Fused View-Time Attention and Feedforward Reconstruction for 4D Scene Generation | Read Paper on Bytez