arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

International Conference on Learning Representations · 会议 · Machine Learning

共收录 9454
2504.06580 2025-12-08 cs.CV cs.AI

Exploring Ordinal Bias in Action Recognition for Instructional Videos

探索教学视频中动作识别的序数偏差

Joochan Kim, Minjoon Jung, Byoung-Tak Zhang

机构 * Korea Institute of Science and Technology(韩国科学技术院) Seoul National University(首尔国立大学)

AI总结 本文提出两种方法探索教学视频中动作识别的序数偏差问题,通过实验揭示模型在面对非常规动作序列时的脆弱性,强调了重新设计评估策略和开发更通用模型的必要性。

Comments Accepted at SCSL @ ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.02031 2025-12-08 cs.CV

Neural Eulerian Scene Flow Fields

神经欧拉场景流场

Kyle Vedder, Neehar Peri, Ishan Khatri, Siyi Li, Eric Eaton, Mehmet Kocamaz, Yue Wang, Zhiding Yu, Deva Ramanan, Joachim Pehserl

机构 * University of Pennsylvania(宾夕法尼亚大学) NVIDIA(英伟达) Carnegie Mellon University(卡内基梅隆大学)

AI总结 EulerFlow通过神经先验估计空间时间微分方程,实现高质量场景流估计,在多个领域表现优异,超越现有方法。

Comments Accepted to ICLR 2025. Winner of CVPR 2024 WoD Argoverse Scene Flow Challenge, Unsupervised Track. Project page at https://vedder.io/eulerflow

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.10062 2025-12-05 cs.CL

HUME: Measuring the Human-Model Performance Gap in Text Embedding Tasks

HUME:文本嵌入任务中人类-模型性能差距的测量

Adnan El Assadi, Isaac Chung, Roman Solomatin, Niklas Muennighoff, Kenneth Enevoldsen

机构 * Carleton University(卡尔顿大学) Zendesk(Zendesk公司) Stanford University(斯坦福大学) Aarhus University(阿arhus大学)

AI总结 HUME通过测量人类与模型在文本嵌入任务中的性能差距,揭示了模型与人类在不同语言资源下的表现差异,并提供了一个可扩展的评估框架。

Comments Submitted to ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.17773 2025-12-05 cs.CV cs.AI cs.CL cs.LG

KiVA: Kid-inspired Visual Analogies for Testing Large Multimodal Models

KiVA:儿童启发的视觉类比用于测试大多模态模型

Eunice Yiu, Maan Qraitem, Anisa Noor Majhi, Charlie Wong, Yutong Bai, Shiry Ginosar, Alison Gopnik, Kate Saenko

机构 * University of California, Berkeley(加州大学伯克利分校) Boston University(波士顿大学) Google DeepMind(谷歌DeepMind) Toyota Technological Institute at Chicago(芝加哥丰田技术研究所)

AI总结 KiVA通过4300个日常物体视觉转换测试大模型的类比推理能力,发现儿童和成人表现优于现有模型,尤其在复杂任务上存在显著差距。

Comments 10 pages. Project website: https://ey242.github.io/kiva.github.io/. Benchmark and code: https://github.com/ey242/KiVA

Journal ref The Thirteenth International Conference on Learning Representations (ICLR), 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.18084 2025-12-04 cs.CV cs.RO

DynamicCity: Large-Scale 4D Occupancy Generation from Dynamic Scenes

DynamicCity: 从动态场景生成大规模4D占用图

Hengwei Bian, Lingdong Kong, Haozhe Xie, Liang Pan, Yu Qiao, Ziwei Liu

机构 * WorldBench Team(WorldBench团队)

AI总结 DynamicCity通过VAE和DiT模型生成高质量动态4D占用图,提升拟合质量与训练效率。

Comments ICLR 2025 Spotlight; 35 pages, 18 figures, 15 tables; Project Page at https://dynamic-city.github.io/

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.00656 2025-12-02 cs.CL cs.CY

Sycophancy Claims about Language Models: The Missing Human-in-the-Loop

语言模型中的趋炎附势主张:缺失的人工智能循环

Jan Batzner, Volker Stocker, Stefan Schmid, Gjergji Kasneci

机构 * Weizenbaum Institute(韦岑鲍姆研究所) Technical University Berlin(柏林技术大学) Technical University Munich(慕尼黑技术大学)

AI总结 本文探讨了大型语言模型中趋炎附势现象的测量挑战,提出五个核心操作化定义,并指出当前研究缺乏对人类感知的评估,为未来研究提供建议。

Comments NeurIPS 2025 Workshop on LLM Evaluation and ICLR 2025 Workshop on Bi-Directional Human-AI Alignment

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.00351 2025-12-02 cs.LG stat.ML

Provable Memory Efficient Self-Play Algorithm for Model-free Reinforcement Learning

可证明的内存高效自博弈算法用于无模型强化学习

Na Li, Yuchen Jiao, Hangguan Shan, Shefeng Yan

机构 * College of Information Science and Electronic Engineering(信息科学与电子工程学院) Zhejiang University(浙江大学) School of Information and Electronics(信息与电子学院) Beijing Institute of Technology(北京理工大学) Institute of Acoustics(声学研究所) Chinese Academy of Sciences(中国科学院)

AI总结 本文提出了一种内存高效的自博弈算法,用于解决多智能体强化学习中的内存效率、样本复杂度和预热成本问题。

Comments ICLR 2024. arXiv admin note: substantial text overlap with arXiv:2110.04645 by other authors

详情

展开后加载摘要…

URL PDF HTML 收藏
2405.00642 2025-12-01 stat.ML cond-mat.dis-nn cond-mat.stat-mech cs.LG

Gaussian Universality in Neural Network Dynamics with Generalized Structured Input Distributions

神经网络动力学中的高斯普遍性与广义结构输入分布

Jaeyong Bae, Hawoong Jeong

机构 * Department of Physics, Korea Advanced Institute of Science and Technology(韩国科学技术院物理系) Center of Complex Systems, Korea Advanced Institute of Science and Technology(韩国科学技术院复杂系统中心)

AI总结 本研究通过将输入建模为高斯混合分布,揭示了神经网络动力学在标准化后表现出高斯普遍性,从而增强了深度学习理论的理解。

Comments Accepted for Bridging the Gap Between Practice and Theory in Deep Learning (BGPT) Workshop at ICLR 2024, [v1] 23 pages, 16 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.11916 2025-11-26 cs.CV

NeuroGaze-Distill: Brain-informed Distillation and Depression-Inspired Geometric Priors for Robust Facial Emotion Recognition

NeuroGaze-Distill:基于脑科学的蒸馏与抑郁启发的几何先验用于鲁棒面部情绪识别

Zilin Li, Weiwei Xu, Xuanqi Zhao, Yiran Zhu

机构 * School of Information and Intelligent Science(信息与智能科学学院) Department of Computer(计算机系) North China Electric Power University (BaoDing)(华北电力大学(保定))

AI总结 NeuroGaze-Distill通过脑科学先验和抑郁启发的几何先验提升面部情绪识别的鲁棒性,采用跨模态蒸馏框架实现无需生物信号的部署。

Comments Preprint. Vision-only deployment; EEG used to form static prototypes. Includes appendix, 7 figures and 3 tables. Considering submission to ICLR 2026. Revision note: This version corrects inaccuracies in the authors' institutional affiliations. No technical content has been modified

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.19434 2025-11-25 cs.CV cs.LG stat.ML

Breaking the Likelihood-Quality Trade-off in Diffusion Models by Merging Pretrained Experts

通过合并预训练专家打破扩散模型中似然-质量权衡

Yasin Esfandiari, Stefan Bauer, Sebastian U. Stich, Andrea Dittadi

机构 * Saarland University(萨尔兰大学) Helmholtz AI(亥姆霍兹人工智能研究所) Technical University of Munich(慕尼黑技术大学) CISPA Helmholtz Center for Information Security(亥姆霍兹信息安全部分研究所) MPI for Intelligent Systems, Tübingen(图宾根智能系统研究所)

AI总结 通过在去噪过程中切换预训练专家,该方法有效打破扩散模型中似然与质量的权衡,提升图像生成质量和似然

Comments ICLR 2025 DeLTa workshop

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.18891 2025-11-25 cs.CL

Reproducibility Study of Large Language Model Bayesian Optimization

大型语言模型贝叶斯优化的可重复性研究

Adam Rychert, Gasper Spagnolo, Evgenii Posashkov

机构 * UL FRI Data Science(UL FRI数据科学)

AI总结 本研究验证了LLAMBO框架在更换语言模型backbone时的鲁棒性,并展示了其在Llama 3.1 70B上的有效性。

Comments 7 pages, 8 figures. Reproducibility study of the LLAMBO framework (ICLR 2024). Code: https://github.com/spagnoloG/llambo-reproducibility

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.15938 2025-11-25 cs.LG cs.AI cs.CL cs.NE

Straight to Zero: Why Linearly Decaying the Learning Rate to Zero Works Best for LLMs

直接到零:为什么线性衰减学习率到零在大语言模型中效果最佳

Shane Bergsma, Nolan Dey, Gurpreet Gosal, Gavia Gray, Daria Soboleva, Joel Hestness

机构 * Cerebras Systems(Cerebras系统)

AI总结 本研究发现线性衰减到零的学习率调度在训练大语言模型时效果最佳,相比其他调度方式具有显著优势。

Comments ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2408.11052 2025-11-25 cs.LG cs.AI

Accelerating Goal-Conditioned RL Algorithms and Research

加速目标条件强化学习算法与研究

Michał Bortkiewicz, Władysław Pałucki, Vivek Myers, Tadeusz Dziarmaga, Tomasz Arczewski, Łukasz Kuciński, Benjamin Eysenbach

机构 * Warsaw University of Technology(华沙技术大学) University of Warsaw(华沙大学) UC Berkeley(伯克利大学) Jagiellonian University(雅盖隆大学) Polish Academy of Sciences(波兰科学院) IDEAS NCBR Princeton University(普林斯顿大学)

AI总结 本文提出JaxGCRL框架,通过高效算法和GPU加速技术显著提升目标条件强化学习的训练效率,并评估对比学习中的关键设计以稳定训练性能。

Comments Published at ICLR 2025 (Spotlight). Website: https://michalbortkiewicz.github.io/JaxGCRL/ Code: https://github.com/MichalBortkiewicz/JaxGCRL

Journal ref International Conference on Learning Representations (ICLR), 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.17622 2025-11-25 cs.LG cs.AI

Neurocircuitry-Inspired Hierarchical Graph Causal Attention Networks for Explainable Depression Identification

受神经回路启发的分层图因果注意网络用于可解释的抑郁症识别

Weidao Chen, Yuxiao Yang, Yueming Wang

机构 * MOE Frontier Science Center for Brain Science and Brain-machine Integration(脑科学与脑机融合前沿科学中心) Nanhu Brain-computer Interface Institute(南湖脑机接口研究院) School of Computer Science and Technology(计算机科学与技术学院) Qiushi Academy for Advanced Studies(启硕高级研究院) State Key Laboratory of Brain-machine Intelligence(脑机智能国家重点实验室) The Department of Neurosurgery, Second Affiliated Hospital, School of Medicine, Zhejiang University Hangzhou, China(浙江大学医学院附属第二医院神经外科)

AI总结 本文提出NH-GCAT网络,通过分层图因果注意机制,结合神经科学知识,实现对抑郁症的可解释识别。

Comments Under review for ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.16924 2025-11-24 math.ST stat.TH

CBMA: Improving conformal prediction through Bayesian model averaging

CBMA:通过贝叶斯模型平均改进符合预测

Pankaj Bhagwat, Linglong Kong, Bei Jiang

AI总结 CBMA通过结合贝叶斯模型平均与符合预测,改进了在模型可能误指定时的预测效率和鲁棒性。

Comments 19 pages

Journal ref ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2503.01478 2025-11-24 cs.CL cs.AI cs.LG

SePer: Measure Retrieval Utility Through The Lens Of Semantic Perplexity Reduction

SePer:通过语义困惑度降低的视角衡量检索效用

Lu Dai, Yijie Xu, Jinhui Ye, Hao Liu, Hui Xiong

机构 * The Hong Kong University of Science and Technology (Guangzhou)(香港科学与技术大学(广州)) The Hong Kong University of Science and Technology(香港科学与技术大学) Carnegie Mellon University(卡内基梅隆大学)

AI总结 SePer通过语义困惑度降低衡量检索效用,提供更精确的RAG评估方法

Comments ICLR 2025 Spotlight

详情

展开后加载摘要…

URL PDF HTML 收藏
2409.06142 2025-11-24 stat.ML cs.LG

Variational Search Distributions

变分搜索分布

Daniel M. Steinberg, Rafael Oliveira, Cheng Soon Ong, Edwin V. Bonilla

机构 * Data61, CSIRO, Australia(Data61,CSIRO,澳大利亚)

AI总结 VSD通过变分推断方法,高效地在稀有类别中生成离散组合设计,实验证明其在蛋白质和DNA/RNA工程任务中的优越性。

Comments Accepted as a poster in the thirteenth International Conference on Learning Representations (ICLR), 2025

Journal ref https://proceedings.iclr.cc/paper_files/paper/2025/hash/055fc19a3ce780b96cff15ffe738c1f1-Abstract-Conference.html

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.01082 2025-11-21 cs.CL

Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs

提升采样温度:用于生成创意且连贯LLM输出的min-p采样

Minh Nhat Nguyen, Andrew Baker, Clement Neo, Allen Roush, Andreas Kirsch, Ravid Shwartz-Ziv

机构 * Apart Research Independent(独立研究者) New York University(纽约大学)

AI总结 min-p采样通过动态截断方法提升LLM生成文本的质量和多样性,尤其在高温度下表现优异,已获多个开源框架采用。

Comments Oral presentation at ICLR 2025. Camera-ready version available at https://iclr.cc/virtual/2025/poster/30358

Journal ref In Proceedings of the 2025 International Conference on Learning Representations (ICLR), 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.10810 2025-11-18 cs.CV

SVBench: A Benchmark with Temporal Multi-Turn Dialogues for Streaming Video Understanding

Zhenyu Yang, Yuhang Hu, Zemin Du, Dizhan Xue, Shengsheng Qian, Jiahong Wu, Fan Yang, Weiming Dong, Changsheng Xu

机构 * Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所) University of Chinese Academy of Sciences(中国科学院大学) Kuaishou Technology(快手科技) Zhengzhou University(郑州大学) ShanghaiTech University(上海科技大学) Peng Cheng Laboratory(鹏城实验室)

Comments ICLR 2025 Accepted (Spotlight)

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.14315 2025-11-17 stat.ML cs.LG

Optimizing importance weighting in the presence of sub-population shifts

Floris Holstege, Bram Wouters, Noud van Giersbergen, Cees Diks

机构 * University of Amsterdam, Department of Quantitative Economics(阿姆斯特丹大学量化经济学系) Tinbergen Institute(廷伯根研究所)

Comments Published at ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.19611 2025-11-14 cs.LG

PRDP: Progressively Refined Differentiable Physics

Kanishk Bhatia, Felix Koehler, Nils Thuerey

机构 * Technical University of Munich(慕尼黑技术大学) Munich Center for Machine Learning(慕尼黑机器学习中心)

Comments Accepted at ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.13595 2025-11-14 cs.CL cs.AI cs.IR

MMTEB: Massive Multilingual Text Embedding Benchmark

Kenneth Enevoldsen, Isaac Chung, Imene Kerboua, Márton Kardos, Ashwin Mathur, David Stap, Jay Gala, Wissam Siblini, Dominik Krzemiński, Genta Indra Winata, Saba Sturua, Saiteja Utpala, Mathieu Ciancone, Marion Schaeffer, Gabriel Sequeira, Diganta Misra, Shreeya Dhakal, Jonathan Rystrøm, Roman Solomatin, Ömer Çağatan, Akash Kundu, Martin Bernstorff, Shitao Xiao, Akshita Sukhlecha, Bhavish Pahwa, Rafał Poświata, Kranthi Kiran GV, Shawon Ashraf, Daniel Auras, Björn Plüster, Jan Philipp Harries, Loïc Magne, Isabelle Mohr, Mariya Hendriksen, Dawei Zhu, Hippolyte Gisserot-Boukhlef, Tom Aarsen, Jan Kostkan, Konrad Wojtasik, Taemin Lee, Marek Šuppa, Crystina Zhang, Roberta Rocca, Mohammed Hamdy, Andrianos Michail, John Yang, Manuel Faysse, Aleksei Vatolin, Nandan Thakur, Manan Dey, Dipam Vasani, Pranjal Chitale, Simone Tedeschi, Nguyen Tai, Artem Snegirev, Michael Günther, Mengzhou Xia, Weijia Shi, Xing Han Lù, Jordan Clive, Gayatri Krishnakumar, Anna Maksimova, Silvan Wehrli, Maria Tikhonova, Henil Panchal, Aleksandr Abramov, Malte Ostendorff, Zheng Liu, Simon Clematide, Lester James Miranda, Alena Fenogenova, Guangyu Song, Ruqiya Bin Safi, Wen-Ding Li, Alessia Borghini, Federico Cassano, Hongjin Su, Jimmy Lin, Howard Yen, Lasse Hansen, Sara Hooker, Chenghao Xiao, Vaibhav Adlakha, Orion Weller, Siva Reddy, Niklas Muennighoff

机构 * Aarhus University(奥胡斯大学) Individual Contributor(个人贡献者) Esker(Esker公司) INSA Lyon(里昂INSA) University of Amsterdam(阿姆斯特丹大学) MBZUAI(穆罕默德·本·拉希德智能技术研究院) Jina AI(Jina AI公司) Microsoft Research(微软研究院) Wikit(Wikit公司)

Comments Accepted for ICLR: https://openreview.net/forum?id=zl3pfz4VCV

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.08207 2025-11-14 cs.CV cs.LG

DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models

Xiaoxiao He, Quan Dao, Ligong Han, Song Wen, Minhao Bai, Di Liu, Han Zhang, Martin Renqiang Min, Felix Juefei-Xu, Chaowei Tan, Bo Liu, Kang Li, Hongdong Li, Junzhou Huang, Faez Ahmed, Akash Srivastava, Dimitris Metaxas

机构 * Rutgers University(新泽西罗格斯大学) MIT-IBM Watson AI Lab(MIT-IBM沃森人工智能实验室) Red Hat AI Innovation(红帽AI创新) Google DeepMind(谷歌DeepMind) NYU(纽约大学) Walmart Global Tech(沃尔玛全球技术) NEC Labs America(NEC美国实验室) Massachusetts Institute of Technology(麻省理工学院) ANU(澳大利亚国立大学) UT Arlington(德克萨斯大学阿灵顿分校)

Comments Project webpage: https://hexiaoxiao-cs.github.io/DICE/. This paper was accepted to CVPR 2025 but later desk-rejected post camera-ready, due to a withdrawal from ICLR made 14 days before reviewer assignment

详情

展开后加载摘要…

URL PDF HTML 收藏
2501.15282 2025-11-13 cs.LG

AutoG: Towards automatic graph construction from tabular data

Zhikai Chen, Han Xie, Jian Zhang, Xiang song, Jiliang Tang, Huzefa Rangwala, George Karypis

机构 * Michigan State University(密歇根州立大学) Amazon(亚马逊)

Comments camera ready version, update meta info,accepted by ICLR 2025 https://openreview.net/forum?id=hovDbX4Gh6

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.08594 2025-11-13 cs.CL

Diverse Preference Learning for Capabilities and Alignment

Stewart Slocum, Asher Parker-Sartori, Dylan Hadfield-Menell

机构 * MIT CSAIL(麻省理工学院计算机科学与人工智能实验室)

Journal ref 13th International Conference on Learning Representations (ICLR 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2412.02181 2025-11-12 cs.LG cs.AI cs.SI

Generalizing Weisfeiler-Lehman Kernels to Subgraphs

Dongkwan Kim, Alice Oh

机构 * KAIST, Republic of Korea(韩国延世大学)

Comments ICLR 2025 Camera Ready (15 pages), with minor typos fixed

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.12982 2025-11-12 cs.LG cs.AI

Flash Inference: Near Linear Time Inference for Long Convolution Sequence Models and Beyond

Costin-Andrei Oncescu, Sanket Purandare, Stratos Idreos, Sham Kakade

机构 * Harvard University(哈佛大学)

Comments Accepted at ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2311.13721 2025-11-12 cs.SE cs.AI

Nova: Generative Language Models for Assembly Code with Hierarchical Attention and Contrastive Learning

Nan Jiang, Chengxiao Wang, Kevin Liu, Xiangzhe Xu, Lin Tan, Xiangyu Zhang, Petr Babkin

机构 * Purdue University(普渡大学) University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校) Lynbrook High School(林布罗克高中) J.P. Morgan AI Research(摩根大通人工智能研究)

Comments Published as a conference paper at ICLR 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.18258 2025-11-11 cs.LG cs.AI

ANO : Faster is Better in Noisy Landscape

Adrien Kegreisz

机构 * Independent Researcher(独立研究者)

Comments Under Review for ICLR 2026, 25 pages total with appendix, 7 figures, 12 tables

详情

展开后加载摘要…

URL PDF HTML 收藏
2405.00646 2025-11-11 cs.CV cs.LG

Learning to Compose: Improving Object Centric Learning by Injecting Compositionality

Whie Jung, Jaehoon Yoo, Sungjin Ahn, Seunghoon Hong

机构 * School of Computing, KAIST(计算学院,韩国科学技术院)

Journal ref International Conference on Learning Representations (ICLR), 2024

详情

展开后加载摘要…

URL PDF HTML 收藏