I run it locally and read the raw thought process, find it very useful (can be ruthless at times) seeing this before it tags on the friendliness.
Then you can see it's planning process to tag on the warmth/friendliness "but the user seems proud of... so I need to acknowledge..."
I don't think Gemini's "thoughts" are the raw CoT process, they're summarized / cleaned up by a small model before returned to you (same as OpenAI models).
That's fascinating. I've been trying to get other models to mimick Gemini 2.5 Pro's thought process, but even with examples, they don't do it very well. Which surprised me, because I think even the original (no RLHF) GPT-3 was pretty good at following formats like that! But maybe there's not enough training data in that format for it to "click".
It does seem similar in structure to Gemini 2.0's output format with the nested bullets though, so I have to assume they trained on synthetic examples.
I run it locally and read the raw thought process, find it very useful (can be ruthless at times) seeing this before it tags on the friendliness.
Then you can see it's planning process to tag on the warmth/friendliness "but the user seems proud of... so I need to acknowledge..."
I don't think Gemini's "thoughts" are the raw CoT process, they're summarized / cleaned up by a small model before returned to you (same as OpenAI models).