LLM judges prefer their own responses 70% more often than humans do

This story was filed as a headline only — the news service holds no English full text for it. Read the original at X: Arena (@arena) →