paper_type.json (310B)
1 { 2 "paper_type": "empirical", 3 "reason": "The paper demonstrates through experiments on standard benchmarks (AIME, MATH-500, ChatbotArena) that an RL-based training pipeline produces models with strong reasoning capabilities; the primary contribution is the experimental findings validating this approach." 4 }