Mirai injects future-token foresight into autoregressive visual generators, accelerating convergence up to 10x and cutting ImageNet FID from 5.34 to 4.34.
Masked autoencoders are scalable vision learners
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Mirai: Autoregressive Visual Generation Needs Foresight
Mirai injects future-token foresight into autoregressive visual generators, accelerating convergence up to 10x and cutting ImageNet FID from 5.34 to 4.34.