Abstract: Zero-pronoun resolution identifies omitted pronouns and their referents, while non-zero coreference resolution clusters and merges expressions that refer to the same entity. This study proposes a continuous knowledge distillation method for jointly handling zero-pronoun resolution and non-zero coreference resolution. The method uses batch-based continuous distillation, where knowledge is passed from the public model to the local model. An end-to-end neural model captures token relationships while preserving the original sequence. This approach leverages the mutually exclusive relationship between zero pronouns and mentions, resulting in more accurate coreference predictions. Experimental results show that the proposed method outperforms baseline models in multiple metrics.