paper_type.json (263B)
1 { 2 "paper_type": "benchmark-creation", 3 "reason": "Introduces GPQA, a new graduate-level multiple-choice benchmark dataset; experiments with experts and models validate the benchmark's difficulty and utility rather than constituting the primary contribution." 4 }