AI Researchers Launch Humanity's Last Exam Benchmark

Researchers launched 'Humanity’s Last Exam' to benchmark advanced artificial intelligence capabilities ^[1][3]. This rigorous assessment, published in the journal Nature, features 2,500 expert-level questions ^[1]. Even the most advanced models currently struggle with these high-level academic challenges ^[1][2]. Texas A&M University researchers noted that the benchmark targets abstract reasoning and specialized knowledge ^[2]. The study provides a clearer picture of how machines compare to human expertise ^[3]. Experts believe these findings are essential for tracking the development and safety of frontier models ^[1].