How we use LLM(GigaChat) for data markup

  • 20 min

We'd like to present the results of using GigaChat's LLM for data markup tasks (dialogues between a virtual assistant and a user) to cheaply increase the amount of markup with a slight decrease in its quality. We will tell how we conducted experiments on connecting LLM to the markup process, to the tasks of binary and multiclass classification of texts, considered the context of the dialogue, added new classes to the model and what results compared to human markup we managed to achieve. 

Comments ({{Comments.length}} )
  • {{comment.AuthorFullName}}
    {{ comment.DateCreated | date: 'dd.MM.yyyy' }}

To leave a feedback you need to

Chat with us, we are online!