MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining

LLM-Core Xiaomi, :, Bingquan Xia +62

We present MiMo-7B, a large language model born for reasoning tasks, with optimization across both pre-training and post-training stages. During pre-training, we enhance the data p…