The scientists built LLMs with nefarious hidden motives and trained them to use lies and deception. The bots were designed to appear honest and harmless during evaluation, then secretly build software ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback