the gears to ascension comments on Large Language Models can Strategically Deceive their Users when Put Under Pressure.

[the gears to ascension]

it could be, but it smells of RLHF to my brain.