KomeijiForce/Cuckoo_Instruct_TuluV3

Cuckoo是一个模仿大型语言模型下一个标记预测范式的小型信息抽取模型，能够利用任何文本资源进行自我增强，特别是免费利用为LLM准备的数据集。README中提供了Cuckoo模型的多个预训练版本，以及在不同任务中的性能表现和微调示例。Cuckoo is a small-scale information extraction model that mimics the next-token prediction paradigm of large language models, capable of self-enhancement using any text resources, especially free-riding on datasets prepared for LLMs. The README provides multiple pre-trained versions of the Cuckoo model, along with their performance on different tasks and examples of fine-tuning.