The paper introduces QA-ViT, a Question Aware Vision Transformer designed to enhance multimodal reasoning by integrating question awareness directly into the vision encoder. This approach addresses the limitation of existing vision-language models, which often fail to align visual features with user queries, leading to suboptimal performance in tasks requiring nuanced image understanding. QA-ViT is model-agnostic and can be integrated into various vision-language architectures, demonstrating its versatility and effectiveness through extensive experiments. The method improves performance on diverse benchmarks, including visual question answering (VQA) and image captioning, across different model sizes and architectures. The paper also provides detailed ablation studies and qualitative results to support the effectiveness of QA-ViT, highlighting its ability to focus visual attention on relevant image aspects based on the provided text.The paper introduces QA-ViT, a Question Aware Vision Transformer designed to enhance multimodal reasoning by integrating question awareness directly into the vision encoder. This approach addresses the limitation of existing vision-language models, which often fail to align visual features with user queries, leading to suboptimal performance in tasks requiring nuanced image understanding. QA-ViT is model-agnostic and can be integrated into various vision-language architectures, demonstrating its versatility and effectiveness through extensive experiments. The method improves performance on diverse benchmarks, including visual question answering (VQA) and image captioning, across different model sizes and architectures. The paper also provides detailed ablation studies and qualitative results to support the effectiveness of QA-ViT, highlighting its ability to focus visual attention on relevant image aspects based on the provided text.