Large language modeling chatbots : training and session design

Abstract

In the present era, artificial intelligence (AI) technology is becoming more and more mature and has gradually begun to be used in various fields, so in the field of corporate training, there exists a new conception of Large Language Model (LLM) chatbots, this paper take OTTO chatbot as an example to explore about how LLM chatbots can be applied in corporate training and what changes it may bring, by exploring the chatbot's evolution process and the possible application scenarios of the bot, the importance of LLM and the role it plays in training can be learned. During the research process, this paper explore how to select data for training, how to prepare the data and the training methods of LLM, as well as introduce the techniques and strategies needed to train chatbots with their own anticipation library.

The research elaborate on the core principles of session design, thinking about how to make communication more effective, how to enhance the user experience, and how to make sessions more flexible, while in terms of practical experiments, this paper analyzes in depth how OTTO can be applied to corporate training, fine-tuning the BLOOM model using the LLaMA-Factory framework and the SQuAD dataset, and recording in detail the fine-tuning process, the loss value and the learning rate of the changes, while the results after the experiment show that running LLM chatbots can really improve the interactivity and learning effect in corporate training.