VideoComp: Advancing Fine-Grained Compositional and Temporal Alignment in Video-Text Models | Read Paper on Bytez