2. The "Adversarial Interrogation" "Now argue against your previous answer. What are the 3 strongest counterarguments?" Models are overconfident by default. This forces intellectual honesty.

→ View original post on X — @godofprompt