Survey on Enhancing Dialogue Agent Alignment via MiniLLM with Targeted Human Assessments

Bhojraj Narware*
Periodicity:January - June'2025

Abstract

This paper presents the development of a compact and effective language model inspired by the LLaMA architecture. Our focus was on constructing this model based on the fundamental principles of LLaMA, which have influenced our architectural decisions and training methods. We sought to explore innovative research avenues and expand the possibilities achievable with limited resources. By leveraging open-source datasets and sophisticated training techniques, we made notable advancements without relying on extensive computational power or proprietary data. Nevertheless, due to resource limitations, the model work in progress. Researchers with access to more substantial computational capabilities could build on this foundation to enhance the model’s performance. We hope this paper encourages others in the field to contribute to the development of more robust language models that are accessible to all. Key Parameters for training include context window size, number of layers, batch size, and model dimensions. Results are evaluated based on epoch count, execution duration, model parameters, and validation loss.

Keywords

BERT, Machine Learning, LLM, NLP, Deep Learning, NN, LM, ToT, AI

How to Cite this Article?

References

If you have access to this article please login to view the article or kindly login to purchase the article

Purchase Instant Access

Single Article

North Americas,UK,
Middle East,Europe
India Rest of world
USD EUR INR USD-ROW
Pdf 40 40 300
Online 15 15 300
Pdf & Online 40 40 300

Options for accessing this content:
  • If you would like institutional access to this content, please recommend the title to your librarian.
    Library Recommendation Form
  • If you already have i-manager's user account: Login above and proceed to purchase the article.
  • New Users: Please register, then proceed to purchase the article.