At this rate, we will have a paper about every single psychological aspect discovered in LLMs. This could have been just a reddit post.
Every phenomenon found massively in training set will eventually pop up in LLMs. I just don't find the discoveries made in these papers very meaningful.
Edit: May be I am being too short sighted. The researchers probably start from "Humans are good at X and the training data had many examples of X. How good is LLM at X?" and X happens to be deception this time.
Every phenomenon found massively in training set will eventually pop up in LLMs. I just don't find the discoveries made in these papers very meaningful.
Edit: May be I am being too short sighted. The researchers probably start from "Humans are good at X and the training data had many examples of X. How good is LLM at X?" and X happens to be deception this time.