Mixing LLM Families in Debate Cuts Harmful Answer-Flipping by 60% — But an Adversarial Peer Erases the Gain Instantly

Mixing LLM Families in Debate Cuts Harmful Answer-Flipping by 60% — But an Adversarial Peer Erases the Gain Instantly | BedrockNews