# Gradients should stay on path: better estimators of the reverse- and forward KL divergence for normalizing flows

> Research article (Machine Learning Science and Technology, 2022) · cited 14× · AI/ML

**Wikidata**: [openalex:W4296907727](https://www.wikidata.org/wiki/openalex:W4296907727)  
**Source**: https://4ort.xyz/entity/gradients-should-stay-on-path-better-estimators-of-the-reverse-and-forward-kl-divergence-for-normalizing-flows