Efficient Generation
Efficient generation focuses on creating high-quality outputs (images, text, meshes, etc.) using minimal computational resources. Current research emphasizes developing parameter-efficient models, such as those based on diffusion processes and generative adversarial networks (GANs), and employing techniques like block-wise generation, sparse attention mechanisms, and adaptive computation to reduce computational complexity. These advancements are crucial for deploying generative models on resource-constrained devices and improving the scalability of applications across various domains, including image synthesis, robotics, and drug discovery.
19papers
Papers
March 20, 2025
NuiScene: Exploring Efficient Generation of Unbounded Outdoor Scenes
Han-Hung Lee, Qinghong Han, Angel X. ChangSimon Fraser University●AmiiSpeCache: Speculative Key-Value Caching for Efficient Generation of LLMs
Shibo Jie, Yehui Tang, Kai Han, Zhi-Hong Deng, Jing HanPeking University●Huawei Noah’s Ark Lab
February 17, 2025
September 26, 2024
February 21, 2024