Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions
机构 * Imperial College London（伦敦帝国学院） ； Canon Inc.（佳能公司）
Comments NeurIPS 2025 (poster). Project page: https://nkiyohara.github.io/nsf-neurips2025/
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
Mechanistic Interpretability of RNNs emulating Hidden Markov Models
机构 * Institute of Neuroinformatics, University of Zurich & ETH Zurich（神经信息学研究所，苏黎世大学及苏黎世联邦理工学院）
Comments Accepted at NeurIPS 2025
Spectral Perturbation Bounds for Low-Rank Approximation with Applications to Privacy
机构 * Yale University（耶鲁大学）
Comments NeurIPS 2025
Subgraph Federated Learning via Spectral Methods
Comments To be presented at The Annual Conference on Neural Information Processing Systems (NeurIPS) 2025
Perturbation Bounds for Low-Rank Inverse Approximations under Noise
机构 * Yale University（耶鲁大学）
Comments NeurIPS 2025
Gradient-Weight Alignment as a Train-Time Proxy for Generalization in Classification Tasks
机构 * Institute for Artifical Intelligence in Medicine（人工智能医学研究所） ； Technical University of Munich（慕尼黑技术大学）
Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Curiosity-driven RL for symbolic equation solving
机构 * Starling Research Institute（星灵研究机构）
Comments Accepted at the NeurIPS 2025 MATH-AI Workshop
SimulMEGA: MoE Routers are Advanced Policy Makers for Simultaneous Speech Translation
机构 * Auditory Cognition and Computational Acoustics Lab（听觉认知与计算声学实验室） ； MoE Key Lab of Artificial Intelligence（MoE人工智能重点实验室） ； AI Institute（人工智能研究院） ； School of Computer Science, Shanghai Jiao Tong University（上海交通大学计算机科学学院）
Comments NeurIPS 2025 poster
Steering Information Utility in Key-Value Memory for Language Model Post-Training
机构 * Dept. of Computer Science Rice University（计算机科学系 瑞士大学）
Comments NeurIPS 2025
Many LLMs Are More Utilitarian Than One
机构 * Forward College（前进学院） ； University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） ； University of Nebraska, Lincoln（内布拉斯加大学林肯分校） ； Technische Universität Berlin（柏林技术大学） ； Humboldt Institute for Internet and Society（洪堡互联网与社会研究所） ； Stony Brook University（石溪大学）
Comments Accepted to the Conference on Neural Information Processing Systems (NeurIPS 2025)
FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering
机构 * Technical University of Berlin（柏林技术大学） ； Technical University of Munich（慕尼黑技术大学） ； CARIAD SE ； Volkswagen AG（大众集团）
Comments Accepted by NeurIPS 2025 - main track. Project page: https://focus-mllm-vqa.github.io/
HAIF-GS: Hierarchical and Induced Flow-Guided Gaussian Splatting for Dynamic Scene
机构 * Institute of Computing Technology, Chinese Academy of Sciences, ICT（中国科学院计算技术研究所） ； University of Chinese Academy of Sciences, UCAS（中国科学院大学） ； The University of Queensland（昆士兰大学） ； Jiangsu University（江苏大学）
Comments Accepted to NeurIPS 2025. Project page: https://echopickle.github.io/HAIF-GS.github.io/
Reinforcement Learning Teachers of Test Time Scaling
机构 * Sakana AI, Japan（日本Sakana AI）
Comments Accepted at NeurIPS 2025
Who You Are Matters: Bridging Topics and Social Roles via LLM-Enhanced Logical Recommendation
机构 * Wuhan University（武汉大学） ； Kuaishou Technology（快手科技）
Comments to be published in NeurIPS 2025
Data-Juicer 2.0: Cloud-Scale Adaptive Data Processing for and with Foundation Models
机构 * Alibaba Group（阿里巴巴集团）
Comments Accepted by NeurIPS 2025 (Spotlight). 43 pages, 16 figures, 4 tables
Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation
机构 * Shanghai Jiao Tong University（上海交通大学） ； Huawei Inc.（华为公司）
Comments Accepted by NeurIPS 2025
U-CAN: Unsupervised Point Cloud Denoising with Consistency-Aware Noise2Noise Matching
Comments Accepted by NeurIPS 2025. Project page: https://gloriasze.github.io/U-CAN/
Selective Learning for Deep Time Series Forecasting
Comments Accepted by NeurIPS 2025
Test-Time Adaptive Object Detection with Foundation Model
机构 * State Key Laboratory of Complex and Critical Software Environment, Beihang University（复杂与关键软件环境国家重点实验室，北京航空航天大学） ； School of Computer Science and Engineering, Beihang University（北京航空航天大学计算机科学与工程学院） ； School of Computer Science and Information Engineering, Hefei University of Technology（合肥工业大学计算机科学与信息工程学院）
Comments Accepted by NeurIPS 2025
EA3D: Online Open-World 3D Object Extraction from Streaming Videos
机构 * Wangxuan Institute of Computer Technology, Peking University（北京大学王轩计算机技术研究所） ； Google DeepMind（谷歌DeepMind） ； University of California, Merced（加州大学梅尔塞德斯分校）
Comments The Thirty-Ninth Annual Conference on Neural Information Processing Systems(NeurIPS 2025)
AtlasGS: Atlanta-world Guided Surface Reconstruction with Implicit Structured Gaussians
机构 * State Key Lab of CAD & CG, Zhejiang University（计算机辅助设计与图形学国家重点实验室，浙江大学）
Comments 18 pages, 11 figures. NeurIPS 2025; Project page: https://zju3dv.github.io/AtlasGS/
Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection
机构 * Korea University（韩国大学） ； Korea Advanced Institute of Science and Technology（韩国科学技术院）
Comments Accepted by NeurIPS 2025
Language Model Behavioral Phases are Consistent Across Architecture, Training Data, and Scale
机构 * Department of Brain and Cognitive Sciences, MIT（麻省理工学院脑科学与认知科学系） ； MIT Libraries CREOS（麻省理工学院图书馆 CREOS） ； Deparmtent of Cognitive Science, UCSD（加州大学圣地亚哥分校认知科学系）
Comments To be presented at NeurIPS 2025
Disaggregation Reveals Hidden Training Dynamics: The Case of Agreement Attraction
机构 * MIT（麻省理工学院） ； EleutherAI
Comments Accepted to the First Workshop on Interpreting Cognition in Deep Learning Models (CogInterp @ NeurIPS 2025)
KAN-GCN: Combining Kolmogorov-Arnold Network with Graph Convolution Network for an Accurate Ice Sheet Emulator
Comments Accept for NeurIPS 2025 Workshop: New Perspectives in Graph Machine Learning
Aggregation Hides Out-of-Distribution Generalization Failures from Spurious Correlations
机构 * Massachusetts Institute of Technology（麻省理工学院）
Comments Accepted as a Spotlight paper at NeurIPS 2025
Augmenting Biological Fitness Prediction Benchmarks with Landscapes Features from GraphFLA
机构 * School of Computer Science and Engineering, University of Electronic Science and Technology of China（电子科技大学计算机科学与工程学院） ； Department of Computer Science, University of Exeter（埃克塞特大学计算机科学系）
Comments 56 apges, 18 figures, 8 tables, accepted as a conference paper at NeurIPS 2025
From Linear to Nonlinear: Provable Weak-to-Strong Generalization through Feature Learning
机构 * KAIST（韩国科学技术院）
Comments NeurIPS 2025 camera-ready version, 70 pages
ProofSketch: Efficient Verified Reasoning for Large Language Models
机构 * Manipal University Jaipur（马普尔大学斋普尔）
Comments Accepted at NeurIPS 2025, ER Workshop
The Underappreciated Power of Vision Models for Graph Structural Understanding
机构 * School of Data Science, The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳）数据科学学院） ； Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所） ； Cheriton School of Computer Science, University of Waterloo（滑铁卢大学切尔顿计算机科学学院）
Comments NeurIPS 2025
