Sparsity Meets Similarity: Leveraging Long-Tail Distribution for Dynamic Optimized Token Representation in Multimodal Large Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910822630948864 |
|---|---|
| author | Yu, Gaotong Chen, Yi Xu, Jian |
| author_facet | Yu, Gaotong Chen, Yi Xu, Jian |
| contents | Recently, multimodal large language models (MM-LLMs) have achieved significant success in various tasks, but their high computational costs limit widespread application. The main computational burden arises from processing concatenated text and visual tokens in the LLM layer, where input token length directly affects efficiency. Our analysis of visual tokens reveals that their similarity to the CLS token follows a long-tail distribution, with only a few showing high similarity. To address this, we propose a dynamic pruning algorithm that identifies the inflection point in the visual CLS token similarity curve, enabling effective trimming of visual markers to accelerate model performance. Additionally, we perform a second round of pruning in the LLM layer, filtering out low-correlation tokens through the interaction between visual and textual features. Experimental results demonstrate that our method achieves performance comparable to the original while utilizing only 22% of the original token quantity. Our source code will be made publicly available upon acceptance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_01162 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Sparsity Meets Similarity: Leveraging Long-Tail Distribution for Dynamic Optimized Token Representation in Multimodal Large Language Models Yu, Gaotong Chen, Yi Xu, Jian Computer Vision and Pattern Recognition Recently, multimodal large language models (MM-LLMs) have achieved significant success in various tasks, but their high computational costs limit widespread application. The main computational burden arises from processing concatenated text and visual tokens in the LLM layer, where input token length directly affects efficiency. Our analysis of visual tokens reveals that their similarity to the CLS token follows a long-tail distribution, with only a few showing high similarity. To address this, we propose a dynamic pruning algorithm that identifies the inflection point in the visual CLS token similarity curve, enabling effective trimming of visual markers to accelerate model performance. Additionally, we perform a second round of pruning in the LLM layer, filtering out low-correlation tokens through the interaction between visual and textual features. Experimental results demonstrate that our method achieves performance comparable to the original while utilizing only 22% of the original token quantity. Our source code will be made publicly available upon acceptance. |
| title | Sparsity Meets Similarity: Leveraging Long-Tail Distribution for Dynamic Optimized Token Representation in Multimodal Large Language Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2409.01162 |