Identifying Multi-modal Knowledge Neurons in Pretrained Transformers via Two-stage Filtering
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913765780357120 |
|---|---|
| author | Sato, Yugen Takagi, Tomohiro |
| author_facet | Sato, Yugen Takagi, Tomohiro |
| contents | Recent advances in large language models (LLMs) have led to the development of multimodal LLMs (MLLMs) in the fields of natural language processing (NLP) and computer vision. Although these models allow for integrated visual and language understanding, they present challenges such as opaque internal processing and the generation of hallucinations and misinformation. Therefore, there is a need for a method to clarify the location of knowledge in MLLMs.
In this study, we propose a method to identify neurons associated with specific knowledge using MiniGPT-4, a Transformer-based MLLM. Specifically, we extract knowledge neurons through two stages: activation differences filtering using inpainting and gradient-based filtering using GradCAM. Experiments on the image caption generation task using the MS COCO 2017 dataset, BLEU, ROUGE, and BERTScore quantitative evaluation, and qualitative evaluation using an activation heatmap showed that our method is able to locate knowledge with higher accuracy than existing methods.
This study contributes to the visualization and explainability of knowledge in MLLMs and shows the potential for future knowledge editing and control. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_22941 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Identifying Multi-modal Knowledge Neurons in Pretrained Transformers via Two-stage Filtering Sato, Yugen Takagi, Tomohiro Artificial Intelligence Machine Learning Multimedia Recent advances in large language models (LLMs) have led to the development of multimodal LLMs (MLLMs) in the fields of natural language processing (NLP) and computer vision. Although these models allow for integrated visual and language understanding, they present challenges such as opaque internal processing and the generation of hallucinations and misinformation. Therefore, there is a need for a method to clarify the location of knowledge in MLLMs. In this study, we propose a method to identify neurons associated with specific knowledge using MiniGPT-4, a Transformer-based MLLM. Specifically, we extract knowledge neurons through two stages: activation differences filtering using inpainting and gradient-based filtering using GradCAM. Experiments on the image caption generation task using the MS COCO 2017 dataset, BLEU, ROUGE, and BERTScore quantitative evaluation, and qualitative evaluation using an activation heatmap showed that our method is able to locate knowledge with higher accuracy than existing methods. This study contributes to the visualization and explainability of knowledge in MLLMs and shows the potential for future knowledge editing and control. |
| title | Identifying Multi-modal Knowledge Neurons in Pretrained Transformers via Two-stage Filtering |
| topic | Artificial Intelligence Machine Learning Multimedia |
| url | https://arxiv.org/abs/2503.22941 |