arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

NeurIPS

Conference on Neural Information Processing Systems · 会议 · Machine Learning

2026-05-28 至 2026-05-28 共收录 4
2605.28111 2026-05-28 cs.LG

Chreode: A Cell World Model for One-Step Temporal Dynamics and Perturbation Prediction

Chreode: 用于一步时间动态和扰动预测的细胞世界模型

Mufan Qiu, Genhui Zheng, Yinuo Xu, Ruichen Zhang, Ying Ding, Qi Long, Tianlong Chen

机构 * University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校) The University of Texas at Austin(德克萨斯大学奥斯汀分校) University of Pennsylvania(宾夕法尼亚大学)

AI总结 提出Chreode,一种基于结构化残差转移算子的单步细胞世界模型,通过预训练和微调实现发育轨迹与扰动预测的统一,在多个基准上取得性能提升。

Comments 25 pages, 3 figures, 14 tables. Submitted to NeurIPS 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.27465 2026-05-28 cs.CV cs.AI

AdaMerge: Salience-Aware Adaptive Token Merging for Training-Free Acceleration of Vision Transformers

AdaMerge: 面向视觉Transformer无训练加速的显著性感知自适应令牌合并

Semi Lee, Hyejin Go, Hyesong Choi

机构 * Electronic Engineering(电子工程) Soongsil University(顺斯大学)

AI总结 提出AdaMerge框架,通过显著性加权相似度和自适应合并强度两个互补机制,在无训练条件下提升令牌合并的精度-计算量帕累托前沿。

Comments 11 pages, 3 figures, 5 tables. Submitted to NeurIPS 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.17720 2026-05-28 cs.LG physics.ao-ph

PEAR: Equal Area Weather Forecasting on the Sphere

PEAR:球面上的等面积天气预报

Hampus Linander, Tage Tykesson, Pietro Rosso, Christoffer Petersson, Daniel Persson, Jan E. Gerken

机构 * VERSES AI Department of Mathematical Sciences(数学科学系) Chalmers University of Technology(楚姆勒斯技术大学) University of Gothenburg(哥德堡大学) Department of Computer Science and Engineering(计算机科学与工程系) Department of Physics and Astronomy(物理与天文学系) Recohere

AI总结 针对球面等角网格在极地分辨率过高的问题,提出基于HEALPix等面积网格的Transformer模型PEAR,实现无计算开销的全球天气预报性能提升。

Comments Extended version of manuscript published in the AI for Science workshop (NeurIPS 2025), 11 pages, 15 pages supplemental

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.21046 2026-05-28 cs.CV cs.RO

CogVLA: Cognition-Aligned Vision-Language-Action Model via Instruction-Driven Routing & Sparsification

CogVLA: 通过指令驱动路由与稀疏化实现认知对齐的视觉-语言-动作模型

Wei Li, Renshan Zhang, Rui Shao, Jie He, Liqiang Nie

机构 * School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen(哈尔滨工业大学深圳校区计算机科学与技术学院)

AI总结 提出CogVLA框架,通过指令驱动路由和稀疏化机制,在LIBERO基准和真实机器人任务上以2.5倍训练成本降低和2.8倍推理延迟降低实现97.4%和70.0%的成功率。

Comments Accepted to NeurIPS 2025, Project Page: https://jiutian-vl.github.io/CogVLA-page

详情

展开后加载摘要…

URL PDF HTML 收藏