The IRMA Community
Newsletters
Research IRM
Click a keyword to search titles using our InfoSci-OnDemand powered search:
|
How Do LLMs Work?: A Deep Dive Into Transformer Models
|
|
Author(s): Satvik Vats (Madan Mohan Malaviya University of Technology (MMMUT), Gorakhpur, India), Vikrant Sharma (Graphic Era Hill University, Dehradun, India), Priya Singh (Dataculture Technologies Private Limited, India), Samriti Thakur (Dataculture Technologies Private Limited, India)and Daksh Rawat (Graphic Era Hill University, Dehradun, India)
Copyright: 2026
Pages: 22
Source title:
Digital Watermarking in Cloud Environments For Copyright Protection
Source Author(s)/Editor(s): Ashwani Kumar (School of Computer Science Engineering & Technology, Bennett University, Greater Noida, UP India), Auzuir Ripardo de Alexandria (Instituto Federal de Educação, Ciência e Tecnologia do Ceará (IFCE), Brazil), Satya Prakash Yadav (Department of Computer Science and Engineering, Madan Mohan Malaviya University of Technology, Gorakhpur, India)and Antonino Galletta (University of Messina, Italy)
DOI: 10.4018/979-8-3373-3785-2.ch004
Purchase
|
Abstract
This chapter provides an in-depth overview of the Large Language Models (LLMs) by putting its lineage in the Transformer architecture- a framework that unites parallelism with long-range dependencies modeling via self-attention. It then introduces the most fundamental interfaces: multi-head attention, positional encoding and feed-forward networks, in a systematic fashion, followed by explanation of pre-training paradigms masked language modeling and causal language modeling. It also focuses on hyper-parameters of better tunes, especially reinforcement learning with human feedback (RLHF). Text generation Decoding routines, such as greedy decoding, beam or nucleus sampling, are discussed, and the limitations of the method are highlighted in terms of such aspects as hallucinations, bias, interpretability, and environmental impact of LLMs. The chapter ends by questioning the ethical aspects of developing the LMM and by mapping future research agendas leading to the ethical discovery of these technologies.
Related Content
|
Fatima Ahmed Mohamed Abdalla, Noor Asiah Rashid.
© 2026.
32 pages.
|
|
Fatima Ahmed Mohamed Abdalla, Noor Asiah Rashid.
© 2026.
32 pages.
|
|
Azana Hafizah Mohd Aman, Wan Muhd Hazwan Azamuddin, Maznifah Salam, Zainab S. Attarbashi.
© 2026.
32 pages.
|
|
Azana Hafizah Mohd Aman, Wan Muhd Hazwan Azamuddin, Maznifah Salam, Zainab S. Attarbashi.
© 2026.
36 pages.
|
|
Salaheldin Mohamed Ibrahim Edam.
© 2026.
42 pages.
|
|
Rubi Kadyan, Sunita Rani, Vinod Kr. Saroha.
© 2026.
46 pages.
|
|
Mamoon M. Saeed, Zeinab E. Ahmed, Rania A. Mokhtar, Rashid A. Saeed.
© 2026.
34 pages.
|
|
|