The recently introduced MIRA benchmark represents a novel advancement in the field of visual reasoning by requiring models to generate intermediate visual representations such as sketches and diagrams. This method aligns closely with human problem-solving strategies, where visualization serves as a foundational step in reasoning processes. By integrating this approach, MIRA aims to enhance the capability of models to reason visually, marking a significant progression in artificial intelligence research. The benchmark’s design encourages models to produce a chain of thought that is visually interpretable, thereby improving transparency and potentially the accuracy of reasoning outcomes. According to recent evaluations, MIRA effectively enhances visual reasoning in AI systems, supporting its intended purpose. This development is documented in a study published on arXiv, underscoring its relevance and contribution to ongoing research in computer vision and AI reasoning. Overall, MIRA sets a new standard for evaluating and improving visual reasoning by embedding visualization as an essential step in the reasoning workflow.