OpenAI 可能会使用 API 响应数据进行训练。
1 分•作者: kissgyorgy•大约 24 小时前
官方文档中，他们首席科学家的这段话颇为可疑：
<p><pre><code> 隐藏的思维链条让我们能够“读取”模型的思维，并理解其思考过程。例如，未来我们可能希望监控思维链条，以发现操纵用户的迹象。
</code></pre>
如果他们不以此进行训练，那他们肯定是在读取推理标记。
https://developers.openai.com/blog/responses-api
查看原文
The quote from their Chief Scientist in the official documentation is quite suspicious:<p><pre><code> the hidden chain of thought allows us to “read the mind” of the model and understand its thought process. For example, in the future we may wish to monitor the chain of thought for signs of manipulating the user.
</code></pre>
If they don't train on it, they are definitely reading the reasoning tokens.<p>https://developers.openai.com/blog/responses-api