What's with these OpenAI charts?

Hallucination & Deception Rate Analysis - GTP5 scored 50 in deception rate, while 03 scored 47.4%, but the bar chart visually misrepresented the data, making GTP5 appear better when it's actually worse [1] - The report lacks clarity on what constitutes a good deception rate, ideally aiming for zero, especially when the model is used for medical advice [2][3] - Chart inconsistencies include boxes of different sizes representing percentages that don't align with their visual representation, such as 69.1%, 30.8%, and 52% [2] Data Presentation Issues - The charts exhibit a lack of consistency, with the highest numbers not always positioned in the highest spot [1] - Random coloring and inconsistent sizing of elements within the charts further contribute to the confusion [2]