trends.barnas.net

method · first seen 20 Oct 2026

differentiable world model

A learned dynamics model that is differentiable so policies can be trained with analytic gradients through imagined rollouts.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

humanoid whole-body controlbackpropagation through timeimagination rolloutKoopman operator learning