This paper focuses on Entailment as Few-shot Learner (EFL), which is a few-shot learning (FSL) model proposed by the academic community for natural language processing (NPL) tasks. Compared with other classic FSL methods, this model generally has better performance in different NPL tasks. This paper conducted research and analysis on EFL, and found that the addition of natural language token to sentence have an impact on the final performance of the model, and found a relatively better way to add token. Due to the excellent performance of EFL model in the field of FSL, this paper applies it to the problem of few-shot one-class classification and finds that the model also has a good performance in this field.