zjunlp/KnowLM

Supports knowledge-augmented inference through RAG integration and hallucination detection modules, alongside dedicated information extraction capabilities via specialized datasets (IEPile, InstructIE) and models. The framework spans the full LLM lifecycle—pre-training, instruction fine-tuning, knowledge editing via EasyEdit, and deployment—with ready-to-use variants like ZhiXi and OneKE optimized for dialogue and extraction tasks across Chinese and English.

1,376 stars. No commits in the last 6 months.