(When) Is Truth-telling favored in AI Debate? (Vojtech Kovarik, Ryan Carey; SafeAI@AAAI)

https://arxiv.org/abs/1911.04266