Comment by monsieurbanana
Comment by monsieurbanana 4 days ago
In which case the model couldn't possibly know that the number was correct.
Comment by monsieurbanana 4 days ago
In which case the model couldn't possibly know that the number was correct.
Claude has some awareness of its CoT. As an experiment, it's easy, for example, to ask Claude to "think of a city, but only reply with the word 'ready' and next to ask "what is the first letter of the city you thought of?"
Oops! I tried a couple experiments after writing this, and I believe I was mistaken, though I don't know how. It appears Claude was simply playing along, and convinced me it could remember the choices it secretly made. I must either have given it a tell, or perhaps it guessed the same answers twice in a row.
I'm also confused by that, but it could just be the model being agreeable. I've seen multiple examples posted online though where it's fairly clear that the COT output is not included in subsequent turns. I don't believe Anthropic is public about it (could be wrong), but I know that the Qwen team specifically recommend against including COT tokensfrom previous inferences.