Google AI Introduces CoverBench: A Challenging Benchmark Focused on Verifying Language Model LM Outputs in Complex Reasoning Settings