SmartCLIP: Modular Vision-language Alignment with Identification Guarantees | Read Paper on Bytez