TRINS: Towards Multimodal Language Models that Can Read | Read Paper on Bytez