机器学习学术速递[5.12]

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733篇

大模型相关(88篇)

【1】Beyond Red-Teaming: Formal Guarantees of LLM Guardrail Classifiers
标题：超越红队：LLM保证分类器的正式保证
链接：https://arxiv.org/abs/2605.10901

作者：Nikita Kezins,Urbas Ekka,Pascal Berrang,Luca Arnaboldi

【2】V4FinBench: Benchmarking Tabular Foundation Models, LLMs, and Standard Methods on Corporate Bankruptcy Prediction
标题：V4 FinBench：对表型基金会模型、LLM和公司破产预测的标准方法进行基准测试
链接：https://arxiv.org/abs/2605.10896

作者：Marcin Kostrzewa,Sebastian Tomczak,Roman Furman,Anna Poberezhna,Michał Furgała,Oleksii Furman,Maciej Zięba

【3】AssayBench: An Assay-Level Virtual Cell Benchmark for LLMs and Agents
标题：AssayBench：针对LLC和代理的Assay-level虚拟单元基准
链接：https://arxiv.org/abs/2605.10876

作者：Edward De Brouwer,Carl Edwards,Alexander Wu,Jenna Collier,Graham Heimberg,Xiner Li,Meena Subramaniam,Ehsan Hajiramezanali,David Richmond,Jan-Christian Hütter,Sara Mostafavi,Gabriele Scalia
备注：22 pages

【4】Compute Where it Counts: Self Optimizing Language Models
标题：在重要的地方计算：自我优化语言模型
链接：https://arxiv.org/abs/2605.10875

作者：Yash Akhauri,Mohamed S. Abdelfattah
备注：Accepted at ICML'26 Code: https://github.com/akhauriyash/SOL

【5】SLIM: Sparse Latent Steering for Interpretable and Property-Directed LLM-Based Molecular Editing
标题：SlIM：可解释和属性指导的LLM分子编辑的稀疏潜在引导
链接：https://arxiv.org/abs/2605.10831

作者：Mingxu Zhang,Yuhan Li,Lujundong Li,Dazhong Shen,Hui Xiong,Ying Sun

【6】LLMs for Secure Hardware Design and Related Problems: Opportunities and Challenges
标题：安全硬件设计及相关问题的法学硕士：机遇与挑战
链接：https://arxiv.org/abs/2605.10807

作者：Johann Knechtel,Ozgur Sinanoglu,Ramesh Karri
备注：Accepted for 2026 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)

【7】ConQuR: Corner Aligned Activation Quantization via Optimized Rotations for LLMs
标题：ConQuR：通过LLM优化旋转的角对齐激活量化
链接：https://arxiv.org/abs/2605.10793

作者：Chayne Thrash,Ali Abbasi,Soheil Kolouri

【8】DynaMiCS: Fine-tuning LLMs with Performance Constraints using Dynamic Mixtures
标题：DynaMiCS：使用动态混合对具有性能约束的LLM进行微调
链接：https://arxiv.org/abs/2605.10770

作者：Eleonora Gualdoni,Sonia Laguna,Louis Bethune,Joao Monteiro,Pierre Ablin,Marco Cuturi

【9】A Single-Layer Model Can Do Language Modeling
标题：单层模型可以进行语言建模
链接：https://arxiv.org/abs/2605.10643

作者：Zanmin Wang
备注：9 pages, 5 figures, 1 table. Code: https://github.com/steve-z-wang/grounded-prediction-network

【10】EnergyLens: Interpretable Closed-Form Energy Models for Multimodal LLM Inference Serving
标题：EnergyLens：用于多模式LLM推理服务的可解释封闭形式能量模型
链接：https://arxiv.org/abs/2605.10556

作者：Vittorio Palladino,Gianluca Palermo,Michael E. Papka,Zhiling Lan
备注：10 pages

【11】Equilibrium Residuals Expose Three Regimes of Matrix-Game Strategic Reasoning in Language Models
标题：均衡剩余揭示语言模型中矩阵博弈策略推理的三种机制
链接：https://arxiv.org/abs/2605.10410

作者：Wenhua Nie, Binhan Luo, Zijie Meng, Jyh-Shing Roger Jang, Ching-Wen Ma

【12】Valid Best-Model Identification for LLM Evaluation via Low-Rank Factorization
标题：通过低等级分解进行LLM评估的有效最佳模型识别
链接：https://arxiv.org/abs/2605.10405

作者：Elad Tolochinsky, Yaniv Tenzer, Yaniv Romano

【13】DP-LAC: Lightweight Adaptive Clipping for Differentially Private Federated Fine-tuning of Language Models
标题：DP-LAC：用于语言模型的差异私有联邦微调的轻量级自适应剪辑
链接：https://arxiv.org/abs/2605.10272

作者：Haaris Mehmood, Jie Xu, Karthikeyan Saravanan, Rogier Van Dalen, Mete Ozay
备注：Accepted at ICASSP 2026

【14】Teaching LLMs to See Graphs: Unifying Text and Structural Reasoning
标题：教法学硕士看图形：统一文本和结构推理
链接：https://arxiv.org/abs/2605.10247

作者：Dario Vajda

【15】Task-Aware Calibration: Provably Optimal Decoding in LLMs
标题：任务感知校准：LLM中可证明的最佳解码
链接：https://arxiv.org/abs/2605.10202

作者：Tim Tomov, Dominik Fuchsgruber, Rajeev Verma, Stephan Günnemann

【16】GELATO: Generative Entropy- and Lyapunov-based Adaptive Token Offloading for Device-Edge Speculative LLM Inference
标题：GELATO：基于生成性信息和Lyapunov的自适应令牌卸载，用于设备边缘推测LLM推理
链接：https://arxiv.org/abs/2605.10124

作者：Zengzipeng Tang, Yuxuan Sun, Wei Chen, Jianwen Ding, Bo Ai
备注：This work has been submitted to the IEEE for possible publication

【17】Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
标题：Metis：通过自我进化的元认知政策优化学习越狱LLC
链接：https://arxiv.org/abs/2605.10067

作者：Huilin Zhou, Jian Zhao, Yilu Zhong, Zhen Liang, Xiuyuan Chen, Yuchen Yuan, Tianle Zhang, Chi Zhang, Lan Zhang, Xuelong Li
备注：Accepted to the 43rd International Conference on Machine Learning (ICML 2026)

【18】TrajDLM: Topology-Aware Block Diffusion Language Model for Trajectory Generation
标题：TrajDLM：用于轨迹生成的具有全局感知的块扩散语言模型
链接：https://arxiv.org/abs/2605.10020

作者：Wilson Wongso, Lihuan Li, Arian Prabowo, Xiachong Lin, Baiyu Chen, Hao Xue, Flora D. Salim

【19】TeleResilienceBench: Quantifying Resilience for LLM Reasoning in Telecommunications
标题：TeleResilienceBench：量化电信LLM推理的弹性
链接：https://arxiv.org/abs/2605.09929

作者：Pranshav Gajjar,Emmanuel Ojo,Vijay K Shah

【20】Verifier-Free RL for LLMs via Intrinsic Gradient-Norm Reward
标题：通过固有的受试者规范奖励，为LLM提供免验证器RL
链接：https://arxiv.org/abs/2605.09920

作者：Xuexiang Wen,Hang Yu,Linchao Zhu,Gaoang Wang
备注：Accepted to Findings of ACL 2026

【21】NaiAD: Initiate Data-Driven Research for LLM Advertising
标题：NaiAD：启动LLM广告的数据驱动研究
链接：https://arxiv.org/abs/2605.09918

作者：Yihang Zhang,Zimeng Huang,Ren Zhai,Yipeng Kang,Tonghan Wang
备注：37 pages, 11 figures

【22】Nautilus Compass: Black-box Persona Drift Detection for Production LLM Agents
标题：Nautilus Compass：适用于生产LLM代理的黑匣子Persona漂移检测
链接：https://arxiv.org/abs/2605.09863

作者：Chunxiao Wang
备注：19 pages, 6 figures. MIT-licensed code + reproduction scripts at this http URL

【23】Concordia: Self-Improving Synthetic Tables for Federated LLMs
标题：Concordia：联邦LLM的自我改进合成表
链接：https://arxiv.org/abs/2605.09855

作者：Jimin Huang, Duanyu Feng, Nuo Chen, Xiaoyu Wang, Zhiqiang Zhang, Xueqing Peng, Mingquan Lin, Prayag Tiwari, Guojun Xiong, Alejandro Lopez-Lira, Sophia Ananiadou
备注：12 pages

【24】Exploration-Driven Optimization for Test-Time Large Language Model Reasoning
标题：测试时大型语言模型推理的探索驱动优化
链接：https://arxiv.org/abs/2605.09853

作者：Changhao Li, Yuchen Zhuang, Chenxiao Gao, Haotian Sun, Rushi Qiang, Chao Zhang, Bo Dai
备注：Accepted by TMLR 2026

【25】The Metacognitive Probe: Five Behavioural Calibration Diagnostics for LLMs
标题：元认知探针：LLM的五种行为校准诊断
链接：https://arxiv.org/abs/2605.09844

作者：Rafael C. T. Oliveira
备注：27 pages, 13 tables. Code, data, prompts, and rubrics released with the paper. OSF deposit pending; DOI in v2

【26】Pretraining large language models with MXFP4
标题：使用MXFP 4预训练大型语言模型
链接：https://arxiv.org/abs/2605.09825

作者：Musa Cim, Poovaiah Palangappa, Miro Hodak, Ravi Dwivedula, Meena Arunachalam, Mahmut Taylan Kandemir

【27】Dystruct: Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference
标题：Dystruct：通过Bayesian推理进行动态结构扩散语言模型解码
链接：https://arxiv.org/abs/2605.09820

作者：Bian Sun, Kevin Zhai, Mubarak Shah, Zhenyi Wang

【28】LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models
标题：LEAD：大型语言模型的长度高效自适应和动态推理
链接：https://arxiv.org/abs/2605.09806

作者：Songtao Wei,Yi Li,Zhikai Li,Xu Hu,Yuede Ji,Guanpeng Li,Feng Chen,Carl Yang,Zhichun Guo,Bingzhe Li

【29】Parameter-Efficient Neuroevolution for Diverse LLM Generation: Quality-Diversity Optimization via Prompt Embedding Evolution
标题：用于多元化LLM生成的参数高效神经进化：通过即时嵌入进化进行质量多样性优化
链接：https://arxiv.org/abs/2605.09781

作者：Dongxin Guo,Jikun Wu,Siu Ming Yiu
备注：11 pages, 3 figures, 7 tables, 1 algorithm, 1 theorem. Accepted to GECCO 2026

【30】EvoPref: Multi-Objective Evolutionary Optimization Discovers Diverse LLM Alignments Beyond Gradient Descent
标题：EvoPref：多目标进化优化发现了超越梯度下降的多样化LLM路线
链接：https://arxiv.org/abs/2605.09777

作者：Dongxin Guo,Jikun Wu,Siu Ming Yiu
备注：10 pages, 2 figures, 6 tables, 1 algorithm. Accepted to GECCO 2026

【31】Metal-Sci: A Scientific Compute Benchmark for Evolutionary LLM Kernel Search on Apple Silicon
标题：Metal-Sci：Apple Silicon上进化LLM核心搜索的科学计算基准
链接：https://arxiv.org/abs/2605.09708

作者：Víctor Gallego
备注：Preprint

【32】Scratchpad Patching: Decoupling Compute from Patch Size in Byte-Level Language Models
标题：Spreadchpad修补：在字节级语言模型中将计算与修补大小脱钩
链接：https://arxiv.org/abs/2605.09630

作者：Lin Zheng,Vasilisa Bashlovkina,Timothy Dozat,Dan Garrette,Laura Rimell,Joshua Maynez
备注：23 pages, 15 figures

【33】SmartEval: A Benchmark for Evaluating LLM-Generated Smart Contracts from Natural Language Specifications
标题：SmartEval：根据自然语言规范评估LLM生成的智能合同的基准
链接：https://arxiv.org/abs/2605.09610

作者：Abhinav Goel,Agostino Capponi,Alfio Gliozzo,Chaitya Shah

【34】Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training
标题：几何冲突：LLM连续后训练中的解释和控制遗忘
链接：https://arxiv.org/abs/2605.09608

作者：Yuanyi Wang,Yifan Yang,Su Lu,Yanggan Gu,Pengkai Wang,Wenjun Wang,Zhaoyi Yan,Congkai Xie,Jianmin Wu,Jialun Cao,Shing-Chi Cheung,Hongxia Yang

【35】LLM-Driven Performance-Space Augmentation for Meta-Learning-Based Algorithm Selection
标题：LLM驱动的元学习算法选择性能空间增强
链接：https://arxiv.org/abs/2605.09518

作者：Darren Zhu,Daren Ler

【36】Beyond Language: Format-Agnostic Reasoning Subspaces in Large Language Models
标题：超越语言：大型语言模型中的不可知推理子空间
链接：https://arxiv.org/abs/2605.09496

作者：Aojie Yuan,Zhiyuan Su
备注：Preprint. 13 pages, 13 figures, 12 tables

【37】Not All Thoughts Need HBM: Semantics-Aware Memory Hierarchy for LLM Reasoning
标题：并非所有想法都需要HBM：LLM推理的语义感知内存层次结构
链接：https://arxiv.org/abs/2605.09490

作者：Aojie Yuan,Tianqi Shen,Dajun Zhang
备注：Preprint. 14 pages + appendix. Under review at AdaptFM Workshop @ ICML 2026

【38】Your Simulation Runs but Solves the Wrong Physics: PDE-Grounded Intent Verification for LLM-Generated Multiphysics Simulation Code
标题：您的模拟解决了错误的物理：LLM生成的多物理场模拟代码的PDE-接地意图验证
链接：https://arxiv.org/abs/2605.09360

作者：Zhenghan Song,Yulong Liu,Cheng Wan,Chenjun Li,Lingfu Liu,Yunyi Li,Congcong Yuan
备注：Preprint

【39】Path-Dependent Denoising: A Non-Conservative Field Perspective on Order Collapse in Diffusion Language Models
标题：路径相关去噪：扩散语言模型中序崩溃的非保守场视角
链接：https://arxiv.org/abs/2605.09303

作者：Jeonseong Kim

【40】Towards Effective Theory of LLMs: A Representation Learning Approach
标题：迈向有效的LLM理论：一种代表性学习方法
链接：https://arxiv.org/abs/2605.09294

作者：Muhammed Ustaomeroglu,Guannan Qu
备注：Project webpage: https://ustaomeroglu.github.io/RET/

【41】The Art of the Jailbreak: Formulating Jailbreak Attacks for LLM Security Beyond Binary Scoring
标题：越狱的艺术：为LLM安全制定超越二进制评分的越狱攻击
链接：https://arxiv.org/abs/2605.09225

作者：Ismail Hossain,Tanzim Ahad,Md Jahangir Alam,Sai Puppala,Syed Bahauddin Alam,Sajedul Talukder
备注：This paper is under review on of top security venues

【42】SMIXAE: Towards Unsupervised Manifold Discovery in Language Models
标题：SMIXAE：迈向语言模型中的无监督Manifold发现
链接：https://arxiv.org/abs/2605.09224

作者：Collin Francel
备注：20 pages, 10 figures, 11 tables. Submitted to Mechanistic Interpretability Workshop, ICML 2026

【43】Flame3D: Zero-shot Compositional Reasoning of 3D Scenes with Agentic Language Models
标题：Flame 3D：使用抽象语言模型的3D场景Zero-Shot合成推理
链接：https://arxiv.org/abs/2605.09218

作者：Sagar Bharadwaj,Ziyong Ma,Anurag Ghosh,Srinivasan Seshan,Anthony Rowe

【44】Emergent Semantic Role Understanding in Language Models
标题：语言模型中的紧急语义角色理解
链接：https://arxiv.org/abs/2605.09187

作者：Carla Griffiths,Mirco Musolesi

【45】Navigating LLM Valley: From AdamW to Memory-Efficient and Matrix-Based Optimizers
标题：航行LLM Valley：从AdamW到内存高效和基于矩阵的优化器
链接：https://arxiv.org/abs/2605.09176

作者：Aditya Ranganath
备注：No figures, 65 pages

【46】Sparse Layers are Critical to Scaling Looped Language Models
标题：稀疏层对于扩展循环语言模型至关重要
链接：https://arxiv.org/abs/2605.09165

作者：Ryan Lee,Jacob Biloki,Edward J. Hu,Jonathan May

【47】A Communication-Theoretic Framework for LLM Agents: Cost-Aware Adaptive Reliability
标题：LLM代理的通信理论框架：成本意识的自适应可靠性
链接：https://arxiv.org/abs/2605.09121

作者：Hamed Omidvar,Vahideh Akhlaghi

【48】Robust Multi-Agent LLMs under Byzantine Faults
标题：拜占庭故障下的鲁棒多Agent LLM
链接：https://arxiv.org/abs/2605.09076

作者：Haejoon Lee,Vincent-Daniel Yun,Hyeonho Oh,Dimitra Panagou,Sai Praneeth Karimireddy

【49】GAMBIT: A Three-Mode Benchmark for Adversarial Robustness in Multi-Agent LLM Collectives
标题：GAMBIT：多智能体LLM群体中对抗鲁棒性的三模式基准
链接：https://arxiv.org/abs/2605.09027

作者：Alexandre Le Mercier,Chris Develder,Thomas Demeester
备注：46 pages, 16 figures

【50】LLiMba: Sardinian on a Single GPU -- Adapting a 3B Language Model to a Vanishing Romance Language
标题：LLiMba：单个图形处理器上的撒丁岛--将3B语言模型适应消失的浪漫语言
链接：https://arxiv.org/abs/2605.09015

作者：Luca Ballore

【51】A Geometric Perspective on Next-Token Prediction in Large Language Models: Three Emerging Phases
标题：大型语言模型中下一个令牌预测的几何视角：三个新兴阶段
链接：https://arxiv.org/abs/2605.09011

作者：Gianfranco Lombardo,Giuseppe Trimigno,Stefano Cagnoni

【52】Large Language Models for Sequential Decision-Making: Improving In-Context Learning via Supervised Fine-Tuning
标题：用于顺序决策的大型语言模型：通过监督微调改善上下文学习
链接：https://arxiv.org/abs/2605.09009

作者：Minmin Zhang,Sina Aghaei,Soroush Saghafian

【53】Relative Kinetic Utility for Reasoning-Aware Structural Pruning in Large Language Models
标题：大型语言模型中推理感知结构剪枝的相对动力学效用
链接：https://arxiv.org/abs/2605.09008

作者：Tianhao Qian
备注：15 pages, 3 figures

【54】Muon-OGD: Muon-based Spectral Orthogonal Gradient Projection for LLM Continual Learning
标题：Muon-OVD：用于LLM连续学习的基于Muon的谱垂直梯度投影
链接：https://arxiv.org/abs/2605.08949

作者：Binghang Lu,Zheyuan Deng,Runyu Zhang,Bing Hu,Yunhan Zhao,Yuan Tian,Changhong Mou,Guang Lin,Xiaomin Li

【55】OTora: A Unified Red Teaming Framework for Reasoning-Level Denial-of-Service in LLM Agents
标题：OTora：LLM代理中用于推理级拒绝服务的统一红色团队框架
链接：https://arxiv.org/abs/2605.08876

作者：Xinyu Li,Ronghui Mu,Lin Li,Tianjin Huang,Gaojie Jin
备注：Accepted to ICML 2026

【56】Unlearners Can Lie: Evaluating and Improving Honesty in LLM Unlearning
标题：未学习者可以撒谎：LLM Unlearning中评估和提高诚实度
链接：https://arxiv.org/abs/2605.08765

作者：Renjie Gu,Jiazhen Du,Yihua Zhang,Sijia Liu
备注：Accepted by ACL 2026

【57】AAAC: Activation-Aware Adaptive Codebooks for 4-bit LLM Weight Quantization
标题：AAAC：用于4位LLM权重量化的激活感知自适应代码簿
链接：https://arxiv.org/abs/2605.08692

作者：Beshr IslamBouli,David Jin

【58】Semantic Voting: Execution-Grounded Consensus for LLM Code Generation
标题：语义投票：LLM代码生成的基于执行的共识
链接：https://arxiv.org/abs/2605.08680

作者：Shan Jiang,Zijian Yi,Chenguang Zhu

【59】PRISM: Fast Online LLM Serving via Scheduling-Memory Co-design
标题：PRism：通过记忆协同设计快速在线LLM服务
链接：https://arxiv.org/abs/2605.08581

作者：Xingyu Qu,Tianhao Lin,Yiqi Li,Zhiyu Chen,Sheng Wang
备注：25 pages, 9 figures, Preprint

【60】Different Prompts, Different Ranks: Prompt-aware Dynamic Rank Selection for SVD-based LLM Compression
标题：不同的预算，不同的排名：基于SVD的LLM压缩的预算感知动态排名选择
链接：https://arxiv.org/abs/2605.08568

作者：Hengyi Zhu,Zhendong Mi,Grace Li Zhang,Shaoyi Huang

【61】Can Revealed Preferences Clarify LLM Alignment and Steering?
标题：揭示的偏好可以澄清LLM对齐和转向吗？
链接：https://arxiv.org/abs/2605.08556

作者：Khurram Yamin,Jingjing Tang,Eric Horvitz,Bryan Wilder

【62】Tokens-per-Parameter Coverage Is Critical for Robust LLM Scaling Law Extrapolation
标题：每个参数的标记覆盖对于鲁棒LLM标度律外推至关重要
链接：https://arxiv.org/abs/2605.08541

作者：Joshua Shay Kricheli,Alexander Lawrence Reid,Soumajyoti Sarkar,Venkata Gandikota,Paulo Shakarian

【63】Human-Inspired Memory Architecture for LLM Agents
标题：LLM代理的人性启发存储架构
链接：https://arxiv.org/abs/2605.08538

作者：Doga Kerestecioglu,Alexei Robsky,Clemens Vasters,Anshul Sharma,Yitzhak Kesselman
备注：10 pages, 4 tables. Preprint; comments welcome

【64】A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models
标题：单个神经元足以绕过大型语言模型中的安全对齐
链接：https://arxiv.org/abs/2605.08513

作者：Hamid Kazemi,Atoosa Chegini,Maria Safi

【65】MathConstraint: Automated Generation of Verified Combinatorial Reasoning Instances for LLMs
标题：MathConstraint：面向LLM的组合推理验证图的自动生成
链接：https://arxiv.org/abs/2605.08498

作者：Viresh Pati,Zhengyu Li,Piyush Jha,Rahul Garg,Yatharth Sejpal,Vijay Ganesh

【66】PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
标题：PYTHALAB-MERA：Frozen-LLM编码代理的验证接地内存、检索和接受控制
链接：https://arxiv.org/abs/2605.08468

作者：Mehmet Iscan
备注：28 pages, 4 figures, 7 tables; local CLI artifact evaluation

【67】CUDAHercules: Benchmarking Hardware-Aware Expert-level CUDA Optimization for LLMs
标题：CUDAHercules：针对LLM的硬件感知专家级CUDA优化基准
链接：https://arxiv.org/abs/2605.08467

作者：Shiyang Li,Zijian Zhang,Guangyan Sun,Yuebo Luo,Winson Chen,Yanzhi Wang,Mingyi Hong,Caiwen Ding

【68】CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging
标题：CUDABeaver：对基于LLM的自动化CUDA收件箱进行基准测试
链接：https://arxiv.org/abs/2605.08455

作者：Shiyang Li,Haoyang Chen,Mattia Fazzini,Caiwen Ding
备注：25 pages, 5 figures

【69】Defense effectiveness across architectural layers: a mechanistic evaluation of persistent memory attacks on stateful LLM agents
标题：跨架构层的防御有效性：对有状态LLM代理的持久内存攻击的机械评估
链接：https://arxiv.org/abs/2605.08442

作者：Jun Wen Leong
备注：9 models, 5,700 runs across 5 experiments, pre-registered comparisons. Code and results: github.com/junwenleong/stateful-agent-security-eval

【70】Interactive Critique-Revision Training for Reliable Structured LLM Generation
标题：用于可靠结构化LLM生成的交互式批评-修订训练
链接：https://arxiv.org/abs/2605.08327

作者：Fei Xu Yu,Zuyuan Zhang,Mahdi Imani,Nathaniel D. Bastian,Tian Lan

【71】LLM Advertisement based on Neuron Auctions
标题：基于Neuron Auctions的LLM广告
链接：https://arxiv.org/abs/2605.08326

作者：Peiran Yun,Wenxin Xu,Jiayuan Liu,Yihang Zhang,Liang Zeng,Lingkai Kong,Tonghan Wang
备注：17 pages, 9 figures, including appendices

【72】LLM Wardens: Mitigating Adversarial Persuasion with Third-Party Conversational Oversight
标题：法学硕士典狱长：通过第三方对话监督减轻敌对说服
链接：https://arxiv.org/abs/2605.08321

作者：Lennart Wachowiak,Scott D. Blain,David Williams-King,Samuele Marro

【73】Seed Hijacking of LLM Sampling and Quantum Random Number Defense
标题：LLM采样和量子随机数防御的种子劫持
链接：https://arxiv.org/abs/2605.08313

作者：Ziyang You,Xiaoke Yang,Zhanling Fan,Feng Guo,Xiaogen Zhou,Xuxing Lu

【74】LLMSYS-HPOBench: Hyperparameter Optimization Benchmark Suite for Real-World LLM Systems
标题：LLMSYS-HPOBench：适用于现实世界LLM系统的超参数优化基准套件
链接：https://arxiv.org/abs/2605.08305

作者：Siyu Wu,Yulong Ye,Zezhen Xiang,Pengzhou Chen,Gangda Xiong,Tao Chen

【75】mHC-SSM: Manifold-Constrained Hyper-Connections for State Space Language Models with Stream-Specialized Adapters
标题：mHC-RSM：具有流专业化适配器的状态空间语言模型的Manifold约束超连接
链接：https://arxiv.org/abs/2605.08300

作者：Abdulvahap Mutlu,Şengül Doğan,Türker Tuncer
备注：28 Pages, 3 Figures, all implementation code available at: https://github.com/abdulvahapmutlu/mhc-slm

【76】Research on Security Enhancement Methods for Adversarial Robust Large Language Model Intelligent Agents for Medical Decision-Making Tasks
标题：对抗性安全增强方法研究医疗决策任务鲁棒大语言模型智能代理
链接：https://arxiv.org/abs/2605.08257

作者：Saisai Hu
备注：5 pages, 2 figures, 1 table.Accepted for oral presentation at AINIT 2026

【77】Can LLMs Predict Polymer Physics Just by Reading Synthesis and Processing Prose?
标题：LLM仅通过阅读合成和处理散文就可以预测聚合物物理吗？
链接：https://arxiv.org/abs/2605.08255

作者：Yuchu Liu,Rui Zhu,Jingwei Xiong,Haixu Tang

【78】Test-Time Training for Visual Foresight Vision-Language-Action Models
标题：视觉预见视觉-语言-动作模型的测试时训练
链接：https://arxiv.org/abs/2605.08215

作者：Sangwu Park,Wonjoong Kim,Yeonjun In,Sein Kim,Hongseok Kang,Chanyoung Park
备注：Preprint. Under review

【79】LLMs with in-context learning for Algorithmic Theoretical Physics
标题：具有数学理论物理背景学习的法学硕士
链接：https://arxiv.org/abs/2605.08212

作者：Anamaria Hell,Leander Thiele
备注：8 pages, 2 figures

【80】Where Reliability Lives in Vision-Language Models: A Mechanistic Study of Attention, Hidden States, and Causal Circuits
标题：视觉语言模型中可靠性所在：注意力、隐藏状态和因果回路的机械学研究
链接：https://arxiv.org/abs/2605.08200

作者：Logan Mann,Ajit Saravanan,Ishan Dave,Shikhar Shiromani,Saadullah Ismail,Yi Xia,Emily Huang
备注：15 pages, 4 figures, 10 tables. Accepted at the ICLR 2026 Workshop on Multimodal Reasoning. Code and probe-training pipelines: https://github.com/itsloganmann/VLM-Reliability-Probe

【81】Understanding Asynchronous Inference Methods for Vision-Language-Action Models
标题：了解视觉-语言-动作模型的同步推理方法
链接：https://arxiv.org/abs/2605.08168

作者：Ayoub Agouzoul

【82】Feature Rivalry in Sparse Autoencoder Representations: A Mechanistic Study of Uncertainty-Driven Feature Competition in LLMs
标题：稀疏自动编码器表示中的特征竞争：LLM中不确定性驱动的特征竞争的机制研究
链接：https://arxiv.org/abs/2605.08149

作者：Harshavardhan
备注：10 pages, 6 figures

【83】Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models
标题：自标题多模式交互调整：扩大鲁棒视觉语言模型的可利用冗余
链接：https://arxiv.org/abs/2605.08145

作者：Yuriel Ryan,Hei Man Ip,Adriel Kuek,Paul Pu Liang,Roy Ka-Wei Lee
备注：Accepted to ICML 2026

【84】Reasoning emerges from constrained inference manifolds in large language models
标题：推理从大型语言模型中的受约束推理流中产生
链接：https://arxiv.org/abs/2605.08142

作者：Yanbiao Ma,Fei Luo,Linfeng Zhang,Chuangxin Zhao,Mingxuan Wang,Yinan Wu,Zhe Qian,Yang Lu,Long Chen,Zhao Cao,Xiaoshuai Hao,Ji-Rong Wen,Jungong Han

【85】Weight Pruning Amplifies Bias: A Multi-Method Study of Compressed LLMs for Edge AI
标题：权重修剪放大了偏差：边缘AI压缩LLM的多方法研究
链接：https://arxiv.org/abs/2605.08137

作者：Plawan Kumar Rath,Rahul Maliakkal
备注：8 pages, 7 figures, 8 tables. Accepted at the 7th Annual World AIIoT Congress (AIIoT 2026). This is the author's accepted version; the version of record will appear in IEEE Xplore

【86】DARE: Diffusion Language Model Activation Reuse for Efficient Inference
标题：DARE：扩散语言模型激活重用以实现高效推理
链接：https://arxiv.org/abs/2605.08134

作者：Natalia Frumkin,Bokun Wang,Hung-Yueh Chiang,Chi-Chih Chang,Mohamed S. Abdelfattah,Diana Marculescu

【87】Federated Language Models Under Bandwidth Budgets: Distillation Rates and Conformal Coverage
标题：带宽预算下的联邦语言模型：蒸馏率和保形覆盖
链接：https://arxiv.org/abs/2605.09986

作者：Prasanjit Dubey,Xiaoming Huo

【88】Large Language Models over Networks: Collaborative Intelligence under Resource Constraints
标题：网络上的大型语言模型：资源约束下的协作智能
链接：https://arxiv.org/abs/2605.08626

作者：Liangqi Yuan,Wenzhi Fang,Shiqiang Wang,H. Vincent Poor,Christopher G. Brinton

Graph相关(图学习|图神经网络|图优化等)(28篇)

【1】On Improving Graph Neural Networks for QSAR by Pre-training on Extended-Connectivity Fingerprints
标题：通过扩展连接性指纹预训练改进QSAR图神经网络
链接：https://arxiv.org/abs/2605.10722

作者：Sam Money-Kyrle,Markus Dablander,Thierry Hanser,Stephane Werner,Charlotte M. Deane,Garrett M. Morris

【2】Reconfigurable Computing Challenge: Real-Time Graph Neural Networks for Online Event Selection in Big Science
标题：可重新配置计算挑战：用于大科学在线事件选择的实时图神经网络
链接：https://arxiv.org/abs/2605.10612

作者：Marc Neu,Frank Baptist,Thomas Lobmaier,Fabio Papagno,Torben Ferber,Jürgen Becker
备注：Accepted to FCCM Reconfigurable Computing Challenge 2026

【3】It's All Connected: Topology-Aware Structural Graph Encoding Improves Performance on Polymer Prediction
标题：一切都是相互关联的：具有布局意识的结构图编码提高了聚合物预测的性能
链接：https://arxiv.org/abs/2605.10551

作者：H. Ibrahim Erdogan,Punith Raviswamy,Nikita Agrawal,Yannik Köster,Stefan Zechel,Ulrich S. Schubert,Ruben Mayer,Christopher Kuenneth
备注：9 pages, 4 figures

【4】Bridging Sequence and Graph Structure for Epigenetic Age Prediction
标题：表观遗传年龄预测的桥梁序列和图结构
链接：https://arxiv.org/abs/2605.10541

作者：Yao Li,Xikun Zhang,Xiaotao Shen,Sonika Tyagi,Xin Zheng,Jiaxing Huang,Feng Xia

【5】PrimeKG-CL: A Continual Graph Learning Benchmark on Evolving Biomedical Knowledge Graphs
标题：PrimeKG-CL：关于不断发展的生物医学知识图的连续图学习基准
链接：https://arxiv.org/abs/2605.10529

作者：Yousef A. Radwan, Yao Li, Qing Qing, Ziqi Xu, Xingtong Yu, Jiaxing Huang, Renqiang Luo, Xikun Zhang

【6】CMKL: Modality-Aware Continual Learning for Evolving Biomedical Knowledge Graphs
标题：CMKL：不断发展的生物医学知识图的模式感知持续学习
链接：https://arxiv.org/abs/2605.10510

作者：Yousef A. Radwan, Yao Li, Qing Qing, Ziqi Xu, Qixin Zhang, Yongcheng Jing, Renqiang Luo, Xikun Zhang

【7】Relations Are Channels: Knowledge Graph Embedding via Kraus Decompositions
标题：关系是渠道：通过克劳斯分解嵌入知识图谱
链接：https://arxiv.org/abs/2605.10317

作者：Sayan Kumar Chaki

【8】One-Step Graph-Structured Neural Flows for Irregular Multivariate Time Series Classification
标题：不规则多元时间序列分类的一步图结构神经流
链接：https://arxiv.org/abs/2605.10179

作者：Mengzhou Gao, Kaiwei Wang, Pengfei Jiao

【9】Learning Graph Foundation Models on Riemannian Graph-of-Graphs
标题：Riemann图上的学习图基础模型
链接：https://arxiv.org/abs/2605.09993

作者：Haokun Liu, Zezhong Ding, Xike Xie
备注：This paper has been accepted by ICML 2026

【10】UFO: A Unified Flow-Oriented Framework for Robust Continual Graph Learning
标题：UFO：用于稳健连续图学习的统一面向流的框架
链接：https://arxiv.org/abs/2605.09862

作者：Danhui Zhang, Zhe Wang, Qing Qing, Jiarui Liu, Wentao Gao, Ziqi Xu, Mingliang Hou, Xikun Zhang, Renqiang Luo

【11】ChaosNetBench: Benchmarking Spatio-Temporal Graph Neural Networks on Chaotic Lattice Dynamics
标题：ChaosNetBench：基于混乱格子动力学的时空图神经网络基准
链接：https://arxiv.org/abs/2605.09676

作者：Henok Tenaw Moges,Charalampos Skokos,Deshendran Moodley
备注：24 pages, 11 figures

【12】End-to-End Keyword Spotting on FPGA Using Graph Neural Networks with a Neuromorphic Auditory Sensor
标题：使用具有神经形态听觉传感器的图神经网络在DSP上进行端到端关键词定位
链接：https://arxiv.org/abs/2605.09570

作者：Wiktor Matykiewicz,Piotr Wzorek,Kamil Jeziorek,Tomás Muñoz,Antonio Rios-Navarro,Angel Jiménez-Fernández,Tomasz Kryjak
备注：Accepted for the ARC 2026 conference

【13】CTQWformer: A CTQW-based Transformer for Graph Classification
标题：CTQWformer：基于CTQW的图形分类Transformer
链接：https://arxiv.org/abs/2605.09486

作者：Zhan Li,Wuqing Yu,Yusen Wu,Chuan Wang

【14】FedCIGAR: A Personalized Reconstruction Approach for Federated Graph-level Anomaly Detection
标题：FedCIGAR：一种用于联邦图级异常检测的个性化重建方法
链接：https://arxiv.org/abs/2605.09428

作者：Yunfeng Zhao,Yixin Liu,Qingfeng Chen,Shiyuan Li,Yue Tan,Shirui Pan
备注：Accepted by IJCAI 2026

【15】GravityGraphSAGE: Link Prediction in Directed Attributed Graphs
标题：GravityGraphSAGE：有向属性图中的链接预测
链接：https://arxiv.org/abs/2605.09408

作者：Riccardo Porcedda,Francesca Chiaromonte,Fabrizio Lillo,Andrea Vandin

【16】Functional Graphs for Predicting and Explaining Goal Failure in Sparse Goal-Conditioned RL
标题：预测和解释稀疏目标条件RL中目标失败的功能图
链接：https://arxiv.org/abs/2605.09335

作者：Shalley Dash
备注：9 pages main, 21 pages appendx, 2 figures in main. 8 figures in appendix, Submitted to a conference

【17】Hierarchical Attention-based Graph Neural Network with Relevance-driven Pruning
标题：具有相关驱动修剪的分层注意力图神经网络
链接：https://arxiv.org/abs/2605.09308

作者：Seungwoo Kum

【18】Dependency-Aware Discrete Diffusion for Scene Graph Generation
标题：场景图生成的依赖性感知离散扩散
链接：https://arxiv.org/abs/2605.09065

作者：Rajalaxmi Rajagopalan,Romit Roy Choudhury

【19】PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation
标题：PACT：用于高效风暴潮模拟的峰值感知交叉注意力图形变形器
链接：https://arxiv.org/abs/2605.09036

作者：Zesheng Liu,Doyup Kwon,Ning Lin,Maryam Rahnemoonfar

【20】Machine Learning-Based Graph Simplification for Symbolic Accelerators
标题：基于机器学习的符号加速器图简化
链接：https://arxiv.org/abs/2605.08996

作者：Tiffany Yu,Rye Stahle-Smith,Darssan Eswaramoorthi,Rasha Karakchi

【21】Structure-Centric Graph Foundation Model via Geometric Bases
标题：通过几何基的结构中心图基础模型
链接：https://arxiv.org/abs/2605.08689

作者：Xiaodong He,Haolan He,Ruiyi Fang,Ming Sun,Zhao Kang
备注：Accepted by ICML 2026

【22】Attention-based graph neural networks: a survey
标题：基于注意力的图神经网络：调查
链接：https://arxiv.org/abs/2605.08679

作者：Chengcheng Sun,Chenhao Li,Xiang Lin,Tianji Zheng,Fanrong Meng,Xiaobin Rui,Zhixiao Wang
备注：This is the accepted manuscript of an article published in Artificial Intelligence Review. The final version is available online at: [10.1007/s10462-023-10577-2](https://link.springer.com/article/10.1007/s10462-023-10577-2)

【23】Belief or Circuitry? Causal Evidence for In-Context Graph Learning
标题：信仰还是循环？上下文图学习的因果证据
链接：https://arxiv.org/abs/2605.08405

作者：Katharine Kowalyshyn,Timothy Duggan,Daniel Little,Michael C Hughes
备注：Under review at ICML Mechanistic Interpretability Workshop 2026

【24】SDG-MoE: Signed Debate Graph Mixture-of-Experts
标题：SDG-MoE：签名辩论图专家混合
链接：https://arxiv.org/abs/2605.08322

作者：Stepan Kulibaba,Kirill Labzin,Artem Dzhalilov,Roman Pakhomov,Oleg Svidchenko,Alexander Gansnikov,Aleksei Shpilman

【25】Graph Computation Meets Circuit Algebra: A Task-Aligned Analysis of Graph Neural Networks for Electronic Design Automation
标题：图计算遇到电路代数：电子设计自动化图神经网络的任务对齐分析
链接：https://arxiv.org/abs/2605.08291

作者：Hyunmog Kim

【26】UMEDA: Unified Multi-modal Efficient Data Fusion for Privacy-Preserving Graph Federated Learning via Spectral-Gated Attention and Diffusion-Based Operator Alignment
标题：UMEDA：通过光谱门控注意力和基于扩散的操作员对齐进行统一多模式高效数据融合，用于隐私保护图联邦学习
链接：https://arxiv.org/abs/2605.08288

作者：Shih-Yu Lai,Hirozumi Yamaguchi,Shang-Tse Chen,Yu-Lun Liu,Bing-Yu Chen

【27】Generalized Category Discovery in Federated Graph Learning
标题：联邦图学习中的广义类别发现
链接：https://arxiv.org/abs/2605.08178

作者：Zhongzheng Yuan,Lianshuai Guo,Xunkai Li,Wenyu Wang,Meixia Qu

【28】Path-Based Gradient Boosting for Graph-Level Prediction
标题：基于路径的梯度提升用于图级预测
链接：https://arxiv.org/abs/2605.08102

作者：Claudio Meggio,Johan Pensar,Riccardo De Bin
备注：20 Pages, 1 figure

Transformer(31篇)

【1】Masked Generative Transformer Is What You Need for Image Editing
标题：图像编辑所需的掩蔽生成Transformer器
链接：https://arxiv.org/abs/2605.10859

作者：Wei Chow,Linfeng Li,Xian Sun,Lingdong Kong,Zefeng Li,Qi Xu,Hang Song,Tian Ye,Xian Wang,Jinbin Bai,Shilin Xu,Xiangtai Li,Junting Pan,Shaoteng Liu,Ran Zhou,Tianshu Yang,Songhua Liu
备注：CVPR 2026 HiGen Workshop; Project Page at https://weichow23.github.io/EditMGT/ GitHub at https://github.com/weichow23/EditMGT

【2】RelFlexformer: Efficient Attention 3D-Transformers for Integrable Relative Positional Encodings
标题：RelFlexformer：用于可积分相对位置编码的高效注意力3D转换器
链接：https://arxiv.org/abs/2605.10706

作者：Byeongchan Kim,Arijit Sehanobish,Avinava Dubey,Min-hwan Oh,Krzysztof Choromanski

【3】DRIFT: Drift-Resilient Invariant-Feature Transformer for DGA Detection
标题：DRIST：用于DGA检测的抗漂移不变特征Transformer
链接：https://arxiv.org/abs/2605.10436

作者：Chaeyoung Lee, Chaeri Jung, Seonghoon Jeong
备注：14 pages, 7 figures, 8 tables. Accepted to appear in Proc. of the 56th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN 2026)

【4】Complex-Valued Phase-Coherent Transformer
标题：复值相一致Transformer
链接：https://arxiv.org/abs/2605.10123

作者：Leona Hioki
备注：26 pages, 17 tables (no figures). Companion Lean 4 formalization of Theorems 1 and 2 at this https URL

【5】Rethinking Random Transformers as Adaptive Sequence Smoothers for Sleep Staging
标题：重新思考随机变形器作为睡眠阶段的自适应序列平滑器
链接：https://arxiv.org/abs/2605.09905

作者：Guisong Liu,Xin Gao,Martin Dresler,Jiansong Zhang,Pengfei Wei

【6】Continuous Latent Contexts Enable Efficient Online Learning in Transformers
标题：连续的潜在上下文实现Transformer中的高效在线学习
链接：https://arxiv.org/abs/2605.09867

作者：Emile Anand, Abdullah Ateyeh, Xinyuan Cao, Max Dabagia
备注：37 pages, 15 figures, 3 tables

【7】The Association of Transformer-based Sentiment Analysis with Symptom Distress and Deterioration in Routine Psychotherapy Care
标题：基于变形者的情绪分析与常规心理治疗护理中症状困扰和恶化的关联
链接：https://arxiv.org/abs/2605.09838

作者：Douglas K. Faust, Peter Awad, Alexandre Vaz, Tony Rousmaniere
备注：20 pages, 4 figures

【8】CALYREX: Cross-Attention LaYeR EXtended Transformers for System Prompt Anchoring
标题：CALLREX：交叉注意LaYeR扩展Transformer，用于系统即时锚定
链接：https://arxiv.org/abs/2605.09737

作者：Li Lixing
备注：Preprint. 25 pages, 4 figures, 9 tables

【9】One for All: A Non-Linear Transformer can Enable Cross-Domain Generalization for In-Context Reinforcement Learning
标题：One for All：非线性Transformer可以为上下文强化学习实现跨域概括
链接：https://arxiv.org/abs/2605.09727

作者：Bowen He,Juncheng Dong,Lin Lin,Xiang Cheng

【10】Benchmarking Transformer and xLSTM for Time-Series Forecasting of Heat Consumption
标题：对标Transformer和xLSTM进行热消耗时间序列预测
链接：https://arxiv.org/abs/2605.09722

作者：Marja Wahl,Daniel R. Bayer,Sven Rausch,Marco Pruckner
备注：Submitted version of the paper submitted to IEEE SusTech, 2026

【11】Spectral Transformer Neural Processes
标题：光谱Transformer神经过程
链接：https://arxiv.org/abs/2605.09498

作者：Xianhe Chen,Hao Chen,Yingzhen Li
备注：37 pages, 10 figures, 18 tables

【12】Sparsity Moves Computation: How FFN Architecture Reshapes Attention in Small Transformers
标题：稀疏性推动计算：FFN架构如何重塑小型Transformer中的注意力
链接：https://arxiv.org/abs/2605.09403

作者：Gabriel Smithline,Chris Mascioli
备注：Preprint

【13】Uncertainty-Aware Token Importance Estimation in Spiking Transformers
标题：尖峰Transformer中的不确定性代币重要性估计
链接：https://arxiv.org/abs/2605.09276

作者：Wenxuan Liu,Zecheng Hao,Tong Bu,Yuran Wang,Zhaofei Yu

【14】RigidFormer: Learning Rigid Dynamics using Transformers
标题：RigidFormer：使用Transformer学习刚性动力学
链接：https://arxiv.org/abs/2605.09196

作者：Zhiyang Dou,Minghao Guo,Haixu Wu,Doug Roble,Tuur Stuyck,Wojciech Matusik
备注：Project Page: https://people.csail.mit.edu/frankzydou/projects/RigidFormer/index.html

【15】VORT: Adaptive Power-Law Memory for NLP Transformers
标题：VORT：用于NLPTransformer的自适应功率定律存储器
链接：https://arxiv.org/abs/2605.08966

作者：Nabil Mlaiki
备注：18 pages, 5 figures

【16】Transformer autoencoder with local attention for sparse and irregular time series with application on risk estimation
标题：局部关注稀疏和不规则时间序列的Transformer自动编码器，并应用于风险估计
链接：https://arxiv.org/abs/2605.08914

作者：Panteleimon Rodis
备注：Under Review

【17】Causal Dimensionality of Transformer Representations: Measurement, Scaling, and Layer Structure
标题：Transformer表示的因果性：测量、标度和层结构
链接：https://arxiv.org/abs/2605.08740

作者：Nilesh Sarkar,Dawar Jyoti Deka
备注：9 pages, 17 figures, 14 tables (excluding references and appendices). Companion short paper under review at the ICML 2026 Mechanistic Interpretability Workshop. Code: https://anonymous.4open.science/r/NeurIPS-Causal-Capacity-in-SAEs-7D20/

【18】Lattice Deduction Transformers
标题：格子演绎Transformer
链接：https://arxiv.org/abs/2605.08605

作者：Liam Davis,Leopold Haller,Alberto Alfarano,Mark Santolucito

【19】Probing the Impact of Scale on Data-Efficient, Generalist Transformer World Models for Atari
标题：探讨规模对Atari数据高效的通才Transformer World模型的影响
链接：https://arxiv.org/abs/2605.08578

作者：Jooyeon Kim

【20】Geometric Flood Depth Estimation: Fusing Transformer-Based Segmentation with Digital Elevation Models
标题：几何洪水深度估计：将基于变换器的分割与数字海拔模型融合
链接：https://arxiv.org/abs/2605.08521

作者：Nhut Le,Ehsan Karimi,Maryam Rahnemoonfar
备注：Accepted by the 2026 IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2026)

【21】Scaling Limits of Long-Context Transformers
标题：长上下文Transformer的扩展限制
链接：https://arxiv.org/abs/2605.08505

作者：Giuseppe Bruno,Shi Chen,Zhengjiang Lin,Yury Polyanskiy,Philippe Rigollet
备注：40 pages, 4 figures

【22】Transformers Can Implement Preconditioned Richardson Iteration for In-Context Gaussian Kernel Regression
标题：Transformer可以为上下文内高斯核回归实现预条件理查森迭代
链接：https://arxiv.org/abs/2605.08475

作者：Mingsong Yan,Dongyang Li,Charles Kulick,Sui Tang

【23】When Attention Beats Fourier: Multi-Scale Transformers for PDE Solving on Irregular Domains
标题：当注意力击败傅里叶时：用于求解不规则域上的PDL的多尺度变换器
链接：https://arxiv.org/abs/2605.08318

作者：Brandon Yee,Pairie Koh,Jack Rodriguez,Mihir Tekal

【24】FlashSVD v1.5: Making Low-Rank Transformers Inference Actually Fast
标题：Flash DID v1.5：快速实现低级Transformer推理
链接：https://arxiv.org/abs/2605.08314

作者：Wenhao Wu,Zishan Shao,Kangning Cui,Jinhee Kim,Yixiao Wang,Hancheng Ye,Danyang Zhuo,Yiran Chen

【25】Priming: Hybrid State Space Models From Pre-trained Transformers
标题：启动：来自预训练Transformer的混合状态空间模型
链接：https://arxiv.org/abs/2605.08301

作者：Aditya Chattopadhyay,Elvis Nunez,Prannay Kaul,Benjamin Bowman,Evan Becker,Luca Zancato,David Thomas,Wei Xia,Stefano Soatto

【26】TTCD:Transformer Integrated Temporal Causal Discovery from Non-Stationary Time Series Data
标题：TTCD：从非平稳时间序列数据中发现Transformer集成时间原因
链接：https://arxiv.org/abs/2605.08111

作者：Omar Faruque,Sahara Ali,Xue Zheng,Jianwu Wang
备注：18 Pages

【27】Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime
标题：低温条件下平均场Transformer的浓度现象的量化
链接：https://arxiv.org/abs/2605.10931

作者：Albert Alcalde,Leon Bungert,Konstantin Riedl,Tim Roith
备注：30 pages, 10 figures

【28】Kinetic theory for Transformers and the lost-in-the-middle phenomenon
标题：Transformer的动力学理论和中间迷失现象
链接：https://arxiv.org/abs/2605.09213

作者：Mitia Duerinckx,Borjan Geshkovski,Stefano Rossi

【29】Learning Theory of Transformers: Local-to-Global Approximation via Softmax Partition of Unity
标题：Transformer学习理论：通过Softmax Unity划分的局部到全局逼近
链接：https://arxiv.org/abs/2605.08811

作者：Zhongjie Shi,Wenjing Liao

【30】Learning the Channel Gain from Anywhere to Anywhere via Cross-environment Transformer Estimators
标题：通过跨环境Transformer估计器了解随时随地的渠道收益
链接：https://arxiv.org/abs/2605.08211

作者：Prasenjit Dhara,Daniel Romero

【31】Domain-Adaptive Arrhythmia Classification Using a Hybrid Transformer on Wearable Heart Signals
标题：使用混合Transformer对可穿戴心脏信号进行区域自适应心律失常分类
链接：https://arxiv.org/abs/2605.08199

作者：Maedeh H. Toosi,Siamak Mohammadi

GAN|对抗|攻击|生成相关(26篇)

【1】Mistake-Bounded Language Generation
标题：有界语言生成
链接：https://arxiv.org/abs/2605.10809

作者：Jon Kleinberg,Charlotte Peale,Omer Reingold

【2】Dynamic Cross-Modal Prompt Generation for Multimodal Continual Instruction Tuning
标题：多模式连续指令调优的动态跨模式提示生成
链接：https://arxiv.org/abs/2605.10765

作者：Tao Hu,Da-Wei Zhou

【3】AllocMV: Optimal Resource Allocation for Music Video Generation via Structured Persistent State
标题：AllocMV：通过结构化持续状态进行音乐视频生成的最佳资源分配
链接：https://arxiv.org/abs/2605.10723

作者：Huimin Wang,Leilei Ouyang,Chang Xia,Yongqi Kang,Yu Fu,Yuqi Ouyang

【4】Not Blind but Silenced: Rebalancing Vision and Language via Adversarial Counter-Commonsense Equilibrium
标题：不是盲目而是沉默：通过对抗性反常识平衡重新平衡视觉和语言
链接：https://arxiv.org/abs/2605.10676

作者：Qingxin Xiao,Peilin Zhao,Yangyang Zhao,Lingwei Dang,Qingyao Wu

【5】Nearly-Optimal Algorithm for Adversarial Kernelized Bandits
标题：对抗性核心化盗贼的近优算法
链接：https://arxiv.org/abs/2605.10299

作者：Shogo Iwazaki
备注：47 pages

【6】Building Korean linguistic resource for NLU data generation of banking app CS dialog system
标题：为银行应用CS对话系统的NLU数据生成构建韩语语言资源
链接：https://arxiv.org/abs/2605.10241

作者：Jeongwoo Yoon, On-yu Park, Changhoe Hwang, Gwanghoon Yoo, Eric Laporte, Jeesun Nam

【7】FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization
标题：FORGE：面向片段的排名和生成，用于上下文感知分子优化
链接：https://arxiv.org/abs/2605.10230

作者：Qingchuan Zhang, He Cao, Hao Li, Yanjun Shao, Zhiyuan Liu, Shihang Wang, Shufang Xie, Shenghua Gao, Xinwu Ye

【8】Fix the Loss, Not the Radius: Rethinking the Adversarial Perturbation of Sharpness-Aware Minimization
标题：修复损失，而不是半径：重新思考锐度感知最小化的对抗性扰动
链接：https://arxiv.org/abs/2605.10183

作者：Jinping Wang, Qinhan Liu, Zhiwu Xie, Zhiqiang Gao
备注：Accepted by ICML2026

【9】Generating Symmetric Materials using Latent Flow Matching
标题：使用潜流匹配生成对称材料
链接：https://arxiv.org/abs/2605.10115

作者：Anmar Karmush, Cedric Mathieu Brandenburg, Soheil Ershadrad, Johanna Rosén, Michael Felsberg, Filip Ekström Kelvinius
备注：Preprint

【10】Geometric 4D Stitching for Grounded 4D Generation
标题：几何4D缝合用于接地4D生成
链接：https://arxiv.org/abs/2605.09984

作者：Sunwoo Park, Taesung Kwon, Jong Chul Ye

【11】G-Zero: Self-Play for Open-Ended Generation from Zero Data
标题：G-Zero：零数据开放式一代的自我游戏
链接：https://arxiv.org/abs/2605.09959

作者：Chengsong Huang,Haolin Liu,Tong Zheng,Runpeng Dai,Langlin Huang,Jinyuan Li,Zongxia Li,Zhepei Wei,Yu Meng,Jiaxin Huang

【12】Generating synthetic electronic health record data using agent-based models to evaluate machine learning robustness under mass casualty incidents
标题：使用基于代理的模型生成合成电子健康记录数据，以评估大规模伤亡事件下的机器学习稳健性
链接：https://arxiv.org/abs/2605.09951

作者：Roben Delos Reyes,Daniel Capurro,Nicholas Geard
备注：14 pages, 1 figure; accepted at CHIL 2026

【13】Discovery of Nonlinear Dynamics with Automated Basis Function Generation
标题：利用自动基函数生成发现非线性动力学
链接：https://arxiv.org/abs/2605.09696

作者：Mohammad Amin Basiri,Charles Nicholson
备注：53 pages, 17 figures. Code available at https://github.com/mabasiri95/AutoSINDy

【14】On Characterizing Learnability for Adversarial Noisy Bandits
标题：对抗性吵闹盗贼的学习能力特征
链接：https://arxiv.org/abs/2605.09200

作者：Steve Hanneke,Kun Wang

【15】Spherical Boltzmann machines: a solvable theory of learning and generation in energy-based models
标题：球形Boltzmann机：基于能量的模型中学习和生成的可解理论
链接：https://arxiv.org/abs/2605.09031

作者：Thomas Tulinski,Simona Cocco,Rémi Monasson,Jorge Fernandez-De-Cossio-Diaz

【16】Enhancing Adversarial Robustness in Network Intrusion Detection: A Layer-wise Adaptive Regularization Approach
标题：增强网络入侵检测中的对抗鲁棒性：分层自适应正规化方法
链接：https://arxiv.org/abs/2605.08910

作者：Hira Nasir,Eiman Javed,Balawal Shabir,Zunera Jalil,Ahmad Mohsin

【17】MDL-GBG: A Non-parametric and Interpretable Granular-Ball Generation Method for Clustering
标题：MDL-GBG：一种非参数且可解释的颗粒球生成方法
链接：https://arxiv.org/abs/2605.08759

作者：Zeqiang Xian,Caihui Liu,Yong Zhang,Wenjing Qiu,Duoqian Miao,Witold Pedrycz
备注：12 pages, 5 figures

【18】MeshFIM: Local Low-Poly Mesh Editing via Fill-in-the-Middle Autoregressive Generation
标题：网格网格：通过中间填充自回归生成进行局部低多边形网格编辑
链接：https://arxiv.org/abs/2605.08744

作者：Dingdong Yang,Jian Liu,Biwen Lei,Haohan Weng,Zhuo Chen,Song Guo,Hao Richard Zhang,Ali Mahdavi Amiri,Chunchao Guo

【19】Structured Recurrent Mixers for Massively Parallelized Sequence Generation
标题：用于大规模并行序列生成的结构化循环混合器
链接：https://arxiv.org/abs/2605.08696

作者：Benjamin L. Badger

【20】Improving Generative Adversarial Networks with Self-Distillation
标题：利用自蒸馏改进生成对抗网络
链接：https://arxiv.org/abs/2605.08577

作者：Antoni Nowinowski,Krzysztof Krawiec

【21】VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation
标题：VeriContest：可验证代码生成的竞争性编程基准
链接：https://arxiv.org/abs/2605.08553

作者：Zichen Xie,Mrigank Pawagi,Yuxin Liu,Aaditi Rai,Lize Shao,John Berberian,Sicong Che,Wenxi Wang

【22】TARO: Temporal Adversarial Rectification Optimization Using Diffusion Models as Purifiers
标题：TARO：使用扩散模型作为净化器的时间对抗纠正优化
链接：https://arxiv.org/abs/2605.08440

作者：Daniel Wesego,Pedram Rooshenas

【23】Alice v1: Distillation-Enhanced Video Generation Surpassing Closed-Source Models
标题：Alice v1：超越闭源模型的蒸馏增强视频生成
链接：https://arxiv.org/abs/2605.08115

作者：Wang Xiaoyu,Phong Nguyen,Chen Zhao

【24】Context-Augmented Code Generation: How Product Context Improves AI Coding Agent Decision Compliance by 49%
标题：上下文增强代码生成：产品上下文如何将人工智能编码代理决策合规性提高49%
链接：https://arxiv.org/abs/2605.08112

作者：Drew Dillon,Kasyap Varanasi
备注：16 pages, 3 figures, 16 tables. Benchmark repository: https://github.com/brief-hq/dcbench

【25】Extended Wasserstein-GAN Approach to Causal Distribution Learning: Density-Free Estimation and Minimax Optimality
标题：因果分布学习的扩展Wasserstein-GAN方法：无密度估计和极小最优性
链接：https://arxiv.org/abs/2605.10206

作者：Shu Tamano,Masaaki Imaizumi

【26】TD3B: Transition-Directed Discrete Diffusion for Allosteric Binder Generation
标题：TD 3B：用于变立体粘合剂生成的转变定向离散扩散
链接：https://arxiv.org/abs/2605.09810

作者：Hanqun Cao,Aastha Pal,Sophia Tang,Yinuo Zhang,Jingjie Zhang,Pheng Ann Heng,Pranam Chatterjee
备注：Published as a Spotlight at ICML 2026 (Proceedings of the 43rd International Conference on Machine Learning, Seoul, South Korea)

半/弱/无/有监督|不确定性|主动学习(19篇)

【1】Likelihood scoring for continuations of mathematical text: a self-supervised benchmark with tests for shortcut vulnerabilities
标题：数学文本延续的可能评分：具有捷径漏洞测试的自我监督基准
链接：https://arxiv.org/abs/2605.10810

作者：Daniel Ranard
备注：13 pages + appendices, 4 figures

【2】An Uncertainty-Aware Resilience Micro-Agent for Causal Observability in the Computing Continuum
标题：一种具有不确定性的弹性微代理，用于计算连续体中因果可观察性
链接：https://arxiv.org/abs/2605.10718

作者：Suvi De Silva,Alfreds Lapkovskis,Alaa Saleh,Sasu Tarkoma,Praveen Kumar Donta

【3】Active Learning for Gaussian Process Regression Under Self-Induced Boltzmann Weights
标题：自诱导Boltzmann权重下高斯过程回归的主动学习
链接：https://arxiv.org/abs/2605.10654

作者：Jixiang Qing,Henry Moss,Matthias Sachs

【4】Self-Attention as a Covariance Readout: A Unified View of In-Context Learning and Repetition
标题：自我注意力作为协方差的解读：背景学习和重复的统一观点
链接：https://arxiv.org/abs/2605.10466

作者：Haoren Xu, Guanhua Fang

【5】DeepLévy: Learning Heavy-Tailed Uncertainty in Highly Volatile Time Series
标题：DeepLévy：在高度波动的时间序列中学习重尾不确定性
链接：https://arxiv.org/abs/2605.10364

作者：Yang Yang, Du Yin, Hao Xue, Flora Salim

【6】Portable Active Learning for Object Detection
标题：用于对象检测的便携式主动学习
链接：https://arxiv.org/abs/2605.10349

作者：Rashi Sharma, Justin Timothy C. Bersamin, Karthikk Subramanian
备注：CVPR 2026(highlight)

【7】When Normality Shifts: Risk-Aware Test-Time Adaptation for Unsupervised Tabular Anomaly Detection
标题：当正态性发生变化时：无监督表格异常检测的风险意识测试时自适应
链接：https://arxiv.org/abs/2605.10242

作者：Wei Huang, Hezhe Qiao, Kailai Zhang, Zaisheng Ye, Yu-Ming Shang, Xiangling Fu
备注：13 pages, 6 figures

【8】Unsupervised Process Reward Models
标题：无监督流程奖励模型
链接：https://arxiv.org/abs/2605.10158

作者：Artyom Gadetsky, Maxim Kodryan, Siba Smarak Panigrahi, Hang Guo, Maria Brbic
备注：preprint

【9】Modeling Atomic Conformational Ensembles of Proteins via Test-Time Supervision of Boltz-2 on Cryo-EM Density Maps
标题：通过Cryo-EM密度图上Boltz-2的测试时间监督对蛋白质的原子骨架进行建模
链接：https://arxiv.org/abs/2605.09832

作者：Jay Shenoy, Miro Astore, Axel Levy, Frédéric Poitevin, Sonya M. Hanson, Gordon Wetzstein
备注：Project page: this https URL

【10】WISTERIA: Learning Clinical Representations from Noisy Supervision via Multi-View Consistency in Electronic Health Records
标题：WISTERIA：通过电子健康记录中的多视图一致性从噪音监督中学习临床代表
链接：https://arxiv.org/abs/2605.09765

作者：Ruan Dong,Yuanyun Zhang,Shi Li

【11】Semi-Supervised Neural Super-Resolution for Mesh-Based Simulations
标题：基于网格的模拟的半监督神经超分辨率
链接：https://arxiv.org/abs/2605.09284

作者：Jiyeon Kim,Youngjoon Hong,Won-Yong Shin
备注：International Conference on Machine Learning (ICML 2026) (to appear) (Please cite our conference version.)

【12】Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories
标题：Self-ReSET：学会从不安全的推理轨迹中自我恢复
链接：https://arxiv.org/abs/2605.08936

作者：Dongcheng Zhang,Yi Zhang,Yuxin Chen,An Zhang,Xiang Wang,Chaochao Lu

【13】TopoGeoScore: A Self-Supervised Source-Only Geometric Framework for OOD Checkpoint Selection
标题：TopoGeocore：用于OOD检查点选择的自我监督的仅源几何框架
链接：https://arxiv.org/abs/2605.08870

作者：Farid Hazratian,Ali Zia,Hien Duy Nguyen

【14】Inpainting physics: self-supervised learning for context-driven fluid simulation
标题：修复物理：上下文驱动的流体模拟的自我监督学习
链接：https://arxiv.org/abs/2605.08832

作者：Jonas Weidner,Yeray Martin-Ruisanchez,Daniel Rückert,Benedikt Wiestler,Julian Suk

【15】The Global Empirical NTK: Self-Referential Bias and Dimensionality of Gradient Descent Learning
标题：全球经验NTK：梯度下降学习的自我参考偏差和主观性
链接：https://arxiv.org/abs/2605.08746

作者：James Hazelden,Laura Driscoll,Eli Shlizerman,Eric Shea-Brown
备注：Submitted to TMLR

【16】SeBA: Semi-supervised few-shot learning via Separated-at-Birth Alignment for tabular data
标题：SeBA：通过表格数据的出生时分离对齐进行半监督少量学习
链接：https://arxiv.org/abs/2605.08519

作者：Kacper Jurek,Wojciech Batko,Marek Śmieja,Marcin Przewięźlikowski

【17】Weakly Supervised Concept Learning for Object-centric Visual Reasoning
标题：用于以对象为中心的视觉推理的弱监督概念学习
链接：https://arxiv.org/abs/2605.08201

作者：Sparsh Tiwari,Bettina Finzel,Gesina Schwalbe

【18】Unified Approach for Weakly Supervised Multicalibration
标题：弱监督多重校准的统一方法
链接：https://arxiv.org/abs/2605.09857

作者：Futoshi Futami,Takashi Ishida

【19】An Explainable Unsupervised-to-Supervised Machine Learning Framework for Dietary Pattern Discovery Using UK National Dietary Survey Data
标题：使用英国国家饮食调查数据进行饮食模式发现的可解释的、无监督到监督的机器学习框架
链接：https://arxiv.org/abs/2605.08242

作者：Wing Yi Yu,Chun Yin Chiu
备注：12 pages, 6 figures, 9 tables. Accepted by the 14th International Conference on Frontiers of Intelligent Computing: Theory and Applications (FICTA 2026)

迁移|Zero/Few/One-Shot|自适应(42篇)

【1】Transcoda: End-to-End Zero-Shot Optical Music Recognition via Data-Centric Synthetic Training
标题：Transcoda：通过以数据为中心的合成训练进行端到端Zero-Shot光学音乐识别
链接：https://arxiv.org/abs/2605.10835

作者：Daniel Dratschuk,Paul Swoboda
备注：13 pages, 7 figures

【2】NoRIN: Backbone-Adaptive Reversible Normalization for Time-Series Forecasting
标题：NoRIN：用于时间序列预测的主干自适应可逆规范化
链接：https://arxiv.org/abs/2605.10823

作者：Shun Zhang,Yuyang Xiao
备注：8 pages, 2 figures

【3】Provable Sparse Inversion and Token Relabel Enhanced One-shot Federated Learning with ViTs
标题：可证明的稀疏倒置和令牌重新标记使用ViT增强的一次性联邦学习
链接：https://arxiv.org/abs/2605.10748

作者：Li Shen,Xiaolei Hao,Qinglun Li,Xiaochun Cao,Zhifeng Hao,Xun Yang
备注：18 Pages

【4】AdaPaD: Adaptive Parallel Deflation for PEFT with Self-Correcting Rank Discovery
标题：AdaPaD：具有自我纠正等级发现的PEFT自适应并行紧缩
链接：https://arxiv.org/abs/2605.10741

作者：Barbara Su,Fangshuo Liao,Anastasios Kyrillidis

【5】Why Zeroth-Order Adaptation May Forget Less: A Randomized Shaping Theory
标题：为什么零阶适应可能会忘记更少：随机成形理论
链接：https://arxiv.org/abs/2605.10658

作者：Yao Shu,Jian Mu,Zhongxiang Dai

【6】Don't Fix the Basis -- Learn It: Spectral Representation with Adaptive Basis Learning for PDEs
标题：不要修复基础--学习它：具有用于PED的自适应基础学习的谱表示
链接：https://arxiv.org/abs/2605.10451

作者：Xuxiang Zhao, Angelica I. Aviles-Rivero
备注：26 pages, 4 figures

【7】Remember to Forget: Gated Adaptive Positional Encoding
标题：记住忘记：门控自适应位置编码
链接：https://arxiv.org/abs/2605.10414

作者：Riccardo Ali, Alessio Borgi, Christopher Irwin, Mario Severino, Pietro Liò

【8】PC3D: Zero-Shot Cooperation Across Variable Rosters via Personalized Context Distillation
标题：PC 3D：通过个性化上下文蒸馏实现可变名册的Zero-Shot合作
链接：https://arxiv.org/abs/2605.10377

作者：Ahmet Onur Akman, Rafał Kucharski

【9】PowerStep: Memory-Efficient Adaptive Optimization via $\ell_p$-Norm Steepest Descent
链接：https://arxiv.org/abs/2605.10335

作者：Yao Lu, Dengdong Fan, Shixun Zhang, Yonghong Tian

【10】LeapTS: Rethinking Time Series Forecasting as Adaptive Multi-Horizon Scheduling
标题：LeapTS：将时间序列预测重新思考为自适应多水平调度
链接：https://arxiv.org/abs/2605.10292

作者：Sheng Pan, Ming Jin, Bo Du, Shirui Pan

【11】OUIDecay: Adaptive Layer-wise Weight Decay for CNNs Using Online Activation Patterns
标题：OUIDecay：使用在线激活模式的CNN自适应分层权重衰减
链接：https://arxiv.org/abs/2605.10161

作者：Alberto Fernández-Hernández, Jose I. Mestre, Cristian Pérez-Corral, Manuel F. Dolz, Jose Duato, Enrique S. Quintana-Ortí

【12】Adaptive Action Chunking via Multi-Chunk Q Value Estimation
标题：通过多块Q值估计的自适应动作分块
链接：https://arxiv.org/abs/2605.10044

作者：Yongjae Shin, Jongseong Chae, Seongmin Kim, Jongeui Park, Youngchul Sung

【13】Continual Harness: Online Adaptation for Self-Improving Foundation Agents
标题：持续调整：自我完善的基金会代理人的在线调整
链接：https://arxiv.org/abs/2605.09998

作者：Seth Karten, Joel Zhang, Tersoo Upaa Jr, Ruirong Feng, Wenzhe Li, Chengshuai Shi, Chi Jin, Kiran Vodrahalli
备注：28 pages, 19 figures, 5 tables

【14】Consolidation-Expansion Operator Mechanics:A Unified Framework for Adaptive Learning
标题：整合-扩展操作员机制：自适应学习的统一框架
链接：https://arxiv.org/abs/2605.09968

作者：Debashis Guha
备注：38 pages

【15】Entropy-informed Decoding: Adaptive Information-Driven Branching
标题：基于信息的解码：自适应信息驱动的分支
链接：https://arxiv.org/abs/2605.09745

作者：Benjamin Patrick Evans,Sumitra Ganesh,Leo Ardon
备注：Accepted at ICML 2026

【16】Adaptive Data Harvesting for Efficient Neural Network Learning with Universal Constraints
标题：自适应数据收集以实现具有普遍约束的高效神经网络学习
链接：https://arxiv.org/abs/2605.09707

作者：Siteng Kang,Xinhua Zhang
备注：Preprint

【17】Learning Multi-Indicator Weights for Data Selection: A Joint Task-Model Adaptation Framework with Efficient Proxies
标题：学习数据选择的多指标权重：具有高效代理的联合任务模型适应框架
链接：https://arxiv.org/abs/2605.09665

作者：Jingze Song,Zihao Chen,Wenqing Chen,Zibin Zheng
备注：This work has been accepted at IJCAI 2026

【18】Adaptive DNN Partitioning and Offloading in Heterogeneous Edge-Cloud Continuum
标题：异类边缘云连续体中的自适应DNN分区和卸载
链接：https://arxiv.org/abs/2605.09623

作者：Akuen Akoi Deng,Eimantas Butkus,Alfreds Lapkovskis,Praveen Kumar Donta

【19】When Adaptation Fails: A Gradient-Based Diagnosis of Collapsed Gating in Vision-Language Prompt Learning
标题：当适应失败时：视觉语言提示学习中门控塌陷的基于对象的诊断
链接：https://arxiv.org/abs/2605.09549

作者：Yunxuan Fang,Ziwei Zhang,Xinhe Wang

【20】D2ACE: Multi-Label Batch Selection Guided by Dual Dynamics and Adaptive Correlation Enhancement
标题：D2 ACE：双动态和自适应相关增强引导的多标签批量选择
链接：https://arxiv.org/abs/2605.09400

作者：Bin Liu,Haoyu Peng,Zhijia Wei,Jiajing Zhang,Grigorios Tsoumakas
备注：18 pages

【21】FLAME: Adaptive Mixture-of-Experts for Continual Multimodal Multi-Task Learning
标题：FLAME：用于连续多模式多任务学习的自适应专家混合
链接：https://arxiv.org/abs/2605.09355

作者：Xing Han,Shravan Chaudhari,Tanvi Ranade,Rama Chellappa,Suchi Saria
备注：37 pages, 25 figures, 6 tables

【22】Neural Cluster First, Route Second: One-Shot Capacitated Vehicle Routing via Differentiable Optimal Transport
标题：神经集群第一，路线第二：通过可区分最优交通的一次性容量车辆路线
链接：https://arxiv.org/abs/2605.09301

作者：Samuel J. K. Chin,Maximilian Schiffer
备注：30 pages, 9 figures

【23】Instance-Adaptive Online Multicalibration
标题：实例自适应在线多重校准
链接：https://arxiv.org/abs/2605.09273

作者：Zhiming Huang,Jamie Morgenstern,Aaron Roth,Claire Jie Zhang

【24】DARE: Difficulty-Adaptive Reinforcement Learning with Co-Evolved Difficulty Estimation
标题：DARE：具有协同进化难度估计的难度自适应强化学习
链接：https://arxiv.org/abs/2605.09188

作者：Yang Zhou,Can Jin,Zihan Dong,Zhepeng Wang,Yanting Yang,Shiyu Zhao,Lei Li,Runxue Bao,Yaochen Xie,Dimitris N. Metaxas

【25】Transfer Learning of Multiobjective Indirect Low-Thrust Trajectories Using Diffusion Models and Markov Chain Monte Carlo
标题：利用扩散模型和马尔科夫链蒙特卡罗进行多目标间接低推力轨迹的迁移学习
链接：https://arxiv.org/abs/2605.09125

作者：Jannik Graebner,Ryne Beeson

【26】AdaPreLoRA: Adafactor Preconditioned Low-Rank Adaptation
标题：AdaPreLoRA：Adafactor预条件低等级适应
链接：https://arxiv.org/abs/2605.08734

作者：Ziyun Liu,Fengmiao Bian,Jian-Feng Cai
备注：27 pages

【27】Robust Server Defense Against Unreliable Clients in One-Shot Fair Collaborative Machine Learning
标题：一次公平协作机器学习中针对不可靠客户端的强大服务器防御
链接：https://arxiv.org/abs/2605.08616

作者：Chia-Yuan Wu,Frank E. Curtis,Daniel P. Robinson
备注：Accepted at the 2nd International Conference on Federated Learning and Intelligent Computing Systems (FLICS 2026)

【28】FLARE: One-Shot PE-Level Fault Localization in Systolic Arrays via Algebraic Test Vectors
标题：MBE：通过代数测试载体在Syrup阵列中进行单次PE级故障定位
链接：https://arxiv.org/abs/2605.08594

作者：Logashree Venkatasubramanian,Zishen Wan,Viveck Cadambe

【29】Beyond Static Bias: Adaptive Multi-Fidelity Bandits with Improving Proxies
标题：超越静态偏见：具有改进代理的自适应多保真盗贼
链接：https://arxiv.org/abs/2605.08558

作者：Muyun Lu,Haoyang Hong,Huazheng Wang,Ying Lin

【30】MC-RFM: Geometry-Aware Few-Shot Adaptation via Mixed-Curvature Riemannian Flow Matching
标题：MC-RFM：通过混合曲线Riemann流匹配的几何感知Few-Shot自适应
链接：https://arxiv.org/abs/2605.08557

作者：Salim Khazem,Ibrahim Mohamed Serouis,Zakaria Ezzahed
备注：Submitted to NeurIPS (Under Review)

【31】HEART: A High-Efficiency Adaptive Real-Time Telemonitoring Framework for Secure Electrocardiogram Signal Transmission Using Chaotic Encryption
标题：HEART：一种使用混乱加密安全心电图信号传输的高效自适应实时远程监控框架
链接：https://arxiv.org/abs/2605.08456

作者：Beyazıt Bestami Yuksel
备注：15 pages, 4 figure, 3 table

【32】Zero-shot Imitation Learning by Latent Topology Mapping
标题：通过潜在布局映射实现Zero-Shot模仿学习
链接：https://arxiv.org/abs/2605.08450

作者：Maxwell J. Jacobson,Yexiang Xue

【33】AdamFLIP: Adaptive Momentum Feedback Linearization Optimization for Hard Constrained PINN Training
标题：AdamFLIP：用于硬约束PINN训练的自适应动量反馈线性化优化
链接：https://arxiv.org/abs/2605.08408

作者：Binghang Lu,Runyu Zhang,Changhong Mou,Na Li,Guang Lin

【34】In-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot Classification
标题：上下文固定：当演示标签在Few-Shot分类中插入语义时
链接：https://arxiv.org/abs/2605.08295

作者：Ming Liu
备注：12 pages (10 main + 2 appendix), 4 figures, 5 tables

【35】Beyond the False Trade-off: Adaptive EWC for Stealthy and Generalizable T2I Backdoors
标题：超越虚假权衡：自适应EWC实现隐形和可推广的T2 I后门
链接：https://arxiv.org/abs/2605.08280

作者：Lu Bowen,Xinyu Tang,Yin Yin Low,Shu-Min Leong

【36】Text-Guided Multi-Scale Frequency Representation Adaptation
标题：文本引导的多尺度频率表示自适应
链接：https://arxiv.org/abs/2605.08181

作者：Weicai Yan,Xinhua Ma,Wang Lin,Tao Jin
备注：ACL 2026 Main

【37】CERSA: Cumulative Energy-Retaining Subspace Adaptation for Memory-Efficient Fine-Tuning
标题：CERSA：累积能量保持子空间自适应以实现内存高效微调
链接：https://arxiv.org/abs/2605.08174

作者：Jingze Ge,Xue Geng,Yun Liu,Wanqi Dong,Wang Zhe Mark,Min Wu,Ngai-Man Cheung,Bharadwaj Veeravalli,Xulei Yang
备注：10 pages, 7 figures, supplementary material included

【38】BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models
标题：BaLoRA：大规模模型的Bayesian低等级适应
链接：https://arxiv.org/abs/2605.08110

作者：Dario Coscia,Sindy Löwe,Max Welling

【39】Kinetic-Optimal Scheduling with Moment Correction for Metric-Induced Discrete Flow Matching in Zero-Shot Text-to-Speech
标题：Zero-Shot文本到语音中公制诱导离散流匹配的带矩修正的动态最优调度
链接：https://arxiv.org/abs/2605.09386

作者：Dong Yang,Yiyi Cai,Haoyu Zhang,Yuki Saito,Hiroshi Saruwatari
备注：Under Review

【40】Quantum Transfer Learning Shows Improved Robustness in Low-Data Regimes
标题：量子转移学习在低数据状态下显示出更好的鲁棒性
链接：https://arxiv.org/abs/2605.09118

作者：Li-An Lo,Li-Yi Hsu,Hsien-Yi Hsieh
备注：22 pages, 5 figures

【41】Transfer Learning for Dead Fuel Moisture Prediction Using Time-Warping Recurrent Neural Networks
标题：使用时间扭曲回归神经网络的传递学习预测死燃料湿度
链接：https://arxiv.org/abs/2605.08379

作者：Jonathon Hirschi,Jan Mandel,Adam Kochanski
备注：Preprint. Related to PhD thesis work that is also available for preprint at https://doi.org/10.48550/arXiv.2604.02474

【42】Rethinking Entropy Minimization in Test-Time Adaptation for Autoregressive Models
标题：重新思考自回归模型测试时自适应中的最小化
链接：https://arxiv.org/abs/2605.08186

作者：Wei-Ping Huang,Chee-En Yu,Guan-Ting Lin,Hung-yi Lee
备注：Submitted to INTERSPEECH 2026

强化学习(28篇)

【1】Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning
标题：用于显着强化学习的动态技能时间表管理
链接：https://arxiv.org/abs/2605.10923

作者：Junhao Shen,Teng Zhang,Xiaoyan Zhao,Hong Cheng
备注：Implementation code is available at https://github.com/ejhshen/SLIM

【2】Policy Gradient Methods for Non-Markovian Reinforcement Learning
标题：非马尔科夫强化学习的策略梯度方法
链接：https://arxiv.org/abs/2605.10816

作者：Avik Kar,Siddharth Chandak,Rahul Singh,Soumitra Sinhahajari,Eric Moulines,Shalabh Bhatnagar,Nicholas Bambos
备注：39 pages, 5 figures, 1 table

【3】Controllability in preference-conditioned multi-objective reinforcement learning
标题：偏好条件多目标强化学习的可控性
链接：https://arxiv.org/abs/2605.10585

作者：Pau de las Heras Molins,Beyazit Yalcinkaya,Lasse Peters,David Fridovich-Keil,Georgios Bakirtzis

【4】Higher Resolution, Better Generalization: Unlocking Visual Scaling in Deep Reinforcement Learning
标题：更高的分辨率、更好的概括：解锁深度强化学习中的视觉缩放
链接：https://arxiv.org/abs/2605.10546

作者：Raphael Trumpp,Ömer Veysel Çağatan,Barış Akgün,Marco Caccamo

【5】Priority-Driven Control and Communication in Decentralized Multi-Agent Systems via Reinforcement Learning
标题：通过强化学习实现分散多智能体系统中优先级驱动的控制和通信
链接：https://arxiv.org/abs/2605.10482

作者：Qingyun Guo, Junyi Shi, Tomasz Piotr Kucner, Dominik Baumann
备注：Accepted to the 23rd IFAC World Congress

【6】Robust Probabilistic Shielding for Safe Offline Reinforcement Learning
标题：用于安全离线强化学习的鲁棒概率屏蔽
链接：https://arxiv.org/abs/2605.10293

作者：Maris F. L. Galesloot, Thomas Rhemrev, Nils Jansen

【7】When Does Non-Uniform Replay Matter in Reinforcement Learning?
标题：非均匀重播什么时候在强化学习中重要？
链接：https://arxiv.org/abs/2605.10236

作者：Michal Korniak, Mikołaj Czarnecki, Yarden As, Piotr Miłoś, Pieter Abbeel, Michal Nauman

【8】Balancing Efficiency and Fairness in Traffic Light Control through Deep Reinforcement Learning
标题：通过深度强化学习平衡红绿灯控制的效率和公平性
链接：https://arxiv.org/abs/2605.10170

作者：Matteo Cederle, Giacomo Scatto, Gian Antonio Susto
备注：Paper accepted to the 2026 IFAC World Congress, held in Busan (KOR), August 23rd-28th, 2026

【9】Learning to Compress Time-to-Control: A Reinforcement Learning Framework for Chronic Disease Management
标题：学会压缩控制时间：慢性病管理的强化学习框架
链接：https://arxiv.org/abs/2605.09818

作者：Prabhjot Singh, Abhishek Gupta, Chris Betz, Abe Flansburg, Brett Ives, Sudeep Lama, Jung Hoon Son
备注：26 pages, 3 figures

【10】Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning
标题：通过强化微调克服视觉连续学习中的灾难性遗忘
链接：https://arxiv.org/abs/2605.09640

作者：Meng Lou,Hanzhong Guo,Linwei Chen,Yizhou Yu

【11】Plan2Cleanse: Test-Time Backdoor Defense via Monte-Carlo Planning in Deep Reinforcement Learning
标题：Plan 2Cleanse：通过深度强化学习中的蒙特卡洛规划进行测试时后门防御
链接：https://arxiv.org/abs/2605.09638

作者：Sze-Ann Chen,Zhi-Yi Chin,Kui-Yuan Chen,Chi-Yu Li,Ping-Chun Hsieh
备注：Published in Transactions on Machine Learning Research (TMLR)

【12】Multi-scale Predictive Representations for Goal-conditioned Reinforcement Learning
标题：目标条件强化学习的多尺度预测表示
链接：https://arxiv.org/abs/2605.09364

作者：Valliappan Chidambaram Adaikkappan,David Meger,Sai Rajeswar,Pietro Mazzaglia

【13】Skill-R1: Agent Skill Evolution via Reinforcement Learning
标题：Skill-R1：通过强化学习的Agent技能进化
链接：https://arxiv.org/abs/2605.09359

作者：Yash Vishe,Rohan Surana,Xunyi Jiang,Zihan Huang,Xintong Li,Nikki Lijing Kuang,Tong Yu,Ryan A. Rossi,Jingbo Shang,Julian McAuley,Junda Wu

【14】Rethinking Ratio-Based Trust Regions for Policy Optimization in Multi-Agent Reinforcement Learning
标题：重新思考基于比率的信任区域以实现多智能体强化学习中的政策优化
链接：https://arxiv.org/abs/2605.09212

作者：Chulabhaya Wijesundara,Andrea Baisero,Zhongheng Li,Gregory Castañón,Alan Carlin,Christopher Amato

【15】BubbleSpec: Turning Long-Tail Bubbles into Speculative Rollout Drafts for Synchronous Reinforcement Learning
标题：BubbleSec：将长尾泡沫转变为同步强化学习的推测性推出草案
链接：https://arxiv.org/abs/2605.08862

作者：Yuhang Xu,Kaibin Tian,Yang Tian,Zhice Yang,Yifeng Yu,Yan Li,Shengzhong Liu,Fan Wu,Guihai Chen

【16】ReLibra: Routing-Replay-Guided Load Balancing for MoE Training in Reinforcement Learning
标题：ReLibra：强化学习中MoE训练的训练-回放-引导负载平衡
链接：https://arxiv.org/abs/2605.08639

作者：Chao Jin,Xinming Wei,Yinmin Zhong,Chengxu Yang,Bingyang Wu,Ruidong Zhu,Zili Zhang,Yuliang Liu,Xin Jin

【17】Quantile-Coupled Flow Matching for Distributional Reinforcement Learning
标题：分布式强化学习的分位耦合流匹配
链接：https://arxiv.org/abs/2605.08515

作者：Michael Groom,Victor-Alexandru Darvariu,Lars Kunze,James Wilson,Nick Hawes

【18】DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards
标题：DUET：通过可验证的奖励优化强化学习的代币预算分配
链接：https://arxiv.org/abs/2605.08441

作者：Haoyu Hu,Xuandong Zhao,Xuhai "Orson'' Xu,Nori Jacoby

【19】SACHI: Structured Agent Coordination via Holistic Information Integration in Multi-Agent Reinforcement Learning
标题：SACHI：多智能体强化学习中通过整体信息集成的结构化智能体协调
链接：https://arxiv.org/abs/2605.08391

作者：Nikunj Gupta,James Zachary Hare,Jesse Milzman,Rajgopal Kannan,Viktor Prasanna

【20】Reinforcement Learning for Scalable and Trustworthy Intelligent Systems
标题：可扩展且值得信赖的智能系统的强化学习
链接：https://arxiv.org/abs/2605.08378

作者：Guangchen Lan
备注：PhD thesis

【21】Path-Coupled Bellman Flows for Distributional Reinforcement Learning
标题：分布式强化学习的路径耦合Bellman流
链接：https://arxiv.org/abs/2605.08253

作者：Boyang Xu,Qing Zou,Siqin Yang,Hao Yan
备注：Accepted to the 43rd International Conference on Machine Learning (ICML 2026)

【22】Beyond Penalization: Diffusion-based Out-of-Distribution Detection and Selective Regularization in Offline Reinforcement Learning
标题：超越惩罚：离线强化学习中基于扩散的分布外检测和选择性正则化
链接：https://arxiv.org/abs/2605.08202

作者：Qingjun Wang,Hongtu Zhou,Hang Yu,Junqiao Zhao,Yanping Zhao,Chen Ye,Ziqiao Wang,Guang Chen
备注：10 pages, 5 figures. Accepted to ICLR 2026

【23】Quantile Geometry Regularization for Distributional Reinforcement Learning
标题：分布强化学习的分位数几何正规化
链接：https://arxiv.org/abs/2605.08182

作者：Zhaofan Zhang,Minghao Yang,Rufeng Chen,Sihong Xie,Hui Xiong

【24】Interactive Inverse Reinforcement Learning of Interaction Scenarios via Bi-level Optimization
标题：基于双层优化的交互场景交互式逆强化学习
链接：https://arxiv.org/abs/2605.08131

作者：Yue Mao,Shicheng Liu,Siyuan Xu,Minghui Zhu

【25】Distributional Reinforcement Learning via the Cramér Distance
标题：通过Cramér距离的分布强化学习
链接：https://arxiv.org/abs/2605.08104

作者：Vanya Aziz,Ivo Nowak,E. M. T Hendrix

【26】Reinforcement learning for inverse structural design and rapid laser cutting of kirigami prototypes
标题：反向结构设计和kirigami原型快速激光切割的强化学习
链接：https://arxiv.org/abs/2605.08098

作者：Milad Yazdani,Shahriar Shalileh,Dena Shahriari

【27】Equivariant Reinforcement Learning for Clifford Quantum Circuit Synthesis
标题：Clifford量子电路综合的等变强化学习
链接：https://arxiv.org/abs/2605.10910

作者：Richie Yeung,Aleks Kissinger,Rob Cornish

【28】Reinforcement Learning Measurement Model
标题：强化学习测量模型
链接：https://arxiv.org/abs/2605.09305

作者：Wenqian Xu,Feng Ji

元学习(4篇)

【1】RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards
标题：RubricEM：具有超越可验证奖励的规则指导政策分解的元RL
链接：https://arxiv.org/abs/2605.10899

作者：Gaotang Li,Bhavana Dalvi Mishra,Zifeng Wang,Jun Yan,Yanfei Chen,Chun-Liang Li,Long T. Le,Rujun Han,George Lee,Hanghang Tong,Chen-Yu Lee,Tomas Pfister
备注：63 pages, 6 figures

【2】From Regression to Inference: Meta-Learning Predictors for Neural Architecture Search
标题：从回归到推理：神经架构搜索的元学习预测器
链接：https://arxiv.org/abs/2605.09290

作者：Liping Deng,MingQing Xiao

【3】PRIM: Meta-Learned Bayesian Root Cause Analysis
标题：PRIM：元学习的Bayesian根本原因分析
链接：https://arxiv.org/abs/2605.08786

作者：Christopher Lohse,Anish Dhir,Amadou Ba,Bradley Eck,Marco Ruffini,Jonas Wahl

【4】NoiseRater: Meta-Learned Noise Valuation for Diffusion Model Training
标题：NoiseRater：扩散模型训练的元学习噪音估值
链接：https://arxiv.org/abs/2605.08144

作者：Fang Wu,Haokai Zhao,Da Xing,Hanqun Cao,Tinson Xu,Yanchao Li,Xiangru Tang,Zehong Wang,Aaron Tu,Kuan Pang,Hanchen Wang,Hongbin Lin,Zeqi Zhou,Yinxi Li,Peng Xia,Li Erran Li,Molei Tao,Jure Leskovec,Aditya Joshi,Yejin Choi

符号|符号学习(1篇)

【1】Additive Atomic Forests for Symbolic Function and Antiderivative Discovery
标题：符号功能和反衍生物发现的可加性原子森林
链接：https://arxiv.org/abs/2605.08130

作者：Reda Belaiche

分层学习(3篇)

【1】WavesFM: Hierarchical Representation Learning for Longitudinal Wearable Sensor Waveforms
标题：WavesFM：纵向可穿戴传感器波形的分层表示学习
链接：https://arxiv.org/abs/2605.09173

作者：Peng Cao,Zhijian Yang,Tennison Liu,Jonathan Wang,Jiang Wu,Magdalena Proszewska,Arvind Pillai,Mingwu Gao,Amir Farjadian,Lawrence Cai,Emily Blanchard,Daniel McDuff,Pramod Rudrapatna,Matthew Thompson,Anupam Pathak,Mark Malhotra,Shwetak Patel,Dina Katabi,Paolo Di Achille,Ming-Zher Poh

【2】Hierarchical Multi-Fidelity Learning for Predicting Three-Dimensional Flame Wrinkling and Turbulent Burning Velocity
标题：预测三维火焰起皱和湍流燃烧速度的分层多保真学习
链接：https://arxiv.org/abs/2605.08232

作者：Saghar Zolfaghari,Yu Xie,Junfeng Yang,Safa Jamali

【3】Performance and Energy Trade-Off Analysis of Hierarchical Federated Learning for Plant Disease Classification
标题：植物病害分类分层联邦学习的性能和能量权衡分析
链接：https://arxiv.org/abs/2605.08121

作者：Athanasios Papanikolaou,Athanasios Tziouvaras,Pavlos Stoikos,Apostolos Xenakis,Shameem A Puthiya Parambath,George Floros,Enrica Zereik,Ivan Petrovic,Fabio Bonsignorio
备注：Accepted for publication at the 2026 ERAS Conference

医学相关(13篇)

【1】Predictive Radiomics for Evaluation of Cancer Immune SignaturE in Glioblastoma: the PRECISE-GBM study
标题：预测放射组学评估胶质母细胞瘤中的癌症免疫SignaturE：PRECISE-GBM研究
链接：https://arxiv.org/abs/2605.10278

作者：Prajwal Ghimire, Junjie Li, Liu Yaou, Marc Modat, Thomas Booth
备注：Abstract : 226; Importance of study: 109; Manuscript: 5690 (excluding references) Figures: 4, Tables: 2 Supplemental File: 1

【2】Voice Biomarkers for Depression and Anxiety
标题：抑郁和焦虑的声音生物标志物
链接：https://arxiv.org/abs/2605.09908

作者：Oleksii Abramenko,Noah D. Stein,Colin Vaz

【3】Efficient Neural Architectures for Real-Time ECG Interpretation on Limited Hardware
标题：在有限硬件上实现实时心电图解释的高效神经架构
链接：https://arxiv.org/abs/2605.09848

作者：Ashery Mbilinyi, Callum O'Riley, Julia Handra, Ashley Moller-Hansen, Jason Andrade, Marc Deyell, Cameron Hague, Nathaniel Hawkins, Kendall Ho, Jonathan Leipsic, Roger Tam
备注：9 pages, 6 figures, 3 tables. Published in: 2025 IEEE International Conference on Big Data (BigData), pp. 3275-3284. DOI: https://doi.org/10.1109/BIGDATA66926.2025.11402097

【4】Quantum Circuit Simulation of Compartmental Drug Dynamics: Leveraging Variational Algorithms for Nonlinear Mixed-Effects Population Pharmacokinetics
标题：隔室药物动力学的量子电路模拟：利用变分算法研究非线性混合效应群体药代动力学
链接：https://arxiv.org/abs/2605.09691

作者：Isshaan Singh,Nandan Patel

【5】CLR-voyance: Reinforcing Open-Ended Reasoning for Inpatient Clinical Decision Support with Outcome-Aware Rubrics
标题：CLR-voyance：用结果感知指标加强住院临床决策支持的开放式推理
链接：https://arxiv.org/abs/2605.09584

作者：Aishik Nagar,Arun-Kumar Kaliya-Perumal,Yu-Hsuan Han,Andrew Sheng-Han Huang,Kristen Kee,Yushi Cao,Yiming Chen,Hongchao Jiang

【6】Biosignal Fingerprinting: A Cross-Modal PPG-ECG Foundation Model
标题：生物信号指纹：跨模式PPG-心电图基础模型
链接：https://arxiv.org/abs/2605.09579

作者：Zhangdaihong Liu,Chang Liu,Fenglin Liu,Yixuan Chen,Yang Yang,David A. Clifton,Xiao Gu
备注：21 pages, 8 figures, 7 tables

【7】A Cross-Layered Multi-Drone Coordination for Medical Supply Delivery during Disaster Response Management
标题：灾难响应管理期间医疗物资交付的跨分层多无人机协调
链接：https://arxiv.org/abs/2605.09342

作者：Aneesh Calyam,Subrahmanya Chandra Bhamidipati,Zack Murry,Sharan Srinivas
备注：18 pages, 14 figures, 3 tables

【8】Evaluating Federated Learning approaches for mammography under breast density heterogeneity
标题：评估乳腺密度不均匀性下乳房X光摄影的联邦学习方法
链接：https://arxiv.org/abs/2605.09137

作者：Gonzalo Iñaki Quintana,Franco Martin Di Maria,Laurence Vancamberg

【9】MedFL-Stress: A Systematic Robustness Evaluation of Federated Brain Tumor Segmentation under Cross-Hospital MRI Appearance Shift
标题：MedFL-Stress：跨医院MRI外观变化下联邦脑肿瘤分割的系统稳健性评估
链接：https://arxiv.org/abs/2605.09025

作者：Kiran Naseer,Naveed Anwer Butt

【10】Shapley Regression for Rare Disease Diagnosis Support: a case study on APDS
标题：罕见疾病诊断支持的Shapley回归：AADS案例研究
链接：https://arxiv.org/abs/2605.08897

作者：Safa Alsaidi,Tomás Brogueira,Nizar Mahlaoui,Marc Vincent,Guilherme Pelegrina,Nicolas Garcelon,Adrien Coulet,Miguel Couceiro
备注：21 pages, 4 figures. Accepted to the AI and Health special track at IJCAI 2026; the first two named authors had equal contribution

【11】MicroDiffuse3D: A Foundation Model for 3D Microscopy Imaging Restoration
标题：Microdiffuse 3D：3D显微镜成像恢复的基础模型
链接：https://arxiv.org/abs/2605.08566

作者：Yongkang Li,Brian Wong,King Wai Chiu,Hanwen Xu,Tangqi Fang,Erin Dunnington,Dan Fu,Sheng Wang

【12】Resource-Aware Evolutionary Neural Architecture Search for Cardiac MRI Segmentation
标题：用于心脏MRI分割的资源感知进化神经架构搜索
链接：https://arxiv.org/abs/2605.08238

作者：Farhana Yasmin,Mahade Hasan,Haipeng Liu,Amjad Ali,Ghulam Muhammad,Yu Xue

【13】Attractor-Vascular Coupling Theory: Formal Grounding and Empirical Validation for AAMI-Standard Cuffless Blood Pressure Estimation from Smartphone Photoplethysmography
标题：吸引器-血管耦合理论：通过智能手机光电体积脉搏成像法估计AAMI标准无袖带血压的正式基础和经验验证
链接：https://arxiv.org/abs/2605.10871

作者：Timothy Oladunni,Farouk Ganiyu Adewumi

蒸馏|知识提取(9篇)

【1】Unmasking On-Policy Distillation: Where It Helps, Where It Hurts, and Why
标题：揭露政策蒸馏：它在哪里有帮助、在哪里有伤害以及为什么
链接：https://arxiv.org/abs/2605.10889

作者：Mohammadreza Armandpour,Fatih Ilhan,David Harrison,Ajay Jaiswal,Duc N. M Hoang,Fartash Faghri,Yizhe Zhang,Minsik Cho,Mehrdad Farajtabar

【2】Locking Pretrained Weights via Deep Low-Rank Residual Distillation
标题：通过深度低级剩余蒸馏锁定预训练权重
链接：https://arxiv.org/abs/2605.10777

作者：Keitaro Sakamoto,Pierre Ablin,Federico Danieli,Marco Cuturi

【3】Step Rejection Fine-Tuning: A Practical Distillation Recipe
标题：分步拒绝微调：实用的蒸馏配方
链接：https://arxiv.org/abs/2605.10674

作者：Igor Slinko,Ilia Zavidnyi,Egor Bogomolov,Yaroslav Zharov

【4】Identified-Set Geometry of Distributional Model Extraction under Top-$K$ Censored API Access
标题：Top-$K$审查API访问下分布模型提取的识别集几何
链接：https://arxiv.org/abs/2605.10407

作者：Wenhua Nie, ZiCheng Zhu, Jianan Wu, Binhan Luo, Haoran Zheng, Jyh-Shing Roger Jang

【5】TRACE: Distilling Where It Matters via Token-Routed Self On-Policy Alignment
标题：TRACE：通过令牌路由的自我政策调整来提炼重要的地方
链接：https://arxiv.org/abs/2605.10194

作者：Jiaxuan Wang, Xuan Ouyang, Zhiyu Chen, Yulan Hu, Zheng Pan, Xin Li, Lan-Zhe Guo
备注：work in progress

【6】GLiNER-Relex: A Unified Framework for Joint Named Entity Recognition and Relation Extraction
标题：GLiNER-Relex：联合命名实体识别和关系提取的统一框架
链接：https://arxiv.org/abs/2605.10108

作者：Ihor Stepanov, Oleksandr Lukashov, Mykhailo Shtopko, Vivek Kalyanarangan
备注：19 pages, 1 figure, 2 tables

【7】CoDistill-GRPO: A Co-Distillation Recipe for Efficient Group Relative Policy Optimization
标题：CoDistill-GRPO：高效集团相对政策优化的共蒸馏配方
链接：https://arxiv.org/abs/2605.08873

作者：Soo Min Kwon,Ziteng Sun,Ananda Theertha Suresh,Himanshu Jain,Sanjiv Kumar

【8】SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training
标题：SlimQwen：探索大型MoE模型预训练中的修剪和蒸馏
链接：https://arxiv.org/abs/2605.08738

作者：Shengkun Tang,Zekun Wang,Bo Zheng,Liangyu Wang,Rui Men,Siqi Zhang,Xiulong Yuan,Zihan Qiu,Zhiqiang Shen,Dayiheng Liu

【9】The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs
标题：近确定性结构性产出的政策上提炼中的推断悬崖
链接：https://arxiv.org/abs/2605.08737

作者：Xin Li,Hao Jiang,Annan Wang,Yichi Zhang,Chau Yuen

推荐(2篇)

【1】LoKA: Low-precision Kernel Applications for Recommendation Models At Scale
标题：LoKA：大规模推荐模型的低精度核心应用程序
链接：https://arxiv.org/abs/2605.10886

作者：Liang Luo,Yinbin Ma,Quanyu Zhu,Vasiliy Kuznetsov,Yuxin Chen,Jian Jiao,Jiecao Yu,Buyun Zhang,Tongyi Tang,Xiaohan Wei,Yanli Zhao,Zeliang Chen,Yuchen Hao,Venkatesh Ranganathan,Sandeep Parab,Yantao Yao,Maxim Naumov,Chunzhi Yang,Shen Li,Ellie Wen,Wenlin Chen,Santanu Kolay,Chunqiang Tang
备注：Accepted to ISCA'26

【2】Compressed Video Aggregator: Content-driven Module for Efficient Micro-Video Recommendation
标题：压缩视频聚合器：内容驱动模块，用于高效微视频推荐
链接：https://arxiv.org/abs/2605.08810

作者：Yang Xiao,Huiyuan Chen,Kaiyuan Deng,Chao Jiang,Zinan Ling,Ruimeng Ye,Xiaolong Ma,Bo Hui
备注：18 pages

聚类(1篇)

【1】PHIDA: Persistence-Guided Node-to-Cluster Mapping for Online Clustering
标题：PHIDA：用于在线集群的持久性引导的节点到集群映射
链接：https://arxiv.org/abs/2605.08673

作者：Naoki Masuyama,Yusuke Nojima,Stefan Wermter,Yuichiro Toda,Hisao Ishibuchi,Chu Kiong Loo
备注：This paper is currently under review

超分辨率|去噪|去模糊|去雾(2篇)

【1】PGID: Progressive Guided Inversion and Denoising for Robust Watermark Detection
标题：PGID：用于鲁棒水印检测的渐进引导反相和去噪
链接：https://arxiv.org/abs/2605.09319

作者：Minh Quoc Duong,Chun Tong Lei,Chun Pong Lau

【2】CASISR: Circular Arbitrary-Scale Image Super-Resolution
标题：CASSR：圆形辅助比例图像超分辨率
链接：https://arxiv.org/abs/2605.08173

作者：Honggui Li,Zhengyang Zhang,Dingtai Li,Sinan Chen,Nahid Md Lokman Hossain,Xinfeng Xu,Yinlu Qin,Ruobing Wang,Hantao Lu,Yuting Feng,Maria Trocan,Dimitri Galayko,Amara Amara,Mohamad Sawan

自动驾驶|车辆|车道检测等(4篇)

【1】Unlocking air traffic flow prediction through microscopic aircraft-state modeling
标题：通过微观飞行器状态建模解锁空中交通流量预测
链接：https://arxiv.org/abs/2605.10083

作者：Bin Wang, Anqi Liu, Jiangtao Zhao, Yanyong Huang, Peilan He, Guiyuan Jiang, Feng Hong, Yanwei Yu, Tianrui Li

【2】TSNN: A Non-parametric and Interpretable Framework for Traffic Time Series Forecasting
标题：TSNN：交通时间序列预测的非参数可解释框架
链接：https://arxiv.org/abs/2605.09208

作者：Bowen Liu,Haijian Lai,Chan-Tong Lam,Junhao Dong,Benjamin Ng,Wei Ke,Sio-Kei Im
备注：Accepted by IEEE Transactions on Knowledge and Data Engineering

【3】Practical Wi-Fi-based Motion Recognition Under Variable Traffic Patterns
标题：可变交通模式下实用的基于Wi-Fi的运动识别
链接：https://arxiv.org/abs/2605.08308

作者：Guolin Yin,Junqing Zhang,Guanxiong Shen,Simon L. Cotton
备注：17 Pages

【4】Efficient Prompt Learning for Traffic Forecasting
标题：交通预测的高效即时学习
链接：https://arxiv.org/abs/2605.08273

作者：Qianru Zhang,Xinyi Gao,Alexander Zhou,Reynold Cheng,Siu-Ming Yiu,Hongzhi Yin
备注：24 pages. This paper is accepted by VLDBJ

点云|SLAM|雷达|激光|深度RGBD相关(2篇)

【1】Sub-Footprint Effect Correction in FW-LiDAR Point Clouds via Intra-Footprint Target Unmixing
标题：通过足迹内目标分解来修正WF-LiDART点云中的子足迹效应
链接：https://arxiv.org/abs/2605.09845

作者：Zhen Xiao, Yanfeng Gu, Xian Li
备注：11 pages,7 figures

【2】A meshfree exterior calculus for generalizable and data-efficient learning of physics from point clouds
标题：无网格外部演算，用于从点云进行可概括且数据高效的物理学习
链接：https://arxiv.org/abs/2605.08436

作者：Benjamin D. Shaffer,Brooks Kinch,M. Ani Hsieh,Nathaniel Trask
备注：25 pages, 13 figures

联邦学习|隐私保护|加密(4篇)

【1】Optimizing Server Placement for Vertical Federated Learning in Dynamic Edge/Fog Networks
标题：优化动态边缘/雾网络中垂直联邦学习的服务器放置
链接：https://arxiv.org/abs/2605.09813

作者：Su Wang, Mung Chiang, H. Vincent Poor
备注：Under revision at IEEE/ACM transactions on networking

【2】Function-Space ADMM for Decentralized Federated Learning: A Control Theoretic Perspective
标题：用于分散式联邦学习的功能空间ADMM：控制理论的角度
链接：https://arxiv.org/abs/2605.09356

作者：Akihito Taya,Yuuki Nishiyama,Kaoru Sezaki
备注：(c) 2026 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

【3】FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning
标题：FedWSSam：缓解敏锐度联邦学习中的平面性不兼容性
链接：https://arxiv.org/abs/2605.09144

作者：Bingnan Xiao,Yuan Gao,Bingcong Li,Wei Ni,Xin Wang,Tony Q. S. Quek

【4】FedGMI: Generative Model-Driven Federated Learning for Probabilistic Mixture Inference
标题：FedGMI：用于概率混合推理的生成模型驱动联邦学习
链接：https://arxiv.org/abs/2605.08760

作者：Qijun Hou,Yuchen Shi,Pingyi Fan,Khaled B. Letaief

推理|分析|理解|解释(40篇)

【1】Variational Inference for Lévy Process-Driven SDEs via Neural Tilting
标题：通过神经倾斜对Lévy过程驱动的SDP进行变分推理
链接：https://arxiv.org/abs/2605.10934

作者：Yaman Kindap,Manfred Opper,Benjamin Dupuis,Umut Simsekli,Tolga Birdal
备注：The associated project page which contains the official implementation can be found in https://circle-group.github.io/research/NeuralTilting/

【2】PhyGround: Benchmarking Physical Reasoning in Generative World Models
标题：PhyGround：生成世界模型中的物理推理基准
链接：https://arxiv.org/abs/2605.10806

作者：Juyi Lin,Arash Akbari,Yumei He,Lin Zhao,Haichao Zhang,Arman Akbari,Xingchen Xu,Zoe Y. Lu,Enfu Nan,Hokin Deng,Edmund Yeh,Sarah Ostadabbas,Yun Fu,Jennifer Dy,Pu Zhao,Yanzhi Wang
备注：Preprint. 56 pages, 39 figures, 40 tables. Project page: https://phyground.github.io/

【3】Rebellious Student: Reversing Teacher Signals for Reasoning Exploration with Self-Distilled RLVR
标题：叛逆的学生：用自我提炼的WLVR逆转老师的推理探索信号
链接：https://arxiv.org/abs/2605.10781

作者：Jeonghye Kim,Jiwon Jeon,Dongsheng Li,Yuqing Yang

【4】Hierarchical Causal Abduction: A Foundation Framework for Explainable Model Predictive Control
标题：分层因果诱拐：可解释模型预测控制的基础框架
链接：https://arxiv.org/abs/2605.10624

作者：Ramesh Arvind Naagarajan,Zühal Wagner,Stefan Streif

【5】ConfoundingSHAP: Quantifying confounding strength in causal inference
标题：混淆SHAP：量化因果推理中的混淆强度
链接：https://arxiv.org/abs/2605.10533

作者：Marie Brockschmidt,Santo M. A. R. Thies,Maresa Schröder,Dennis Frauen,Valentyn Melnychuk,Maximilian Muschalik,Eyke Hüllermeier,Stefan Feuerriegel

【6】Causal Explanations from the Geometric Properties of ReLU Neural Networks
标题：从ReLU神经网络的几何性质进行因果推理
链接：https://arxiv.org/abs/2605.10396

作者：Hector Woods, Philippa Ryan, Rob Alexander
备注：7 pages, 0 figures, Accepted for presentation at the Yorkshire Innovation in Science and Engineering Conference

【7】Foundations of Reliable Inference: Reliability-Efficiency Co-Design
标题：可靠性推理的基础：可靠性-效率协同设计
链接：https://arxiv.org/abs/2605.10351

作者：Jiayi Huang
备注：PhD Thesis

【8】Qwen Goes Brrr: Off-the-Shelf RAG for Ukrainian Multi-Domain Document Understanding
标题：Qwen Goes Brrr：乌克兰多领域文档理解的现成RAG
链接：https://arxiv.org/abs/2605.10296

作者：Anton Bazdyrev, Ivan Bashtovyi, Ivan Havlytskyi, Oleksandr Kharytonov, Artur Khodakovskyi
备注：Accepted to The Fifth Ukrainian Natural Language Processing Conference (UNLP 2026)

【9】Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration
标题：打破奖励障碍：通过投机探索加速思想树推理
链接：https://arxiv.org/abs/2605.10195

作者：Shuzhang Zhong, Haochen Huang, Shengxuan Qiu, Pengfei Zuo, Runsheng Wang, Meng Li
备注：OSDI 2026

【10】APEX: Audio Prototype EXplanations for Classification Tasks
标题：APEX：分类任务的音频原型解释
链接：https://arxiv.org/abs/2605.10153

作者：Piotr Kawa, Kornel Howil, Piotr Borycki, Miłosz Adamczyk, Przemysław Spurek, Piotr Syga

【11】Fairness of Explanations in Artificial Intelligence (AI): A Unifying Framework, Axioms, and Future Direction toward Responsible AI
标题：人工智能（AI）中解释的公平性：统一框架、公理和负责任人工智能的未来方向
链接：https://arxiv.org/abs/2605.09852

作者：Gideon Popoola, John Sheppard
备注：53 pages, 1 figure

【12】Free Energy Manifold: Score-Based Inference for Hybrid Bayesian Networks
标题：自由能Manifold：混合Bayesian网络的基于分数的推理
链接：https://arxiv.org/abs/2605.09839

作者：Cheol Young Park, Shou Matsumoto

【13】FreeMOCA: Memory-Free Continual Learning for Malicious Code Analysis
标题：FreeMOCA：用于恶意代码分析的无内存持续学习
链接：https://arxiv.org/abs/2605.09664

作者：Zahra Asadi,Haeseung Jeon,Sohyun Han,Md Mahmuduzzaman Kamol,Se Eun Oh,Mohammad Saidur Rahman
备注：17 pages, 5 figures, 12 tables

【14】Hidden Error Awareness in Chain-of-Thought Reasoning: The Signal Is Diagnostic, Not Causal
标题：思想链推理中隐藏的错误意识：信号是诊断性的，而不是因果性的
链接：https://arxiv.org/abs/2605.09502

作者：Aojie Yuan,Zhiyuan Julian Su,Haiyue Zhang,Yi Nian,Yue Zhao
备注：10 pages, 5 figures, 10 tables.Mechanistic Interpretability @ ICML 2026

【15】PoHAR: Understanding Hyperlocal Human Activities with Pollution Sensor Networks
标题：PoHAR：利用污染传感器网络了解超本地人类活动
链接：https://arxiv.org/abs/2605.09434

作者：Prasenjit Karmakar,Karthik Reddy,Sandip Chakraborty
备注：8 pages, 8 figures, accepted to IEEE DCOSS-IoT 2026

【16】Split CNN Inference on Networked Microcontrollers
标题：CNN对网络微控制器的分离推理
链接：https://arxiv.org/abs/2605.09357

作者：Junyu Lu,Shashwath Suresh,Hao Liu,Qi Hong,Qing Wang
备注：10 pages

【17】The Trap of Trajectory: Towards Understanding and Mitigating Spurious Correlations in Agentic Memory
标题：轨迹陷阱：了解和减轻记忆中的虚假相关性
链接：https://arxiv.org/abs/2605.09330

作者：Luoxi Tang,Rupali Rajendra Vaje,Yuqiao Meng,Sakshi Sunil Narkar,Weicheng Ma,Zeyu Ding,Dazheng Zhang,Zhaohan Xi

【18】Memorize Theorems, Not Instances: Probing SFT Generalization through Mathematical Reasoning
标题：简化定理，而不是简化：通过数学推理探索SFT推广
链接：https://arxiv.org/abs/2605.09270

作者：Ruiying Peng,Mengyu Yang,Jing Lei,Xiaohui Li,Xueyu Wu,Xinlei Chen

【19】A Quantum Inspired Variational Kernel and Explainable AI Framework for Cross Region Solar and Wind Energy Forecasting
标题：跨区域太阳能和风能预测的量子启发变分核和可解释人工智能框架
链接：https://arxiv.org/abs/2605.09032

作者：Pavan Manjunath,Thomas Prufer

【20】PMCTS: Particle Monte Carlo Tree Search for Principled Parallelized Inference Time Scaling
标题：PMCTS：粒子蒙特卡罗树搜索原则平行化推理时间缩放
链接：https://arxiv.org/abs/2605.08982

作者：Yaniv Oren,Viliam Vadocz,Joery A. de Vries,Wendelin Böhmer,Matthijs T. J. Spaan,Hendrik Baier

【21】MicroFuse: Protein-to-Genome Expert Fusion for Microbial Operon Reasoning
标题：微生物学家：蛋白质到基因组专家融合，用于微生物操纵子推理
链接：https://arxiv.org/abs/2605.08815

作者：Seungik Cho

【22】Curvature-Aware Captioning:Leveraging Geodesic Attention for 3D Scene Understanding
标题：曲线感知字幕：利用测地注意力来理解3D场景
链接：https://arxiv.org/abs/2605.08808

作者：Ziyao He,Yingjie Liu,ZhangYangRui,Mingsong Chen,Xuan Tang,Xian Wei
备注：CVPR2026 Highlight!

【23】LAQuant: A Simple Overhead-free Large Reasoning Model Quantization by Layer-wise Lookahead Loss
标题：LAQuant：一个简单的无管理费用的大型推理模型，通过逐层前瞻损失进行量化
链接：https://arxiv.org/abs/2605.08755

作者：Euntae Choi,Sumin Song,Sungjoo Yoo

【24】Sketch-and-Verify: Structured Inference-Time Scaling via Program Sketching
标题：绘制并验证：通过程序绘制的结构化推断时间缩放
链接：https://arxiv.org/abs/2605.08658

作者：Shan Jiang,Zijian Yi,Chenguang Zhu

【25】Reasoning-Aware Training for Time Series Forecasting
标题：时间序列预测的推理感知训练
链接：https://arxiv.org/abs/2605.08625

作者：Md Atik Ahamed,Mihir Parmar,Palash Goyal,Chun-Liang Li,Qiang Cheng,Tomas Pfister,Jinsung Yoon

【26】When Independent Sampling Outperforms Agentic Reasoning
标题：当独立抽样优于抽象推理时
链接：https://arxiv.org/abs/2605.08478

作者：Yihe Dong,Boris Shigida

【27】Convergence Analysis of Newton's Method for Neural Networks in the Overparameterized Limit
标题：超参数化极限下神经网络牛顿方法的收敛性分析
链接：https://arxiv.org/abs/2605.08352

作者：Konstantin Riedl,Konstantinos Spiliopoulos,Justin Sirignano

【28】Private Vertical Federated Inference for Time-Series
标题：时间序列的私人垂直联合推理
链接：https://arxiv.org/abs/2605.08343

作者：Lucas Fenaux,Larris Xie,Aditya Bang,Alex Zhang,Kevin Wilson,Florian Kerschbaum

【29】Social Determinants of Health and Fentanyl Overdose Mortality Across US Counties: An XGBoost and SHAP Analysis Identifying Silent Risk Counties and Treatment Deserts
标题：美国各县健康和芬太尼过量死亡率的社会决定因素：XGBOP和SHAP分析，确定沉默风险县和治疗落后
链接：https://arxiv.org/abs/2605.08230

作者：Kabi Raj Tiruwa,Abhisan Ghimire,Anuj Kumar Shah
备注：21 pages, 7 figures, 4 tables

【30】A Simulated Federated Analysis of MS-Induced Brain Lesions
标题：MS诱导脑损伤的模拟联邦分析
链接：https://arxiv.org/abs/2605.08223

作者：Evelyn Trautmann,Joël Federer-Gsponer,Markus C. Elze,José-Tomás Prieto
备注：Accepted for publication at The 39th IEEE International Symposium on Computer-Based Medical Systems

【31】NoisyCoconut: Counterfactual Consensus via Latent Space Reasoning
标题：NoisyCoconut：通过潜在空间推理的反事实共识
链接：https://arxiv.org/abs/2605.08221

作者：Michael Jerge,David Evans

【32】Towards Universal Gene Regulatory Network Inference: Unlocking Generalizable Regulatory Knowledge in Single-cell Foundation Models
标题：走向通用基因调节网络推理：解锁单细胞基础模型中的可推广调节知识
链接：https://arxiv.org/abs/2605.08128

作者：Jiaxin Qi,Hang Li,Yan Cui,Yuhua Zheng,Jianqiang Huang
备注：Accepted to the 43rd International Conference on Machine Learning (ICML 2026)

【33】Statistical Inference and Quality Measures of KV Cache Quantisations Inspired by TurboQuant
标题：受TurboQuant启发的KV缓存量化的统计推断和质量衡量
链接：https://arxiv.org/abs/2605.08114

作者：Paolo D'Alberto
备注：23 pages, 7 Figures, multiple tables, the process is highly assisted by AI

【34】Amortizing Causal Sensitivity Analysis via Prior Data-Fitted Networks
标题：通过先验数据匹配网络进行摊销因果敏感性分析
链接：https://arxiv.org/abs/2605.10590

作者：Emil Javurek,Dennis Frauen,Marie Brockschmidt,Jonas Schweisthal,Stefan Feuerriegel

【35】Regret Analysis of Guided Diffusion for Black-Box Optimization over Structured Inputs
标题：结构化输入黑箱优化引导扩散的遗憾分析
链接：https://arxiv.org/abs/2605.10385

作者：Masaki Adachi,Anita Yang,Yakun Wang,Song Liu
备注：48 pages, 12 figures

【36】Scalable Gaussian process inference via neural feature maps
标题：通过神经特征图的可扩展高斯过程推断
链接：https://arxiv.org/abs/2605.10285

作者：Anthony Stephenson
备注：27 pages

【37】Supercharging Bayesian Inference with Reliable AI-Informed Priors
标题：利用可靠的人工智能先验数据增强Bayesian推理
链接：https://arxiv.org/abs/2605.09834

作者：Jongwoo Choi,Sean O'Hagan

【38】Survey-aware Machine Learning: A Guideline for Valid Population Health Inference based on Scoping Review
标题：调查感知机器学习：基于范围界定审查的有效人口健康推断指南
链接：https://arxiv.org/abs/2605.08963

作者：YongKyung Oh,Henry W. Zheng,Jeffrey Feng,Alex A. T. Bui

【39】Energy-based models for diagnostic reconstruction and analysis in a laboratory plasma device
标题：实验室等离子体设备中用于诊断重建和分析的基于能量的模型
链接：https://arxiv.org/abs/2605.08645

作者：Phil Travis,Troy Carter
备注：15 pages, 10 figures

【40】Active Multiple-Prediction-Powered Inference
标题：主动多重预测推理
链接：https://arxiv.org/abs/2605.08429

作者：Nicholas Brawand,Nima Leclerc,Anhthy Ngo,Matthew Peterson,Sriram Vishwanath,Laith Alhussein,Ben Wellner

检测相关(17篇)

【1】Conditional anomaly detection methods for patient-management alert systems
标题：患者管理警报系统的条件异常检测方法
链接：https://arxiv.org/abs/2605.10847

作者：Michal Valko,Gregory Cooper,Amy Seybert,Shyam Visweswaran,Melissa Saul,Miloš Hauskrecht
备注：Published at Workshop on Machine Learning in Health Care Applications ICML-2008 - MLHealth

【2】MARGIN: Margin-Aware Regularized Geometry for Imbalanced Vulnerability Detection
标题：MARGIN：用于不平衡漏洞检测的边缘感知规则化几何
链接：https://arxiv.org/abs/2605.10240

作者：Yuteng Zhang, Huifang Ma, Jiahui Wei, Qingqing Li, Yafei Yang
备注：12 pages.9 figures, 4 tables

【3】CrossVL: Complexity-Aware Feature Routing and Paired Curriculum for Cross-View Vision-Language Detection
标题：CrossVL：用于跨视图视觉语言检测的复杂性感知特征路由和配对课程
链接：https://arxiv.org/abs/2605.09802

作者：Zhipeng Liu,Chunbo Luo
备注：Accepted to CVPR 2026. Code available at https://github.com/1nyourlife/Crossvl_cvpr2026

【4】Multi-Tier Labeling and Physics-Informed Learning for Orbital Anomaly Detection at Scale
标题：用于大规模轨道异常检测的多层标签和物理知情学习
链接：https://arxiv.org/abs/2605.09790

作者：Yong Fu

【5】Learning Unified Representations of Normalcy for Time Series Anomaly Detection
标题：学习用于时间序列异常检测的正常性统一表示
链接：https://arxiv.org/abs/2605.09685

作者：Prithul Sarker,Sushmita Sarker,Nicholas G. Murray,Alireza Tavakkoli

【6】Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts
标题：微缺陷暴露宏造假：通过局部分布漂移检测人工智能生成的图像
链接：https://arxiv.org/abs/2605.09296

作者：Boxuan Zhang,Jianing Zhu,Qifan Wang,Jiang Liu,Ruixiang Tang
备注：41 pages, 10 figures

【7】Towards Trustworthy Audio Deepfake Detection: A Systematic Framework for Diagnosing and Mitigating Gender Bias
标题：迈向值得信赖的音频Deepfake检测：诊断和缓解性别偏见的系统框架
链接：https://arxiv.org/abs/2605.09087

作者：Aishwarya Fursule,Shruti Kshirsagar,Anderson R. Avila
备注：Submitted to SMC 2026 conference

【8】Diagnosing and Mitigating Domain Shift in Permission-Based Android Malware Detection
标题：诊断和缓解基于许可的Android恶意软件检测中的域转移
链接：https://arxiv.org/abs/2605.09028

作者：Md Rafid Islam

【9】Outlier detection for patient monitoring and alerting
标题：用于患者监控和警报的离群值检测
链接：https://arxiv.org/abs/2605.08955

作者：Miloš Hauskrecht,Iyad Batal,Michal Valko,Shyam Visweswaran,Gregory F. Cooper,Gilles Clermont
备注：Published at JBI 2013

【10】Max-pooling Network Revisited: Analyzing the Role of Semantic Probability in Multiple Instance Learning for Hallucination Detection
标题：重温最大池网络：分析语义概率在幻觉检测的多实例学习中的作用
链接：https://arxiv.org/abs/2605.08863

作者：Shota Fujikawa,Issei Sato

【11】Privacy-Aware Video Anomaly Detection through Orthogonal Subspace Projection
标题：通过垂直子空间投影检测隐私意识的视频异常
链接：https://arxiv.org/abs/2605.08651

作者：Lei Wang,Wenxiang Diao,Andrew Busch,Jun Zhou,Yongsheng Gao
备注：Accepted as a Spotlight paper at the Forty-Third International Conference on Machine Learning (ICML 2026)

【12】Beyond Toy Benchmarks: A Systematic Evaluation of OOD Detection Methods For Plant Pathology Classification
标题：超越玩具基准：用于植物病理分类的OOD检测方法的系统评估
链接：https://arxiv.org/abs/2605.08618

作者：Devesh Shah

【13】Post-hoc Selective Classification for Reliable Synthetic Image Detection
标题：用于可靠合成图像检测的事后选择性分类
链接：https://arxiv.org/abs/2605.08574

作者：Kaixiang Zheng,Jacob H. Seidman

【14】Smart Railway Obstruction Detection System using IoT and Computer Vision
标题：使用物联网和计算机视觉的智能铁路障碍物检测系统
链接：https://arxiv.org/abs/2605.08246

作者：Pravin Kumar,Mritunjay Shall Peelam,Ramakant Kumar,Sanjay Kumar,Vinay Chamola

【15】FQPDR: Federated Quantum Neural Network for Privacy-preserving Early Detection of Diabetic Retinopathy
标题：CLARPDR：联邦量子神经网络，用于保护糖尿病视网膜病变的隐私早期检测
链接：https://arxiv.org/abs/2605.08324

作者：Debashis De,Mahua Nandy Pal,Dipankar Hazra

【16】Decentralized Conformal Novelty Detection via Quantized Model Exchange
标题：通过量化模型交换进行分散共形新奇检测
链接：https://arxiv.org/abs/2605.08263

作者：Kyle Loh,Yu Xiang

【17】Towards Interpretable Damage Detection based on Aerodynamic Pressure Measurements
标题：基于空气动力压力测量的可解释损伤检测
链接：https://arxiv.org/abs/2605.08187

作者：Philip Franz,Max von Danwitz,Gregory Duthé,Alexander Popp,Eleni Chatzi
备注：28 pages, 30 figures

分类|识别(6篇)

【1】Revisiting Policy Gradients for Restricted Policy Classes: Escaping Myopic Local Optima with $k$-step Policy Gradients
标题：重新审视受限制政策类别的政策要素：用$k$-步骤政策要素摆脱短视的本地优化
链接：https://arxiv.org/abs/2605.10909

作者：Alex DeWeese,Guannan Qu

【2】DANCE: Detect and Classify Events in EEG
标题：舞蹈：检测和分类脑电中的事件
链接：https://arxiv.org/abs/2605.10688

作者：Jarod Lévy,Hubert Banville,Jérémy Rapin,Jean-Remi King,Thomas Moreau,Stéphane d'Ascoli
备注：29 pages

【3】Discriminative Span as a Predictor of Synthetic Data Utility via Classifier Reconstruction
标题：通过分类器重建将区分范围作为合成数据效用的预测器
链接：https://arxiv.org/abs/2605.09697

作者：Radhika Amar Desai,Modigari Narendra
备注：15 pages, 17 tables

【4】Causal Parametric Drift Simulation: A Digital Twin Framework for Classifier Robustness Evaluation
标题：因果参数漂移模拟：分类器稳健性评估的数字孪生框架
链接：https://arxiv.org/abs/2605.09663

作者：Julien Lafrance,Richard Khoury,Véronique Tremblay
备注：34 pages, 13 figures, 14 tables

【5】Classification-Head Bias in Class-Level Machine Unlearning: Diagnosis, Mitigation, and Evaluation
标题：分类级机器学习中的分类头偏差：诊断、缓解和评估
链接：https://arxiv.org/abs/2605.08730

作者：Weidong Zheng,Kongyang Chen,Yuanwei Guo,Yatie Xiao

【6】Modular Retrieval-Augmented Generalization for Human Action Recognition
标题：基于模块检索的增强泛化人体动作识别
链接：https://arxiv.org/abs/2605.08117

作者：Peng Liao,Shangsong Liang,Lin Chen,Peijia Zheng
备注：ICME 2026

表征(9篇)

【1】On periodic distributed representations using Fourier embeddings
标题：使用傅里叶嵌入的周期性分布表示
链接：https://arxiv.org/abs/2605.10818

作者：Jakeb Chouinard

【2】Elucidating Representation Degradation Problem in Diffusion Model Training
标题：扩散模型训练中的表示退化问题的解析
链接：https://arxiv.org/abs/2605.10790

作者：Zhipeng Yao,Dazhou Li,Zitong Zhang,Durude Mahee,Fan Zhu,Wenbin Zhang,Xinwei He,Yeying Jin,Rui Yu

【3】Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
标题：张量产品表示探针揭示线性方向上的共享结构
链接：https://arxiv.org/abs/2605.09967

作者：Andrew Lee, Fernanda Viégas, Martin Wattenberg

【4】SEMASIA: A Large-Scale Dataset of Semantically Structured Latent Representations
标题：SEMASIA：语义结构潜在表示的大规模数据集
链接：https://arxiv.org/abs/2605.09485

作者：Mario Edoardo Pandolfo,Enrico Grimaldi,Lorenzo Marinucci,Leonardo Di Nino,Simone Fiorellino,Sergio Barbarossa,Paolo Di Lorenzo

【5】Repeated-Token Counting Reveals a Dissociation Between Representations and Outputs
标题：重复代币计数揭示了表示和输出之间的脱节
链接：https://arxiv.org/abs/2605.09239

作者：Sohan Venkatesh
备注：Code is available at https://github.com/sohv/counting-failure

【6】Anchoring the Eigengap: Cross-Modal Spectral Stabilization for Sample-Efficient Representation Learning
标题：支撑特征间隙：跨模式谱稳定化以实现样本高效的表示学习
链接：https://arxiv.org/abs/2605.08764

作者：Nikhil J. Dhinagar,Vidhi Chhatbar,Chirag Jagad,Pavithra Senthilkumar,Sophia I. Thomopoulos,Mahir H. Khan,Sook-Lei Liew,the ENIGMA-Stroke Recovery Working Group,Paul M. Thompson

【7】MoMo: Conditioned Contrastive Representation Learning for Preference-Modulated Planning
标题：MoMo：偏好调制规划的条件对比表示学习
链接：https://arxiv.org/abs/2605.08512

作者：Yusuf Syed,Viraj Parimi,Brian Williams

【8】Echo-LoRA: Parameter-Efficient Fine-Tuning via Cross-Layer Representation Injection
标题：Echo-LoRA：通过跨层表示注入进行参数高效的微调
链接：https://arxiv.org/abs/2605.08177

作者：Yihang Peng,Peng Jin,Jie Gong,Xingyuan Chen,Lingjiao Xu,Ning Su,Yan Ran

【9】PoDAR: Power-Disentangled Audio Representation for Generative Modeling
标题：PoDART：用于生成式建模的功率分解音频表示
链接：https://arxiv.org/abs/2605.10084

作者：Alejandro Luebs,Mithilesh Vaidya,Ishaan Kumar,Sumukh Badam,Stephen W. Bailey,Matthew Bendel,Jose Sotelo,Xingzhe He
备注：9 pages, 3 figures

3D|3D重建等相关(2篇)

【1】Predicting 3D structure by latent posterior sampling
标题：通过潜在后验抽样预测3D结构
链接：https://arxiv.org/abs/2605.10830

作者：Azmi Haider,Dan Rosenbaum

【2】Beyond Spatial Compression: Interface-Centric Generative States for Open-World 3D Structure
标题：超越空间压缩：开放世界3D结构的以界面为中心的生成状态
链接：https://arxiv.org/abs/2605.10438

作者：Xiang Chen, Alexander Binder

编码器(4篇)

【1】Scalable Mamba-Based Message-Passing Neural Decoder for Error-Correcting Codes
标题：用于错误纠正码的可扩展基于Mamba的消息传递神经解码器
链接：https://arxiv.org/abs/2605.10681

作者：Rostislav Gusev,Nikita Aleksandrov,Artem Solomkin,Dmitry Artemasov
备注：This work has been submitted to the IEEE for possible publication

【2】The Geometric Wall: Manifold Structure Predicts Layerwise Sparse Autoencoder Scaling Laws
标题：几何墙：多管齐结构预测分层稀疏自动编码器缩放定律
链接：https://arxiv.org/abs/2605.09887

作者：Eslam Zaher, Maciej Trzaskowski, Quan Nguyen, Fred Roosta

【3】Single-Thread JPEG Decoder Benchmarks Mis-Evaluate ML Data Loaders
标题：单线程JPEG解码器基准错误评估ML数据加载器
链接：https://arxiv.org/abs/2605.08731

作者：Vladimir Iglovikov
备注：9 pages, 4 figures. Code and data: https://github.com/ternaus/imread_benchmark

【4】What Cohort INRs Encode and Where to Freeze Them
标题：哪些队列IPR编码以及在哪里冻结它们
链接：https://arxiv.org/abs/2605.08298

作者：Vasiliki Sideri-Lampretsa,Sophie Starck,Robbie Holland,Julian McGinnis,Daniel Rueckert
备注：9 content pages plus appendix

优化|敛散性(38篇)

【1】Optimal and Scalable MAPF via Multi-Marginal Optimal Transport and Schrödinger Bridges
标题：通过多边缘最优传输和薛定格桥实现最优且可扩展的MAPF
链接：https://arxiv.org/abs/2605.10917

作者：Usman A. Khan,Joseph W. Durham
备注：Accepted in ICML 2026 as a spotlight paper

【2】Muown: Row-Norm Control for Muon Optimization
标题：Muown：μ子优化的行规范控制
链接：https://arxiv.org/abs/2605.10797

作者：Kai Lion,Florian Hübler,Bingcong Li,Antonio Orvieto,Niao He

【3】MASS-DPO: Multi-negative Active Sample Selection for Direct Policy Optimization
标题：MASS-DPO：用于直接政策优化的多负主动样本选择
链接：https://arxiv.org/abs/2605.10784

作者：Rohan Surana,Xintong Li,Sheldon Yu,Yiran Jenny Shen,Chuhan Wang,Tong Yu,Prithviraj Ammanabrolu,Jingbo Shang,Julian McAuley,Junda Wu

【4】Compander-Aligned Query Geometry for Quantized Zeroth-Order Optimization
标题：用于量化零阶优化的缩扩器对齐查询几何
链接：https://arxiv.org/abs/2605.10673

作者：Yao Shu,Zilin Zhu

【5】Online Sharp-Calibrated Bayesian Optimization
标题：在线精确校准的Bayesian优化
链接：https://arxiv.org/abs/2605.10572

作者：Marshal Arijona Sinaga,Julien Martinelli,Teemu Turpeinen,Samuel Kaski

【6】BROS: Bias-Corrected Randomized Subspaces for Memory-Efficient Single-Loop Bilevel Optimization
标题：Bros：用于内存高效的单循环二层优化的偏置纠正随机子空间
链接：https://arxiv.org/abs/2605.10288

作者：Hengrui Zhang, Boao Kong, Engao Zhang, Kun Yuan

【7】Chebyshev Center-Based Direction Selection for Multi-Objective Optimization and Training PINNs
标题：基于切比雪夫中心的多目标优化和训练PINN方向选择
链接：https://arxiv.org/abs/2605.09975

作者：Hoyeol Yoon, Seoungbin Bae, Nam Ho-Nguyen, Dabeen Lee

【8】Bayesian Optimization with Structured Measurements: A Vector-Valued RKHS Framework
标题：具有结构化测量的Bayesian优化：一个Vector值RKHS框架
链接：https://arxiv.org/abs/2605.09775

作者：Wenbin Wang,Colin N. Jones

【9】Workspace Optimization: How to Train Your Agent
标题：工作空间优化：如何训练您的代理
链接：https://arxiv.org/abs/2605.09650

作者：Elad Sarafian,Gal Kaplun,Ron Banner,Daniel Soudry,Boris Ginsburg

【10】Learning-Augmented Scalable Linear Assignment Problem Optimization via Neural Dual Warm-Starts
标题：通过神经双温启动的学习增强可扩展线性指派问题优化
链接：https://arxiv.org/abs/2605.09382

作者：Ilay Yavlovich,Jad Agbaria,Muhamed Mhamed,Jose Yallouz,Nir Weinberger
备注：Accepted to ICML 2026. 20 pages, 13 figures

【11】Near-Optimal Last-Iterate Convergence for Zero-Sum Games with Bandit Feedback and Opponent Actions
标题：具有Bandit反馈和对手行动的零和博弈的近最优最后迭代收敛
链接：https://arxiv.org/abs/2605.09363

作者：Soumita Hait,Ping Li,Haipeng Luo,Mengxiao Zhang

【12】dFlowGRPO: Rate-Aware Policy Optimization for Discrete Flow Models
标题：dFlowGRPO：离散流模型的速率感知策略优化
链接：https://arxiv.org/abs/2605.09291

作者：Zhengyan Wan,Yidong Ouyang,Panwen Hu,Qiang Sun

【13】Intrinsic Muon: Spectral Optimization on Riemannian Matrix Manifolds
标题：本质μ子：Riemann矩阵上的谱优化
链接：https://arxiv.org/abs/2605.09238

作者：Yibang Li,Bihari Lal Pandey,Ravi Sah,Andi Han,Cyrus Mostajeran,Pratik Jawanpuria,Bamdev Mishra
备注：Code: https://github.com/1bang118/manifold-intrinsic-muon

【14】Cosine-Gated Adam-Decay: Drop-In Staleness-Aware Outer Optimization for Decoupled DiLoCo
标题：Cosine门控Adam-Decay：针对去耦合DiLoCo的临时停滞感知外部优化
链接：https://arxiv.org/abs/2605.09126

作者：Vatsal Shah,Jiahao Sun

【15】A Tale of Two Problems: Multi-Task Bilevel Learning Meets Equality Constrained Multi-Objective Optimization
标题：两个问题的故事：多任务二层学习满足平等约束的多目标优化
链接：https://arxiv.org/abs/2605.09094

作者：Zhiyao Zhang,Myeung Suk Oh,Zhen Qin,Jiaxiang Li,Xin Zhang,Jia Liu

【16】Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration
标题：利用乘方迭代的部分迭代加速零阶谱优化
链接：https://arxiv.org/abs/2605.09034

作者：Jiahe Chen,Ziye Ma

【17】MolWorld: Molecule World Models for Actionable Molecular Optimization
标题：MolWorld：可操作分子优化的分子世界模型
链接：https://arxiv.org/abs/2605.08954

作者：Yang Qiao,Bo Pan,Hao-Wei Pang,Peter Zhiping Zhang,Liying Zhang,Liang Zhao

【18】When and Why Grouping Attention Heads Accelerates Muon Optimization
标题：何时以及为何取消注意力头加速μ子优化
链接：https://arxiv.org/abs/2605.08933

作者：Hongtao Zhang,Wenjie Zhou,Wei Chen,Xueqi Cheng
备注：16 pages, 4 figures

【19】Discrete Flow Matching: Convergence Guarantees Under Minimal Assumptions
标题：离散流匹配：最低假设下的收敛保证
链接：https://arxiv.org/abs/2605.08882

作者：Le-Tuyet-Nhi Pham,Giovanni Conforti,Zhenjie Ren,Alain Durmus

【20】Geometrically Constrained Stenosis Editing in Coronary Angiography via Entropic Optimal Transport
标题：通过熵最优传输进行冠状动脉造影中的几何约束狭窄编辑
链接：https://arxiv.org/abs/2605.08851

作者：Jialin Li,Zhuo Zhang,Yue Cao,Guipeng Lan,Jiabao Wen,Shuai Xiao,Jiachen Yang
备注：Accepted to ICML 2026

【21】cuRegOT: A GPU-Accelerated Solver for Entropic-Regularized Optimal Transport
标题：cuRegOT：一个用于熵正规化最优传输的运算器
链接：https://arxiv.org/abs/2605.08793

作者：Yixuan Qiu

【22】Learning Polyhedral Conformal Sets for Robust Optimization
标题：鲁棒优化的多面体共形集学习
链接：https://arxiv.org/abs/2605.08506

作者：Shuyi Chen,Wenbin Zhou,Shixiang Zhu

【23】CDS4RAG: Cyclic Dual-Sequential Hyperparameter Optimization for RAG
标题：CDS 4RAG：RAG的循环双序列超参数优化
链接：https://arxiv.org/abs/2605.08333

作者：Pengzhou Chen,Tao Chen
备注：Accepted by main track at IJCAI 2026

【24】Reflective Prompted Policy Optimization: Trajectory-Grounded Revision and Salience Bias
标题：反思性预算政策优化：基于轨迹的修订和显着偏差
链接：https://arxiv.org/abs/2605.08315

作者：Rahaf Abu Hara,Vaibbhav Murarri,Claudio Zito

【25】Hierarchical Mixture-of-Experts with Two-Stage Optimization
标题：具有两阶段优化的分层专家混合
链接：https://arxiv.org/abs/2605.08292

作者：Gleb Molodtsov,Alexander Miasnikov,Aleksandr Beznosikov

【26】Toward Optimal Regret in Robust Pricing: Decoupling Corruption and Time
标题：走向稳健定价中的最佳遗憾：腐败与时间脱钩
链接：https://arxiv.org/abs/2605.08290

作者：Kalana Kalupahana,Francesco Emanuele Stradi,Matteo Castiglioni,Alberto Marchesi

【27】HTPO: Towards Exploration-Exploitation Balanced Policy Optimization via Hierarchical Token-level Objective Control
标题：HTPO：通过分层代币级目标控制实现探索-开发平衡政策优化
链接：https://arxiv.org/abs/2605.08283

作者：Xincheng Yao,Ruoqi Li,Cheng Chen,Daoxin Zhang,Yi Wu,Yao Hu,Chongyang Zhang
备注：29 pages

【28】Fixed-Point Neural Optimal Transport without Implicit Differentiation
标题：无隐式分化的定点神经最优传输
链接：https://arxiv.org/abs/2605.10792

作者：Yesom Park,Eric Gelphman,Stanley Osher,Samy Wu Fung
备注：37 pages, submitted to SIAM Journal on Mathematical Data Science (currently under review)

【29】On the global convergence of gradient descent for wide shallow models with bounded nonlinearities
标题：具有有界非线性的宽浅模型梯度下降的全局收敛性
链接：https://arxiv.org/abs/2605.10775

作者：Romain Petit,Clarice Poon,Gabriel Peyré

【30】Sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks
标题：宽网络中的尖锐特征学习转变和Bayes最优神经缩放定律
链接：https://arxiv.org/abs/2605.10395

作者：Minh-Toan Nguyen,Jean Barbier

【31】Quantitative Local Convergence of Mean-Field Stein Variational Gradient Flow
标题：平均场Stein变分梯度流的定量局部收敛
链接：https://arxiv.org/abs/2605.09456

作者：Lénaïc Chizat,Maria Colombo,Roberto Colombo,Xavier Fernández-Real

【32】Optimal Regret for Single Index Bandits
标题：单指数盗贼的最佳遗憾
链接：https://arxiv.org/abs/2605.09454

作者：Devdan Dey,Sujoy Bhore,Avishek Ghosh
备注：27 pages, 9 figures

【33】Mutual Information Optimal Density Control of Linear Systems and Generalized Schrödinger Bridges with Reference Refinement
标题：线性系统和带参考细化的广义Schrödinger桥的互信息最优密度控制
链接：https://arxiv.org/abs/2605.09349

作者：Shoju Enami,Kenji Kashima
备注：19 pages, 5 figures

【34】Optimality of Sub-network Laplace Approximations: New Results and Methods
标题：子网络拉普拉斯逼近的最佳性：新结果和方法
链接：https://arxiv.org/abs/2605.09075

作者：Swarnali Raha,Kshitij Khare,Rohit K Patra
备注：34 Pages, 8 Figures, 2 Tables

【35】Rennala MVR: Improved Time Complexity for Parallel Stochastic Optimization via Momentum-Based Variance Reduction
标题：Rennala MVR：通过基于动量的方差约减提高并行随机优化的时间复杂性
链接：https://arxiv.org/abs/2605.08871

作者：Zhirayr Tovmasyan,Artavazd Maranjyan,Peter Richtárik

【36】Tight Generalization Bounds for Noiseless Inverse Optimization
标题：无噪声逆优化的紧广义界
链接：https://arxiv.org/abs/2605.08866

作者：Pouria Fatemi,Hoomaan Maskan,Suvrit Sra,Peyman Mohajerin Esfahani
备注：29 pages, 2 figures

【37】Local LMO: Constrained Gradient Optimization via a Local Linear Minimization Oracle
标题：局部LMO：通过局部线性最小化Oracle进行约束梯度优化
链接：https://arxiv.org/abs/2605.08850

作者：Peter Richtárik,Kaja Gruntkowska,Hanmin Li
备注：71 pages, 8 figures

【38】Sinkhorn Treatment Effects: A Causal Optimal Transport Measure
标题：辛克霍恩治疗效果：因果最佳运输措施
链接：https://arxiv.org/abs/2605.08485

作者：Medha Agarwal,Alex Luedtke
备注：55 pages, 6 figures

预测|估计(29篇)

【1】Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories
标题：Clin-JEPA：用于对EHR患者轨迹进行联合嵌入预测预训练的多阶段联合训练框架
链接：https://arxiv.org/abs/2605.10840

作者：Yixuan Yang,Mehak Arora,Ryan Zhang,Baraa Abed,Junseob Kim,Tilendra Choudhary,Md Hassanuzzaman,Kevin Zhu,Ayman Ali,Chengkun Yang,Alasdair Edward Gent,Victor Moas,Rishikesan Kamaleswaran
备注：17 pages, 4 figures, 8 tables. Code: https://github.com/YeungYathin/Clin-JEPA

【2】Benchmarking Sensor-Fault Robustness in Forecasting
标题：基准传感器-预测中的故障鲁棒性
链接：https://arxiv.org/abs/2605.10822

作者：Alexander Windmann,Philipp Wittenberg,Gianluca Manca,Marcel Dix,Jens U. Brandt,Oliver Niggemann

【3】A Spectral Framework for Closed-Form Relative Density Estimation
标题：封闭相对密度估计的谱框架
链接：https://arxiv.org/abs/2605.10668

作者：Francis Bach

【4】AxiomOcean: Forecasting the Three-Dimensional Structure of the Upper Ocean
标题：公理海洋：预测上层海洋的三维结构
链接：https://arxiv.org/abs/2605.10455

作者：Sensen Wu, Yifan Chen, Guantao Pu, Xiaoyao Sun, Yijun Chen, Jin Qi, Ming Kong, Keyi Yang, Lichen Xu, Wenguan Wang, Xiaofeng Li, Zhenhong Du

【5】Real vs. Semi-Simulated: Rethinking Evaluation for Treatment Effect Estimation
标题：真实与半模拟：重新思考治疗效果估计的评估
链接：https://arxiv.org/abs/2605.10430

作者：George Panagopoulos

【6】Set Prediction for Next-Day Active Fire Forecasting
标题：集合预测法在次日火灾预报中的应用
链接：https://arxiv.org/abs/2605.10298

作者：Yuchen Bai, Georgios Athanasiou, Xin Yu, Diogenis Antonopoulos, Ioannis Papoutsis, Stijn Hantson, Nuno Carvalhais

【7】Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators
标题：通过潜在结构化光谱相关器进行稳定的长期偏东方程预测
链接：https://arxiv.org/abs/2605.10154

作者：Xiaoxiao Lu, Ye Yuan, Jiahao Shi

【8】PixelFlowCast: Latent-Free Precipitation Nowcasting via Pixel Mean Flows
标题：PixelFlowCast：通过像素平均流量进行无潜伏降水实时预报
链接：https://arxiv.org/abs/2605.10046

作者：Yufeng Zhu, Chunlei Shi, Yongchao Feng, Dan Niu
备注：26 pages, 7 figures

【9】Nectar: Neural Estimation of Cached-Token Attention via Regression
标题：Nectar：通过回归对缓存代币注意力的神经估计
链接：https://arxiv.org/abs/2605.09778

作者：João Monteiro,Michal Klein,Pierre Ablin,Marco Cuturi

【10】Adversary-Robust Learning from Fully Asynchronous Directional Derivative Estimates
标题：来自完全同步方向求导估计的对抗鲁棒学习
链接：https://arxiv.org/abs/2605.09337

作者：Anik Kumar Paul,Nibedita Roy,Nagesh Talagani,Swetha Ganesh,Gugan Thoppe,Alexandre Reiffers-Masson

【11】Prediction Bottlenecks Don't Discover Causal Structure (But Here's What They Actually Do)
标题：预测瓶颈不会发现因果结构（但这是他们实际做的）
链接：https://arxiv.org/abs/2605.09169

作者：Ankit Hemant Lade,Sai Krishna Jasti,Indar Kumar,Aman Chadha
备注：6 pages, 3 tables. Code: https://github.com/ankitlade12/ssm-causal

【12】Predicting Large Model Test Losses with a Noisy Quadratic System
标题：用带噪声的二次系统预测大模型试验损失
链接：https://arxiv.org/abs/2605.09154

作者：Chuning Li,Chris J. Maddison
备注：ICML 2026

【13】Predicting Plasticity in Deep Continual Learning: A Theoretical Perspective
标题：预测深度持续学习中的可塑性：一个理论视角
链接：https://arxiv.org/abs/2605.09044

作者：Jiuqi Wang,Jayanth Srinivasa,Claire Chen,Shuze Daniel Liu,Ali Payani,Shangtong Zhang
备注：21 pages, 4 figures, 2 tables

【14】Learning predictive models for combinations of heterogeneous proteomic data sources
标题：学习用于异类蛋白质组数据源组合的预测模型
链接：https://arxiv.org/abs/2605.08958

作者：Michal Valko,Richard Pelikan,Miloš Hauskrecht
备注：Published at in AMIA Summit on Translational Bioinformatics (STB 2008

【15】PnP-Corrector: A Universal Correction Framework for Coupled Spatiotemporal Forecasting
标题：PnP修正器：一种用于耦合时空预测的通用修正框架
链接：https://arxiv.org/abs/2605.08935

作者：Hao Wu,Fan Xu,Yuxu Lu,Penghao Zhao,Fan Zhang,Hao Jia,Yuxuan Liang,Ruijian Gou,Qingsong Wen,Xian Wu,Xiaomeng Huang,Yuan Gao

【16】Higher-Order Equilibrium Tracking for EM-Compressible Online Estimation
标题：EM可压缩在线估计的高级均衡跟踪
链接：https://arxiv.org/abs/2605.08864

作者：ZhiMing Li,Yue Song
备注：41 pages, 6 figures

【17】RareCP: Regime-Aware Retrieval for Efficient Conformal Prediction
标题：RareCP：用于高效保形预测的区域感知检索
链接：https://arxiv.org/abs/2605.08857

作者：Manuel Heurich,Maximilian Granz,Tim Landgraf

【18】Optimised Support Vector Regression for California Housing Price Prediction: The Critical Role of Feature Engineering and Hyperparameter Tuning
标题：加州房价预测的优化支持载体回归：特征工程和超参数调整的关键作用
链接：https://arxiv.org/abs/2605.08660

作者：Emmanuel Adutwum
备注：25 pages, 13 figures, 10 tables

【19】A Deep Risk Estimator for Known Operator Learning
标题：已知操作员学习的深度风险估计
链接：https://arxiv.org/abs/2605.08517

作者：Andreas Maier,Md Hasan,Paulina Conrad,Paula Andrea Perez-Toro
备注：In Review

【20】GNN for Structural Displacement Prediction
标题：GNN用于结构位移预测
链接：https://arxiv.org/abs/2605.08303

作者：Hung-Fu Chang,Tzu-Kang Lin,Yung-Li Cheng
备注：12 pages

【21】What If We Let Forecasting Forget? A Sparse Bottleneck for Cross-Variable Dependencies
标题：如果我们让预测忘记怎么办？跨变量附属关系的稀疏瓶颈
链接：https://arxiv.org/abs/2605.08289

作者：Fan Zhang,Shiming Fan,Hua Wang

【22】Retrieval Mechanisms Surpass Long-Context Scaling in Time Series Forecasting
标题：时间序列预测中的检索机制超越了长上下文缩放
链接：https://arxiv.org/abs/2605.08217

作者：Rishi Ahuja,Kumar Prateek,Simranjit Singh,Vijay Kumar

【23】Geometry-free prediction of inertial lift forces in microfluidic devices using deep learning
标题：使用深度学习对微流体设备中的惯性提升力进行无几何学预测
链接：https://arxiv.org/abs/2605.08109

作者：Jesse Ward-Bond,Ali Mashadian,Timothy C. Y. Chan,Edmond W. K. Young

【24】Fast Training of Mixture-of-Experts for Time Series Forecasting via Expert Loss Integration
标题：通过专家损失集成快速训练用于时间序列预测的混合专家
链接：https://arxiv.org/abs/2605.10330

作者：Btissame El Mahtout,Florian Ziel

【25】A Market-Rule-Informed Neural Network for Efficient Imbalance Electricity Price Forecasting
标题：基于市场规则的神经网络高效不平衡电价预测
链接：https://arxiv.org/abs/2605.09061

作者：Runyao Yu,Julia Lin,Derek W. Bunn,Jochen Stiasny,Wentao Wang,Yujie Chen,Tara Esterl,Peter Palensky,Jochen L. Cremer
备注：10 pages, 3 figures, 3 tables

【26】Beyond the Black Box: An Interpretable Machine Learning Framework for Predicting Electronic Structure Microdescriptors and Structure-Performance Relationships in Fe-based Catalytic Systems
标题：超越黑匣子：用于预测铁基催化体系电子结构微描述符和结构-性能关系的可解释机器学习框架
链接：https://arxiv.org/abs/2605.08994

作者：Oyinkansola Romiluyi
备注：27 pages, 10 figures

【27】CONTRA: Conformal Prediction Region via Normalizing Flow Transformation
标题：CONTRA：通过标准化流变换的保形预测区域
链接：https://arxiv.org/abs/2605.08561

作者：Zhenhan Fang,Aixin Tan,Jian Huang
备注：18 pages, 7 figures and 5 tables

【28】Neural Posterior Estimation of Terrain Parameters from Radar Sounder Data
标题：雷达探测器数据地形参数的神经后验估计
链接：https://arxiv.org/abs/2605.08179

作者：Jordy Dal Corso,Annalena Kofler,Marco Cortellazzi,Lorenzo Bruzzone,Bernhard Schölkopf
备注：5 pages, 3 figures; accepted at IGARSS 2026, 9 - 14 August 2026, Washington D.C., USA

【29】Forecasting Source Stability in Scientific Experiments using Temporal Learning Models: A Case Study from Tritium Monitoring
标题：使用时间学习模型预测科学实验中的源稳定性：来自钍监测的案例研究
链接：https://arxiv.org/abs/2605.08140

作者：Nicholas Tan Jerome,Nadia Aouadi,Christoph Koehler,Suren Chilingaryan,Andreas Kopmann

其他神经网络|深度学习|模型|建模(99篇)

【1】DataMaster: Towards Autonomous Data Engineering for Machine Learning
标题：DataMaster：迈向机器学习的自主数据工程
链接：https://arxiv.org/abs/2605.10906

作者：Yaxin Du,Xiyuan Yang,Zhifan Zhou,Wanxu Liu,Zixing Lei,Zimeng Chen,Fenyi Liu,Haotian Wu,Yuzhu Cai,Zexi Liu,Xinyu Zhu,WenHao Wang,Linfeng Zhang,Chen Qian,Siheng Chen

【2】BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data
标题：BEACON：用于从游戏数据中学习行为指纹的多模式数据集
链接：https://arxiv.org/abs/2605.10867

作者：Ishpuneet Singh,Gursmeep Kaur,Uday Pratap Singh Atwal,Guramrit Singh,Gurjot Singh,Maninder Singh

【3】Reinforce Adjoint Matching: Scaling RL Post-Training of Diffusion and Flow-Matching Models
标题：加强伴随匹配：扩展扩散和流匹配模型的RL训练后
链接：https://arxiv.org/abs/2605.10759

作者：Andreas Bergmeister,Stefanie Jegelka,Nikolas Nüsken,Carles Domingo-Enrich,Jakiw Pidstrigach

【4】Kernel-Gradient Drifting Models
标题：核梯度漂移模型
链接：https://arxiv.org/abs/2605.10727

作者：Maria Esteban-Casadevall,Jorge Carrasco-Pollo,Max Welling,Jan-Willem van de Meent,Erik J. Bekkers,Floor Eijkelboom

【5】Heteroscedastic Diffusion for Multi-Agent Trajectory Modeling
标题：多主体轨迹建模的异方差扩散
链接：https://arxiv.org/abs/2605.10717

作者：Guillem Capellera,Antonio Rubio,Luis Ferraz,Antonio Agudo
备注：Accepted to IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). Extended version of arXiv:2503.18589 (CVPR 2025)

【6】A Recursive Decomposition Framework for Causal Structure Learning in the Presence of Latent Variables
标题：存在潜在变量时因果结构学习的递进分解框架
链接：https://arxiv.org/abs/2605.10651

作者：Zheng Li,Feng Xie,Shenglan Nie,Xichen Guo,Ruxin Wang,Hao Zhang

【7】A Random-Matrix Criterion for Initializing Gated Recurrent Neural Networks
标题：门控回归神经网络初始化的随机矩阵准则
链接：https://arxiv.org/abs/2605.10650

作者：Tommaso Fioratti,Riccardo Marcaccioli,Francesco Casola
备注：10 pages, 5 figures, 2 appendices

【8】Hierarchical End-to-End Taylor Bounds for Complete Neural Network Verification
标题：用于完整神经网络验证的分层端到端泰勒界
链接：https://arxiv.org/abs/2605.10621

作者：Taha Entesari,Mahyar Fazlyab

【9】MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Image
标题：MulTaBench：通过文本和图像对多模式制表学习进行基准测试
链接：https://arxiv.org/abs/2605.10616

作者：Alan Arazi,Eilam Shapira,Shoham Grunblat,Mor Ventura,Elad Hoffer,Gioia Blayer,David Holzmüller,Lennart Purucker,Gaël Varoquaux,Frank Hutter,Roi Reichart

【10】PhysEDA: Physics-Aware Learning Framework for Efficient EDA With Manhattan Distance Decay
标题：PhysEDA：一个物理感知的学习框架，用于具有曼哈顿距离衰减的高效EDA
链接：https://arxiv.org/abs/2605.10547

作者：Zetao Yang
备注：9 pages, 4 figures, plus appendix. Code and data to be released upon publication

【11】Formally Verifying Analog Neural Networks Under Process Variations Using Polynomial Zonotopes
标题：使用多项区位在工艺变化下形式化模拟神经网络
链接：https://arxiv.org/abs/2605.10474

作者：Yasmine Abu-Haeyeh, Tobias Ladner, Matthias Althoff, Lars Hedrich

【12】Can Muon Fine-tune Adam-Pretrained Models?
标题：Muon可以微调亚当预训练模型吗？
链接：https://arxiv.org/abs/2605.10468

作者：Xingyu Qu, Peigeng Huang, Samuel Horvath

【13】QT-Net: Rethinking Evaluation of AI Models in Atomic Chemical Space
标题：QT-Net：原子化学空间人工智能模型的重新思考评估
链接：https://arxiv.org/abs/2605.10458

作者：Pablo Martínez Crespo, Stefano Ribes, Martin Rahm, Richard Beckmann, Robert S. Jordan, Marisa Gliege, Santiago Miret, Vijay Kris Narasimhan, Rocío Mercado

【14】The Polynomial Counting Capabilities of Message Passing Neural Networks
标题：消息传递神经网络的多项计数能力
链接：https://arxiv.org/abs/2605.10393

作者：Marco Sälzer, Pascal Bergsträßer, Anthony W. Lin

【15】Sample-Mean Anchored Thompson Sampling for Offline-to-Online Learning with Distribution Shift
标题：样本均值锚定Thompson抽样，用于具有分布转移的线下到在线学习
链接：https://arxiv.org/abs/2605.10289

作者：Bochao Li, Yao Fu, Wei Chen, Fang Kong

【16】Generalization Error Bounds for Picard-Type Operator Learning in Nonlinear Parabolic PDEs
标题：非线性方程组中Picard型运算符学习的广义误差界
链接：https://arxiv.org/abs/2605.10277

作者：Koichi Taniguchi, Sho Sonoda
备注：39 pages

【17】Empty SPACE: Cross-Attention Sparsity for Concept Erasure in Diffusion Models
标题：空白空间：扩散模型中概念擦除的交叉注意稀疏性
链接：https://arxiv.org/abs/2605.10198

作者：Nicola Novello, Andrea M. Tonello

【18】jNO: A JAX Library for Neural Operator and Foundation Model Training
标题：jNO：用于神经操作员和基础模型训练的JAX库
链接：https://arxiv.org/abs/2605.10159

作者：Leon Armbruster, Rathan Ramesh, Georg Kruse, Christopher Straub

【19】Learning to Sparsify Stochastic Linear Bandits
标题：学习稀疏随机线性盗贼
链接：https://arxiv.org/abs/2605.10151

作者：Zhengmiao Wang, Ming Chi, Zhi-Wei Liu, Lintao Ye, Carla Fabiana Chiasserini
备注：Include all the omitted details and proofs from the conference paper accepted to IJCAI 2026

【20】Per-Loss Adapters for Gradient Conflict in Physics-Informed Neural Networks
标题：物理信息神经网络中梯度冲突的按损失适配器
链接：https://arxiv.org/abs/2605.10136

作者：Bum Jun Kim, Gnankan Landry Regis N'guessan
备注：49 pages, 10 figures

【21】Explainability of Recurrent Neural Networks for Enhancing P300-based Brain-Computer Interfaces
标题：用于增强基于P300的脑机接口的回归神经网络的可解释性
链接：https://arxiv.org/abs/2605.10121

作者：Christian Oliva, Vinicio Changoluisa, Francisco B Rodríguez, Luis F Lago-Fernández

【22】CFSPMNet: Cross-subject Fourier-guided Spatial-Patch Mamba Network for EEG Motor Imagery Decoding in Stroke Patients
标题：CFSPMNet：用于中风患者脑电运动图像解码的跨学科傅里叶引导空间补丁Mamba网络
链接：https://arxiv.org/abs/2605.10111

作者：Xiangkai Wang, Yun Zhao, Dongyi He, Qingling Xia, Gen Li, Xinlai Xing, Yuchi Pan, Bin Jiang

【23】Rethinking Loss Reweighting for Imbalance Learning as an Inverse Problem: A Neural Collapse Point of View
标题：重新思考不平衡学习的损失重权作为逆问题：神经崩溃的观点
链接：https://arxiv.org/abs/2605.10047

作者：Jinping Wang, Zixin Tong, Zhiwu Xie, Zhiqiang Gao
备注：Accepted by ICML2026

【24】The two clocks and the innovation window: When and how generative models learn rules
标题：两个时钟和创新窗口：生成模型何时以及如何学习规则
链接：https://arxiv.org/abs/2605.10019

作者：Binxu Wang, Emma Lucia Byrnes Finn, Bingbin Liu
备注：48 pages, 28 figures. Earlier versions are presented in NeurIPS2025 SPIGM workshop as oral presentation this https URL

【25】Lakestream: A Consistent and Brokerless Data Plane for Large Foundation Model Training
标题：Lakestream：用于大型基础模型训练的一致且无经纪人的数据平面
链接：https://arxiv.org/abs/2605.09994

作者：Ting Sun, Junjie Zhang, Xiao Yan, Songxin Zhang, Zhuoyang Song, Jingyi Xi, Zunyao Mao, Bingyi Jing, Jiaxing Zhang, Zejian Xie

【26】Attention Drift: What Autoregressive Speculative Decoding Models Learn
标题：注意力漂移：自回归投机解码模型学到了什么
链接：https://arxiv.org/abs/2605.09992

作者：Doğaç Eldenk, Payal Mohapatra, Yigitcan Comlek, Kaan Oktay, Hongyang Zhang, Stephen Xia

【27】Prospective Compression in Human Abstraction Learning
标题：人类抽象学习中的前瞻性压缩
链接：https://arxiv.org/abs/2605.09985

作者：Leonardo Hernandez Cano, Ivan Zareski, Luisa El Amouri, Pinzhe Zhao, Max Mascini, Emanuele Sansone, Yewen Pu, Bonan Zhao, Marta Kryven
备注：under review at neurips 2026

【28】From Syntax to Semantics: Unveiling the Emergence of Chirality in SMILES Translation Models
标题：从字形到语义学：揭开SMILES翻译模型中手握性的出现
链接：https://arxiv.org/abs/2605.09949

作者：Zehao Li,Yasuhiro Yoshikai,Shumpei Nemoto,Hiroyuki Kusuhara,Tadahaya Mizuno

【29】Deep Learning under Fractional-Order Differential Privacy
标题：分数阶差异隐私下的深度学习
链接：https://arxiv.org/abs/2605.09890

作者：Mohammad Partohaghighi,Roummel Marcia

【30】Flag Varieties: A Geometric Framework for Deep Network Alignment
标题：旗帜品种：深度网络对齐的几何框架
链接：https://arxiv.org/abs/2605.09861

作者：Jingchuan Xiao, Xinyi Sui, Cihan Ruan

【31】Learning from Acceptance: Cumulative Regret in the Game of Coding
标题：从接受中学习：编码游戏中累积的遗憾
链接：https://arxiv.org/abs/2605.09754

作者：Hanzaleh Akbari Nodehi,Parsa Moradi,Mohammad Ali Maddah-Ali

【32】TIDES: Implicit Time-Awareness in Selective State Space Models
标题：潮汐：选择性状态空间模型中的隐性时间意识
链接：https://arxiv.org/abs/2605.09742

作者：Taylan Soydan,Miguel A. Bessa,Dirk Mohr,Rui Barreira
备注：Preprint submitted for peer-review

【33】Model Capacity Determines Grokking through Competing Memorisation and Generalisation Speeds
标题：模型容量通过竞争的小型化和通用化速度决定Groking
链接：https://arxiv.org/abs/2605.09724

作者：Yiding Song,Hanming Ye
备注：23 pages, 10 figures, 12 tables

【34】Do multimodal models imagine electric sheep?
标题：多模式模型会想象电动羊吗？
链接：https://arxiv.org/abs/2605.09693

作者：Santhosh Kumar Ramakrishnan,Carl Vondrick,Raja Giryes,Philipp Krähenbühl,Vladlen Koltun

【35】Minimal Filling Architectures of Polynomial Neural Networks: Counterexamples, Frontier Search, and Defects
标题：多项神经网络的最小填充架构：反例、前沿搜索和缺陷
链接：https://arxiv.org/abs/2605.09609

作者：Kevin Dao,Jose Israel Rodriguez

【36】Online Set Learning from Precision and Recall Feedback
标题：从精确度和召回反馈中在线集学习
链接：https://arxiv.org/abs/2605.09565

作者：Lee Cohen,Yishay Mansour,Shay Moran,Han Shao

【37】Doubly Robust Proxy Causal Learning with Neural Mean Embeddings
标题：神经均值嵌入的双重鲁棒代理因果学习
链接：https://arxiv.org/abs/2605.09514

作者：Bariscan Bozkurt,Alexandre Galashov,Dimitri Meunier,Zikai Shen,Arthur Gretton,Houssam Zenati

【38】Position: AI Security Policy Should Target Systems, Not Models
标题：立场：人工智能安全政策应针对系统，而不是模型
链接：https://arxiv.org/abs/2605.09504

作者：Michael A. Riegler,Inga Strümke

【39】Kintsugi: Learning Policies by Repairing Executable Knowledge Bases
标题：近杉：通过修复可执行知识库来学习政策
链接：https://arxiv.org/abs/2605.09487

作者：Teng Cao,Yu Deng,Hikaru Shindo,Quentin Delfosse,Lanxi Wen,Suli Wang,Jannis Blüml,Christopher Tauchmann,Kristian Kersting

【40】Learning to Bid with Unknown Private Values in Budget-Constrained First-Price Auctions
标题：学会在预算限制的第一价格拍卖中以未知的私人价值竞标
链接：https://arxiv.org/abs/2605.09448

作者：Zihao Hu,Yuxiao Wen,Yuan Yao,Jiheng Zhang,Zhengyuan Zhou

【41】Tabular Foundation Model for Generative Modelling
标题：生成建模的表格基础模型
链接：https://arxiv.org/abs/2605.09424

作者：Xiangjian Jiang,Mingxuan Liu,Nikola Simidjievski,Tassilo Klein,Mateja Jamnik

【42】Improving Generalization by Permutation Routing Across Model Copies
标题：通过模型副本之间的排列路由来改进概括
链接：https://arxiv.org/abs/2605.09256

作者：Shuhei Kashiwamura,Timothee Leleu

【43】Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models
标题：Sub-JEPA：稳定端到端世界模型的子空间高斯正规化
链接：https://arxiv.org/abs/2605.09241

作者：Kai Zhao,Dongliang Nie,Yuchen Lin,Zhehan Luo,Yixiao Gu,Deng-Ping Fan,Dan Zeng
备注：https://github.com/intcomp/Sub-JEPA

【44】On Variance Reduction in Learning Mean Flows
标题：学习平均流中的方差约减
链接：https://arxiv.org/abs/2605.09235

作者：Juanwu Lu,Ziran Wang
备注：25 pages, 7 figures, 6 tables

【45】Privacy-Preserving Distributed Learning in IoT Systems: A Unified Threat Model and Evaluation Framework
标题：物联网系统中保护隐私的分布式学习：统一的威胁模型和评估框架
链接：https://arxiv.org/abs/2605.09232

作者：John Cartmell,Alexander Williams
备注：14 pages, 6 figures

【46】Learning the Preferences of a Learning Agent
标题：学习学习代理的首选项
链接：https://arxiv.org/abs/2605.09217

作者：Karim Abdel Sadek,Mark Bedaywi,Rhys Gould,Stuart Russell
备注：Published at ICLR 2026, Workshop on Multi-Agent Learning and Its Opportunities in the Era of Generative AI. 9 pages main text

【47】Learning When to Stop: Selective Imitation Learning Under Arbitrary Dynamics Shift
标题：学习何时停止：任意动态变化下的选择性模仿学习
链接：https://arxiv.org/abs/2605.09183

作者：Surbhi Goel,Jonathan Pei,James Wang

【48】Objective-Specific Privileged Bases via Full-Prefix Matryoshka Learning
标题：通过全前置Matryoshka学习的特定对象的加密基础
链接：https://arxiv.org/abs/2605.09160

作者：Arghamitra Talukder,Philippe Chlenski,Itsik Pe'er

【49】AlphaExploitem: Going Beyond the Nash Equilibrium in Poker by Learning to Exploit Suboptimal Play
标题：AlphaExplotify：通过学习利用次优玩法来超越扑克中的纳什均衡
链接：https://arxiv.org/abs/2605.09150

作者：Vlad Murgoci,Matthijs Spaan,Yaniv Oren

【50】Contextual Plackett-Luce: An Efficient Neural Model for Probabilistic Sequence Selection under Ambiguity
标题：上下文Plackett-Luce：模糊情况下概率序列选择的有效神经模型
链接：https://arxiv.org/abs/2605.09112

作者：Noam Mizrachi,Nadav Har-Tuv,Shai Shalev-Shwartz
备注：22 pages, 5 figures

【51】Bridging Spectral Operator Learning and U-Net Hierarchies: SpectraNet for Stable Autoregressive PDE Surrogates
标题：桥连谱运算符学习和U-Net层次结构：SpectraNet用于稳定自回归偏出方程
链接：https://arxiv.org/abs/2605.09096

作者：Enrique Hernández Noguera,Md Meftahul Ferdaus,Elias Ioup,Mahdi Abdelguerfi,Julian Simeonov
备注：29 pages, 9 figures. Code: https://github.com/Enrikkk/spectranet

【52】FactoryNet: A Large-Scale Dataset toward Industrial Time-Series Foundation Models
标题：FactoryNet：面向工业时间序列基础模型的大规模数据集
链接：https://arxiv.org/abs/2605.09081

作者：Karim Othman,Jonas Petersen,Matei Ignuta-Ciuncanu,Riccardo Maggioni,Camilla Mazzoleni,Federico Martelli,Philipp Petersen
备注：8 pages, 4 figures, 5 tables. Submitted to ICML 2026 Workshop on AI for Physics (AI4Physics)

【53】UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence
标题：Uxsid：超长序列的语义感知用户兴趣建模
链接：https://arxiv.org/abs/2605.09040

作者：Hongwei Zhang,Qiqiang Zhong,Jiangxia Cao,Yiyang Lv,Huanjie Wang,Liwei Guan,Jing Yao,Yiyu Wang,Junfeng Shu,Zhaojie Liu,Han Li
备注：Work in progress

【54】Non-Parametric Rehearsal Learning via Conditional Mean Embeddings
标题：通过条件平均嵌入的非参数排练学习
链接：https://arxiv.org/abs/2605.08999

作者：Wen-Bo Du,Tian-Zuo Wang,Han-Jia Ye,Zhi-Hua Zhou

【55】When More Parameters Hurt: Foundation Model Priors Amplify Worst-Client Disparity Under Extreme Federated Heterogeneity
标题：当更多参数受到伤害时：基础模型先验放大极端联邦异类下最差客户差异
链接：https://arxiv.org/abs/2605.08992

作者：Kiran Naseer,Umar Shoaib
备注：7 pages, 5 figures. Submitted to FL@FM-IJCAI 2026 Workshop

【56】Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials
标题：机器学习原子间潜力的基准组合概括
链接：https://arxiv.org/abs/2605.08988

作者：Amir Masoud Nourollah,Irtaza Khalid,Stefano Leoni,Steven Schockaert

【57】Trustworthy AI: Ensuring Reliability and Accountability from Models to Agents
标题：值得信赖的人工智能：确保从模型到代理的可靠性和问责制
链接：https://arxiv.org/abs/2605.08964

作者：Carol Xuan Long
备注：PhD thesis

【58】A Single Deep Preference-Conditioned Policy for Learning Pareto Coverage Sets
标题：学习帕累托覆盖集的单一深度偏好条件策略
链接：https://arxiv.org/abs/2605.08946

作者：Akihiro Kubo,Kosuke Nakanishi,Shin Ishii

【59】FragileFlow: Spectral Control of Correct-but-Fragile Predictions for Foundation Model Robustness
标题：FragileFlow：基础模型稳健性的正确但脆弱预测的频谱控制
链接：https://arxiv.org/abs/2605.08896

作者：Zhuoyun Li,Boxuan Wang,Jinwei Hu,Xiaowei Huang,Yi Dong

【60】Machine Learning Research Has Outpaced Its Communication Norms and NeurIPS Should Act
标题：机器学习研究已超出其通信规范，NeurIPS应该采取行动
链接：https://arxiv.org/abs/2605.08889

作者：Ajay Mandyam Rangarajan,Jeyashree Krishnan
备注：9 pages, 11 figures, 7 tables

【61】Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning
标题：通过结构修剪的紧凑SO（3）等变原子基础模型
链接：https://arxiv.org/abs/2605.08885

作者：Chen Wang,Siyu Hu,Guangming Tan,Weile Jia

【62】ProcVLM: Learning Procedure-Grounded Progress Rewards for Robotic Manipulation
标题：ProcVLM：机器人操纵以学习过程为基础的进步奖励
链接：https://arxiv.org/abs/2605.08774

作者：Youhe Feng,Hansen Shi,Haoyang Li,Xinlei Guo,Yang Wang,Chengyang Zhang,Jinkai Zhang,Xiaohan Zhang,Jie Tang,Jing Zhang

【63】Latent Geometry Beyond Search: Amortizing Planning in World Models
标题：超越搜索的潜在几何：世界模型中的摊销规划
链接：https://arxiv.org/abs/2605.08732

作者：Hoang Nguyen,Xiaohao Xu,Xiaonan Huang
备注：31 pages

【64】Supersampling Stable Diffusion and More: An Approach for Interpolating Neural Networks Using Common Interpolation Methods
标题：超稳定扩散及更多：使用常用插值方法进行神经网络插值的方法
链接：https://arxiv.org/abs/2605.08698

作者：Md Abu Obaida Zishan,Jannatun Noor,Annajiat Alim Rasel

【65】Event Fields: Learning Latent Event Structure for Waveform Foundation Models
标题：事件字段：学习Waveform Foundation模型的潜在事件结构
链接：https://arxiv.org/abs/2605.08685

作者：Li Na,Yuanyun Zhang,Shi Li

【66】Fitting Multilinear Polynomials for Logic Gate Networks
标题：逻辑门网络的多线性多边形的匹配
链接：https://arxiv.org/abs/2605.08657

作者：Youngsung Kim

【67】Uncovering Intra-expert Activation Sparsity for Efficient Mixture-of-Expert Model Execution
标题：揭示专家内激活稀疏性以高效混合专家模型执行
链接：https://arxiv.org/abs/2605.08575

作者：Jongseok Park,Sunga Kim,Zhenyu Gu,Ion Stoica,Alvin Cheung

【68】Biological Plausibility and Representational Alignment of Feedback Alignment in Convolutional Networks
标题：卷积网络中反馈对齐的生物合理性和表示对齐
链接：https://arxiv.org/abs/2605.08564

作者：Jake Lance,Larry Kieu

【69】A Call to Lagrangian Action: Learning Population Mechanics from Temporal Snapshots
标题：呼吁拉格朗日行动：从时间快照学习人口机制
链接：https://arxiv.org/abs/2605.08550

作者：Vincent Guan,Lazar Atanackovic,Kirill Neklyudov
备注：Accepted at ICML 2026 (spotlight)

【70】Continuity Laws for Sequential Models
标题：序列模型的连续性定律
链接：https://arxiv.org/abs/2605.08539

作者：Annan Yu,Dongwei Lyu,N. Benjamin Erichson

【71】The Propagation Field: A Geometric Substrate Theory of Deep Learning
标题：传播场：深度学习的几何基底理论
链接：https://arxiv.org/abs/2605.08529

作者：Xingrui Gu
备注：Technical notes on exploring the nature of deep learning propagation, Under review by the ICML 4th Workshop on High-dimensional Learning Dynamics (HiLD) 2026

【72】NeuralBench: A Unifying Framework to Benchmark NeuroAI Models
标题：NeuralBench：一个基准NeuroAI模型的统一框架
链接：https://arxiv.org/abs/2605.08495

作者：Hubert Banville,Stéphane d'Ascoli,Simon Dahan,Jérémy Rapin,Marlène Careil,Yohann Benchetrit,Jarod Lévy,Saarang Panchavati,Antoine Ratouchniak,Mingfang,Zhang,Elisa Cascardi,Katelyn Begany,Teon Brooks,Jean-Rémi King
备注：31 pages, 9 figures

【73】The Geometric Structure of Models Learning Sparse Data
标题：学习稀疏数据的模型的几何结构
链接：https://arxiv.org/abs/2605.08464

作者：Thomas Walker,T. Mitchell Roddenberry,Ahmed Imtiaz Humayun,Randall Balestriero,Richard Baraniuk
备注：27 pages, 7 figures, 5 tables

【74】Geometry-Aware Discretization Error of Diffusion Models
标题：扩散模型的几何感知离散误差
链接：https://arxiv.org/abs/2605.08392

作者：Samuel Hurault,Thomas Moreau,Gabriel Peyré

【75】Revitalizing the Beginning: Avoiding Storage Dependency for Model Merging in Continual Learning
标题：重振开始：避免持续学习中模型合并的存储依赖
链接：https://arxiv.org/abs/2605.08311

作者：Xi Wang,Cheng Deng

【76】A Qualitative Test-Risk Mechanism for Scaling Behavior in Normalized Residual Networks
标题：规范化剩余网络中缩放行为的定性测试风险机制
链接：https://arxiv.org/abs/2605.08297

作者：Daning Cheng,Zeyu Liu,Jun Sun,Fen Xia,Boyang Zhang,Dongping Liu,Yunquan Zhang

【77】LaWM: Least Action World Models for Long-Horizon Physical Consistency from Visual Observations
标题：LaWM：来自视觉观察的长期物理一致性的最小行动世界模型
链接：https://arxiv.org/abs/2605.08279

作者：Qixin Xiao,Maani Ghaffari

【78】HyperTransport: Amortized Conditioning of T2I Generative Models
标题：HyperTransport：T2 I生成模型的摊销条件反射
链接：https://arxiv.org/abs/2605.08254

作者：Valentino Maiorca,Eleonora Gualdoni,Xavier Suau,Marco Cuturi,Luca Zappella,Pau Rodríguez

【79】Deep Dreams Are Made of This: Visualizing Monosemantic Features in Diffusion Models
标题：深度梦想由此构成：将扩散模型中的单一语义特征可视化
链接：https://arxiv.org/abs/2605.08218

作者：Adam Szokalski,Mateusz Modrzejewski

【80】Learngene Search Across Multiple Datasets for Building Variable-Sized Models
标题：Learngene在多个数据集中搜索以构建可变大小模型
链接：https://arxiv.org/abs/2605.08209

作者：Boyu Shi,Junbo Zhou,Chang Liu,Xu Yang,Qiufeng Wang,Xin Geng

【81】ExecuTorch -- A Unified PyTorch Solution to Run AI Models On-Device
标题：ExecuTorch --在设备上运行人工智能模型的统一PyTorch解决方案
链接：https://arxiv.org/abs/2605.08195

作者：Mergen Nachin,Digant Desai,Sicheng Stephen Jia,Chen Lai,Mengwei Liu,Jacob Szwejbka,Raziel Alvarez,RJ Ascani,Dave Bort,Manuel Candales,Andrew Caples,Yanan Cao,Zhengxu Chen,Soumith Chintala,Gregory Comer,Tanvir Islam,Songhao Jia,Tarun Karuturi,Jack Khuu,Abhinay Kukkadapu,Tugsbayasgalan Manlaibaatar,Andrew Or,Kimish Patel,Siddartha Pothapragada,Lucy Qiu,Supriya Rao,Orion Reblitz-Richardson,Max Ren,Scott Roy,Anthony Shoumikhin,Scott Wolchok,Guang Yang,Angela Yi,Martin Yuan,Hansong Zhang,Jack Zhang,Jerry Zhang,Shunting Zhang,C. Cagatay Bilgin

【82】Physics-Modeled Neural Networks
标题：物理模型神经网络
链接：https://arxiv.org/abs/2605.08176

作者：Raul Felipe-Sosa,Angel Martin del Rey,Maria Flores Ceballos

【83】Augmented Equivariant Mesh Networks for Anatomical Segmentation
标题：用于解剖分割的增强等变网格网络
链接：https://arxiv.org/abs/2605.08172

作者：Daniel Saragih
备注：21 pages, 7 figures, 14 tables

【84】Communication Dynamics Neural Networks: FFT-Diagonalized Layers for Improved Hessian Conditioning at Reduced Parameter Count
标题：通信动力学神经网络：在减少参数计数的情况下改进黑森条件反射的快速傅里叶变换层
链接：https://arxiv.org/abs/2605.08171

作者：Lurong Pan

【85】A PyTorch Library of Turing-Complete Neural Networks
标题：图灵完全神经网络的PyTorch库
链接：https://arxiv.org/abs/2605.08150

作者：Jonathan Bates

【86】HoReN: Normalized Hopfield Retrieval for Large-Scale Sequential Model Editing
标题：HoReN：用于大规模序列模型编辑的规范化Hopfield检索
链接：https://arxiv.org/abs/2605.08143

作者：Yuan Fang,Yi Xie,Xuming Ran
备注：30 pages, 10 figures

【87】Dendritic Neural Networks with Equilibrium Propagation
标题：具有平衡传播的树状神经网络
链接：https://arxiv.org/abs/2605.08135

作者：Yoshimasa Kubo
备注：8 pages

【88】Feature Repulsion and Spectral Lock-in: An Empirical Study of Two-Layer Network Grokking
标题：特征排斥和谱锁定：两层网络Grokking的实证研究
链接：https://arxiv.org/abs/2605.08119

作者：Yongzhong Xu
备注：11 pages, 4 figures

【89】The Safety-Aware Denoiser for Text Diffusion Models
标题：文本扩散模型的安全意识降噪器
链接：https://arxiv.org/abs/2605.08116

作者：Amman Yusuf,Zhejun Jiang,Mijung Park
备注：27 pages, 12 figures. Code available at: https://github.com/ammanyusuf/SAD

【90】Do Foundation Model Embeddings Improve Cross-Country Crop Yield Generalisation? A Leave-One-Country-Out Evaluation in Sub-Saharan Africa
标题：基础模型嵌入能否改善跨国作物产量的普遍化？撒哈拉以南非洲的留一国评估
链接：https://arxiv.org/abs/2605.08113

作者：Yaw Osei Adjei
备注：9 pages, 10 figures, appendix, code and processed results released publicly

【91】Coarsening Linear Non-Gaussian Causal Models with Cycles
标题：带循环的线性非高斯因果模型粗化
链接：https://arxiv.org/abs/2605.10163

作者：Francisco Madaleno,Francisco C Pereira,Alex Markham

【92】PFN-TS: Thompson Sampling for Contextual Bandits via Prior-Data Fitted Networks
标题：PFN-TS：通过先验数据匹配网络对上下文带宽进行Thompson采样
链接：https://arxiv.org/abs/2605.10137

作者：Yan Shuo Tan,Kenyon Ng,Ruizhe Deng,Sumetha Loganathan,Qiong Zhang,Bibhas Chakraborty

【93】Learning stochastic multiscale models through normalizing flows
标题：通过规范化流学习随机多尺度模型
链接：https://arxiv.org/abs/2605.09718

作者：Anan Saha,Arnab Ganguly
备注：17 pages, 4 figures

【94】Metropolis-Adjusted Diffusion Models
标题：大都市调整扩散模型
链接：https://arxiv.org/abs/2605.09654

作者：Kevin H. Lam,Tyler Farghly,Christopher Williams,Jun Yang,Yee Whye Teh,Arnaud Doucet

【95】Nonlinear GENERIC Informed Neural Networks (N-GINNs): learning GENERIC dynamics with non-quadratic dissipation potentials
标题：非线性GENERIC知情神经网络（N-GINN）：学习具有非二次消散势的GENERIC动态
链接：https://arxiv.org/abs/2605.09058

作者：Vojtěch Votruba,Zequn He,Weilun Qiu,Celia Reina,Michal Pavelka
备注：26 pages, 7 figures, 4 tables

【96】Learning Pure Quantum States in Any Dimension (Almost) Without Regret
标题：在任何维度（几乎）毫无遗憾地学习纯量子状态
链接：https://arxiv.org/abs/2605.09019

作者：Josep Lumbreras,Marco Tomamichel
备注：43 pages

【97】CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models
标题：CrystalREPA：将物理先验从通用MLIP转移到晶体生成模型
链接：https://arxiv.org/abs/2605.08960

作者：Chengqian Zhang,Yucheng Jin,Duo Zhang,Tiejun Li,Han Wang

【98】On Observation Time for Recovering Latent Hawkes Networks
标题：恢复潜在霍克斯网络的观察时间
链接：https://arxiv.org/abs/2605.08400

作者：Jonas Linkerhägner,Michele Bortolasi,Lorenzo Baldassari,Maarten V. de Hoop,Ivan Dokmanić

【99】Lecture Notes on Statistical Physics and Neural Networks
标题：统计物理和神经网络课堂笔记
链接：https://arxiv.org/abs/2605.06394

作者：Olaf Hohm
备注：56 pages, 7 figures, based on a course given at Humboldt University Berlin

其他(182篇)

【1】ELF: Embedded Language Flows
标题：ELF：嵌入式语言流
链接：https://arxiv.org/abs/2605.10938

作者：Keya Hu,Linlu Qiu,Yiyang Lu,Hanhong Zhao,Tianhong Li,Yoon Kim,Jacob Andreas,Kaiming He
备注：Tech Report. Project webpage: https://github.com/lillian039/ELF

【2】DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices
标题：DICO：在端端设备上具有密集可比性能的稀疏专家混合
链接：https://arxiv.org/abs/2605.10933

作者：Chenyang Song,Weilin Zhao,Xu Han,Chaojun Xiao,Yingfa Chen,Zhiyuan Liu
备注：14 pages, 11 figures, 11 tables

【3】Neural Weight Norm = Kolmogorov Complexity
标题：神经权重规范= Kolmogorov复杂性
链接：https://arxiv.org/abs/2605.10878

作者：Tiberiu Musat

【4】The Generalized Turing Test: A Foundation for Comparing Intelligence
标题：广义图灵测试：比较智力的基础
链接：https://arxiv.org/abs/2605.10851

作者：Daniel Mitropolsky,Susan S. Hong,Riccardo Neumarker,Emanuele Rimoldi,Tomaso Poggio

【5】MaD Physics: Evaluating information seeking under constraints in physical environments
标题：MaD物理：评估物理环境约束下的信息搜索
链接：https://arxiv.org/abs/2605.10820

作者：Moksh Jain,Mehdi Bennani,Johannes Bausch,Yuri Chervonyi,Bogdan Georgiev,Simon Osindero,Nenad Tomašev
备注：64 pages, 10 figures. Project page: https://mad-physics.github.io/

【6】The Last Word Often Wins: A Format Confound in Chain-of-Thought Corruption Studies
标题：最后一句话常常获胜：思想链腐败研究中的一种混乱形式
链接：https://arxiv.org/abs/2605.10799

作者：Gabriel Garcia
备注：34 pages, 6 figures, 13 tables. Submitted to NeurIPS 2026. Code and data: https://github.com/Gpgabriel25/LastWordWinsCoT

【7】XQCfD: Accelerating Fast Actor-Critic Algorithms with Prior Data and Prior Policies
标题：XQCfD：利用先验数据和先验策略加速快速行动者批评算法
链接：https://arxiv.org/abs/2605.10734

作者：Daniel Palenicek,Florian Vogt,Joe Watson,Ingmar Posner,Danica Kragic,Jan Peters
备注：22 pages, 10 figures, 2 tables

【8】What should post-training optimize? A test-time scaling law perspective
标题：训练后应该优化什么？测试时间缩放定律的角度
链接：https://arxiv.org/abs/2605.10716

作者：Muheng Li,Jian Qian,Wenlong Mou

【9】The finite expression method for turbulent dynamics with high-order moment recovery
标题：具有高次矩恢复的湍流动力学有限表达方法
链接：https://arxiv.org/abs/2605.10687

作者：Xingjian Xu,Di Qi,Chunmei Wang
备注：20 pages, 8 figures, 1 table

【10】Is Data Shapley Not Better than Random in Data Selection? Ask NASH
标题：Data Shapley在数据选择方面不比随机好吗？询问NASH
链接：https://arxiv.org/abs/2605.10684

作者：Xiao Tian,Jue Fan,Rachael Hwee Ling Sim,Zixuan Wang,Nancy F. Chen,Bryan Kian Hsiang Low
备注：Accepted to the 43rd International Conference on Machine Learning (ICML-26) as a Spotlight paper

【11】Exact Unlearning from Proxies Induces Closeness Guarantees on Approximate Unlearning
标题：代理的精确遗忘导致近似遗忘的接近保证
链接：https://arxiv.org/abs/2605.10680

作者：Virgile Dine,Teddy Furon

【12】Natural Policy Gradient as Doubly Smoothed Policy Iteration: A Bellman-Operator Framework
标题：作为双重平滑政策迭代的自然政策梯度：贝尔曼-操作员框架
链接：https://arxiv.org/abs/2605.10671

作者：Phalguni Nanda,Zaiwei Chen

【13】BCJR-QAT: A Differentiable Relaxation of Trellis-Coded Weight Quantization
标题：BCJR-QAT：网格编码权重量化的可区分放松
链接：https://arxiv.org/abs/2605.10655

作者：Venugopalan Iyengar
备注：26 pages, 4 figures, 4 tables. Code at https://github.com/Venugopalan2610/quant-2bit. Model weights and trajectory snapshots at https://huggingface.co/Venugopalan2610/BCJR-QAT-Llama-3.2-1B-2bit

【14】Composing diffusion priors with explicit physical context via generative Gibbs sampling
标题：通过生成吉布斯采样构建具有明确物理背景的扩散先验
链接：https://arxiv.org/abs/2605.10642

作者：Weizhou Wang,Jonathan Weare,Aaron R. Dinner
备注：31 pages, 11 figures

【15】Fairness vs Performance: Characterizing the Pareto Frontier of Algorithmic Decision Systems
标题：公平与绩效：数学决策系统的帕累托前沿特征
链接：https://arxiv.org/abs/2605.10604

作者：Mieke Wilms,Christoph Heitz
备注：23 pages, The 2026 ACM conference on Fairness, Accountability, and Transparency (FAccT'26)

【16】A Resilient Solution for Sewer Overflow Monitoring across Cloud and Edge
标题：跨云和边缘污水溢流监控的弹性解决方案
链接：https://arxiv.org/abs/2605.10592

作者：Vipin Singh,Tianheng Ling,Peter Ghaly,Felix Grimmeisen,Gregor Schiele,Felix Biessmann
备注：3 pages, 6 figures, accepted at 35th International Joint Conference on Artificial Intelligence 2026 (IJCAI-ECAI 2026), Demonstrations Track. URL: https://riwwer.demo.calgo-lab.de

【17】Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims
标题：受理卡：安全微调防御索赔的四诊断标准
链接：https://arxiv.org/abs/2605.10575

作者：Phongsakon Mark Konrad,Toygar Tanyel,Serkan Ayvaz

【18】HH-SAE: Discovering and Steering Hierarchical Knowledge of Complex Manifolds
标题：HH-AE：发现和指导复杂流体的分层知识
链接：https://arxiv.org/abs/2605.10536

作者：Honghan Wu,Tianyan Wang,Jiacong Mi,Zhoyang Jiang,Yunsoo Kim

【19】Regret Minimization in Bilateral Trade With Perturbed Markets
标题：市场扰动下双边贸易中最大限度地减少遗憾
链接：https://arxiv.org/abs/2605.10475

作者：Anna Lunghi, Matteo Castiglioni, Alberto Marchesi

【20】SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding
标题：SlimSec：用于加速投机解码的低级别草案LM-Head
链接：https://arxiv.org/abs/2605.10453

作者：Anton Plaksin, Sergei Krutikov, Sergei Skvortsov, Alexander Samarin

【21】Active Tabular Augmentation via Policy-Guided Diffusion Inpainting
标题：通过政策引导的扩散修复进行主动表格增强
链接：https://arxiv.org/abs/2605.10315

作者：Zheyu Zhang, Shuo Yang, Bardh Prenkaj, Gjergji Kasneci
备注：Accepted for publication at ICML 2026

【22】Signature Approach for Contextual Bandits with Nonlinear and Path-dependent Rewards
标题：具有非线性和路径相关奖励的背景盗贼签名方法
链接：https://arxiv.org/abs/2605.10313

作者：Xin Guo, Grace He, Xinyu Li

【23】Follow the Mean: Reference-Guided Flow Matching
标题：遵循平均值：参考引导的流量匹配
链接：https://arxiv.org/abs/2605.10302

作者：Pedro M. P. Curvo, Maksim Zhdanov, Floor Eijkelboom, Jan-Willem van de Meent

【24】DeepLog: A Software Framework for Modular Neurosymbolic AI
标题：DeepLog：模块化神经符号人工智能的软件框架
链接：https://arxiv.org/abs/2605.10279

作者：Robin Manhaeve, Stefano Colamonaco, Vincent Derkinderen, Rik Adriaensen, Lucas Van Praet, Luc De Raedt, Giuseppe Marra
备注：Preprint accepted at IJCAI2026 Demo Track

【25】E-TCAV: Formalizing Penultimate Proxies for Efficient Concept Based Interpretability
标题：E-TCAB：形式化倒数第二个代理，以实现高效的基于概念的可解释性
链接：https://arxiv.org/abs/2605.10261

作者：Hasib Aslam, Muhammad Ali Chattha, Muhammad Taha Mukhtar, Muhammad Imran Malik, Andreas Dengel, Sheraz Ahmed

【26】The Benefits of Temporal Correlations: SGD Learns k-Juntas from Random Walks Efficiently
标题：时间相关性的好处：Singapore从随机漫步中有效学习k-Juntas
链接：https://arxiv.org/abs/2605.10237

作者：Elisabetta Cornacchia, Dan Mikulincer, Elchanan Mossel
备注：10 pages main body, 3 figures

【27】Unveiling High-Probability Generalization in Decentralized SGD
标题：揭示分散式新元中的高概率概括
链接：https://arxiv.org/abs/2605.10205

作者：Jiahuan Wang, Ping Luo, Ziqing Wen, Dongsheng Li, Tao Sun

【28】Many Needles in a Haystack: Active Hit Discovery for Perturbation Experiments
标题：干草堆中的许多针：微扰实验的积极发现
链接：https://arxiv.org/abs/2605.10196

作者：Andrea Rubbi, Arpit Merchant, Samuel Ogden, Amir Akbarnejad, Pietro Liò, Sattar Vakili, Mo Lotfollahi
备注：To be published in International Conference on Machine Learning (ICML) 2026

【29】ProteinOPD: Towards Effective and Efficient Preference Alignment for Protein Design
标题：ProteinOPD：实现蛋白质设计的有效和高效的偏好比对
链接：https://arxiv.org/abs/2605.10189

作者：Yulin Zhang, He Cao, Zihao Jiang, Chenyi Zi, Zhipeng Zhou, Zijing Liu, Yu Li, Jia Li, Ziqi Gao

【30】Hyperparameter Transfer for Dense Associative Memories
标题：密集联想记忆的超参数转移
链接：https://arxiv.org/abs/2605.10164

作者：Roi Holtzman, Dmitry Krotov, Boris Hanin

【31】Rethinking Constraint Awareness for Efficient State Embedding of Neural Routing Solver
标题：重新思考约束意识以实现神经路由求解器的高效状态嵌入
链接：https://arxiv.org/abs/2605.10122

作者：Canhong Yu, Changliang Zhou, Rongsheng Chen, Zhenkun Wang, Yu Zhou

【32】Scaling the Memory of Balanced Adam
标题：扩展平衡亚当的记忆
链接：https://arxiv.org/abs/2605.10119

作者：Alberto Fernández-Hernández, Cristian Pérez-Corral, Jose I. Mestre, Manuel F. Dolz, Enrique S. Quintana-Ortí

【33】TopoU-Net: a U-Net architecture for topological domains
标题：TopoU-Net：一种用于布局域的U-Net架构
链接：https://arxiv.org/abs/2605.10091

作者：Gaurav Gaurav, Ibrahem ALJabea, Yaroslav Zakomornyy, Eric Frank, Mohamed Elhamdadi, Theodore Papamarkou, Mustafa Hajij

【34】The Value of Mechanistic Priors in Sequential Decision Making
标题：机械先验在顺序决策中的价值
链接：https://arxiv.org/abs/2605.10018

作者：Itai Shufaro, Gal Benor, Shie Mannor

【35】Anchor-guided Hypergraph Condensation with Dual-level Discrimination
标题：具有双重区分的锚点引导超图凝聚
链接：https://arxiv.org/abs/2605.10001

作者：Fan Li, Xiaoyang Wang, Chen Chen, Wenjie Zhang
备注：This paper has been accepted by ICML 2026

【36】Optimizer-Induced Mode Connectivity: From AdamW to Muon
标题：优化器诱导的模式连接性：从AdamW到Muon
链接：https://arxiv.org/abs/2605.09991

作者：Fangzhao Zhang, Sungyoon Kim, Erica Zhang, Yiqi Jiang, Mert Pilanci

【37】The Truth Lies Somewhere in the Middle (of the Generated Tokens)
标题：真相就在（生成的代币的）中间某处
链接：https://arxiv.org/abs/2605.09969

作者：Sophie L. Wang, Phillip Isola, Brian Cheung

【38】Novel GPU Boruta algorithms for feature selection from high-dimensional data
标题：用于从多维数据中选择特征的新型GDPBoruta算法
链接：https://arxiv.org/abs/2605.09950

作者：Xurui Li,Zhiguo Gan,Jiaming Zhang,Zheng Liu,Diannan Lu
备注：This paper has been submitted to the journal Data Mining and Knowledge Discovery, and a preprint is available for the authors' records

【39】Selection of the Best Policy under Fairness Constraints for Subpopulations
标题：公平约束下亚群体最佳政策的选择
链接：https://arxiv.org/abs/2605.09945

作者：Tingyu Zhu,Yuhang Wu,Zeyu Zheng

【40】Urban-ImageNet: A Large-Scale Multi-Modal Dataset and Evaluation Framework for Urban Space Perception
标题：Urban-ImageNet：城市空间感知的大规模多模式数据集和评估框架
链接：https://arxiv.org/abs/2605.09936

作者：Yiwei Ou,Chung Ching Cheung,Jun Yang Ang,Xiaobin Ren,Ronggui Sun,Guansong Gao,Kaiqi Zhao,Manfredo Manfredini

【41】Key-Value Means
标题：关键值意味着
链接：https://arxiv.org/abs/2605.09877

作者：Daniel Goldstein, Eugene Cheah

【42】Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions
标题：通过模拟器生成的干预分布发现基于干预的时间序列原因
链接：https://arxiv.org/abs/2605.09870

作者：Tsuyoshi Okita
备注：54 pages, 6 figures

【43】DA-SegFormer: Damage-Aware Semantic Segmentation for Fine-Grained Disaster Assessment
标题：DA-SegFormer：用于细粒度灾难评估的损害感知语义分割
链接：https://arxiv.org/abs/2605.09864

作者：Kevin Zhu, William Tang, Raphael Hay Tene, Zesheng Liu, Nhut Le, Maryam Rahnemoonfar
备注：Accepted for 2026 IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2026)

【44】Cross-Domain Lossy Compression via Constrained Minimum Entropy Coupling
标题：基于约束最小熵耦合的跨域有损压缩
链接：https://arxiv.org/abs/2605.09833

作者：Nam Nguyen, Hassan Tavakoli, An Vuong, Thinh Nguyen, Bella Bose

【45】On Uniform Error Bounds for Kernel Regression under Non-Gaussian Noise
标题：非高斯噪音下核回归的一致误差界
链接：https://arxiv.org/abs/2605.09757

作者：Johannes Teutsch,Oleksii Molodchyk,Marion Leibold,Timm Faulwasser,Armin Lederer
备注：This paper has been accepted at the 43rd International Conference on Machine Learning (ICML) 2026

【46】Accelerating Power Method with Fast Sketching for Stronger Low-Rank Approximation
标题：快速绘制加速功率方法以获得更强的低阶逼近
链接：https://arxiv.org/abs/2605.09755

作者：Shabarish Chenakkod,Michał Dereziński

【47】Sequential Feature Selection for Efficient Landslide Segmentation from Multi-Spectral Data
标题：从多光谱数据中有效分割滑坡的序列特征选择
链接：https://arxiv.org/abs/2605.09746

作者：Arsalaan Ahmad,Oktay Karakus,Paul L. Rosin
备注：In Process of Submission to Frontiers in Remote Sensing. Keywords: landslide segmentation, multispectral remote sensing, feature selection, explainability, Landslide4Sense

【48】RubricRefine: Improving Tool-Use Agent Reliability with Training-Free Pre-Execution Refinement
标题：RubricRefine：通过免训练的执行前细化提高工具使用代理的可靠性
链接：https://arxiv.org/abs/2605.09730

作者：Will LeVine,Brendan Evers,Sam Saltwick,Abhay Venkatesh

【49】Make Each Token Count: Towards Improving Long-Context Performance with KV Cache Eviction
标题：让每个令牌都有意义：通过KV缓存驱逐来提高长上下文性能
链接：https://arxiv.org/abs/2605.09649

作者：Ngoc Bui,Hieu Trung Nguyen,Arman Cohan,Rex Ying
备注：A learnable KV eviction method for large language models

【50】Efficient Ensemble Selection from Binary and Pairwise Feedback
标题：从二进制和成对反馈中高效选择
链接：https://arxiv.org/abs/2605.09588

作者：Tzeh Yuan Neoh,Nicholas Teh,Je Qin Chooi,Paul W. Goldberg,Milind Tambe

【51】HS-FNO: History-Space Fourier Neural Operator for Non-Markovian Partial Differential Equations
标题：HS-FNO：非马尔科夫偏微方程的历史空间傅里叶神经运算符
链接：https://arxiv.org/abs/2605.09523

作者：Lennon J. Shikhman
备注：29 pages, 4 figures, 1 table. Under review. Code at https://github.com/lennonshikhman/hs-fno/

【52】Mixture of Layers with Hybrid Attention
标题：具有混合注意力的混合层
链接：https://arxiv.org/abs/2605.09516

作者：Ivan Ternovtsii,Yurii Bilak

【53】A Cognitively Grounded Bayesian Framework for Misinformation Susceptibility
标题：错误信息易感性的认知基础Bayesian框架
链接：https://arxiv.org/abs/2605.09483

作者：Pranava Madhyastha
备注：work in progress

【54】Positional LSH: Binary Block Matrix Approximation for Attention with Linear Biases
标题：位置LSH：具有线性偏差的注意力的二进制块矩阵逼近
链接：https://arxiv.org/abs/2605.09472

作者：Daniel Wolfson,Tal Wagner

【55】RAwR: Role-Aware Rewiring via Approximate Equitable Partition
标题：RAwR：通过大致公平分区实现角色感知重新布线
链接：https://arxiv.org/abs/2605.09457

作者：Riccardo Porcedda,Giuseppe Squillace,Bastian Epping,Andrea Vandin,Michael Schaub,Mirco Tribastone,Francesca Chiaromonte

【56】Inverse Design for Conditional Distribution Matching
标题：条件分布匹配的逆设计
链接：https://arxiv.org/abs/2605.09439

作者：Ori Meidler,Shaul Tolkovsky,Or Zuk

【57】fmxcoders: Factorized Masked Crosscoders for Cross-Layer Feature Discovery
标题：fmxCoders：用于跨层特征发现的分解掩蔽交叉器
链接：https://arxiv.org/abs/2605.09438

作者：Andreas D. Demou,Panagiotis Koromilas,James Oldfield,Yannis Panagakis,Mihalis A. Nicolaou

【58】A Controlled Diagnostic Study of Hardware-Induced Distortions in Hardware-Aware Training
标题：硬件感知训练中硬件诱发失真的对照诊断研究
链接：https://arxiv.org/abs/2605.09416

作者：Yunxuan Fang,Xinhe Wang

【59】Let the Target Select for Itself: Data Selection via Target-Aligned Paths
标题：让目标自行选择：通过目标对齐路径选择数据
链接：https://arxiv.org/abs/2605.09404

作者：Huitao Yang,Hengzhi He,Guang Cheng

【60】Universal Feature Selection with Noisy Observations and Weak Symmetry Conditions
标题：具有噪音观察和弱对称性条件的通用特征选择
链接：https://arxiv.org/abs/2605.09396

作者：Dier Tang,Guangyue Han
备注：6 pages, 0 figures. This work has been submitted to the 2026 IEEE Information Theory Workshop (ITW) for possible publication

【61】Selection Plateau and a Sparsity-Dependent Hierarchy of Pruning Features
标题：选择平台和稀疏依赖的剪枝特征层次
链接：https://arxiv.org/abs/2605.09345

作者：Guangqi Li,Yongxin Li
备注：22 pages, 3 figures, 5 tables. Empirical study + framework hypothesis on ViT-Small/CIFAR-10. Cross-domain validation (vision token pruning, KV cache compression, MoE routing) and cross-architecture extensions deferred to follow-up work

【62】Dimension-Free Saddle-Point Escape in Muon
标题：μ子中无冲突马鞍点逃生
链接：https://arxiv.org/abs/2605.09331

作者：Yanlin Long,Yufei Gu,Zeke Xie
备注：33 pages, 5 figures. Preprint

【63】Test-Time Speculation
标题：测试时间猜测
链接：https://arxiv.org/abs/2605.09329

作者：Avinash Kumar,Sujay Sanghavi,Poulami Das

【64】Mem-W: Latent Memory-Native GUI Agents
标题：Mem-W：潜在内存原生图形用户界面代理
链接：https://arxiv.org/abs/2605.09317

作者：Guibin Zhang,Yaohui Ling,Fanci Meng,Kun Wang,Shuicheng Yan

【65】Teaching Molecular Dynamics to a Non-Autoregressive Ionic Transport Predictor
标题：非自回归离子输运预测者的分子动力学教授
链接：https://arxiv.org/abs/2605.09311

作者：Jiyeon Kim,Byungju Lee,Won-Yong Shin
备注：International Conference on Machine Learning (ICML 2026) (to appear) (Please cite our conference version.)

【66】Discrete Langevin-Inspired Posterior Sampling
标题：离散Langevin启发的后验抽样
链接：https://arxiv.org/abs/2605.09302

作者：Chaitanya Amballa,Sattwik Basu,Jorge Vančo Sampedro,Romit Roy Choudhury

【67】LagrangianSplats: Divergence-Free Transport of Gaussian Primitives for Fluid Reconstruction
标题：拉格朗日Splats：用于流体重建的高斯基元的无分歧传输
链接：https://arxiv.org/abs/2605.09299

作者：Ningxiao Tao,Baoquan Chen,Mengyu Chu

【68】MC$^2$: Monte Carlo Correction for Fast Elliptic PDE Solving
标题：MC $' 2 $：快速椭圆形偏东方程求解的蒙特卡罗修正
链接：https://arxiv.org/abs/2605.09288

作者：Ethan Hsu,Hong Meng Yam,Ivan Ge

【69】TileQ: Efficient Low-Rank Quantization of Mixture-of-Experts with 2D Tiling
标题：TileQ：一种高效的混合专家低秩量化算法
链接：https://arxiv.org/abs/2605.09281

作者：Hongyaoxing Gu,Xinzhe Chen,Lijuan Hu,Fangfang Liu

【70】First Worst-Case Regret Bounds for Combinatorial Thompson Sampling in Sleeping Semi-Bandits
标题：睡眠半强盗组合Thompson抽样的第一个最坏情况遗憾界限
链接：https://arxiv.org/abs/2605.09277

作者：Zhiming Huang,Bingshan Hu,Jianping Pan
备注：Accepted by INFOCOM 26 on Dec 2025

【71】DiffATS: Diffusion in Aligned Tensor Space
标题：差异：对齐张量空间中的扩散
链接：https://arxiv.org/abs/2605.09275

作者：Jinhua Lyu,Tianmin Yu,Brian Kim,Lizhuo Zhou,Chanwook Park,Naichen Shi

【72】ProactBench: Beyond What The User Asked For
标题：ProactBench：超越用户要求的内容
链接：https://arxiv.org/abs/2605.09228

作者：Sepehr Harfi,Ahmad Salimi,Dongming Shen,Alex Smola

【73】The Pokémon Theorem and other Fairness Impossibility Results
标题：神奇宝贝定理和其他公平不可能结果
链接：https://arxiv.org/abs/2605.09221

作者：Daniel Matsui Smola,Alex Smola

【74】Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability
标题：单一政策集中性下，通过向前KL正规化的离线上下文盗贼快速费率
链接：https://arxiv.org/abs/2605.09214

作者：Qingyue Zhao,Kaixuan Ji,Heyang Zhao,Quanquan Gu
备注：31 pages, comments are welcome

【75】LBI: Parallel Scan Backpropagation via Latent Bounded Interfaces
标题：LBI：通过潜在有界接口的并行扫描反向传播
链接：https://arxiv.org/abs/2605.09204

作者：Shaun Christopher Lee,Sangeetha Abdu Jyothi

【76】Practical Scaling Laws: Converting Compute into Performance in a Data-Constrained World
标题：实用的缩放定律：在数据受限的世界中将计算转化为性能
链接：https://arxiv.org/abs/2605.09189

作者：Christopher M. Bryant,Hao Liu

【77】CIVeX: Causal Intervention Verification for Language Agents
标题：CIVeX：语言代理的因果干预验证
链接：https://arxiv.org/abs/2605.09168

作者：Fabio Rovai
备注：16 pages, 3 figures. Includes Causal-ToolBench, IHDP, ZOZO Open Bandit, and LaLonde NSW evaluations

【78】WorldSpeech: A Multilingual Speech Corpus from Around the World
标题：WorldSpeech：来自世界各地的多语言语音库
链接：https://arxiv.org/abs/2605.09167

作者：Antonis Asonitis,Luca A. Lanzendörfer,Frédéric Berdoz,Roger Wattenhofer

【79】Revisiting Mixture Policies in Entropy-Regularized Actor-Critic
标题：重新审视《论》中的混合政策
链接：https://arxiv.org/abs/2605.09157

作者：Jiamin He,Samuel Neumann,Jincheng Mei,Adam White,Martha White

【80】Personalized Alignment Revisited: The Necessity and Sufficiency of User Diversity
标题：重新审视个性化一致：用户多样性的必要性和充分性
链接：https://arxiv.org/abs/2605.09119

作者：Enoch Hyunwook Kang

【81】When Style Similarity Scores Fail: Diagnosing Raw CSD Cosine in Artist-Style Evaluation
标题：当风格相似性得分失败时：在艺术家风格评估中诊断原始的CPD Cosine
链接：https://arxiv.org/abs/2605.09030

作者：Jörg Frochte
备注：24 pages, 7 figures, 19 tables

【82】Evolutionary Ensemble of Agents
标题：代理人的进化联盟
链接：https://arxiv.org/abs/2605.09018

作者：Zongmin Yu,Liu Yang

【83】CATO: Charted Attention for Neural PDE Operators
标题：CATO：神经PDE算子的注意力图表
链接：https://arxiv.org/abs/2605.09016

作者：Chun-Wun Cheng,Sifan Wang,Carola-Bibiane Schönlieb,Angelica I. Aviles-Rivero

【84】Muon Does Not Converge on Convex Lipschitz Functions
标题：μ子在凸Lipschitz函数上不收敛
链接：https://arxiv.org/abs/2605.08980

作者：Tetiana Parshakova,Ahmed Khaled,Michael Crawshaw,Guillaume Garrigos,Robert M. Gower

【85】Can We Formally Verify Neural PDE Surrogates? SMT Compilation of Small Fourier Neural Operators
标题：我们可以正式验证神经PDL替代品吗？小傅里叶神经运算符的Smart编译
链接：https://arxiv.org/abs/2605.08938

作者：Ali Baheri,David Millard,Ignacio Laguna Peralta

【86】From Mechanistic to Compositional Interpretability
标题：从机械解释到成分解释
链接：https://arxiv.org/abs/2605.08934

作者：Ward Gauderis,Thomas Dooms,Steven T. Holmer,Kola Ayonrinde,Geraint A. Wiggins

【87】Physics-Informed Neural PDE Solvers via Spatio-Temporal MeanFlow
标题：通过时空平均流实现物理知识的神经PED求解器
链接：https://arxiv.org/abs/2605.08915

作者：Hanru Bai,Yuncheng Zhou,Difan Zou

【88】Non-Monotonic Latency in Apple MPS Decoding: KV Cache Interactions and Execution Regimes
标题：Apple MPS解码中的非单调延迟：KV缓存交互和执行机制
链接：https://arxiv.org/abs/2605.08913

作者：Willy Fitra Hendria
备注：9 pages, 5 figures, 6 tables

【89】Bilinear autoencoders find interpretable manifolds
标题：双线性自动编码器找到可解释的管形
链接：https://arxiv.org/abs/2605.08891

作者：Thomas Dooms,Ward Gauderis,Geraint Wiggins,Jose Oramas

【90】Controlling Transient Amplification Improves Long-horizon Rollouts
标题：控制瞬时放大改进了长视野滚动
链接：https://arxiv.org/abs/2605.08856

作者：Adeel Pervez,Francesco Locatello

【91】M$^3$: Reframing Training Measures for Discretized Physical Simulations
标题：M$#3 $：重新制定离散物理模拟的训练措施
链接：https://arxiv.org/abs/2605.08843

作者：Yuan Mei,Xingyu Song,Xiaowen Song,Naoya Takeishi

【92】From pre-training to downstream performance: Does domain-specific pre-training make sense?
标题：从预训练到下游性能：特定领域的预训练有意义吗？
链接：https://arxiv.org/abs/2605.08819

作者：Felix Krones

【93】AgentSlimming: Towards Efficient and Cost-Aware Multi-Agent Systems
标题：AgentSlimming：迈向高效且成本意识的多代理系统
链接：https://arxiv.org/abs/2605.08813

作者：Yulang Chen,Haoxuan Peng,Jinyan Liu,Zichen Wen,Dongrui Liu,Linfeng Zhang

【94】Data-driven transport modelling without overfit
标题：数据驱动的交通建模，没有过度适合
链接：https://arxiv.org/abs/2605.08801

作者：Peter Vanya,Katarína Šimková,Rastislav Farkaš
备注：6 pages, 6 figures

【95】Deterministic Decomposition of Stochastic Generative Dynamics
标题：随机生成动力学的确定性分解
链接：https://arxiv.org/abs/2605.08794

作者：Xingyu Song,Yuan Mei,Naoya Takeishi
备注：submitted to NeruIPS 2026

【96】Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking
标题：并非所有回合都很重要：多回合越狱的信用分配
链接：https://arxiv.org/abs/2605.08778

作者：Zhida He,Xiaoyu Wen,Han Qi,Ziyuan Zhou,Peng Yu,Xingcheng Xu,Dongrui Liu,Xia Hu,Chaochao Lu,Qiaosheng Zhang
备注：41 pages, 10 figures

【97】Omni-DeepSearch: A Benchmark for Audio-Driven Omni-Modal Deep Search
标题：Omni-DeepSearch：音频驱动的Omni-Modal深度搜索的基准
链接：https://arxiv.org/abs/2605.08762

作者：Tao Yu,yiming ding,Shenghua Chai,Minghui Zhang,Zhongtian Luo,Xinming Wang,Xinlong Chen,Zhaolu Kang,Junhao Gong,Yuxuan Zhou,Haopeng Jin,Zhiqing Cui,Jiabing Yang,YiFan Zhang,Hongzhu Yi,Zheqi He,Xi Yang,Yan Huang,Liang Wang
备注：43 pages

【98】Beyond the All-in-One Agent: Benchmarking Role-Specialized Multi-Agent Collaboration in Enterprise Workflows
标题：超越一体化代理：企业工作流中角色专业化多代理协作基准
链接：https://arxiv.org/abs/2605.08761

作者：Tao Yu,Hao Wang,Changyu Li,Shenghua Chai,Minghui Zhang,Zhongtian Luo,Yuxuan Zhou,Haopeng Jin,Zhaolu Kang,Jiabing Yang,YiFan Zhang,Xinming Wang,Hongzhu Yi,Zheqi He,Jing-Shu Zheng,Xi Yang,Yan Huang,Liang Wang
备注：45 pages

【99】Communicating Sound Through Natural Language
标题：通过自然语言传达声音
链接：https://arxiv.org/abs/2605.08750

作者：Emanuele Rossi,Emanuele Rodolà
备注：Includes link to demo page

【100】The Wristband Gaussian Loss: Deterministic, Composable Latents via a Sphere-Interval Decomposition
标题：腕带高斯损失：通过球区间分解的确定性、可组合的潜伏
链接：https://arxiv.org/abs/2605.08749

作者：Mikhail Parakhin,André M. Carvalho,Patrick Haluptzok
备注：preprint

【101】Generative Actor-Critic with Soft Bridge Policies
标题：具有软桥政策的生成性演员批评家
链接：https://arxiv.org/abs/2605.08733

作者：Ke He,Le He,Shunpu Tang,Yafei Wang,Lisheng Fan

【102】Control Your View: High-Resolution Global Semantic Manipulation in Learned Image Compression
标题：控制您的视图：习得图像压缩中的高分辨率全局语义操纵
链接：https://arxiv.org/abs/2605.08727

作者：Jiaming Liang,Chi-Man Pun,Weisi Lin,Greta Seng Peng Mok

【103】Bias by Necessity: Impossibility Theorems for Sequential Processing with Convergent AI and Human Validation
标题：必要性偏差：融合人工智能和人类验证的顺序处理的不可能性定理
链接：https://arxiv.org/abs/2605.08716

作者：Jikun Wu,Dongxin Guo,Siu-Ming Yiu
备注：6 pages, 3 figures, 5 tables. Accepted to CogSci 2026

【104】RewardHarness: Self-Evolving Agentic Post-Training
标题：奖励：自我进化的强化训练后
链接：https://arxiv.org/abs/2605.08703

作者：Yuxuan Zhang,Penghui Du,Bo Li,Cong Wei,Junwen Miao,Huaisong Zhang,Songcheng Cai,Yubo Wang,Dongfu Jiang,Yuyu Zhang,Ping Nie,Wenhu Chen,Changqian Yu,Kelsey R. Allen
备注：Project page: https://rewardharness.com

【105】METBRA25Y: Brazil Surface Meteorology Archive with Harmonized Variables and Quality Control
标题：METBRA25 Y：具有协调变量和质量控制的巴西地表气象档案
链接：https://arxiv.org/abs/2605.08701

作者：Matheus Lima Castro,William Dantas Vichete,Leopoldo Lusquino Filho
备注：12 pages, 5 figures. Dataset paper describing METBRA25Y, a harmonized archive of hourly Brazilian surface meteorological observations derived from INMET records. Dataset available at Zenodo: 10.5281/zenodo.19964979

【106】MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI
标题：MLS长凳：对人工智能系统构建更好人工智能的全面而严格的评估
链接：https://arxiv.org/abs/2605.08678

作者：Bohan Lyu,Yucheng Yang,Siqiao Huang,Jiaru Zhang,Qixin Xu,Xinghan Li,Xinyang Han,Yicheng Zhang,Huaqing Zhang,Runhan Huang,Kaicheng Yang,Zitao Chen,Wentao Guo,Junlin Yang,Xinyue Ai,Wenhao Chai,Yadi Cao,Ziran Yang,Kun Wang,Dapeng Jiang,Huan-ang Gao,Shange Tang,Chengshuai Shi,Simon S. Du,Max Simchowitz,Jiantao Jiao,Dawn Song,Chi Jin

【107】The Cancellation Hypothesis in Critic-Free RL: From Outcome Rewards to Token Credits
标题：无批判RL中的取消假设：从结果奖励到代币积分
链接：https://arxiv.org/abs/2605.08666

作者：Tianhao Cheng,Zeyu Huang,Zihan Qiu,Yu Cheng,Edoardo Ponti,Yinghui Xu,Ivan Titov,Zenglin Xu

【108】FLUX: Geometry-Aware Longitudinal Flow Matching with Mixture of Experts
标题：FLOX：具有几何意识的纵向流匹配专家混合
链接：https://arxiv.org/abs/2605.08648

作者：Josue Ortega Caro,Yongxu Zhang,Hannah M Batchelor,Sizhuang He,Jessica Cardin,Shreya Saxena

【109】AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators
标题：AgentCollabBench：诊断好的代理何时会成为坏的合作者
链接：https://arxiv.org/abs/2605.08647

作者：Aritra Mazumder,Shubhashis Roy Dipta,Nusrat Jahan Lia,Tanzila Khan,Kainat Raisa Hossain,Nehaa Shri,Shubhrangshu Debsarkar,Humayra Tasnim,Gour Gupal Talukder Shawon,Debjoty Mitra,Sumaiya Ahmed Rani,Al Jami Islam Anik,Al Nafeu Khan

【110】PAAC: Privacy-Aware Agentic Device-Cloud Collaboration
标题：PAAC：隐私意识的大型设备云协作
链接：https://arxiv.org/abs/2605.08646

作者：Liangqi Yuan,Wenzhi Fang,Shiqiang Wang,Christopher G. Brinton

【111】Kaczmarz Linear Attention
标题：卡茨马尔兹线性注意力
链接：https://arxiv.org/abs/2605.08587

作者：Jiaxuan Zou,Ruifeng Ren,Yong Liu

【112】Finer is Better (with the Right Scaling)
标题：越细越好（具有正确的比例）
链接：https://arxiv.org/abs/2605.08565

作者：Clemens Schaefer,Gil Tabak

【113】Skill-CMIB: Multimodal Agent Skill for Consistent Action via Conditional Multimodal Information Bottleneck
标题：Skill-CMIB：通过条件多模式信息瓶颈实现一致行动的多模式代理技能
链接：https://arxiv.org/abs/2605.08526

作者：Zihan Huang,Junda Wu,Tong Yu,Qianqi Yan,Rohan Surana,Uttaran Bhattacharya,Lina Yao,Xin Eric Wang,Julian McAuley

【114】FlashEvolve: Accelerating Agent Self-Evolution with Asynchronous Stage Orchestration
标题：Flash Evolve：通过同步阶段规划加速代理自我进化
链接：https://arxiv.org/abs/2605.08520

作者：Zhengding Hu,Mingge Lu,Zhen Wang,Jixuan Ruan,Chang Chen,Zaifeng Pan,Yue Guan,Ruiyi Wang,Zhongkai Yu,Chao Zhang,Yufei Ding

【115】ShifaMind: A Multiplicative Concept Bottleneck for Interpretable ICD-10 Coding
标题：ShifaMind：可解释ICD-10编码的相乘概念瓶颈
链接：https://arxiv.org/abs/2605.08482

作者：Mohammed Sameer Syed,Xuan Lu

【116】Neurally-plausible radial basis kernels using distributed Fourier embeddings
标题：使用分布式傅里叶嵌入的神经合理的放射基核
链接：https://arxiv.org/abs/2605.08458

作者：Jakeb Chouinard

【117】Recovering Physical Dynamics from Discrete Observations via Intrinsic Differential Consistency
标题：通过内在差异一致性从离散观测恢复物理动力学
链接：https://arxiv.org/abs/2605.08454

作者：Yuxiang Luo,Andrew Perrault

【118】Sink vs. diagonal patterns as mechanisms for attention switch and oversmoothing prevention
标题：下沉模式与对角线模式作为注意力切换和过度平滑预防机制
链接：https://arxiv.org/abs/2605.08453

作者：Peter Súkeník,Cristina López Amado,Christoph H. Lampert,Marco Mondelli

【119】RubiConv -- Efficient Boundary-Respecting Convolutions
标题：RubiConv --有效的尊重边界的卷积
链接：https://arxiv.org/abs/2605.08451

作者：Linda Friso,Annie Marsden,Xinyi Chen,Arushi Gupta,Peter Bartlett,Mark Braverman,Elad Hazan
备注：19 pages, 12 figures

【120】Direct Bethe Free Energy Minimization for Bayesian Neural Ne twork
标题：Bayesian神经网络的直接Bethe自由能最小化
链接：https://arxiv.org/abs/2605.08446

作者：Pavel Prochazka
备注：Submited to conference

【121】The Attacker in the Mirror: Breaking Self-Consistency in Safety via Anchored Bipolicy Self-Play
标题：镜子中的攻击者：通过锚定双政策自我游戏打破安全上的自我一致性
链接：https://arxiv.org/abs/2605.08427

作者：Gabriele La Malfa,Emanuele La Malfa,Saar Cohen,Jie M. Zhang,Michael Luck,Michael Wooldridge,Elizabeth Black

【122】Generalized Wasserstein Flow Matching: Transport Plans, Everywhere, All at Once
标题：广义Wasserstein流量匹配：交通计划，无处不在，同时进行
链接：https://arxiv.org/abs/2605.08424

作者：Moritz Piening,Richard Duong,Gabriele Steidl

【123】Queryable LoRA: Instruction-Regularized Routing Over Shared Low-Rank Update Atoms
标题：可查询LoRA：共享低等级更新原子上的指令正规化路由
链接：https://arxiv.org/abs/2605.08423

作者：Omatharv Bharat Vaidya,Connor T. Jerzak,Nhat Ho,Chandrajit Bajaj

【124】Central Limit Theorem for Two-Time-Scale Approximate Distributionally Robust RL
标题：两时间尺度逼近分布鲁棒RL的中心极限定理
链接：https://arxiv.org/abs/2605.08417

作者：Shengbo Wang,Zexi Zhang

【125】Alignment as Jurisprudence
标题：作为法理学的一致
链接：https://arxiv.org/abs/2605.08416

作者：Nicholas Caputo

【126】Exploring and Exploiting Stability in Latent Flow Matching
标题：探索和利用潜在流匹配中的稳定性
链接：https://arxiv.org/abs/2605.08398

作者：Rania Briq,Michael Kamp,Ohad Fried,Sarel Cohen,Stefan Kesselheim
备注：Accepted at ICML 2026

【127】The Power of Second Order Methods for Sequence Preconditioning
标题：序列预处理二阶方法的力量
链接：https://arxiv.org/abs/2605.08390

作者：Annie Marsden,Elad Hazan
备注：14 pages, 5 figures

【128】Embedding Dimension Lower Bounds for Universality of Deep Sets and Janossy Pooling
标题：深集普适性的嵌入维下界和Janossy Pooling
链接：https://arxiv.org/abs/2605.08377

作者：Ali Syed,Aditya Nambiar,Jonathan W. Siegel

【129】On Distinguishing Capability Elicitation from Capability Creation in Post-Training: A Free-Energy Perspective
标题：关于区分后训练中的能力激发与能力创造：自由能源的视角
链接：https://arxiv.org/abs/2605.08368

作者：Yuhao Li,Shengchao Liu

【130】SWE Atlas: Benchmarking Coding Agents Beyond Issue Resolution
标题：SWE Atlas：超越问题解决的编码代理基准
链接：https://arxiv.org/abs/2605.08366

作者：Mohit Raghavendra,Soham Dan,Miguel Romero Calvo,Yannis Yiming He,Johannes Baptist Mols,Gautam Anand,Cole McCollum,Edgar Arakelyan,Vijay Bharadwaj,Andrew Park,Jeff Da,MohammadHossein Rezaei,Bing Liu,Brad Kenstler,Yunzhong He
备注：10 pages

【131】What Time Is It? How Data Geometry Makes Time Conditioning Optional for Flow Matching
标题：几点了？数据如何使几何时间调节可用于流量匹配
链接：https://arxiv.org/abs/2605.08344

作者：Alec Helbling,Sebastian Gutierrez Hernandez,Benjamin Hoover,Duen Horng Chau,Parikshit Ram

【132】P-Flow: Proxy-gradient Flows for Linear Inverse Problems
标题：P-Flow：线性反问题的代理梯度流
链接：https://arxiv.org/abs/2605.08328

作者：Zehua Jiang,Fenghao Zhu,Xinquan Wang,Chongwen Huang,Zhaoyang Zhang

【133】The Reciprocity Gradient
标题：互易梯度
链接：https://arxiv.org/abs/2605.08323

作者：Yue Lin,Pascal Poupart,Shuhui Zhu,Dan Qiao,Wenhao Li,Yuan Liu,Hongyuan Zha,Baoxiang Wang

【134】RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache
标题：RDKV：用于联合驱逐和量化的速率失真位分配
链接：https://arxiv.org/abs/2605.08317

作者：Junkai Zhang,Hang Guo,Luca Benini,Yawei Li

【135】SGC-RML: A reliable and interpretable longitudinal assessment for PD in real-world DNS
标题：SGC-RML：现实世界DNS中PD的可靠且可解释的纵向评估
链接：https://arxiv.org/abs/2605.08302

作者：Wenbin Wei,Ruixiang Gao,Suyuan Yao,Xuanzhen Zhao,Cheng Huang,Hen-Wei Huang
备注：Preprint. The first five authors contributed equally. Corresponding author: Hen-Wei Huang. 9 pages main text + appendix; 4 figures, 5 tables in main text

【136】Multi-Armed Bandits With Best-Action Queries
标题：拥有最佳动作按钮的多臂强盗
链接：https://arxiv.org/abs/2605.08287

作者：Francesco Bacchiocchi,Matteo Castiglioni,Alberto Marchesi,Francesco Emanuele Stradi

【137】Diagnosing Spectral Ceilings in Equivariant Neural Force Fields
标题：诊断等变神经力场中的光谱Cephase
链接：https://arxiv.org/abs/2605.08286

作者：Hyunmog Kim

【138】Exactness Matters for Physical Rule Enforcement
标题：严格执行物理规则很重要
链接：https://arxiv.org/abs/2605.08285

作者：Bum Jun Kim
备注：28 pages, 6 figures

【139】Trapping Attacker in Dilemma: Examining Internal Correlations and External Influences of Trigger for Defending GNN Backdoors
标题：陷入困境的攻击者：审视捍卫GNN后门触发器的内部相关性和外部影响
链接：https://arxiv.org/abs/2605.08278

作者：Fan Yang,Binyan Xu,Di Tang,Kehuan Zhang

【140】GPU-Accelerated Synthesis of Mixed-Boolean Arithmetic: Beyond Caching
标题：混合布尔算法的混合布尔算法的混合合成：超越缓存
链接：https://arxiv.org/abs/2605.08243

作者：Gabriel Bathie,Baptiste Mouillon,Nathanaël Fijalkow

【141】Distributional Spectral Diagnostics for Localizing Grokking Transitions
标题：局部化Grokking转变的分布光谱诊断
链接：https://arxiv.org/abs/2605.08237

作者：Ziyue Wang,Yufeng Ying,Takafumi Kanamori

【142】When Does Value-Aware KV Eviction Help? A Fixed-Contract Diagnostic for Non-Monotone Cache Compression
标题：意识价值的KV驱逐何时有所帮助？非单调缓存压缩的固定合同诊断
链接：https://arxiv.org/abs/2605.08234

作者：Ruijie Zhang,Haozhe Liang,Da Chang,Li Hu,Fanqi Kong,Huaxiao Yin,Yu Li

【143】TRAM: Training Approximate Multiplier Structures for Low-Power AI Accelerators
标题：TRAM：训练低功耗人工智能加速器的近似乘数结构
链接：https://arxiv.org/abs/2605.08231

作者：Chang Meng,Hanyu Wang,Yuyang Ye,Mingfei Yu,Wayne Burleson,Giovanni De Micheli

【144】FairHealth: An Open-Source Python Library for Trustworthy Healthcare AI in Low-Resource Settings
标题：FairHealth：一个开源Python库，用于低资源环境中值得信赖的医疗保健人工智能
链接：https://arxiv.org/abs/2605.08198

作者：Farjana Yesmin
备注：8 pages, open-source Python library

【145】ReplaySCM: A Benchmark for Executable Causal Mechanism Induction from Interventions
标题：ReplaySCP：干预诱导可执行因果机制的基准
链接：https://arxiv.org/abs/2605.08197

作者：Serafim Batzoglou

【146】NeurIPS Should Require Reproducibility Standards for Frontier AI Safety Claims
标题：NeurIPS应要求前沿人工智能安全声明的再现性标准
链接：https://arxiv.org/abs/2605.08192

作者：Varad Vishwarupe,Nigel Shadbolt,Marina Jirotka,Ivan Flechais
备注：Preprint

【147】Synergistic Simplex: Cooperative Runtime Assurance for Safety-Critical Autonomous Systems
标题：协同简单：安全关键自治系统的合作运行时间保证
链接：https://arxiv.org/abs/2605.08190

作者：Ayoosh Bansal,Mikael Yeghiazaryan,Artyom Khachatryan,Tianyi Zhu,Hunmin Kim,Naira Hovakimyan,Lui Sha

【148】Sparsity Hurts: Simple Linear Adapter Can Boost Generalized Category Discovery
标题：稀疏性伤害：简单的线性适配器可以促进广义类别发现
链接：https://arxiv.org/abs/2605.08183

作者：Bo Ye,Kai Gan,Tong Wei,Min-Ling Zhang
备注：Submitted to IEEE TPAMI

【149】Information Density as a Quantitative Measure for AI-enabled Virtual Sensing: Feasibility and Limits
标题：信息密度作为人工智能虚拟感知的量化指标：可行性和局限性
链接：https://arxiv.org/abs/2605.08180

作者：Hrishikesh Dutta,Roberto Minerva,Reza Farahbakhsh,Noel Crespi
备注：IEEE Transactions on Sustainable Computing (2026)

【150】Quantitative Sobolev Approximation Bounds for Neural Operators with Empirical Validation on Burgers Equation
标题：神经运算符的定量Sobolev逼近界及其Burgers方程的经验验证
链接：https://arxiv.org/abs/2605.08170

作者：Nicole Hao

【151】Temporal-Decay Shapley: A Time-Aware Data Valuation Framework for Time-Series Data
标题：时间衰变Shapley：时间序列数据的时间感知数据估值框架
链接：https://arxiv.org/abs/2605.08153

作者：Chuwen Pang,Bing Mi,Kongyang Chen

【152】DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis
标题：DataArc-SynData-Tools：用于多路径、多模式和多语言数据合成的统一闭环框架
链接：https://arxiv.org/abs/2605.08138

作者：Zhichao Shi,Cehao Yang,Hao Zhou,Xiaojun Wu,Huajie Li,Xuhui Jiang,Chengjin Xu,Yuanzhuo Wang,Jian Guo
备注：6 pages

【153】Towards Customized Multimodal Role-Play
标题：迈向定制多模式角色扮演
链接：https://arxiv.org/abs/2605.08129

作者：Chao Tang,Jianzong Wu,Qingyu Shi,Ye Tian,Aixi Zhang,Hao Jiang,Jiangning Zhang,Yunhai Tong
备注：Code available at https://github.com/Tangc03/UniCharacter Project page available at https://tangc03.github.io/UniCharacter.github.io/

【154】Block-Wise Differentiable Sinkhorn Attention: Tail-Refinement Gradients with a Gap-Aware Dustbin Bridge
标题：区块差异化Sinkhorn注意力：尾部细化与缝隙感知垃圾箱桥相结合
链接：https://arxiv.org/abs/2605.08123

作者：Dylan Forde

【155】GONE: Structural Knowledge Unlearning via Neighborhood-Expanded Distribution Shaping
标题：GONE：通过邻居扩展分布塑造消除结构知识
链接：https://arxiv.org/abs/2603.12275

作者：Chahana Dahal,Ashutosh Balasubramaniam,Zuobin Xiong

【156】Factual recall in linear associative memories: sharp asymptotics and mechanistic insights
标题：线性联想记忆中的事实回忆：敏锐的渐进学和机械学见解
链接：https://arxiv.org/abs/2605.10795

作者：Alessio Giorlandino,Sebastian Goldt,Antoine Maillard

【157】Price of Quality: Sufficient Conditions for Sparse Recovery using Mixed-Quality Data
标题：质量代价：使用混合质量数据进行稀疏恢复的充分条件
链接：https://arxiv.org/abs/2605.10713

作者：Youssef Chaabouni,David Gamarnik
备注：Published as a conference paper at ICLR 2026

【158】Exact Fixed-Point Constraints in Neural-ODEs with Provable Universality
标题：具有可证明普遍性的神经元ODE中的精确定点约束
链接：https://arxiv.org/abs/2605.10613

作者：Feliciano Giuseppe Pacifico,Duccio Fanelli,Lorenzo Buffoni,Lorenzo Chicchi,Diego Febbe,Raffaele Marino
备注：15 pages, 3 figures

【159】Affine Tracing: A New Paradigm for Probabilistic Linear Solvers
标题：仿射追踪：概率线性求解器的新范式
链接：https://arxiv.org/abs/2605.10566

作者：Disha Hegde,Marvin Pförtner,Jon Cockayne

【160】Multifidelity Gaussian process regression for solving nonlinear partial differential equations
标题：求解非线性偏微方程的多保真度高斯过程回归
链接：https://arxiv.org/abs/2605.10383

作者：Fatima-Zahrae El-Boukkouri,Josselin Garnier,Olivier Roustant
备注：31 pages, 20 figures

【161】Characterizing the Generalization Error of Random Feature Regression with Arbitrary Data-Augmentation
标题：用任意数据扩充来描述随机特征回归的推广误差
链接：https://arxiv.org/abs/2605.10290

作者：Lucas Morisset,Alain Durmus,Adrien Hardy

【162】Stellar Age Compression Reshapes Interpretations of the Milky Way Thick-Disk Formation History
标题：恒星时代压缩重塑银河系厚盘形成历史的解释
链接：https://arxiv.org/abs/2605.10220

作者：Zhipeng Zhang

【163】Parameterized Complexity of Stationarity Testing for Piecewise-Affine Functions and Shallow CNN Losses
标题：分段仿射函数和浅层CNN损失平稳性测试的参数化复杂性
链接：https://arxiv.org/abs/2605.10219

作者：Yuhan Ye
备注：32 pages, 1 figure, 1 table

【164】Joint sparse coding and temporal dynamics support context reconfiguration
标题：联合稀疏编码和时间动态支持上下文重新配置
链接：https://arxiv.org/abs/2605.10178

作者：Qianqian Shi,Yue Che,Faqiang Liu,Hongyi Li,Mingkun Xu,Sandra Reinert,Pieter M. Goltstein,Rong Zhao,Luping Shi
备注：37 pages, 6 figures, 6 extended data figures. Preprint version

【165】A Stability Benchmark of Generative Regularizers for Inverse Problems
标题：反问题生成正则化器的稳定性基准
链接：https://arxiv.org/abs/2605.10076

作者：Alexander Denker,Johannes Hertrich,Sebastian Neumayer

【166】Differentially Private Sampling from Distributions via Wasserstein Projection
标题：通过Wasserstein投影从分布中进行差异私人抽样
链接：https://arxiv.org/abs/2605.10015

作者：Shokichi Takakura,Seng Pei Liew,Satoshi Hasegawa

【167】Total Generalized Variation regularization closes the gap between neural-eld and classical methods in seismic travel-time tomography
标题：总广义变分正规化缩小了地震走时断层扫描中神经场方法与经典方法之间的差距
链接：https://arxiv.org/abs/2605.09960

作者：Isao Kurosawa
备注：15 pages, 6 figures. Manuscript submitted to Geophysical Journal International

【168】The Observable Wasserstein Distance
标题：可观测的沃瑟斯坦距离
链接：https://arxiv.org/abs/2605.09916

作者：Edivaldo Lopes dos Santos,Leandro Vicente Mauri,Washington Mio,Tom Needham

【169】Dissecting Jet-Tagger Through Mechanistic Interpretability
标题：通过机械解释剖析喷气式飞机标签
链接：https://arxiv.org/abs/2605.09881

作者：Saurabh Rai,Sanmay Ganguly
备注：40 pages, 14 figures, 12 tables. Comments are welcome

【170】A Real-Calibrated Synthetic-First Data Engine
标题：真实校准的综合优先数据引擎
链接：https://arxiv.org/abs/2605.09699

作者：Yukang Shen
备注：7 pages, 6 figures

【171】Phases of Muon: When Muon Eclipses SignSGD
标题：Muon的阶段：当Muon宣布SignSingapore时
链接：https://arxiv.org/abs/2605.09552

作者：Elliot Paquette,Noah Marshall,Lucas Benigni,Guangyuan Wang,Atish Agarwala,Courtney Paquette

【172】Empirical Bayes 1-bit matrix completion
标题：经验Bayes 1位矩阵完成
链接：https://arxiv.org/abs/2605.09509

作者：Takeru Matsuda

【173】Enabling Structure-Only Initialization and Out-of-Distribution Generalization in GNN-based Molecular Dynamics Simulators
标题：在基于GNN的分子动力学模拟器中实现纯结构的扩展和非分布概括
链接：https://arxiv.org/abs/2605.09495

作者：S. A. Shteingolts,Salman N. Salman,Dan Mendels
备注：10 pages, 7 figures

【174】Measuring and Decomposing Mode Separation via the Canonical Diffusion
标题：通过典型扩散测量和分解模式分离
链接：https://arxiv.org/abs/2605.08777

作者：Shaul Tolkovsky,Ori Meidler,Or Zuk

【175】Core-Halo Decomposition: Decentralizing Large-Scale Fixed-Point Problems
标题：核心晕分解：分散化大规模不动点问题
链接：https://arxiv.org/abs/2605.08681

作者：Haixiang,Yang Xu,Jiefu Zhang,Xudong Wu,Zihan Zhou,Jun He,Jiayu Chen

【176】Structure-Preserving Reconstruction of Convex Lipschitz Functionals on Hilbert Spaces from Finite Samples
标题：由有限样本重建Hilbert空间上凸Lipschitz函式的保结构重建
链接：https://arxiv.org/abs/2605.08559

作者：Anastasis Kratsios

【177】Learnability and Competition in High-Dimensional Multi-Component ICA
标题：多维多分量ICA的可学习性和竞争性
链接：https://arxiv.org/abs/2605.08552

作者：Eser Ilke Genc,Samet Demir,Zafer Dogan
备注：56 pages, 9 figures

【178】Sliced Inner Product Gromov-Wasserstein Distances
标题：切片内积Gromovov-Wasserstein距离
链接：https://arxiv.org/abs/2605.08546

作者：Xiaoyun Gong,Gabriel Rioux,Ziv Goldfeld
备注：49 pages, 8 figures

【179】A Unified Lyapunov-IQC Framework for Uniform Stability of Smooth Quadratic First-Order Accelerated Optimizers
标题：光滑二次一阶加速优化器一致稳定性的统一Lyapunov-IQC框架
链接：https://arxiv.org/abs/2605.08488

作者：Don Li,Dacian Daescu

【180】CAMAL: Improving Attention Alignment and Faithfulness with Segmentation Masks
标题：CAMAL：用分割口罩提高注意力一致性和忠诚度
链接：https://arxiv.org/abs/2605.08325

作者：Rajdeep Singh Hundal,Yan Xiao,Jin Song Dong,Manuel Rigger

【181】Non-intrusive Body Composition Assessment from Full-body mmWave Scans
标题：通过全身毫米波扫描进行非侵入性身体成分评估
链接：https://arxiv.org/abs/2605.08306

作者：Miriam Senne,Benjamin D. Killeen,Tony Wang,Nassir Navab

【182】Inverse Design of Multi-Layer Sub-Pixel-Resolution RF Passives Through Grayscale Diffusion with Flexible S-Parameter Conditioning
标题：通过具有灵活S参数条件的灰度扩散进行多层亚像素分辨率RF无源的逆设计
链接：https://arxiv.org/abs/2605.08233

作者：Tommaso Dreossi,Christopher M. Bryant,Hao Liu,Nathan Mirman,Noah Kessler,Michael Frei,Harish Krishnaswamy

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