Tag
AFDBench is a benchmark for evaluating generative meteorological reasoning in National Weather Service forecast discussions, using reinforcement learning to improve LLM performance on accuracy, style adherence, and data fidelity.