Hugging Face Daily Papers·· 2026-10-16AI score24
PPMem: enhancing video world models with persistent planar memory
Persistent Planar Memory for Video World Models
AI briefing
PPMem is a video world model that uses planar primitives as persistent 3D memory. By adding a plane prediction head and a Depth Token Fusion (DTF) module to the video diffusion transformer, each autoregressive chunk generates both RGB video and planar geometry.
Open original arxiv.org