If there's an architecture that could be 10% better, I think it's highly unlikely that we would conduct ablation experiments, since it's too expensive, It's re

If there's an architecture that could be 10% better, I think it's highly unlikely that we would conduct ablation experiments, since it's too expensive, It's really a pity compared to CV.