A Multi-hop Problem Generation Model Fine-tuned Based on the Open-source Large Language Model PiSSA
With the rapid development of machine learning technology,the importance of"problem-solving"tasks has become increasingly prominent.In order to train a good problem-solving model,a large number of questions described in natural language and corresponding answer labels are required.In order to improve the reasoning ability of the model,the problem must also be related to multiple pieces of information in the context.This article proposes an efficient multi hop problem generation scheme using a large language model combined with PiSSA fine-tuning techniques.BERTScore and BLEU metrics are used for evaluation,and the proposed model demonstrates significant advantages over traditional models in multi hop problem generation tasks.
PiSSA fine-tuningLoRA fine-tuningmulti-hop problem generation
万方数据
维普
