Results 291 to 300 of about 29,508,989 (334)

Mean Flows for One-step Generative Modeling

Neural Information Processing Systems
We propose a principled and effective framework for one-step generative modeling. We introduce the notion of average velocity to characterize flow fields, in contrast to instantaneous velocity modeled by Flow Matching methods.
Zhengyang Geng, Mingyang Deng, Xingjian Bai, J. Kolter, Kai-Ming He   +4 more
semanticscholar   +1 more source

Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion

Neural Information Processing Systems
This paper presents Diffusion Forcing, a new training paradigm where a diffusion model is trained to denoise a set of tokens with independent per-token noise levels.
Boyuan Chen, Diego Marti Monso, Yilun Du, M. Simchowitz, Russ Tedrake, Vincent Sitzmann   +5 more
semanticscholar   +1 more source

Simplified and Generalized Masked Diffusion for Discrete Data

Neural Information Processing Systems
Masked (or absorbing) diffusion is actively explored as an alternative to autoregressive models for generative modeling of discrete data. However, existing work in this area has been hindered by unnecessarily complex model formulations and unclear ...
Jia-Xin Shi, Kehang Han, Zhe Wang, Arnaud Doucet, Michalis K. Titsias   +4 more
semanticscholar   +1 more source

DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving

Computer Vision and Pattern Recognition
Recently, the diffusion model has emerged as a powerful generative technique for robotic policy learning, capable of modeling multi-mode action distributions. Leveraging its capability for end-to-end autonomous driving is a promising direction.
Ben-Cheng Liao, Shaoyu Chen, Haoran Yin, Bo Jiang, Cheng Wang, Si-Xu Yan, Xinbang Zhang, Xiangyu Li, Ying Zhang, Qian Zhang, Xinggang Wang   +10 more
semanticscholar   +1 more source

Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data

International Conference on Learning Representations
Discrete diffusion models with absorbing processes have shown promise in language modeling. The key quantities to be estimated are the ratios between the marginal probabilities of two transitive states at all timesteps, called the concrete score. In this
Jingyang Ou, Shen Nie, Kaiwen Xue, Fengqi Zhu, Jiacheng Sun, Zhenguo Li, Chongxuan Li   +6 more
semanticscholar   +1 more source

Drift-Diffusion Models

2017
In 1950, van Roosbroeck introduced the fundamental semiconductor device equations as a system of three nonlinearly coupled partial differential equations (PDEs). They describe the semiclassical transport of free electrons and holes in a self-consistent electric field using a drift-diffusion approximation.
Farrell P., Rotundo N., Doan D. H., Kantner M., Fuhrmann J., Koprucki T.   +5 more
openaire   +2 more sources

Scaling Diffusion Language Models via Adaptation from Autoregressive Models

International Conference on Learning Representations
Diffusion Language Models (DLMs) have emerged as a promising new paradigm for text generative modeling, potentially addressing limitations of autoregressive (AR) models.
Shansan Gong, Shivam Agarwal, Yizhe Zhang, Jiacheng Ye, Lin Zheng, Mukai Li, Chenxin An, Peilin Zhao, Wei Bi, Jiawei Han, Hao Peng, Lingpeng Kong   +11 more
semanticscholar   +1 more source

A Survey on Diffusion Models for Inverse Problems

arXiv.org
Diffusion models have become increasingly popular for generative modeling due to their ability to generate high-quality samples. This has unlocked exciting new possibilities for solving inverse problems, especially in image restoration and reconstruction,
G. Daras, Hyungjin Chung, Chieh-Hsin Lai, Yuki Mitsufuji, Jong Chul Ye, P. Milanfar, A. Dimakis, M. Delbracio   +7 more
semanticscholar   +1 more source

Application of the thermodynamic extremal principle to diffusion-controlled phase transformations in Fe-C-X alloys: Modeling and applications

Acta Materialia, 2018
Diffusion-controlled phase transformations are of singular importance in controlling microstructures and mechanical properties but are difficult to model and calculate for Fe-C-X alloys because of the large difference in the diffusivities of the ...
Wang-Wang Kuang, Hai-Feng Wang, Xin Li, Jianbao Zhang, Qing Zhou, Yuhong Zhao   +5 more
semanticscholar   +1 more source

Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling

International Conference on Learning Representations
Masked diffusion models (MDMs) have emerged as a popular research topic for generative modeling of discrete data, thanks to their superior performance over other discrete diffusion models, and are rivaling the auto-regressive models (ARMs) for language ...
Kaiwen Zheng, Yongxin Chen, Hanzi Mao, Mingying Liu, Jun Zhu, Qinsheng Zhang   +5 more
semanticscholar   +1 more source