Method for optimizing large model in reading field based on fine tuning and method for generating reading accompanying manuscript

The invention discloses a reading field large model optimization method and a reading accompanying manuscript generation method based on fine tuning, and the method comprises the steps: obtaining a manuscript generation demand of a reading accompanying manuscript and a local corpus, the manuscript generation demand being used for representing a requirement for generating the reading accompanying manuscript, and the local corpus being used for generating the reading accompanying manuscript; the local corpus is constructed based on reading accompanying text data in the field of children language; according to the manuscript generation requirement, parameters to be fine-tuned are determined from the large language model, and the parameters to be fine-tuned are partial parameters suitable for the manuscript generation requirement in the large language model; according to the training corpus and the to-be-fine-tuned parameters in the local corpus, the large language model is fine-tuned and optimized, and a pre-tra