less than 1 minute read

Authors: Y. Lu, Q. Sun, X. Wang*, Z. Jiang, H. Zhao, K. He
Status: Submitted to CVPR 2026.
Preprint: arXiv: 2512.10953

Bidirectional Normalizing Flow (BiFlow) explores architectures that strengthen both forward and inverse mappings without sacrificing tractable likelihoods. The model emphasizes stable Jacobian computation and expressive couplings to improve sampling quality and inference accuracy simultaneously. The goal is to make flows more competitive on high-dimensional generative tasks while keeping exact log-likelihood training.

You may also enjoy

Empty Blog

1 minute read

Tags:

Normalizing Flows (NFs) are generative models that keep every layer invertible. They learn a mapping (f: x \rightarrow z) that pushes complex data (x) into a...