Talking Heads: Understanding Inter-layer Communication in Transformer Language Models
机构 * Department of Computer Science（计算机科学系） ； Brown University（布朗大学） ； School of Medicine（医学院） ； University of Tübingen（图宾根大学）
Comments Neurips 2024
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
Pretraining with Random Noise for Fast and Robust Learning without Weight Transport
机构 * Department of Brain and Cognitive Sciences（脑科学与认知科学系） ； Graduate School of Data Science（数据科学研究生院） ； Kim Jaechul Graduate School of AI（金在哲人工智能研究生院）
Journal ref Advances in Neural Information Processing Systems 37, 13748-13768, 2024
Active learning of neural population dynamics using two-photon holographic optogenetics
机构 * University of California, Berkeley（加州大学伯克利分校） ； Georgia Tech（佐治亚理工学院） ； Allen Institute for Neural Dynamics（神经动态研究所） ； Allen Institute for Brain Science（脑科学研究所） ； University of Washington（华盛顿大学）
Comments NeurIPS 2024
Learning from Convolution-based Unlearnable Datasets
机构 * University of Maryland, College Park（马里兰大学 College Park分校）
Comments Accepted to AdvML-Frontiers Workshop at NeurIPS 2024. Code available at https://github.com/aseriesof-tubes/RSK
SceneCraft: Layout-Guided 3D Scene Generation
机构 * Shanghai Jiao Tong University（上海交通大学） ； University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Comments NeurIPS 2024. Code: https://github.com/OrangeSodahub/SceneCraft Project Page: https://orangesodahub.github.io/SceneCraft
Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding
机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） ； Carnegie Mellon University（卡内基梅隆大学）
Comments NeurIPS 2024. Project page: https://yunzeman.github.io/lexicon3d Github: https://github.com/YunzeMan/Lexicon3D
Noise-Aware Differentially Private Regression via Meta-Learning
机构 * University of Helsinki（赫尔辛基大学） ； University of Cambridge（剑桥大学） ； Microsoft Research AI for Science（微软研究院人工智能科学部）
Comments NeurIPS 2024
Rejection via Learning Density Ratios
机构 * Amazon（亚马逊公司） ； The Australian National University（澳大利亚国立大学）
Comments Compared to published version, a typo in Appendix Section E has been fixed
Journal ref Published at The Thirty-eighth Annual Conference on Neural Information Processing Systems 2024
Queueing Matching Bandits with Preference Feedback
Comments NeurIPS 2024
Optimal Transport for Treatment Effect Estimation
机构 * Department of XXX, University of YYY, Location, Country（XXX系，YYYY大学，地点，国家） ； School of ZZZ, Institute of WWW, Location, Country（ZZZ学院，WWW研究所，地点，国家）
Comments Accepted as NeurIPS 2023 Poster
Efficient Vocabulary-Free Fine-Grained Visual Recognition in the Age of Multimodal LLMs
Comments preprint; earlier version accepted at NeurIPS 2024 Workshop on Adaptive Foundation Models
DivShift: Exploring Domain-Specific Distribution Shifts in Large-Scale, Volunteer-Collected Biodiversity Datasets
Comments Published at AAAI-25 AI for Social Impact Track (https://ojs.aaai.org/index.php/AAAI/article/view/35060) Presented at NeurIPS 2024 Workshop on Tackling Climate Change with Machine Learning (https://www.climatechange.ai/papers/neurips2024/43)
Journal ref Proceedings of the AAAI Conference on Artificial Intelligence 39(27) 28386-28396 (2025)
Deep Reinforcement Learning Policies for Underactuated Satellite Attitude Control
机构 * Argotec srl（Argotec公司） ； IDSIA, Università della Svizzera italiana（IDSIA，瑞士大学）
Comments Originally presented at the NeurIPS 2021 Workshop on Deployable Decision-Making in Embodied Systems
Journal ref NeurIPS 2021 Workshop on Deployable Decision-Making in Embodied Systems
SustainDC: Benchmarking for Sustainable Data Center Control
Comments Accepted at Advances in Neural Information Processing Systems 2024 (NeurIPS 2024)
Journal ref Advances in Neural Information Processing Systems 37 (NeurIPS 2024)
How to Solve Contextual Goal-Oriented Problems with Offline Datasets?
机构 * University of Wisconsin-Madison（威斯康星大学麦迪逊分校） ； ByteDance Research（字节跳动研究） ； Microsoft Research（微软研究院）
Comments NeurIPS 2024
Performativity and Prospective Fairness
Comments Presented at NeurIPS 2023 Workshop "Algorithmic Fairness through the Lens of Time''
An exactly solvable model for emergence and scaling laws in the multitask sparse parity problem
机构 * Rudolf Peierls Centre for Theoretical Physics, University of Oxford（鲁道夫·皮尔士理论物理中心，牛津大学） ； Center for Quantum Structures in Modules and Spaces, Seoul National University（模块与空间中的量子结构中心，首尔国立大学）
Comments Accepted at NeurIPS 2024 Conference
Mitigating Bias in Facial Recognition Systems: Centroid Fairness Loss Optimization
机构 * LTCI, Télécom Paris Institut Polytechnique de Paris（电信巴黎理工学院）
Comments Accepted at both the AFME and RegML Workshops at NeurIPS 2024. A preliminary version has been accepted for publication by Springer Nature, in the context of the ICPR 2024 conference
Leveraging Multi-Modal Saliency and Fusion for Gaze Target Detection
机构 * Elm Company（埃尔姆公司）
Comments accepted at NeurIPS 2023 Gaze Meets ML Workshop
Building a stable classifier with the inflated argmax
机构 * Department of Statistics, University of Chicago（统计学系，芝加哥大学） ； Department of Computer Science, University of Chicago（计算机科学系，芝加哥大学）
Comments NeurIPS 2024
Nearest Neighbor Speculative Decoding for LLM Generation and Attribution
机构 * Cohere ； Meta FAIR ； University of Chicago（芝加哥大学） ； Carnegie Mellon University（卡内基梅隆大学） ； University of Waterloo（滑铁卢大学）
Journal ref Advances in Neural Information Processing Systems (2024), vol. 37, page 80987-81015
Multiple-Instance, Cascaded Classification for Keyword Spotting in Narrow-Band Audio
机构 * Voicera
Comments Published in the proceedings of NeurIPS 2017 Workshop: Machine Learning on the Phone and other Consumer Devices
Diff4Steer: Steerable Diffusion Prior for Generative Music Retrieval with Semantic Guidance
机构 * University of Toronto（多伦多大学） ； Google Research（谷歌研究） ； Google DeepMind（谷歌DeepMind） ； Seoul National University（首尔国立大学）
Comments NeurIPS 2024 Creative AI Track
Journal ref Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2025
Invariant Learning with Annotation-free Environments
机构 * University of Marburg（马尔堡大学） ； University of Mannheim（曼海姆大学）
Comments Accepted at NeurIPS 2024 Workshop UniReps
Listenable Maps for Zero-Shot Audio Classifiers
机构 * Fondazione Bruno Kessler（布鲁诺·科塞拉基金会） ； Mila, Québec AI Institute（魁北克人工智能研究所） ； Concordia University（康科德大学） ； Université Laval（拉瓦尔大学）
Comments Accepted to NeurIPS 2024
Efficient Document Retrieval with G-Retriever
机构 * University of Stuttgart（斯图加特大学）
Comments Extended version of a paper presented at NeurIPS 2024 (arXiv:2402.07630)
PK-YOLO: Pretrained Knowledge Guided YOLO for Brain Tumor Detection in Multiplanar MRI Slices
Comments References updated; for example, papers in NeurIPS 2024 proceedings appeared on 6 Feb 2025 and AAAI 2025 one on 11 Apr 2025
Journal ref In WACV (2025) 3732--3741
Dynamic Context Adaptation and Information Flow Control in Transformers: Introducing the Evaluator Adjuster Unit and Gated Residual Connections
机构 * Brain Corporation（Brain公司）
Comments 10 pages, 2 figures, 4 experiments
Journal ref NeurIPS 2024 Submission
On Tractable $Φ$-Equilibria in Non-Concave Games
机构 * Yale University（耶鲁大学） ； MIT CSAIL（麻省理工学院计算机科学与人工智能实验室） ； University of Southern California（南加州大学） ； University of Virginia（弗吉尼亚大学）
Comments 59 pages. The abstract has been shortened to meet the arXiv requirement. A preliminary version of the paper has been accepted to NeurIPS 2024. Compared to the last version, this version contains updated references
Recognize Any Regions
机构 * University of Surrey（萨里大学） ； The University of Hong Kong（香港大学） ； ByteDance（字节跳动）
Comments NeurIPS 2024. Github: https://github.com/Surrey-UPLab/Recognize-Any-Regions
