Google and researchers from Google DeepMind and Stanford University have introduced a new method called Search-Augmented Factuality Evaluator (SAFE) to evaluate the factuality of large language models (LLMs). The method combines LLMs as annotators and search engines to create a benchmark for assessing long-form factuality. The research shows that LLM agents can achieve superhuman rating performance in fact checking and are more cost-effective than human annotators. The study introduces a new dataset, evaluation method, and aggregation metric for assessing the factuality of LLMs like Gemini, GPT, and Claude.