[Sharlin]

The gap between LLM and human cases was greater in the deceptive case. This may, of course, simply reflect the fact that random humans are bad at lying.