Information-Theoretical Analysis of a Transformer-Based Generative AI Model
Large Language models have shown a remarkable ability to “converse” with humans in a natural language across myriad topics. Despite the proliferation of these models, a deep understanding of how they work under the hood remains elusive. The core of these Generative AI models is composed of layers of...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-05-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/27/6/589 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849432610598551552 |
|---|---|
| author | Manas Deb Tokunbo Ogunfunmi |
| author_facet | Manas Deb Tokunbo Ogunfunmi |
| author_sort | Manas Deb |
| collection | DOAJ |
| description | Large Language models have shown a remarkable ability to “converse” with humans in a natural language across myriad topics. Despite the proliferation of these models, a deep understanding of how they work under the hood remains elusive. The core of these Generative AI models is composed of layers of neural networks that employ the Transformer architecture. This architecture learns from large amounts of training data and creates new content in response to user input. In this study, we analyze the internals of the Transformer using Information Theory. To quantify the amount of information passing through a layer, we view it as an information transmission channel and compute the capacity of the channel. The highlight of our study is that, using Information-Theoretical tools, we develop techniques to visualize on an Information plane how the Transformer encodes the relationship between words in sentences while these words are projected into a high-dimensional vector space. We use Information Geometry to analyze the high-dimensional vectors in the Transformer layer and infer relationships between words based on the length of the geodesic connecting these vector distributions on a Riemannian manifold. Our tools reveal more information about these relationships than attention scores. In this study, we also show how Information-Theoretic analysis can help in troubleshooting learning problems in the Transformer layers. |
| format | Article |
| id | doaj-art-f36e3e6642254eb3a6ef73e69eeedb31 |
| institution | Kabale University |
| issn | 1099-4300 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj-art-f36e3e6642254eb3a6ef73e69eeedb312025-08-20T03:27:18ZengMDPI AGEntropy1099-43002025-05-0127658910.3390/e27060589Information-Theoretical Analysis of a Transformer-Based Generative AI ModelManas Deb0Tokunbo Ogunfunmi1Department of Electrical and Computer Engineering, Santa Clara University, Santa Clara, CA 95053, USADepartment of Electrical and Computer Engineering, Santa Clara University, Santa Clara, CA 95053, USALarge Language models have shown a remarkable ability to “converse” with humans in a natural language across myriad topics. Despite the proliferation of these models, a deep understanding of how they work under the hood remains elusive. The core of these Generative AI models is composed of layers of neural networks that employ the Transformer architecture. This architecture learns from large amounts of training data and creates new content in response to user input. In this study, we analyze the internals of the Transformer using Information Theory. To quantify the amount of information passing through a layer, we view it as an information transmission channel and compute the capacity of the channel. The highlight of our study is that, using Information-Theoretical tools, we develop techniques to visualize on an Information plane how the Transformer encodes the relationship between words in sentences while these words are projected into a high-dimensional vector space. We use Information Geometry to analyze the high-dimensional vectors in the Transformer layer and infer relationships between words based on the length of the geodesic connecting these vector distributions on a Riemannian manifold. Our tools reveal more information about these relationships than attention scores. In this study, we also show how Information-Theoretic analysis can help in troubleshooting learning problems in the Transformer layers.https://www.mdpi.com/1099-4300/27/6/589machine learninggenerative AItransformerinformation theorymutual information estimationinformation geometry |
| spellingShingle | Manas Deb Tokunbo Ogunfunmi Information-Theoretical Analysis of a Transformer-Based Generative AI Model Entropy machine learning generative AI transformer information theory mutual information estimation information geometry |
| title | Information-Theoretical Analysis of a Transformer-Based Generative AI Model |
| title_full | Information-Theoretical Analysis of a Transformer-Based Generative AI Model |
| title_fullStr | Information-Theoretical Analysis of a Transformer-Based Generative AI Model |
| title_full_unstemmed | Information-Theoretical Analysis of a Transformer-Based Generative AI Model |
| title_short | Information-Theoretical Analysis of a Transformer-Based Generative AI Model |
| title_sort | information theoretical analysis of a transformer based generative ai model |
| topic | machine learning generative AI transformer information theory mutual information estimation information geometry |
| url | https://www.mdpi.com/1099-4300/27/6/589 |
| work_keys_str_mv | AT manasdeb informationtheoreticalanalysisofatransformerbasedgenerativeaimodel AT tokunboogunfunmi informationtheoreticalanalysisofatransformerbasedgenerativeaimodel |