GPQA Diamond ：一个困难的智力基准，用于测试化学、物理和生物学方面的专业知识。

CoT（思维链）

• 用户输入问题后，o1 相比之前 GPT 系列模型多使用一个叫做“推理标记”的东西，你可以理解为它学会了像人一样选择在什么时候进行思考，并输出当前的想法，而这些“推理标记”中的“思考”的内容，并不会展示出来。这也是为什么有体验者提到，模型的等待时间比较久，正是因为思考的过程并不会显式地展示；
• 在新一轮的对话中（用户第二次输入），上一轮“思考”的内容全部被删除，开始全新的“思考”；
• 依此类推，当对话到达128k Tokens的上限时，模型会给出一个“删减版”答案，也避免了我们白白等待却碰到了上下文上限。

Life can only be understood backward, but it must be lived forward - Søren Kierkegaard

(Quiet-STaR 在论文的 Abstract 引用了这句话，当时觉得挺有意境的)

官方演示视频

总结（o1 很强，但不要尬吹）

According to these evaluations, o1-preview hallucinates less frequently than GPT-4o, and o1-mini hallucinates less frequently than GPT-4o-mini. However, we have received anecdotal feedback that o1-preview and o1-mini tend to hallucinate more than GPT-4o and GPT-4o-mini. More work is needed to understand hallucinations holistically, particularly in domains not covered by our evaluations (e.g., chemistry). Additionally, red teamers have noted that o1-preview is more convincing in certain domains than GPT-4o given that it generates more detailed answers. This potentially increases the risk of people trusting and relying more on hallucinated generation.