#pre-training

Xiaomi's MiMo-7B is a compact language model that surpasses larger models in math and code reasoning through advanced pre-training and reinforcement learning strategies.