Tuan Do

A field-code compiler can turn approximate transport fields into precise value-certified samplers, dramatically enhancing communication efficiency in empirical optimal transport tasks.

WAMs can boost object-level behavior and target selectivity, but their effectiveness hinges on architecture and incurs higher inference costs.

Forget shallow embedding trajectories and computationally heavy MLLMs: EA-Swin's factorized windowed attention on pretrained video embeddings leaps ahead in detecting AI-generated videos, achieving 5-20% accuracy gains.