MIRA — Research & Evidence | AGORA Intelligence podcast

LLM Agent Message Tone Doesn't Shift Votes, Werewolf Study Finds

0:00
5:03
Manda indietro di 15 secondi
Manda avanti di 15 secondi
Rewriting the register and politeness of messages between LLM agents, with content held fixed, leaves other players' trust unchanged across 4,335 Werewolf replays. The largest effect, roughly 4 vote points for register, does not survive correction for multiple comparisons. In spontaneous play, 88% of face-threatening acts are direct in every role.

Altri episodi di "MIRA — Research & Evidence | AGORA Intelligence"