1 paper
Elad Amrani, Leonid Karlinsky, Alex Bronstein
We introduce XTRA, a vision model pre-trained with a novel auto-regressive objective that significantly enhances both sample and parameter efficiency compared to previous auto-regr…