DuoGPT: Training-free Dual Sparsity through Activation-aware Pruning in LLMs
机构 * Yale University（耶鲁大学）
Comments NeurIPS 2025. The code is available on Github (see hyperlink in the paper)
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
Caption This, Reason That: VLMs Caught in the Middle
机构 * Integrated Program in Neuroscience (IPN) McGill University（神经科学联合计划 麦吉尔大学） ； Mila, University of Montreal（蒙特利尔大学Mila） ； Microsoft Research USA（微软研究院美国总部） ； Department of Physiology McGill University（生理学系 麦吉尔大学）
Comments Paper accepted by nips 2025
Regression Trees Know Calculus
机构 * Department of Mathematics and Statistics（数学与统计学系） ； University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）
Comments Camera ready version for Neurips 2025
Robust and Diverse Multi-Agent Learning via Rational Policy Gradient
机构 * UC Berkeley（加州大学伯克利分校） ； Google Deepmind（谷歌DeepMind）
Comments Published at NeurIPS 2025
AdaptDel: Adaptable Deletion Rate Randomized Smoothing for Certified Robustness
Comments 33 pages, 7 figures, camera ready version for NeurIPS 2025
Compact Memory for Continual Logistic Regression
机构 * RIKEN Center for AI Project（RIKEN人工智能研究中心） ； Kookmin University（韩国庆熙大学） ； Imperial College London（伦敦帝国理工学院） ； KAIST（韩国科学技术院）
Journal ref NeurIPS 2025
Robust Sampling for Active Statistical Inference
机构 * Department of Statistics（统计学系） ； Stanford Data Science（斯坦福数据科学） ； Department of Mathematics（数学系） ； Stanford University（斯坦福大学）
Comments NeurIPS 2025
The Double Contingency Problem: AI Recursion and the Limits of Interspecies Understanding
机构 * UC San Diego, Synthesis Program（UC圣地亚哥大学，合成计划）
Comments 5 pages, no figures, to be published in the NeurIPS 2025: AI for Non-Human Animal Communication Workshop Proceedings
Re-envisioning Euclid Galaxy Morphology: Identifying and Interpreting Features with Sparse Autoencoders
机构 * Space Telescope Science Institute（太空望远镜科学研究所） ； Johns Hopkins University（约翰霍普金斯大学） ； University of Toronto（多伦多大学）
Comments Authors contributed equally to this work. Accepted to NeurIPS Machine Learning and the Physical Sciences Workshop. See trained model at https://huggingface.co/mwalmsley/euclid-rr2-mae, HuggingFace demo at https://huggingface.co/spaces/mwalmsley/euclid_masked_autoencoder, and code at https://github.com/jwuphysics/euclid-galaxy-morphology-saes
Heterogeneous Point Set Transformers for Segmentation of Multiple View Particle Detectors
机构 * NOvA Collaboration（NOvA合作组）
Comments Camera-ready version for Machine Learning and the Physical Sciences Workshop (ML4PS) at NeurIPS 2025
Vicinity-Guided Discriminative Latent Diffusion for Privacy-Preserving Domain Adaptation
Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Act to See, See to Act: Diffusion-Driven Perception-Action Interplay for Adaptive Policies
机构 * University of Alberta（阿尔伯塔大学） ； Huazhong University of Science and Technology（华中科技大学）
Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
P-DRUM: Post-hoc Descriptor-based Residual Uncertainty Modeling for Machine Learning Potentials
机构 * Massachusetts Institute of Technology（麻省理工学院） ； Preferred Networks
Comments Accepted to 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Machine Learning and the Physical Sciences (ML4PS)
Radio Astronomy in the Era of Vision-Language Models: Prompt Sensitivity and Adaptation
机构 * University of Geneva（日内瓦大学）
Comments Machine Learning and the Physical Sciences Workshop, NeurIPS 2025
The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?
Comments NeurIPS 2025 (Spotlight)
What Do Latent Action Models Actually Learn?
机构 * Microsoft Research（微软研究院） ； Tsinghua University（清华大学）
Comments Accepted by NeurIPS-25
Dynamic Diffusion Schrödinger Bridge in Astrophysical Observational Inversions
机构 * LIX, École Polytechnique, IP Paris（巴黎高等理工学院LIX研究所） ； Canadian Institute for Theoretical Astrophysics (CITA), University of Toronto（多伦多大学加拿大理论天体物理研究所） ； Google DeepMind（谷歌DeepMind） ； Department of Astronomy, University of Virginia（弗吉尼亚大学天文学系） ； Department of Space, Earth & Environment, Chalmers University of Technology（楚科夫斯基技术大学空间、地球与环境系）
Comments Accepted to NeurIPS 2025. Code available at https://github.com/L-YeZhu/AstroDSB. Updated funding information from last version
MoE-Gyro: Self-Supervised Over-Range Reconstruction and Denoising for MEMS Gyroscopes
机构 * Southeast University（东南大学） ； Harbin Institute of Technology（哈尔滨工业大学）
Comments Accepted to the NeurIPS 2025 Main Track
Military AI Needs Technically-Informed Regulation to Safeguard AI Research and its Applications
机构 * Harvard Medical School（哈佛医学院） ； Harvard University（哈佛大学） ； Kempner Institute（凯普纳研究所） ； U.S. Army War College（美国陆军战争学院） ； Kennedy School of Government（肯尼迪政府学院）
Comments Published at NeurIPS 2025, 10 pages, 2 tables, 1 figure
Trajectory Bellman Residual Minimization: A Simple Value-Based Method for LLM Reasoning
机构 * UW-Madison（威斯康星大学麦迪逊分校） ； MIT（麻省理工学院）
Comments NeurIPS 2025
Learning conformational ensembles of proteins based on backbone geometry
机构 * Max Planck Institute for Polymer Research（马克斯·普朗克聚合物研究所） ； Heidelberg Institute for Theoretical Studies（海德堡理论研究所） ； IWR, Heidelberg University（海德堡大学IWR中心） ； SciLifeLab and DBB at Stockholm University（斯德哥尔摩大学SciLifeLab和DBB中心） ； Karlsruhe Institute of Technology（卡尔斯鲁厄技术大学）
Comments To be published in proceedings of NeurIPS 2025
Contextual Thompson Sampling via Generation of Missing Data
Comments NeurIPS 2025
CHOICE: Benchmarking the Remote Sensing Capabilities of Large Vision-Language Models
机构 * State Key Lab. LIESMARS, Wuhan University（武汉大学遥感信息与地学实验室国家重点实验室）
Comments Accepted by NeurIPS 2025 Track on Datasets and Benchmarks
Mixture of Scope Experts at Test: Generalizing Deeper Graph Neural Networks with Shallow Variants
机构 * University of Southern California（南加州大学） ； DEVCOM ARL Army Research Office（DEVCOM ARL陆军研究办公室）
Comments NeurIPS 2025
Adaptive Data Analysis for Growing Data
Comments 28 pages, 6 figures, camera ready version for NeurIPS 2025. Addresses a bug in the computation of the JLNRSS and JLNRSS+ curves in Fig 2 of the previous version. Generalizes the results to non-uniform differential privacy
Symbolic Regression Is All You Need: From Simulations to Scaling Laws in Binary Neutron Star Mergers
Comments Accepted at Machine Learning and the Physical Sciences Workshop, NeurIPS 2025
Chord-conditioned Melody and Bass Generation
机构 * CUNY Graduate Center（CUNY研究生中心） ； Brooklyn College（布鲁克林学院）
Comments To appear at NeurIPS 2025 Workshop on AI for Music (AI4Music)
Data-driven Feynman-Kac Discovery with Applications to Prediction and Data Generation
Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Generative AI in Finance
Prompt-Based Safety Guidance Is Ineffective for Unlearned Text-to-Image Diffusion Models
机构 * KAIST（韩国科学技术院）
Comments Accepted at NeurIPS 2025 Workshop on Generative and Protective AI for Content Creation
Polar Sparsity: High Throughput Batched LLM Inferencing with Scalable Contextual Sparsity
机构 * Texas A&M University（德克萨斯大学） ； NVIDIA（英伟达） ； Lawrence Livermore National Laboratory（劳伦斯利弗莫尔国家实验室）
Comments NeurIPS 2025, 10 pages, 7 figures
