中文
Libraries & frameworks

opd

thunlp/opd

Research code companion for a paper on on-policy distillation of large language models, implementing training recipes for studying distillation phenomena, mechanisms, and practical configurations.