ForgeWM Progressive Causal Training for Video World Models
August 14, 2026
ForgeWM enables low-latency, action-conditioned video generation for interactive world models. It uses causal consistency distillation and teacher-forced causal training to align discrete controls with temporally compressed latent chunks.
HOW THIS AFFECTS YOU
●
builderYou can develop faster, more responsive interactive environments for gaming or simulation.
●
researcherYou can bridge the gap between bidirectional video generation and causal, interactive world models.