#1-shot reinforcement learning

Researchers reveal that training large language models with just one example using 1-shot reinforcement learning significantly enhances their math reasoning abilities, matching results from large datasets.