<ul data-eligibleForWebStory="true"><li>Researchers proposed a new method, ODE_t(ODE_l), to enhance sampling efficiency in continuous normalizing flows and diffusion models.</li><li>The method involves controlling the tradeoff between quality and complexity by adjusting time steps and the length of the neural network.</li><li>By using this approach, sampling can be done with varying time steps and transformer blocks, reducing latency and memory usage.</li><li>Experiments on image generation datasets demonstrated up to a 3x latency reduction and a 3.5-point FID score improvement compared to the previous state of the art.</li></ul>

$\textrm{ODE}_t \left(\textrm{ODE}_l \right)$: Shortcutting the Time and Length in Diffusion and Flow Models for Faster Sampling

Discover more