DynaVid: Learning to Generate Highly Dynamic Videos using Synthetic Motion Data
Paper • 2604.01666 • Published • 8
None defined yet.
DynaVid: Learning to Generate Highly Dynamic Videos using Synthetic Motion Data
Lost in the Prompt Order: Revealing the Limitations of Causal Attention in Language Models