This paper introduces TalkFunny, a large-scale Chinese explainable humor response dataset designed to enhance the humorous response generation capabilities of pre-trained language models (PLMs). The dataset includes 4,116 context-response pairs with chain-of-humor annotations and humor mind maps, providing a comprehensive understanding of how humorous responses are generated. The authors also design auxiliary tasks, such as humor sentiment-style classification and humor rewriting, to further improve PLMs' humor response performance. Extensive evaluations using both automatic and manual metrics demonstrate that the proposed dataset and auxiliary tasks effectively enhance the quality of humorous responses generated by PLMs. The study highlights the importance of rich knowledge and commonsense in humor generation and lays the groundwork for future research in this area.This paper introduces TalkFunny, a large-scale Chinese explainable humor response dataset designed to enhance the humorous response generation capabilities of pre-trained language models (PLMs). The dataset includes 4,116 context-response pairs with chain-of-humor annotations and humor mind maps, providing a comprehensive understanding of how humorous responses are generated. The authors also design auxiliary tasks, such as humor sentiment-style classification and humor rewriting, to further improve PLMs' humor response performance. Extensive evaluations using both automatic and manual metrics demonstrate that the proposed dataset and auxiliary tasks effectively enhance the quality of humorous responses generated by PLMs. The study highlights the importance of rich knowledge and commonsense in humor generation and lays the groundwork for future research in this area.