Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM
Large language models (LLMs) like those that power OpenAI’s ChatGPT and Google’s Gemini have played a major part in the recent wave of machine learning and artificial intelligence advancements. However, interpreting LLMs and visualizing their components is extremely difficult due to the incredible s...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-12-01
|
| Series: | Graphical Models |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1524070324000262 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846138587307835392 |
|---|---|
| author | Ollie Woodman Zhen Wen Hui Lu Yiwen Ren Minfeng Zhu Wei Chen |
| author_facet | Ollie Woodman Zhen Wen Hui Lu Yiwen Ren Minfeng Zhu Wei Chen |
| author_sort | Ollie Woodman |
| collection | DOAJ |
| description | Large language models (LLMs) like those that power OpenAI’s ChatGPT and Google’s Gemini have played a major part in the recent wave of machine learning and artificial intelligence advancements. However, interpreting LLMs and visualizing their components is extremely difficult due to the incredible scale and high dimensionality of model data. NeuronautLLM introduces a visual analysis system for identifying and visualizing influential neurons in transformer-based language models as they relate to user-defined prompts. Our approach combines simple, yet information-dense visualizations as well as neuron explanation and classification data to provide a wealth of opportunities for exploration. NeuronautLLM was reviewed by two experts to verify its efficacy as a tool for practical model interpretation. Interviews and usability tests with five LLM experts demonstrated NeuronautLLM’s exceptional usability and its readiness for real-world application. Furthermore, two in-depth case studies on model reasoning and social bias highlight NeuronautLLM’s versatility in aiding the analysis of a wide range of LLM research problems. |
| format | Article |
| id | doaj-art-2c3f332f14a44681a410af153fa8cb67 |
| institution | Kabale University |
| issn | 1524-0703 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Graphical Models |
| spelling | doaj-art-2c3f332f14a44681a410af153fa8cb672024-12-07T08:25:15ZengElsevierGraphical Models1524-07032024-12-01136101238Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLMOllie Woodman0Zhen Wen1Hui Lu2Yiwen Ren3Minfeng Zhu4Wei Chen5State Key Lab of CAD&CG, Zhejiang University, ChinaState Key Lab of CAD&CG, Zhejiang University, ChinaZhejiang University, Hangzhou, 310000, ChinaZhejiang University, Hangzhou, 310000, ChinaZhejiang University, Hangzhou, 310000, China; Corresponding author.State Key Lab of CAD&CG, Zhejiang University, ChinaLarge language models (LLMs) like those that power OpenAI’s ChatGPT and Google’s Gemini have played a major part in the recent wave of machine learning and artificial intelligence advancements. However, interpreting LLMs and visualizing their components is extremely difficult due to the incredible scale and high dimensionality of model data. NeuronautLLM introduces a visual analysis system for identifying and visualizing influential neurons in transformer-based language models as they relate to user-defined prompts. Our approach combines simple, yet information-dense visualizations as well as neuron explanation and classification data to provide a wealth of opportunities for exploration. NeuronautLLM was reviewed by two experts to verify its efficacy as a tool for practical model interpretation. Interviews and usability tests with five LLM experts demonstrated NeuronautLLM’s exceptional usability and its readiness for real-world application. Furthermore, two in-depth case studies on model reasoning and social bias highlight NeuronautLLM’s versatility in aiding the analysis of a wide range of LLM research problems.http://www.sciencedirect.com/science/article/pii/S1524070324000262Artificial intelligenceLarge language modelsMechanistic interpretabilityData visualization |
| spellingShingle | Ollie Woodman Zhen Wen Hui Lu Yiwen Ren Minfeng Zhu Wei Chen Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM Graphical Models Artificial intelligence Large language models Mechanistic interpretability Data visualization |
| title | Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM |
| title_full | Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM |
| title_fullStr | Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM |
| title_full_unstemmed | Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM |
| title_short | Exploring the neural landscape: Visual analytics of neuron activation in large language models with NeuronautLLM |
| title_sort | exploring the neural landscape visual analytics of neuron activation in large language models with neuronautllm |
| topic | Artificial intelligence Large language models Mechanistic interpretability Data visualization |
| url | http://www.sciencedirect.com/science/article/pii/S1524070324000262 |
| work_keys_str_mv | AT olliewoodman exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm AT zhenwen exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm AT huilu exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm AT yiwenren exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm AT minfengzhu exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm AT weichen exploringtheneurallandscapevisualanalyticsofneuronactivationinlargelanguagemodelswithneuronautllm |