Opens in a new window
Improvements or additions to documentation
，更多细节参见新收录的资料
蒸馏是模仿，学强模型的输出，把它的「答案形状」复制过来；RL 是探索，模型必须大量自己推理、自己生成、在错误里反复迭代，从试错中提炼能力。
Monaco hit a low at the Bernabéu at the end of January. Their 6-1 defeat to Real Madrid was their heaviest in European competition and followed a run of seven defeats in eight games in Ligue 1, the worst record in the club’s history. After their humbling defeat in Madrid, the squad remained in the city until the afternoon of the following day to come to terms with the deepening crisis. The club’s coaches and staff held a meeting to talk things through. The players also gathered to thrash things out. “We thought it was important to have one as players, to be open, to try to find solutions,” said Folarin Balogun. “It was positive.”