@ma_nanye
phd @nyu_courant | prev intern @Google @nvidia
AI
A user says the work avoids MeanFlow’s unstable JVP and DMD’s heavy memory overhead.