Training Language Models to Win Debates with Self-Play Improves Judge Accuracy

ELM
Re-assign community