🧵 4. Feedback highlights o1's tendency to "hallucinate" and not admit lack of answers as often as GPT-4o, emphasizing limitations of the new model.