[99112000]

TL;DR BuzzFeed man performs statistical analysis on LLM output, in an attempt to determine hidden internally encoded motives.

Next up, for a more clickbaity titel, BuzzFeed man pretends to be therapist to uncover LLM's dark secret.

I only wrote this snarky comment because 90% of the authors job is to evaluate the effectiveness of their clickbaity titles, or am I wrong?

[minimaxir]

Yes, you're wrong.

[99112000]

Breaking News, BuzzFeed man can take a joke amd fires back.

I appreciate defining a clear hypothesis and the exploring an LLM using statistics. I feel like the analysis could benefit from prompts that contain neutral consequenses as well. You have given it clear positive rewards, clear negative ones and no reward. Neutral consequences may be a better baseline than no reward.