Discover how coarse-to-fine ordered tokens improve test-time search efficiency and scalability in autoregressive generative models for image synthesis.
Discover how instance-adaptive parametrization improves variational autoencoders, reducing the amortization gap for better inference accuracy and efficienc...