图像生成 - arXivDaily 专题

2512.03125 2025-12-04 cs.LG cs.AI 50%

Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models

缓解统一多模态模型持续学习中的模态内和模态间遗忘

Xiwen Wei, Mustafa Munir, Radu Marculescu

专题命中效率与蒸馏：image generation(abstract)

AI总结本文提出MoDE，通过解耦模态以缓解统一多模态模型中的模态内和模态间遗忘问题。

Comments NeurIPS 2025

2511.22853 2025-12-01 cs.LG 50%

TARFVAE: Efficient One-Step Generative Time Series Forecasting via TARFLOW based VAE

TARFVAE：通过TARFLOW基于VAE实现高效的单步生成时间序列预测

Jiawen Wei, Lan Jiang, Pengbo Wei, Ziwen Ye, Teng Song, Chen Chen, Guangrui Ma

机构 * Meituan（美团）

专题命中效率与蒸馏：image synthesis(abstract)

AI总结 TARFVAE结合Transformer自回归流与VAE，实现高效单步生成时间序列预测，提升预测效率与性能。

2506.18165 2025-11-26 cs.LG cs.AI 50%

Non-equilibrium Annealed Adjoint Sampler

非平衡退火共轭采样器

Jaemoo Choi, Yongxin Chen, Molei Tao, Guan-Horng Liu

机构 * Georgia Institute of Technology（佐治亚理工学院）； FAIR at Meta（Meta的FAIR）

专题命中效率与蒸馏：diffusion(abstract)

AI总结非平衡退火共轭采样器通过引入退火参考动力学，提升扩散模型在学习控制的稳定性和效率，适用于多种采样任务。

Comments 26 pages, 8 figures

2511.19594 2025-11-26 astro-ph.IM 50%

Pixellated Posterior Sampling of Point Spread Functions in Astronomical Images

像素化的点扩散函数后验采样在天文学图像中

Connor Stone, Ronan Legin, Alexandre Adam, Nikolay Malkin, Gabriel Missael Barco, Laurence Perreaul-Levasseur, Yashar Hezaveh

专题命中效率与蒸馏：diffusion(abstract)

AI总结本文提出了一种基于生成扩散模型的像素级PSF后验采样方法，通过结合高斯似然与生成模型先验，实现对天文学图像中点扩散函数的高精度建模，提升后续分析的不确定性传播能力。

Comments 17 pages, 10 figures

2511.17898 2025-11-25 cs.RO 50%

L1 Sample Flow for Efficient Visuomotor Learning

L1样本流用于高效视觉-运动学习

Weixi Song, Zhetao Chen, Tao Xu, Xianchao Zeng, Xinyu Zhou, Lixin Yang, Donglin Wang, Cewu Lu, Yong-Lu Li

机构 * Zhejiang University（浙江大学）； Shanghai Innovation Institute（上海创新研究院）； Westlake University（西湖大学）； Shanghai Jiao Tong University（上海交通大学）

专题命中效率与蒸馏：diffusion(abstract)

AI总结 L1流通过结合去噪模型的多模分布捕捉能力与L1回归的高效性，实现高效的视觉-运动学习。

2509.05093 2025-11-24 physics.bio-ph 50%

Physical interactions enable energy-efficient Turing patterns

物理相互作用使能效 Turing 模式

Cathelijne ter Burg, David Zwicker

专题命中效率与蒸馏：diffusion(abstract)

AI总结本研究通过引入热力学一致的 Turing 系统模型，发现排斥相互作用显著降低能量需求，揭示物理相互作用在形成自然模式中的关键作用。

Comments 11 pages, 10 figures

2511.15042 2025-11-20 cond-mat.mtrl-sci 50%

Integrating Atomic Scale Catalyst Design with Transport Engineering for Stable and Efficient CO2 Electrolysis to CO in a Membrane Electrode Assembly

Zahra Teimouri, Mahtab Masouminia, Ashkan Irannezhad, Reza Eslami, Joseph Deering, Navid Noor, Shunquan Tan, Amirhossein Foroozan, Shayan Angizi, Sung-Fu Hun, Drew Higgins

专题命中效率与蒸馏：diffusion(abstract)

Comments 32 pages, 7 figures

2511.12603 2025-11-18 cs.LG cs.AI cs.SY eess.SY 50%

PID-controlled Langevin Dynamics for Faster Sampling of Generative Models

Hongyi Chen, Jianhai Shu, Jingtao Ding, Yong Li, Xiao-Ping Zhang

机构 * Shenzhen Key Laboratory of Ubiquitous Data Enabling Laboratory, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China（深圳 ubiquitous 数据赋能实验室，深圳国际研究生院，清华大学，深圳）； Department of Electronic Engineering, Tsinghua University, Beijing 100084, China（电子工程系，清华大学，北京）； Pengcheng Laboratory, Shenzhen 518055, China（鹏城实验室，深圳）

专题命中效率与蒸馏：image generation(abstract)

Comments NeurIPS 2025 poster paper

2511.12411 2025-11-18 math.NA cs.NA 50%

An Efficient Unconditionally Energy-Stable Numerical Scheme for Bose--Einstein Condensate

Jing Guo, Cheng Wang, Dong Wang

专题命中效率与蒸馏：diffusion(abstract)

2511.10208 2025-11-14 cs.LG cs.AI math.DS math.PR physics.bio-ph 50%

Fractional neural attention for efficient multiscale sequence processing

Cheng Kevin Qu, Andrew Ly, Pulin Gong

机构 * School of Physics, University of Sydney（物理学院，悉尼大学）

专题命中效率与蒸馏：diffusion(abstract)

2504.12190 2025-11-14 math.ST stat.CO stat.TH 50%

Rebalancing Markov jump processes for non-reversible continuous-time sampling

Erik Jansson, Moritz Schauer, Ruben Seyer, Akash Sharma

专题命中效率与蒸馏：diffusion(abstract)

Comments 25 pages, 13 figures; completely revised

2511.08648 2025-11-13 q-bio.BM cs.LG 50%

Compact Artificial Neural Network Models for Predicting Protein Residue -- RNA Base Binding

Stanislav Selitskiy

机构 * University of Bedfordshire（贝福德郡大学）

专题命中效率与蒸馏：image generation(abstract)

2511.07869 2025-11-12 cs.DS cs.DC cs.LG math.PR 50%

Parallel Sampling via Autospeculation

Nima Anari, Carlo Baronio, CJ Chen, Alireza Haqi, Frederic Koehler, Anqi Li, Thuy-Duong Vuong

专题命中效率与蒸馏：diffusion(abstract)

2501.14331 2025-11-12 astro-ph.HE 50%

Cosmic ray transport and acceleration in an evolving shock landscape

Sophie Aerdker, Roark Habegger, Lukas Merten, Ellen Zweibel, Julia Becker Tjus

专题命中效率与蒸馏：diffusion(abstract)

Journal ref A&A 703, A51 (2025)

2511.04641 2025-11-07 cs.LG 50%

Efficient probabilistic surrogate modeling techniques for partially-observed large-scale dynamical systems

Hans Harder, Abhijeet Vishwasrao, Luca Guastoni, Ricardo Vinuesa, Sebastian Peitz

专题命中效率与蒸馏：diffusion(abstract)

2507.02392 2025-11-06 math.NA cs.NA 50%

An efficient asymptotic preserving Monte Carlo method for frequency-dependent radiative transfer equations

Yiyang Hong, Yi Shi, Yi Cai, Tao Xiong

专题命中效率与蒸馏：diffusion(abstract)

2011.10648 2025-11-06 math.NA cs.MS cs.NA 50%

Efficient space-time reduced order model for linear dynamical systems in Python using less than 120 lines of code

Youngkyu Kim, Karen May Wang, Youngsoo Choi

专题命中效率与蒸馏：diffusion(abstract)

Comments 24 pages, 18 figures

Journal ref Mathematics, 9(14):1690, 2021

2508.15069 2025-11-04 stat.CO stat.ME stat.ML 50%

Sampling by averaging: A multiscale approach to score estimation

Paula Cordero-Encinar, Andrew B. Duncan, Sebastian Reich, O. Deniz Akyildiz

专题命中效率与蒸馏：diffusion(abstract)

2503.17468 2025-11-04 eess.IV eess.SP 50%

Anatomically Guided Motion Correction for Placental IVIM Parameter Estimation with Accelerated Sampling Method

Mbaimou Auxence Ngremmadji, Freddy Odille, Charline Bertholdt, Marine Beaumont, Olivier Morel, Bailiang Chen

专题命中效率与蒸馏：diffusion(abstract)

Comments 11 pages, 6 figures

2511.00700 2025-11-04 cs.LG 50%

Privacy-Aware Time Series Synthesis via Public Knowledge Distillation

Penghang Liu, Haibei Zhu, Eleonora Kreacic, Svitlana Vyetrenko

机构 * JPMorgan AI Research（摩根大通人工智能研究）

专题命中效率与蒸馏：diffusion(abstract)

Comments Published on Transactions on Machine Learning Research (TMLR)

2504.06312 2025-11-04 cs.LG cs.AI 50%

DMol: A Highly Efficient and Chemical Motif-Preserving Molecule Generation Platform

Peizhi Niu, Yu-Hsiang Wang, Vishal Rana, Chetan Rupakheti, Abhishek Pandey, Olgica Milenkovic

机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）； AbbVie（阿比唯）

专题命中效率与蒸馏：diffusion(abstract)

2502.12395 2025-11-04 cs.LG 50%

Efficient Neural SDE Training using Wiener-Space Cubature

Luke Snow, Vikram Krishnamurthy

专题命中效率与蒸馏：diffusion(abstract)

2510.26892 2025-11-03 cs.LG cs.AI 50%

BI-DCGAN: A Theoretically Grounded Bayesian Framework for Efficient and Diverse GANs

Mahsa Valizadeh, Rui Tuo, James Caverlee

专题命中效率与蒸馏：diffusion(abstract)

2506.15653 2025-11-03 cond-mat.stat-mech physics.chem-ph 50%

Fokker-Planck Score Learning: Efficient Free-Energy Estimation under Periodic Boundary Conditions

Daniel Nagel, Tristan Bereau

专题命中效率与蒸馏：diffusion(abstract)

Comments 11 pages, 4 figures

2506.01158 2025-10-31 cs.LG cs.AI stat.ML 50%

Efficient Regression-Based Training of Normalizing Flows for Boltzmann Generators

Danyal Rehman, Oscar Davis, Jiarui Lu, Jian Tang, Michael Bronstein, Yoshua Bengio, Alexander Tong, Avishek Joey Bose

机构 * Mila – Québec AI Institute（魁北克AI研究所）； Université de Montréal（蒙特利尔大学）； Massachusetts Institute of Technology（麻省理工学院）； University of Oxford（牛津大学）； AITHYRA ； HEC Montréal（蒙特利尔HEC）

专题命中效率与蒸馏：diffusion(abstract)

Comments Preprint; ICML GenBio Best Paper Award 2025

2410.12652 2025-10-31 cs.LG cs.AI eess.SP 50%

Constrained Posterior Sampling: Time Series Generation with Hard Constraints

Sai Shankar Narasimhan, Shubhankar Agarwal, Litu Rout, Sanjay Shakkottai, Sandeep P. Chinchali

机构 * Department of Electrical and Computer Engineering, The University of Texas at Austin（电气与计算机工程系，德克萨斯大学奥斯汀分校）

专题命中效率与蒸馏：diffusion(abstract)

2510.23700 2025-10-29 astro-ph.HE 50%

Shock Acceleration in the Intracluster Medium: Implications of Micromirror Confinement

Rebecca Diesing, Ludwig Böss, Damiano Caprioli

专题命中效率与蒸馏：diffusion(abstract)

Comments 10 pages, 5 figures, submitted to ApJ

2510.03095 2025-10-29 cs.LG cs.AI stat.ML 50%

Distilled Protein Backbone Generation

Liyang Xie, Haoran Zhang, Zhendong Wang, Wesley Tansey, Mingyuan Zhou

机构 * The University of Texas at Austin（德克萨斯大学奥斯汀分校）； The Memorial Sloan Kettering Cancer Center（纪念斯隆凯特林癌症中心）

专题命中效率与蒸馏：diffusion(abstract)

Comments PyTorch implementation: https://github.com/LY-Xie/SiD_Protein

2509.20753 2025-10-27 stat.ML cs.LG 50%

RAPTOR-GEN: RApid PosTeriOR GENerator for Bayesian Learning in Biomanufacturing

Wandi Xu, Wei Xie

机构 * Department of Mechanical and Industrial Engineering, Northeastern University（机械与工业工程系，东北大学）

专题命中效率与蒸馏：diffusion(abstract)

Comments 72 pages, 6 figures

2506.07254 2025-10-27 cs.LG 50%

A Stable Whitening Optimizer for Efficient Neural Network Training

Kevin Frans, Sergey Levine, Pieter Abbeel

机构 * UC Berkeley（伯克利大学）

专题命中效率与蒸馏：diffusion(abstract)

1. 效率与蒸馏 2544 篇

Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models

TARFVAE: Efficient One-Step Generative Time Series Forecasting via TARFLOW based VAE

Non-equilibrium Annealed Adjoint Sampler

Pixellated Posterior Sampling of Point Spread Functions in Astronomical Images

L1 Sample Flow for Efficient Visuomotor Learning

Physical interactions enable energy-efficient Turing patterns

Integrating Atomic Scale Catalyst Design with Transport Engineering for Stable and Efficient CO2 Electrolysis to CO in a Membrane Electrode Assembly

PID-controlled Langevin Dynamics for Faster Sampling of Generative Models

An Efficient Unconditionally Energy-Stable Numerical Scheme for Bose--Einstein Condensate

Fractional neural attention for efficient multiscale sequence processing

Rebalancing Markov jump processes for non-reversible continuous-time sampling

Compact Artificial Neural Network Models for Predicting Protein Residue -- RNA Base Binding

Parallel Sampling via Autospeculation

Cosmic ray transport and acceleration in an evolving shock landscape

Efficient probabilistic surrogate modeling techniques for partially-observed large-scale dynamical systems

An efficient asymptotic preserving Monte Carlo method for frequency-dependent radiative transfer equations

Efficient space-time reduced order model for linear dynamical systems in Python using less than 120 lines of code

Sampling by averaging: A multiscale approach to score estimation

Anatomically Guided Motion Correction for Placental IVIM Parameter Estimation with Accelerated Sampling Method

Privacy-Aware Time Series Synthesis via Public Knowledge Distillation

DMol: A Highly Efficient and Chemical Motif-Preserving Molecule Generation Platform

Efficient Neural SDE Training using Wiener-Space Cubature

BI-DCGAN: A Theoretically Grounded Bayesian Framework for Efficient and Diverse GANs

Fokker-Planck Score Learning: Efficient Free-Energy Estimation under Periodic Boundary Conditions

Efficient Regression-Based Training of Normalizing Flows for Boltzmann Generators

Constrained Posterior Sampling: Time Series Generation with Hard Constraints

Shock Acceleration in the Intracluster Medium: Implications of Micromirror Confinement

Distilled Protein Backbone Generation

RAPTOR-GEN: RApid PosTeriOR GENerator for Bayesian Learning in Biomanufacturing

A Stable Whitening Optimizer for Efficient Neural Network Training