Fetching the paper…

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