灾难性遗忘（Catastrophic Forgetting）指大模型在微调过程中过度适应新任务数据，导致预训练阶段学到的通用知识严重退化的问题。解决方案需结合数据策略、训练策略和模型架构优化三方面： 数据策略：混合通用与领域数据，控制数据比例；动态重采样。

+ +## 动态 +- 2023/05/27 [CPM-Bee](https://github.com/OpenBMB/CPM-Bee) 发布了！ +- 2023/04/12 CPM-Ant 可以在[HuggingFace Transformers](https://huggingface.co ...

Haibin Ling, a leading AI researcher who designed the core algorithm behind the world's first mobile plant-identification app, has left his U.S. professorship to join Westlake University, one of China ...