后转化率
链接：https://arxiv.org/abs/2510.04816
备注：This work has been accepted for publication at the IEEE International Conference on Data Mining (ICDM) 2025
【5】MetaMP: Seamless Metadata Enrichment and AI Application Framework for Enhanced Membrane Protein Visualization and Analysis
标题：MetaMP：用于增强膜蛋白可视化和分析的无缝元数据丰富和AI应用框架
链接：https://arxiv.org/abs/2510.04776
【6】COSMIR: Chain Orchestrated Structured Memory for Iterative Reasoning over Long Context
标题：COSMIR：用于长上下文迭代推理的链结构化记忆
链接：https://arxiv.org/abs/2510.04568
【7】DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization
标题：DRPO：通过脱钩奖励政策优化进行高效推理
链接：https://arxiv.org/abs/2510.04474
备注：20 pages, 7 figures
【8】Quantifying Ambiguity in Categorical Annotations: A Measure and Statistical Inference Framework
标题：量化类别注释中的模糊性：衡量和统计推理框架
链接：https://arxiv.org/abs/2510.04366
备注：Preprint, 20 pages in total, 7 figures
【9】PABSA: Hybrid Framework for Persian Aspect-Based Sentiment Analysis
标题：PABSA：基于波斯语的情绪分析混合框架
链接：https://arxiv.org/abs/2510.04291
备注：8 pages
【10】PolyKAN: A Polyhedral Analysis Framework for Provable and Minimal KAN Compression
标题：PolyKAN：可证明且最小KAN压缩的多边形分析框架
链接：https://arxiv.org/abs/2510.04205
备注：10
【11】CALM Before the STORM: Unlocking Native Reasoning for Optimization Modeling
标题：风暴前的CALM：解锁优化建模的原生推理
链接：https://arxiv.org/abs/2510.04204
备注：Work in progress
【12】Finite Time Analysis of Constrained Natural Critic-Actor Algorithm with Improved Sample Complexity
标题：提高样本复杂性的约束自然批评者算法的有限时间分析
链接：https://arxiv.org/abs/2510.04189
【13】Variational Diffusion Unlearning: A Variational Inference Framework for Unlearning in Diffusion Models under Data Constraints
标题：变分扩散去学习：数据约束下扩散模型去学习的变分推理框架
链接：https://arxiv.org/abs/2510.04058
【14】Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment
标题：探索可操纵多元对齐的思想链推理
链接：https://arxiv.org/abs/2510.04045
备注：ACL EMNLP 2025
【15】The Debate on RLVR Reasoning Capability Boundary: Shrinkage, Expansion, or Both? A Two-Stage Dynamic View
标题：关于WLVR推理能力边界的争论：收缩、扩张，还是两者兼而有之？两阶段动态视图
链接：https://arxiv.org/abs/2510.04028
【16】What Is The Performance Ceiling of My Classifier? Utilizing Category-Wise Influence Functions for Pareto Frontier Analysis
标题：我的分类器的性能上限是多少？利用类别影响函数进行帕累托前沿分析
链接：https://arxiv.org/abs/2510.03950
【17】Trajectory prediction for heterogeneous agents: A performance analysis on small and imbalanced datasets
标题：异类代理的轨迹预测：小型且不平衡数据集的性能分析
链接：https://arxiv.org/abs/2510.03776
备注：This paper has been accepted to the IEEE Robotics and Automation Letters journal and presented at the 40th Anniversary of the IEEE International Conference on Robotics and Automation, which was held in Rotterdam, Netherlands on 23-26 September, 2024
【18】Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders
标题：更高的可解释性是否意味着更好的实用性？稀疏自动编码器的成对分析
链接：https://arxiv.org/abs/2510.03659
备注：24 pages
【19】Understanding the Role of Training Data in Test-Time Scaling
标题：了解训练数据在测试时间缩放中的作用
链接：https://arxiv.org/abs/2510.03605
备注：24 pages, 4 figures
【20】Exploring the Hierarchical Reasoning Model for Small Natural-Image Classification Without Augmentation
标题：探索无增强的小型自然图像分类的分层推理模型
链接：https://arxiv.org/abs/2510.03598
【21】Reasoning-based Anomaly Detection Framework: A Real-time, Scalable, and Automated Approach to Anomaly Detection Across Domains
标题：基于推理的异常检测框架：跨领域异常检测的实时、可扩展和自动化方法
链接：https://arxiv.org/abs/2510.03486
备注：11 pages, 7 figures
【22】Spatial-ViLT: Enhancing Visual Spatial Reasoning through Multi-Task Learning
标题：Spatial-ViLT：通过多任务学习增强视觉空间推理
链接：https://arxiv.org/abs/2510.03441
备注：12 pages, 5 figures
【23】Inference-Time Search using Side Information for Diffusion-based Image Reconstruction
标题：使用边信息进行基于扩散的图像重建的推理时搜索
链接：https://arxiv.org/abs/2510.03352
【24】AgentCaster: Reasoning-Guided Tornado Forecasting
标题：AgentCaster：推理引导的龙卷风预测
链接：https://arxiv.org/abs/2510.03349
【25】Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data
标题：前加载推理：训练前和训练后数据之间的协同作用
链接：https://arxiv.org/abs/2510.03264
【26】TCR-EML: Explainable Model Layers for TCR-pMHC Prediction
标题：TCR-EML：TCR-pMHC预测的可解释模型层
链接：https://arxiv.org/abs/2510.04377
【27】Simulation-based inference via telescoping ratio estimation for trawl processes
标题：拖网过程的伸缩比估计基于模拟的推理
链接：https://arxiv.org/abs/2510.04042
【28】Mathematically rigorous proofs for Shapley explanations
标题：Shapley解释的数学严格证明
链接：https://arxiv.org/abs/2510.03281
检测相关(10篇)
【1】SPEGNet: Synergistic Perception-Guided Network for Camouflaged Object Detection
标题：SPEGNet：用于伪装对象检测的协同感知引导网络
链接：https://arxiv.org/abs/2510.04472
【2】Detection of retinal diseases using an accelerated reused convolutional network
标题：使用加速重复使用的卷积网络检测视网膜疾病
链接：https://arxiv.org/abs/2510.04232
备注：None
【3】On the Empirical Power of Goodness-of-Fit Tests in Watermark Detection
标题：水印检测中的适合度检验的经验力量
链接：https://arxiv.org/abs/2510.03944
备注：Accepted at NeurIPS 2025 as a spotlight
【4】THEMIS: Unlocking Pretrained Knowledge with Foundation Model Embeddings for Anomaly Detection in Time Series
标题：THEMIS：通过基础模型嵌入解锁预训练知识，用于时间序列中的异常检测
链接：https://arxiv.org/abs/2510.03911
备注：Oral Presentation. AI4TS Workshop, IJCAI'25
【5】Detecting Invariant Manifolds in ReLU-Based RNNs
标题：在基于ReLU的RNN中检测不变Manifle
链接：https://arxiv.org/abs/2510.03814
【6】6G-Enabled Digital Twin Framework for Real-Time Cyber-Physical Systems: An Experimental Validation with Industrial Bearing Fault Detection
标题：用于实时网络物理系统的支持6G的数字双胞胎框架：工业轴承故障检测的实验验证
链接：https://arxiv.org/abs/2510.03807
【7】Road Damage and Manhole Detection using Deep Learning for Smart Cities: A Polygonal Annotation Approach
标题：使用深度学习实现智慧城市的道路损坏和检修孔检测：多边形注释方法
链接：https://arxiv.org/abs/2510.03797
备注：13 pages
【8】Consistent Kernel Change-Point Detection under m-Dependence for Text Segmentation
标题：m依赖下文本分割的一致核变化点检测
链接：https://arxiv.org/abs/2510.03437
【9】LogAction: Consistent Cross-system Anomaly Detection through Logs via Active Domain
标题：LogAction：通过Active域通过收件箱进行一致的跨系统异常检测
链接：https://arxiv.org/abs/2510.03288
备注：The 40th IEEE/ACM International Conference on Automated Software Engineering, ASE 2025
【10】Variational Autoencoders-based Detection of Extremes in Plant Productivity in an Earth System Model
标题：基于变分自动编码器的地球系统模型中植物生产力极端检测
链接：https://arxiv.org/abs/2510.03266
分类|识别(12篇)
【1】BenthiCat: An opti-acoustic dataset for advancing benthic classification and habitat mapping
标题：BenthiCat：用于推进底生物分类和栖息地绘图的光学声学数据集
链接：https://arxiv.org/abs/2510.04876
备注：Article under review by IJRR
【2】EVaR-Optimal Arm Identification in Bandits
标题：EVaR--盗贼中的最佳手臂识别
链接：https://arxiv.org/abs/2510.04728
【3】A Study on the Data Distribution Gap in Music Emotion Recognition
标题：音乐情感识别中的数据分布差距研究
链接：https://arxiv.org/abs/2510.04688
备注：Accepted at the 17th International Symposium on Computer Music Multidisciplinary Research (CMMR) 2025
【4】Score-based Greedy Search for Structure Identification of Partially Observed Linear Causal Models
标题：基于分数的贪婪搜索部分观察线性因果模型的结构识别
链接：https://arxiv.org/abs/2510.04378
【5】Critical appraisal of artificial intelligence for rare-event recognition: principles and pharmacovigilance case studies
标题：人工智能用于罕见事件识别的批判性评估：原则和药物警戒案例研究
链接：https://arxiv.org/abs/2510.04341
备注：28 pages, 2 figures
【6】From Segments to Concepts: Interpretable Image Classification via Concept-Guided Segmentation
标题：从片段到概念：通过概念引导分割的可解释图像分类
链接：https://arxiv.org/abs/2510.04180
【7】SPEAR: Soft Prompt Enhanced Anomaly Recognition for Time Series Data
标题：SPECTRA：时间序列数据的软提示增强异常识别
链接：https://arxiv.org/abs/2510.03962
备注：Accepted to 2025 IEEE International Conference on Autonomous and Trusted Computing (ATC 2025)
【8】Sequential decoder training for improved latent space dynamics identification
标题：用于改进潜在空间动态识别的顺序解码器训练
链接：https://arxiv.org/abs/2510.03535
【9】Know Thyself? On the Incapability and Implications of AI Self-Recognition
标题：了解你自己？论人工智能自我识别的无能及其影响
链接：https://arxiv.org/abs/2510.03399
备注：Our code is available, see this https URL
【10】A Unified Optimization Framework for Multiclass Classification with Structured Hyperplane Arrangements
标题：具有结构化超平面排列的多类分类统一优化框架
链接：https://arxiv.org/abs/2510.05047
备注：28 pages, 2 tables, 9 figures
【11】Set to Be Fair: Demographic Parity Constraints for Set-Valued Classification
标题：公平：集值分类的人口平价约束
链接：https://arxiv.org/abs/2510.04926
【12】Drax: Speech Recognition with Discrete Flow Matching
标题：Drax：具有离散流匹配的语音识别
链接：https://arxiv.org/abs/2510.04162
表征(2篇)
【1】ONNX-Net: Towards Universal Representations and Instant Performance Prediction for Neural Architectures
标题：ONNX-Net：走向神经架构的通用表示和即时性能预测
链接：https://arxiv.org/abs/2510.04938
备注：Our code is available at: this https URL
【2】Decrypt Modality Gap in Multimodal Contrastive Learning: From Convergent Representation to Pair Alignment
标题：解密多模式对比学习中的情态差距：从收敛表示到配对对齐
链接：https://arxiv.org/abs/2510.03268
3D|3D重建等相关(1篇)
【1】Physics-Inspired All-Pair Interaction Learning for 3D Dynamics Modeling
标题：3D动力学建模的受物理启发的全对交互学习
链接：https://arxiv.org/abs/2510.04233
优化|敛散性(15篇)
【1】Counterfactual Credit Guided Bayesian Optimization
标题：反事实信用引导的Bayesian优化
链接：https://arxiv.org/abs/2510.04676
【2】Closed-Form Last Layer Optimization
标题：封闭式最后一层优化
链接：https://arxiv.org/abs/2510.04606
【3】Stochastic Approximation Methods for Distortion Risk Measure Optimization
标题：失真风险度量优化的随机逼近方法
链接：https://arxiv.org/abs/2510.04563
【4】Scale-Invariant Regret Matching and Online Learning with Optimal Convergence: Bridging Theory and Practice in Zero-Sum Games
标题：规模不变的遗憾匹配和具有最佳收敛性的在线学习：零和游戏中的桥梁理论与实践
链接：https://arxiv.org/abs/2510.04407
【5】Challenge on Optimization of Context Collection for Code Completion
标题：优化上下文收集以实现代码完成的挑战
链接：https://arxiv.org/abs/2510.04349
备注：7 pages, 3 figures, 5 tables. A report on the Context Collection Workshop co-located with ASE'25
【6】Optimal Scaling Needs Optimal Norm
标题：最佳规模需要最佳规范
链接：https://arxiv.org/abs/2510.03871
【7】HOFLON: Hybrid Offline Learning and Online Optimization for Process Start-Up and Grade-Transition Control
标题：HOFLON：混合离线学习和在线优化，用于流程启动和年级过渡控制
链接：https://arxiv.org/abs/2510.03830
备注：31 pages, 15 figures, submitted to Computers and Chemical Engineering
【8】Distributed Low-Communication Training with Decoupled Momentum Optimization
标题：具有解耦动量优化的分布式低通信训练
链接：https://arxiv.org/abs/2510.03371
备注：NeurIPS 2025 - DynaFront 2025: Dynamics at the Frontiers of Optimization, Sampling, and Games Workshop
【9】Learning Pareto-Optimal Pandemic Intervention Policies with MORL
标题：用MORL学习帕累托最优流行病干预政策
链接：https://arxiv.org/abs/2510.03340
【10】Revoking Amnesia: RL-based Trajectory Optimization to Resurrect Erased Concepts in Diffusion Models
标题：撤销错觉：基于RL的轨迹优化以恢复扩散模型中被删除的概念
链接：https://arxiv.org/abs/2510.03302
备注：21 pages, 10 figures
【11】Single-Core Superscalar Optimization of Clifford Neural Layers
标题：Clifford神经层的单核超标量优化
链接：https://arxiv.org/abs/2510.03290
备注：9 pages
【12】Zeroth-Order Methods for Stochastic Nonconvex Nonsmooth Composite Optimization
标题：随机非凸非光滑组合优化的零阶方法
链接：https://arxiv.org/abs/2510.04446
【13】Optimal Computation from Fluctuation Responses
标题：波动响应的最优计算
链接：https://arxiv.org/abs/2510.03900
备注：10 pages, 6 figures; this https URL
【14】Composite Optimization with Error Feedback: the Dual Averaging Approach
标题：具有误差反馈的复合优化：双重平均方法
链接：https://arxiv.org/abs/2510.03507
【15】Quantile-Scaled Bayesian Optimization Using Rank-Only Feedback
标题：基于等级反馈的分位数贝叶斯优化
链接：https://arxiv.org/abs/2510.03277
备注：28 pages, 7 figures
预测|估计(22篇)
【1】ResCP: Reservoir Conformal Prediction for Time Series Forecasting
标题：ResCP：时间序列预测的储层保形预测
链接：https://arxiv.org/abs/2510.05060
【2】Feasibility-Aware Decision-Focused Learning for Predicting Parameters in the Constraints
标题：用于预测约束参数的基于灵活性的以决策为中心的学习
链接：https://arxiv.org/abs/2510.04951
【3】Benchmarking M-LTSF: Frequency and Noise-Based Evaluation of Multivariate Long Time Series Forecasting Models
标题：基准M-LTSF：基于频率和噪音的多元长时间序列预测模型评估
链接：https://arxiv.org/abs/2510.04900
备注：Number of pages: 13 Number of figures: 16 Number of Tables: 1 Submitted to: IEEE Transactions on Signal Processing
【4】Beyond the Seen: Bounded Distribution Estimation for Open-Vocabulary Learning
标题：超越所见：开放词汇学习的有界分布估计
链接：https://arxiv.org/abs/2510.04770
【5】Real-time Prediction of Urban Sound Propagation with Conditioned Normalizing Flows
标题：条件规范化流实时预测城市声音传播
链接：https://arxiv.org/abs/2510.04510
【6】Learning to Predict Chaos: Curriculum-Driven Training for Robust Forecasting of Chaotic Dynamics
标题：学习预测混乱：课程驱动的混乱动力学稳健预测训练
链接：https://arxiv.org/abs/2510.04342
备注：MIT URTC Technical Paper (Oral), 5 pages, 4 figures
【7】Crash Severity Prediction Using Deep Learning Approaches: A Hybrid CNN-RNN Framework
标题：使用深度学习方法进行碰撞严重度预测：一个混合的CNN-RNN框架
链接：https://arxiv.org/abs/2510.04316
【8】Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy
标题：使用Softmax熵的累积量展开探索下一个代币预测的几何形状
链接：https://arxiv.org/abs/2510.04285
备注：14 pages, 7 figures. Poster at HiLD 2025: 3rd Workshop on High-dimensional Learning Dynamics
【9】PhaseFormer: From Patches to Phases for Efficient and Effective Time Series Forecasting
标题：PhaseFormer：从补丁到阶段，实现高效且有效的时间序列预测
链接
：https://arxiv.org/abs/2510.04134
【10】Incorporating Multivariate Consistency in ML-Based Weather Forecasting with Latent-space Constraints
标题：具有潜空间约束的ML天气预报中的多元一致性计算
链接：https://arxiv.org/abs/2510.04006
【11】Optimising Battery Energy Storage System Trading via Energy Market Operator Price Forecast
标题：通过能源市场运营商价格预测优化电池储能系统交易
链接：https://arxiv.org/abs/2510.03657
【12】Is it Bigger than a Breadbox: Efficient Cardinality Estimation for Real World Workloads
标题：它比面包盒还大吗：现实世界工作负载的有效基数估计
链接：https://arxiv.org/abs/2510.03386
【13】Cross-Modal Reconstruction Pretraining for Ramp Flow Prediction at Highway Interchanges
标题：高速公路立交匝道流量预测的跨模式重建预训练
链接：https://arxiv.org/abs/2510.03381
【14】Physics-informed Neural-operator Predictive Control for Drag Reduction in Turbulent Flows
标题：基于物理信息的神经操作员预测控制用于湍流中减阻
链接：https://arxiv.org/abs/2510.03360
【15】High Cycle S-N curve prediction for Al 7075-T6 alloy using Recurrent Neural Networks (RNNs)
标题：利用回归神经网络（RNN）预测Al 7075-T6合金的高周S-N曲线
链接：https://arxiv.org/abs/2510.03355
【16】Numerion: A Multi-Hypercomplex Model for Time Series Forecasting
标题：Numerion：时间序列预测的多超复杂模型
链接：https://arxiv.org/abs/2510.03251
【17】Real-Time Brain Biomechanics Prediction with Neural Operators: Toward Clinically Deployable Traumatic Brain Injury Models
标题：使用神经运算符进行实时脑生物力学预测：迈向临床可部署的创伤性脑损伤模型
链接：https://arxiv.org/abs/2510.03248
【18】VIFO: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion
标题：VIFO：通过跨模式融合实现视觉特征赋予的多元时间序列预测
链接：https://arxiv.org/abs/2510.03244
【19】A Noise Resilient Approach for Robust Hurst Exponent Estimation
标题：鲁棒Hurst指数估计的抗噪方法
链接：https://arxiv.org/abs/2510.04811
【20】Predictive economics: Rethinking economic methodology with machine learning
标题：预测经济学：用机器学习重新思考经济方法论
链接：https://arxiv.org/abs/2510.04726
备注：8 pages
【21】A Universal Deep Learning Force Field for Molecular Dynamic Simulation and Vibrational Spectra Prediction
标题：用于分子动力学模拟和振动光谱预测的通用深度学习场
链接：https://arxiv.org/abs/2510.04227
备注：19 pages, 5 figures
【22】Improving S&P 500 Volatility Forecasting through Regime-Switching Methods
标题：通过制度转换方法改进标准普尔500指数波动率预测
链接：https://arxiv.org/abs/2510.03236
其他神经网络|深度学习|模型|建模(60篇)
【1】ResMimic: From General Motion Tracking to Humanoid Whole-body Loco-Manipulation via Residual Learning
标题：Resmimic：从一般运动跟踪到通过剩余学习的仿人全身部位操纵
链接：https://arxiv.org/abs/2510.05070
备注：9 pages, 8 figures
【2】Modeling Student Learning with 3.8 Million Program Traces
标题：用380万个项目轨迹模拟学生学习
链接：https://arxiv.org/abs/2510.05056
【3】Focused Skill Discovery: Learning to Control Specific State Variables while Minimizing Side Effects
标题：重点技能发现：学会控制特定状态变量，同时最大限度地减少副作用
链接：https://arxiv.org/abs/2510.04901
备注：Reinforcement Learning Journal 2025
【4】On the Hardness of Learning Regular Expressions
标题：论学习规则式的难度
链接：https://arxiv.org/abs/2510.04834
【5】Directional Sheaf Hypergraph Networks: Unifying Learning on Directed and Undirected Hypergraphs
标题：有向层超图网络：统一有向和无向超图的学习
链接：https://arxiv.org/abs/2510.04727
【6】How does the optimizer implicitly bias the model merging loss landscape?
标题：优化器如何隐式地偏向模型合并损失格局？
链接：https://arxiv.org/abs/2510.04686
备注：preprint
【7】IMLP: An Energy-Efficient Continual Learning Method for Tabular Data Streams
标题：IMLP：表格数据流的节能持续学习方法
链接：https://arxiv.org/abs/2510.04660
【8】Compressed Concatenation of Small Embedding Models
标题：小型嵌入模型的压缩级联
链接：https://arxiv.org/abs/2510.04626
【9】Improved probabilistic regression using diffusion models
标题：使用扩散模型改进的概率回归
链接：https://arxiv.org/abs/2510.04583
【10】SONA: Learning Conditional, Unconditional, and Mismatching-Aware Discriminator
标题：SONA：学习有条件、无条件和错配意识辨别器
链接：https://arxiv.org/abs/2510.04576
备注：24 pages with 9 figures
【11】Categorical Invariants of Learning Dynamics
标题：学习动力学的类别不变量
链接：https://arxiv.org/abs/2510.04376
【12】GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks
标题：GDPval：评估人工智能模型在现实世界具有经济价值的任务上的性能
链接：https://arxiv.org/abs/2510.04374
【13】Arithmetic-Mean $μ$P for Modern Architectures: A Unified Learning-Rate Scale for CNNs and ResNets
标题：现代架构的算术平均$μ$P：CNN和ResNet的统一学习率量表
链接：https://arxiv.org/abs/2510.04327
备注：Preprint. Under review at ICLR 2026
【14】Influence branching for learning to solve mixed-integer programs online
标题：影响分支学习在线求解混合整数规划
链接：https://arxiv.org/abs/2510.04273
备注：11 pages
【15】Spectral Alignment as Predictor of Loss Explosion in Neural Network Training
标题：频谱对齐作为神经网络训练中损失爆炸的预测因素
链接：https://arxiv.org/abs/2510.04202
备注：18 pages, 8 figures
【16】Modeling Time Series Dynamics with Fourier Ordinary Differential Equations
标题：时间序列动力学的Fourier常微分方程建模
链接：https://arxiv.org/abs/2510.04133
备注：8 pages, 7 figures, conference
【17】Learning-Based Hashing for ANN Search: Foundations and Early Advances
标题：ANN搜索的基于学习的哈希：基础和早期进展
链接：https://arxiv.org/abs/2510.04127
【18】On the Statistical Query Complexity of Learning Semiautomata: a Random Walk Approach
标题：学习半自动机的统计查询复杂性：随机游走方法
链接：https://arxiv.org/abs/2510.04115
备注：42 pages
【19】Wasserstein projection distance for fairness testing of regression models
标题：回归模型公平性测试的沃瑟斯坦投影距离
链接：https://arxiv.org/abs/2510.04114
【20】Why Cannot Neural Networks Master Extrapolation? Insights from Physical Laws
标题：为什么神经网络无法掌握外推？物理定律的见解
链接：https://arxiv.org/abs/2510.04102
【21】Early-Warning of Thunderstorm-Driven Power Outages with a Two-Stage Machine Learning Model
标题：基于两阶段机器学习模型的雷暴停电预警
链接：https://arxiv.org/abs/2510.03959
备注：23 pages (main), 70 pages incl. appendices; figures & tables as in manuscript. Code (main figure, synthetic data): this https URL License: CC BY 4.0 (preprint)
【22】Transductive and Learning-Augmented Online Regression
标题：转化和学习增强在线回归
链接：https://arxiv.org/abs/2510.03917
【23】Fair Minimum Labeling: Efficient Temporal Network Activations for Reachability and Equity
标题：公平的最低标签：有效的时间网络激活以实现可达性和公平性
链接：https://arxiv.org/abs/2510.03899
备注：Accepted at NeurIPS 2025
【24】Cellular Learning: Scattered Data Regression in High Dimensions via Voronoi Cells
标题：细胞学习：通过Voronoi细胞进行多维分散数据回归
链接：https://arxiv.org/abs/2510.03810
备注：15 pages + 2 pages references; 3 figures; 4 tables; 1 algorithm
【25】Bridging the Gap Between Multimodal Foundation Models and World Models
标题：弥合多模式基础模型与世界模型之间的差距
链接：https://arxiv.org/abs/2510.03727
备注：PhD thesis
【26】Mapping Rio de Janeiro's favelas: general-purpose vs. satellite-specific neural networks
标题：绘制里约热内卢贫民窟地图：通用神经网络与卫星专用神经网络
链接：https://arxiv.org/abs/2510.03725
备注：6 pages, 1 figure, 1 table. Presented at the 21st Brazilian Symposium on Remote Sensing (SBSR 2025)
【27】Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models
标题：LAION-400 M以人为本的注释：审计偏见及其向模型的转移
链接：https://arxiv.org/abs/2510.03721
备注：48 pages
【28】From Moments to Models: Graphon Mixture-Aware Mixup and Contrastive Learning
标题：从时刻到模型：Graphon混合感知混合和对比学习
链接：https://arxiv.org/abs/2510.03690
【29】Implicit Models: Expressive Power Scales with Test-Time Compute
标题：隐式模型：具有测试时间计算的表现能力量表
链接：https://arxiv.org/abs/2510.03638
【30】Latent Mixture of Symmetries for Sample-Efficient Dynamic Learning
标题：用于样本高效动态学习的潜在对称混合
链接：https://arxiv.org/abs/2510.03578
备注：30 pages, 6 figures
【31】BEKAN: Boundary condition-guaranteed evolutionary Kolmogorov-Arnold networks with radial basis functions for solving PDE problems
标题：BEKAN：具有用于解决PCE问题的辐射基函数的边界条件保证进化Kolmogorov-Arnold网络
链接：https://arxiv.org/abs/2510.03576
备注：29 pages, 22 figures
【32】Longitudinal Flow Matching for Trajectory Modeling
标题：用于弹道建模的纵向流匹配
链接：https://arxiv.org/abs/2510.03569
【33】How to Set $β_1, β_2$ in Adam: An Online Learning Perspective
标题：如何在Adam中设置$β_1、β_2$：在线学习的角度
链接：https://arxiv.org/abs/2510.03478
备注：15 pages
【34】Paris: A Decentralized Trained Open-Weight Diffusion Model
标题：巴黎：分散训练的开放权重扩散模型
链接：https://arxiv.org/abs/2510.03434
【35】Training Variation of Physically-Informed Deep Learning Models
标题：物理知情深度学习模型的训练变体
链接：https://arxiv.org/abs/2510.03416
【36】Studying the Korean Word-Chain Game with RLVR:Mitigating Reward Conflicts via Curriculum Learning
标题：使用WLVR研究韩国字链游戏：通过课程学习缓解奖励冲突
链接：https://arxiv.org/abs/2510.03394
备注：10 pages
【37】Provenance Networks: End-to-End Exemplar-Based Explainability
标题：起源网络：端到端基于示例的解释性
链接：https://arxiv.org/abs/2510.03361
【38】Pilot selection in the era of Virtual reality: algorithms for accurate and interpretable machine learning models
标题：虚拟现实时代的飞行员选择：准确且可解释的机器学习模型的算法
链接：https://arxiv.org/abs/2510.03345
【39】Machine Learning Workflows in Climate Modeling: Design Patterns and Insights from Case Studies
标题：气候建模中的机器学习工作流程：来自案例研究的设计模式和见解
链接：https://arxiv.org/abs/2510.03305
备注：Supplement
【40】MemMamba: Rethinking Memory Patterns in State Space Model
标题：MemMamba：重新思考状态空间模型中的记忆模式
链接：https://arxiv.org/abs/2510.03279
【41】QuadEnhancer: Leveraging Quadratic Transformations to Enhance Deep Neural Networks
标题：QuadEnhancer：利用二次变换增强深度神经网络
链接：https://arxiv.org/abs/2510.03276
备注：39th Conference on Neural Information Processing Systems (NeurIPS 2025)
【42】MindCraft: How Concept Trees Take Shape In Deep Models
标题：MindCraft：概念树如何在深度模型中成形
链接：https://arxiv.org/abs/2510.03265
【43】Memory Self-Regeneration: Uncovering Hidden Knowledge in Unlearned Models
标题：记忆自我再生：揭露未学习模型中隐藏的知识
链接：https://arxiv.org/abs/2510.03263
【44】Data-Driven Temperature Modelling of Machine Tools by Neural Networks: A Benchmark
标题：通过神经网络对机械工具进行数据驱动的温度建模：基准
链接：https://arxiv.org/abs/2510.03261
【45】Light Differentiable Logic Gate Networks
标题：轻可微逻辑门网络
链接：https://arxiv.org/abs/2510.03250
【46】Frequency-Aware Model Parameter Explorer: A new attribution method for improving explainability
标题：频率感知模型参数资源管理器：一种用于提高可解释性的新归因方法
链接：https://arxiv.org/abs/2510.03245
备注：Preprint
【47】Neural Network Surrogates for Free Energy Computation of Complex Chemical Systems
标题：复杂化学体系自由能计算的神经网络替代方法
链接：https://arxiv.org/abs/2510.01396
备注：6 pages, 4 figures. This work has already been accepted for presentation in The 29th International Computer Science and Engineering Conference (ICSEC) 2025, Chiang Mai, Thailand, and will be published in IEEE Xplore
【48】Bayesian Distributional Models of Executive Functioning
标题：执行职能的Bayesian分布模型
链接：https://arxiv.org/abs/2510.00387
备注：42 pages, 8 figures, 1 table
【49】Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning
标题：拥抱离散搜索：因果结构学习的合理方法
链接：https://arxiv.org/abs/2510.04970
【50】Gini-based Model Monitoring: A General Framework with an Application to Non-life Insurance Pricing
标题：基于基尼的模型监控：应用于非寿险定价的通用框架
链接：https://arxiv.org/abs/2510.04556
【51】Learning Linear Regression with Low-Rank Tasks in-Context
标题：在上下文中使用低级别任务学习线性回归
链接：https://arxiv.org/abs/2510.04548
【52】Quantum generative model on bicycle-sharing system and an application
标题：共享自行车系统的量子生成模型及其应用
链接：https://arxiv.org/abs/2510.04512
备注：8 pages, 11 figures
【53】Benchmarking atmospheric circulation variability in an AI emulator, ACE2, and a hybrid model, NeuralGCM
标题：在人工智能模拟器ACE 2和混合模型NeuralGCM中对大气环流变化进行基准测试
链接：https://arxiv.org/abs/2510.04466
备注：12 pages, 4 main figures, 6 supplementary figures
【54】Inverse Mixed-Integer Programming: Learning Constraints then Objective Functions
标题：反向混合时间表编程：学习约束，然后是目标函数
链接：https://arxiv.org/abs/2510.04455
备注：33 pages
【55】Learning Survival Models with Right-Censored Reporting Delays
标题：学习具有右审查报告延迟的生存模型
链接：https://arxiv.org/abs/2510.04421
备注：21 pages, 3 figures, 4 tables
【56】Quantizer Design for Finite Model Approximations, Model Learning, and Quantized Q-Learning for MDPs with Unbounded Spaces
标题：具有无界空间的MPP的有限模型逼近、模型学习和量化Q学习的量化器设计
链接：https://arxiv.org/abs/2510.04355
【57】A Contextual Quality Reward Model for Reliable and Efficient Best-of-N Sampling
标题：可靠有效的N最佳抽样的上下文质量奖励模型
链接：https://arxiv.org/abs/2510.04087
【58】Spectral Thresholds for Identifiability and Stability:Finite-Sample Phase Transitions in High-Dimensional Learning
标题：可识别性和稳定性的谱阈值：多维学习中的伪样本相转变
链接：https://arxiv.org/abs/2510.03809
【59】Quantum feature-map learning with reduced resource overhead
标题：量子特征图学习，减少资源负担
链接：https://arxiv.org/abs/2510.03389
备注：17 pages, 9 figures
【60】Machine Learning and Control: Foundations, Advances, and Perspectives
标题：机器学习和控制：基础、进展和前景
链接：https://arxiv.org/abs/2510.03303
其他(93篇)
【1】MICROTRIPS: MICRO-geography TRavel Intelligence and Pattern Synthesis
标题：MicroTRIPS：微地理TRavel智能和模式合成
链接：https://arxiv.org/abs/2510.05080
【2】Rethinking Langevin Thompson Sampling from A Stochastic Approximation Perspective
标题：从随机逼近的角度重新思考Langevin Thompson抽样
链接：https://arxiv.org/abs/2510.05023
备注：39 pages, 3 figures, 2 tables
【3】Think Then Embed: Generative Context Improves Multimodal Embedding
标题：思考然后嵌入：生成上下文改进多模式嵌入
链接：https://arxiv.org/abs/2510.05014
【4】Power Transform Revisited: Numerically Stable, and Federated
标题：重新审视权力转型：数字稳定和联邦
链接：https://arxiv.org/abs/2510.04995
备注：25 pages
【5】Federated Computation of ROC and PR Curves
标题：ROC和PR曲线的联邦计算
链接：https://arxiv.org/abs/2510.04979
备注：23 pages
【6】StructuralDecompose: A Modular Framework for Robust Time Series Decomposition in R
标题：StructalDecomose：R中鲁棒时间序列分解的模块化框架
链接：https://arxiv.org/abs/2510.04974
备注：8 pages, 4 figures. Part of the R package StructuralDecompose (this https URL)
【7】On Structured State-Space Duality
标题：论结构化状态空间二元性
链接：https://arxiv.org/abs/2510.04944
【8】Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking
标题：平等主义梯度下降：加速Grokking的简单方法
链接：https://arxiv.org/abs/2510.04930
【9】Glocal Information Bottleneck for Time Series Imputation
标题：时间序列插补的GlLocal信息瓶颈
链接：https://arxiv.org/abs/2510.04910
【10】DP-HYPE: Distributed Differentially Private Hyperparameter Search
标题：DP-PHPE：分布式差异专用超参数搜索
链接：https://arxiv.org/abs/2510.04902
【11】CLEAR-IR: Clarity-Enhanced Active Reconstruction of Infrared Imagery
标题：Clear-IR：红外图像的清晰度增强主动重建
链接：https://arxiv.org/abs/2510.04883
备注：8 pages, 8 figures
【12】Distributionally Robust Causal Abstractions
标题：分布稳健的因果抽象
链接：https://arxiv.org/abs/2510.04842
【13】Bond-Centered Molecular Fingerprint Derivatives: A BBBP Dataset Study
标题：以键为中心的分子指纹衍生物：BBBP数据集研究
链接：https://arxiv.org/abs/2510.04837
备注：14 pages, 10 figures, 1 table
【14】When Do Credal Sets Stabilize? Fixed-Point Theorems for Credal Set Updates
标题：Credal套装什么时候稳定？Credal集更新的不稳定点定理
链接：https://arxiv.org/abs/2510.04769
【15】Provable Affine Identifiability of Nonlinear CCA under Latent Distributional Priors
标题：潜在分布先验下非线性PCA的可证仿射可识别性
链接：https://arxiv.org/abs/2510.04758
【16】ViTs: Teaching Machines to See Time Series Anomalies Like Human Experts
标题：ViTS：教机器像人类专家一样看到时间序列异常
链接：https://arxiv.org/abs/2510.04710
备注：13 pages
【17】Multilingual Routing in Mixture-of-Experts
标题：混合专家中的多语言路由
链接：https://arxiv.org/abs/2510.04694
【18】Semantic Channel Equalization Strategies for Deep Joint Source-Channel Coding
标题：深度联合源通道编码的语义通道均衡策略
链接：https://arxiv.org/abs/2510.04674
备注：Proceedings of IEEE Globecom 2025 Workshops
【19】Fairness in Repeated Matching: A Maximin Perspective
标题：重复匹配中的公平性：最大限度的视角
链接：https://arxiv.org/abs/2510.04624
【20】Busemann Functions in the Wasserstein Space: Existence, Closed-Forms, and Applications to Slicing
标题：Wasserstein空间中的Busemann函数：存在性、封闭形式及其在切片中的应用
链接：https://arxiv.org/abs/2510.04579
【21】Challenger-Based Combinatorial Bandits for Subcarrier Selection in OFDM Systems
标题：基于队列的组合带宽用于CDMA系统中子载体选择
链接：https://arxiv.org/abs/2510.04559
备注：6 pages
【22】Post-training quantization of vision encoders needs prefixing registers
标题：视觉编码器的训练后量化需要添加寄存器
链接：https://arxiv.org/abs/2510.04547
【23】Expand Neurons, Not Parameters
标题：扩展神经元，而不是参数
链接：https://arxiv.org/abs/2510.04500
备注：10 pages, 6 figures
【24】Deep vs. Shallow: Benchmarking Physics-Informed Neural Architectures on the Biharmonic Equation
标题：深层与浅层：双调和方程上的物理信息神经架构基准测试
链接：https://arxiv.org/abs/2510.04490
备注：16 Pages, 7 Figures and 1 Table. Submitted and accepted at Machine Learning and the Physical Sciences Workshop at the 39th conference on Neural Information Processing Systems (NeurIPS)
【25】Forking-Sequences
标题：分叉序列
链接：https://arxiv.org/abs/2510.04487
【26】Domain Generalization: A Tale of Two ERMs
标题：领域概括：两个ERM的故事
链接：https://arxiv.org/abs/2510.04441
【27】Partial Information Decomposition via Normalizing Flows in Latent Gaussian Distributions
标题：通过潜高斯分布的正规化流进行部分信息分解
链接：https://arxiv.org/abs/2510.04417
备注：NeurIPS 2025
【28】Time Is Effort: Estimating Human Post-Editing Time for Grammar Error Correction Tool Evaluation
标题：时间就是努力：估计语法错误纠正工具评估的人类后期编辑时间
链接：https://arxiv.org/abs/2510.04394
备注：Accepted for publication in the 4th HCI+NLP Workshop (Fourth Workshop on Bridging Human-Computer Interaction and Natural Language Processing; part of EMNLP 2025)
【29】Improving Consistency in Retrieval-Augmented Systems with Group Similarity Rewards
标题：通过组相似性奖励提高检索增强系统的一致性
链接：https://arxiv.org/abs/2510.04392
备注：Accepted at NeurIPS 2025 Workshop on Reliable ML from Unreliable Data
【30】Environment-Aware Indoor LoRaWAN Path Loss: Parametric Regression Comparisons, Shadow Fading, and Calibrated Fade Margins
标题：环境意识室内LoRawan路径损失：参数回归比较、阴影褪色和校准的褪色边缘
链接：https://arxiv.org/abs/2510.04346
备注：Code: this https URL
【31】Pitch-Conditioned Instrument Sound Synthesis From an Interactive Timbre Latent Space
标题：从交互式音色潜在空间进行音调调节乐器声音合成
链接：https://arxiv.org/abs/2510.04339
备注：8 pages, accepted to the Proceedings of the 28-th Int. Conf. on Digital Audio Effects (DAFx25) - demo: this https URL
【32】Towards Fast Option Pricing PDE Solvers Powered by PIELM
标题：迈向快速期权定价由PIELM支持的DOE求解器
链接：https://arxiv.org/abs/2510.04322
备注：6 Pages, 5 Figures, 3 Tables
【33】Activation Steering with a Feedback Controller
标题：具有反馈控制器的激活转向
链接：https://arxiv.org/abs/2510.04309
备注：9 pages in the main text. Under Review
【34】Wave-PDE Nets: Trainable Wave-Equation Layers as an Alternative to Attention
标题：Wave-PCE网络：可训练的波动方程层作为注意力的替代方案
链接：https://arxiv.org/abs/2510.04304
备注：PRICAI 2025 Oral, 9 pages, 3 figures
【35】Efficient Latent Variable Causal Discovery: Combining Score Search and Targeted Testing
标题：高效的潜在变量因果发现：结合分数搜索和目标测试
链接：https://arxiv.org/abs/2510.04263
备注：30 pages, 23 figures, 6 tables
【36】Truncated Kernel Stochastic Gradient Descent with General Losses and Spherical Radial Basis Functions
标题：具有一般损失的截断核随机梯度下降和球面径向基函数
链接：https://arxiv.org/abs/2510.04237
备注：54 pages, 20 figures
【37】COSMO-RL: Towards Trustworthy LMRMs via Joint Safety and Stability
标题：COSMO-RL：通过关节安全性和稳定性实现可信赖的LMRM
链接：https://arxiv.org/abs/2510.04196
【38】Fine Tuning Methods for Low-resource Languages
标题：低资源语言的微调方法
链接：https://arxiv.org/abs/2510.04139
【39】Efficient Manifold-Constrained Neural ODE for High-Dimensional Datasets
标题：用于多维数据集的高效的Manifold约束神经ODE
链接：https://arxiv.org/abs/2510.04138
备注：8 pages; 7 figures; conference IJCNN
【40】On the Limitations and Capabilities of Position Embeddings for Length Generalization
标题：关于长度概括位置嵌入的局限性和能力
链接：https://arxiv.org/abs/2510.04130
【41】What Scales in Cross-Entropy Scaling Law?
标题：交叉熵标度定律中有哪些标度？
链接：https://arxiv.org/abs/2510.04067
【42】Sharp Lower Bounds for Linearized ReLU^k Approximation on the Sphere
标题：球体上线性化ReLU ' k逼近的尖锐下限
链接：https://arxiv.org/abs/2510.04060
【43】Multi-Class Support Vector Machine with Differential Privacy
标题：具有差异隐私的多类支持向量机
链接：https://arxiv.org/abs/2510.04027
备注：NeurIPS 2025
【44】Zephyrus: An Agentic Framework for Weather Science
标题：Zephyrus：天气科学的抽象框架
链接：https://arxiv.org/abs/2510.04017
【45】Replacing Softmax Similarity with a Sharpened Angular Similarity: Theory and Practice of Scaling To Billion-Context Attention
标题：用尖锐的角度相似性取代Softmax相似性：扩大十亿背景关注度的理论与实践
链接：https://arxiv.org/abs/2510.04008
备注：28 pages, 7 figures
【46】Multi-Modal Multi-Task Semantic Communication: A Distributed Information Bottleneck Perspective
标题：多模式多任务语义通信：分布式信息瓶颈视角
链接：https://arxiv.org/abs/2510.04000
【47】Beyond Static Evaluation: Rethinking the Assessment of Personalized Agent Adaptability in Information Retrieval
标题：超越静态评估：重新思考信息检索中个性化代理适应性评估
链接：https://arxiv.org/abs/2510.03984
备注：None
【48】Beyond Softmax: A New Perspective on Gradient Bandits
标题：超越Softmax：梯度盗贼的新视角
链接：https://arxiv.org/abs/2510.03979
【49】What Can You Do When You Have Zero Rewards During RL?
标题：当RL期间的奖励为零时，你能做什么？
链接：https://arxiv.org/abs/2510.03971
【50】Generalized Fitted Q-Iteration with Clustered Data
标题：具有简化数据的广义拟适Q迭代
链接：https://arxiv.org/abs/2510.03912
【51】The Hidden Game Problem
标题：隐藏的游戏问题
链接：https://arxiv.org/abs/2510.03845
【52】Smart Paste: Automatically Fixing Copy/Paste for Google Developers
标题：智能粘贴：为Google开发人员自动修复复制/粘贴
链接：https://arxiv.org/abs/2510.03843
备注：11 pages
【53】Technical note on Fisher Information for Robust Federated Cross-Validation
标题：关于用于稳健联邦交叉验证的Fisher信息的技术说明
链接：https://arxiv.org/abs/2510.03838
【54】Proximal Diffusion Neural Sampler
标题：近端扩散神经采样器
链接：https://arxiv.org/abs/2510.03824
备注：31 pages, 12 figures
【55】Robust Batched Bandits
标题：强大的批量强盗
链接：https://arxiv.org/abs/2510.03798
备注：39 pages
【56】Neural Low-Discrepancy Sequences
标题：神经低方差序列
链接：https://arxiv.org/abs/2510.03745
【57】Cost Efficient Fairness Audit Under Partial Feedback
标题：部分反馈下的成本高效公平审计
链接：https://arxiv.org/abs/2510.03734
备注：Accepted at NeurIPS 2025 RegML Workshop; Reliable ML Workshop
【58】REG: A Regularization Optimizer for Robust Training Dynamics
标题：REG：鲁棒训练动态的正规化优化器
链接：https://arxiv.org/abs/2510.03691
【59】Group Policy Gradient
标题：团体政策梯度
链接：https://arxiv.org/abs/2510.03679
【60】Towards Sampling Data Structures for Tensor Products in Turnstile Streams
标题：转向转门流中张量产品的采样数据结构
链接：https://arxiv.org/abs/2510.03678
【61】In-Vivo Training for Deep Brain Stimulation
标题：脑部深层刺激的体内训练
链接：https://arxiv.org/abs/2510.03643
【62】Neural Bayesian Filtering
标题：神经Bayesian过滤
链接：https://arxiv.org/abs/2510.03614
【63】Explore the Loss space with Hill-ADAM
标题：与Hill-ADAM一起探索损失空间
链接：https://arxiv.org/abs/2510.03613
备注：14-15 pages
【64】D2 Actor Critic: Diffusion Actor Meets Distributional Critic
标题：D2演员评论家：扩散演员会见发行评论家
链接：https://arxiv.org/abs/2510.03508
【65】Trajectory Data Suffices for Statistically Efficient Policy Evaluation in Finite-Horizon Offline RL with Linear $q^π$-Realizability and Concentrability
标题：轨迹数据足以在具有线性$q & pi $-可实现性和集中性的虚拟地平线离线RL中进行统计有效的政策评估
链接：https://arxiv.org/abs/2510.03494
【66】The Argument is the Explanation: Structured Argumentation for Trust in Agents
标题：论点就是解释：对代理人信任的结构化论点
链接：https://arxiv.org/abs/2510.03442
备注：8 pages, 4 figures, 6 tables, submitted to IAAI-26
【67】Generalized Orders of Magnitude for Scalable, Parallel, High-Dynamic-Range Computation
标题：可扩展、并行、高动态范围计算的广义量级
链接：https://arxiv.org/abs/2510.03426
备注：18 pages, 4 figures (main text). 14 pages, 21 figures (appendix)
【68】Constant in an Ever-Changing World
标题：在不断变化的世界中保持不变
链接：https://arxiv.org/abs/2510.03330
备注：in Chinese language
【69】The View From Space: Navigating Instrumentation Differences with EOFMs
标题：从太空看：用EOFM导航仪器差异
链接：https://arxiv.org/abs/2510.03316
【70】Decomposing Attention To Find Context-Sensitive Neurons
标题：分解注意力以找到上下文敏感神经元
链接：https://arxiv.org/abs/2510.03315
备注：10 pages, 7 figures. Submitted to the Mechanistic Interpretability Workshop at NeurIPS 2025
【71】Multimodal Arabic Captioning with Interpretable Visual Concept Integration
标题：具有可解释视觉概念集成的多模式阿拉伯语字幕
链接：https://arxiv.org/abs/2510.03295
【72】From Score Distributions to Balance: Plug-and-Play Mixture-of-Experts Routing
标题：从分数分布到平衡：即插即用混合专家路由
链接：https://arxiv.org/abs/2510.03293
【73】Why mask diffusion does not work
标题：为什么口罩扩散不起作用
链接：https://arxiv.org/abs/2510.03289
【74】WAREX: Web Agent Reliability Evaluation on Existing Benchmarks
标题：WAREX：基于现有基准的Web代理可靠性评估
链接：https://arxiv.org/abs/2510.03285
【75】Quantifying constraint hierarchies in Bayesian PINNs via per-constraint Hessian decomposition
标题：通过逐约束Hessian分解量化Bayesian PINN中的约束层次结构
链接：https://arxiv.org/abs/2510.03278
备注：5 pages, 2 figures
【76】General Exploratory Bonus for Optimistic Exploration in RLHF
标题：RL HF乐观探索的一般探索奖金
链接：https://arxiv.org/abs/2510.03269
【77】Rethinking Inter-LoRA Orthogonality in Adapter Merging: Insights from Orthogonal Monte Carlo Dropout
标题：重新思考适配器合并中的LoRA间跨性：来自Orthogonal Monte Carlo Dropout的见解
链接：https://arxiv.org/abs/2510.03262
【78】Triple-BERT: Do We Really Need MARL for Order Dispatch on Ride-Sharing Platforms?
标题：Triple-BERT：我们真的需要MARL在拼车平台上进行订单调度吗？
链接：https://arxiv.org/abs/2510.03257
【79】Universal Multi-Domain Translation via Diffusion Routers
标题：通过扩散路由器的通用多域翻译
链接：https://arxiv.org/abs/2510.03252
【80】StructPrune: Structured Global Pruning asymptotics with $\mathcal{O}(\sqrt{N})$ GPU Memory
链接：https://arxiv.org/abs/2510.03246
【81】Causal Abstractions, Categorically Unified
标题：因果抽象，分类统一
链接：https://arxiv.org/abs/2510.05033
【82】Curiosity-Driven Co-Development of Action and Language in Robots Through Self-Exploration
标题：好奇心驱动通过自我探索实现机器人动作和语言的联合开发
链接：https://arxiv.org/abs/2510.05013
备注：26 pages, 14 pages of supplementary material
【83】Pivotal CLTs for Pseudolikelihood via Conditional Centering in Dependent Random Fields
标题：通过相依随机场条件定中心的伪随机迭代的关键CLT
链接：https://arxiv.org/abs/2510.04972
备注：73 pages, 1 figure
【84】Kernel ridge regression under power-law data: spectrum and generalization
标题：乘势数据下的核岭回归：谱和推广
链接：https://arxiv.org/abs/2510.04780
【85】Fisher-Bingham-like normalizing flows on the sphere
标题：球体上类似费舍尔-宾厄姆的正常化流动
链接：https://arxiv.org/abs/2510.04762
【86】Computing Wasserstein Barycenters through Gradient Flows
标题：通过梯度流计算沃瑟斯坦重心
链接：https://arxiv.org/abs/2510.04602
备注：4 Figures, 3 Tables, under review
【87】Perspectives on Stochastic Localization
标题：随机本地化的观点
链接：https://arxiv.org/abs/2510.04460
【88】spd-metrics-id: A Python Package for SPD-Aware Distance Metrics in Connectome Fingerprinting and Beyond
标题：spd-metrics-id：Connectome指纹识别及其他领域中SPD感知距离收件箱的Python包
链接：https://arxiv.org/abs/2510.04438
【89】Relative Information Gain and Gaussian Process Regression
标题：相对信息增益与高斯过程回归
链接：https://arxiv.org/abs/2510.04277
备注：28 pages
【90】Self-Speculative Masked Diffusions
标题：自我思考的蒙面扩散
链接：https://arxiv.org/abs/2510.03929
备注：32 pages, 7 figures, 3 tables
【91】The analogy theorem in Hoare logic
标题：霍尔逻辑中的类比定理
链接：https://arxiv.org/abs/2510.03685
【92】Bias and Coverage Properties of the WENDy-IRLS Algorithm
标题：WENDy-IRLS算法的偏差和覆盖特性
链接：https://arxiv.org/abs/2510.03365
【93】Assessing the impact of contact time on leachate chemistry from recycled concrete aggregates
标题：评估接触时间对再生混凝土骨料沥滤液化学性质的影响
链接：https://arxiv.org/abs/2510.03344
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
Python社区是高质量的Python/Django开发社区
本文地址：http://www.python88.com/topic/187594
本文地址：http://www.python88.com/topic/187594