Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates

Re-assign community