A compact open-source recipe trains action-conditioned streaming video world models from scratch on one 8-GPU server, using a block-causal DiT with Diffusion Forcing style noise schedules and a rolling KV cache.
Dreamdojo: A generalist robot world model from large-scale human videos,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
MiniWorld: Democratizing the Training of Video World Models from Scratch
A compact open-source recipe trains action-conditioned streaming video world models from scratch on one 8-GPU server, using a block-causal DiT with Diffusion Forcing style noise schedules and a rolling KV cache.